1 Aims, objectives and scope

1.1 What this book is about

If a writing system is understood as a particular language written down in a particular script subject to certain rules, this book is on the various writing systems of the Sindhi language. Native to the South Asian region of Sindh — today a province of Pakistan — Sindhi is an Indo-Aryan language spoken by around 33 million people, the vast majority of whom reside in Pakistan and India. Over the last millennium, Sindhi has appeared in various writing systems, some better known than others. This monograph is the first to analyse the various past and present writing systems of Sindhi, both individually and comparatively, using contemporary grapholinguistic research methods.

This monograph forms part of the series Grapholinguistics and its Applications. A previous volume from this series concluded with the following message:

[A] comprehensive evaluation of an entire [writing] system would be an important test for the theory as it likely highlights gaps in the theory or where it requires revision. Such case studies of the naturalness of different […] writing systems that are carried out in the same theoretical framework could, in a further step, also be straightforwardly compared, which would allow attaining a deeper understanding of universal, typological, and system-specific aspects of writing.

(Meletis, 2020, p. 404)

This is where the present work comes in. Harnessing and building on the theoretical approaches outlined in Meletis (2020), this book will describe, analyse and theorise the various writing systems used for the Sindhi language. In doing so, the book will also scrutinise the theoretical approaches themselves, and suggest practicable refinements.

In terms of scope, Meletis’ (2020) work was self-admittedly not “a minute description of any specific writing system, nor a detailed comparison of any two writing systems” (p. 10). Rather, it aimed to be a broad-based ‘horizontal’ treatment of writing systems in general. The present book is something of the opposite in featuring in-depth or ‘vertical’ analyses of specific writing systems and their salient features, followed by explanations or hypotheses for the existence of these features. More importantly, since the present volume describes several writing systems of various types using a common theoretical approach and terminology, robust comparisons of features from two or more of these writing systems will become possible in future analyses. To adopt Meletis’ (2020, p. 391) terms, the holistic approach of this volume will facilitate atomistic approaches towards the same subject matter in future. In this regard, the present monograph is comparable to another from the Grapholinguistics and its Applications series — Honda’s (forthcoming) monograph on Japanese writing systems.

1.2 What this book is not about

Being the first monograph dedicated to multiple writing systems of the Sindhi language, this work necessarily deals with linguistics and related domains. At the same time, there are certain areas of investigation that lie outside the purview of this book. Below is a list of salient domains and subject areas that readers may suspect this book covers, but which actually lie outside its scope.

Sindhi language, literature and society

This book is on the writing systems of the Sindhi language, and not on the Sindhi language itself. To this end, the book’s primary focus is not the phonology, morphology, syntax, lexicon, semantics or pragmatics of Sindhi. Nor does it treat in any detail issues of Sindhi-language education, government policy or politics. While these topics are briefly mentioned in Chapter 4, it is only with the aim of providing a foundation to facilitate a deeper understanding of Sindhi’s writing systems. Similarly, this book does not analyse or critique Sindhi-language literature, nor does it focus on Sindhi society per se. Just as Chapter 4 provides an overview of the language to orient the reader, Chapter 3 provides a brief historical and sociological background of Sindhi society. That said, certain sections of the book, particularly Chapters 5, 13 and 15, have an explicit sociolinguistic bent to them, albeit within the overarching domain of grapholinguistics.

Some readers may instinctively associate the concept of Sindhi’s writing systems with the symbols found on archaeological artefacts excavated from sites of the Indus Valley Civilisation (3300–1100 BC). These symbols have received considerable academic and popular attention since their discovery in the early twentieth century. However, there is as yet no consensus on what the symbols represent, let alone whether they represent the Sindhi language. For this reason, and aside from brief allusions in Sections 3.1 and 5.1, the book does not investigate in any detail the purported meaning and significance of the Indus Valley symbols.

Script pedagogy or advocacy

This book is situated in the domain of grapholinguistics, and is not directly associated with the subject areas of literacy and pedagogy. It describes and analyses various past and present writing systems used for the Sindhi language. It is not a reader, primer or self-instructional guide on ‘How to Read Sindhi’. Rather, what the book does is make clear how the idea of ‘reading Sindhi’ only makes sense when a particular writing system is also specified.

The book also does not advocate for one writing system or the other. Viewed against the ideologically charged and contested debate on Sindhi-language writing systems in post-Partition India (§5.3), it is emphasised that this book only describes the various writing systems (and others) in question. It does not prescribe. That said, the book cannot be understood as ideologically ‘neutral’, since a scholarly work situated in an ideological vacuum may well be an epistemological impossibility. What can be asserted, though, is the absence of an ideological stance favouring one writing system over another.

Typography and calligraphy

Although concerned with the writing systems of the Sindhi language, matters of typography and calligraphy are not central to the focus of this book. Consequently, the graphical and design aspects of the scripts in question are examined only to the extent necessary for a holistic analysis of a particular writing system. Thus, the book analyses the visual differences between unstandardised and standardised Sindhi-in-Khudawadi (Chapter 8), but does not concentrate on the visual differences between the Khudawadi, Khojki and Gurmukhi scripts. The latter type of analysis leans towards the typographical end of the grapholinguistic spectrum, which falls outside the scope of the book.

1.3 Limitations

Content and focus

This book is the first scholarly monograph to focus at once on multiple writing systems of the Sindhi language. Understandably, the reader may be sceptical about the intrinsic coherence of the book’s content, and whether the various themes in this work ‘go together’ in an intuitive manner. The content of the book may also be critiqued as skewing towards description rather than explanation. However, the seemingly eclectic nature of the content, and the surfeit of description, are natural and unavoidable outcomes of the book being a first mover in its discipline. As Meletis (2020, p. 178) puts it, the descriptive ‘what’ or ‘how’ needs to precede the explanatory ‘why’. Moreover, we still lack comprehensive and widely accepted terminology to accurately describe various grapholinguistic phenomena, let alone having at our disposal unambiguous metrics to measure and compare these phenomena. While recent scholarship in grapholinguistics (Gnanadesikan, 2017a; Meletis, 2020; Sproat & Gutkin, 2021) has made great strides in plugging the terminological gap, much conceptual and terminological development remains pending. This gap is acutely felt when having to consistently describing cross-linguistic grapholinguistic phenomena across typologically distinct writing systems, not to mention writing systems that straddle typological boundaries, such as unstandardised Sindhi-in-Khudawadi (Chapter 8). This means that the very task of description has required terminological innovation in order to enable subsequent analysis and explanation.

Nevertheless, and as mentioned in Section 1.1, an initial description of a phenomenon is likely to be broad-based and holistic. Atomistic analyses, whether individualised or comparative, can only be reasonably attempted once the initial groundwork has been laid — which is what this book aims to do. Consequently, the book contains a greater proportion of descriptive content than analytical. That said, acting as a counterweight to the description in Parts One and Two is the analysis in Part Three, which is dedicated solely to explanations and theory based on the descriptions preceding. I hope this book serves as the impetus and inspiration for deeper, more sophisticated graphematic and sociolinguistic analyses of Sindhi’s writing systems, and of writing systems in general.

The focus on Sindhi has also necessitated excluding writing systems where the language might be considered distinct from Sindhi proper. It is for this reason that writing systems of Kutchi — which is grammatically related to Sindhi but sociolinguistically seen as a separate variety — have not been analysed in any detail in this book. The exception to this rule is Kutchi written in the Khojki script (Chapter 10), justified based on historical-sociolinguistic factors.

Epistemological and logistical

The book’s first mover status meant that it was written in an epistemological silo of sorts, especially since works comprising a holistic treatment of a single writing system, such as Honda (forthcoming), were unavailable for comparison and ideation at the time of writing. Due to the epistemological vacuum this book was faced with, it may be critiqued as providing a “fragmentary picture” (Meletis, 2020, p. 403). As described in the previous section, the descriptive and analytical tools required for the job at hand are still a work in progress. Yet, imperfect as this book may be, it should still prove useful.

In terms of focus, my greater first-hand familiarity with the Indian and diasporic Sindhisphere, and my inability thus far to visit Pakistan and engage further with the Pakistani Sindhisphere, may have caused me to unwittingly foreground the Indian dimension of the language and its writing systems. This should not count as a critical shortcoming, though, since the smaller Indian Sindhisphere has been far less researched than the larger Pakistani one. In any event, since the geopolitical boundary between Pakistan and India only came into being in 1947, older data from the nineteenth and early twentieth century remains unaffected.

Further limitations of the book may arise from my own proficiency — or lack thereof — in certain subject areas, and the physical or material inaccessibility of specific sources of information. For instance, I was able to obtain and consult scholarly literature and data on the subject in certain languages and writing systems, including Sindhi, English, Hindi-Urdu, Marathi, Gujarati, German and French. However, my inability in Russian prevented me from taking advantage of the body of Russian-language scholarship on Sindhi. Besides, certain sources of information were not accessible at all, whether in print or digital format. To top it all, much of this book was written during the COVID-19 pandemic, rendering in-person travel to physical archives and data sources onerous or even impossible. I remain an optimist, though, and view the omissions and gaps in this book as a serendipitous reminder that there remains much work to be done, and many more grapholinguistic treasures to be unearthed.

1.4 Structure of the book

The book contains fifteen Chapters organised into three Parts. Part One, comprising the first four Chapters, lays down the conceptual and epistemological foundations for the reminder of the book. Chapter 2 explains the theoretical approaches and establishes the terminology used in the book. Chapters 3 and 4 provide an overview of the Sindh region, the Sindhi people and their history, and the Sindhi language, thereby setting the stage for the remainder of the book. Part Two constitutes the descriptive core of the book, starting off with a chronology of written Sindhi and its various forms in Chapter 5. This is followed by descriptions and evaluations of seven individual writing systems of Sindhi, concluding with a sociolinguistic investigation of attitudes towards the three most prominent systems in present times. In Part Three, Chapters 14 and 15 encapsulate the findings of this book and explore how the lessons offered by Sindhi’s writing systems can help advance the state of the grapholinguistic art.

☙ ❧

2 Theoretical foundations and terminology

Unpacking the core subject matter of Sindhi writing systems requires fit-for-purpose terminology that allows us to describe and analyse the grapholinguistic phenomena in focus. Since grapholinguistics remains an emerging discipline, it is still grappling with conflicting, vague or even a complete lack of terminology to describe certain phenomena. Establishing and setting out the requisite terminology or meta-language to enable clear and unambiguous discussion necessarily entails a slight departure from the Sindhisphere, which is what this Chapter will do.

The terms and definitions described in this chapter are sourced from a variety of academic and industry publications. The industry publications consulted primarily comprise specifications issued by the Unicode Consortium, which is behind the eponymous “standard for digital representation of the characters used in writing all of the world’s languages” (Unicode, 2025a). Also consulted are sources intended for a popular audience but compiled by qualified linguists, such as the ScriptSource website (ScriptSource, 2022a) created and maintained by the Christian missionary organisation SIL International.

2.1 Writing

Probably the most elusive and hard-to-define concept in the present context is that of writing itself. Early monographs on the subject offered up terse, laconic definitions of the term. Diringer (1948, p. 20) conceived of writing as “the graphic counterpart of speech”, in which every graphic element corresponded to one or more specific speech elements. Inherent in this definition were the fundamental assumptions of writing directly representing spoken human language, while also being secondary to it. Gelb (1963, p. 7) repackaged these fundamental assumptions and portrayed writing as “signs [that] become ultimately secondary symbols for notions of linguistic value”. In so saying, Gelb drew a distinction between semasiography or ‘meaning writing’ and phonography or ‘speech writing’ (p. 11), emphasising that only the latter is considered “full writing” (p. 191). Recent scholarship has reiterated Gelb’s notion of writing as ‘speech writing’ (but see Brokaw (2022)), while also refining it further. Coulmas (1991, p. 17) identifies writing in terms of three basic characteristics:

1. consisting of artificial graphical marks on a durable surface;
2. intending to communicate something;
3. achieving its purpose of communication thanks to the marks’ conventional relation to human language.

Subsequently, Coulmas (1996a, p. 555) combines these characteristics and portrays writing as “a system of recording language by means of visible or tactile marks which relate in a systematic way to units of speech”. In the now-classic reference work The World’s Writing Systems, Daniels (1996a, p. 3) echoes Coulmas’ sentiment in defining writing as “a system of more or less permanent marks used to represent an utterance in such a way that it can be recovered more or less exactly without the intervention of the utterer”. By explicitly stating that “writing is bound up with language” and that “pictography is not writing”, Daniels reiterates the notion of writing essentially being ‘speech writing’ and not ‘meaning writing’. Sampson (2015, p. 21 ff.) retains Gelb’s term of semasiography but eschews Gelb’s problematic term of phonography, replacing it with the more justifiable glottography or ‘language writing’.¹ This book will adopt the now-established definition of writing as glottography, and restrict itself only to writing systems whose elements correspond more or less to identifiable linguistic units in spoken Sindhi. Such an epistemological position further justifies the preclusion of the Indus Valley signs from this book, whose glottographic status remains unclear (§5.1).

Gelb also classified human communication — both linguistic and non-linguistic — into categories based on how such communication is received or decoded: through the visual, auditory or tactile route (1963, p. 8). Even if restricted to linguistic communication, Gelb’s categorisation proves insightful and has been built upon by subsequent scholarship. Joyce (2011, p. 69) encapsulates the essence of such views as language being conceptualised as “an abstract entity, where speech and writing, as well as sign, are different mediums for expressing language”. Joyce’s statement may be expanded into a useful paradigm of the various modalities used in human linguistic communication, shown in Table 2.1.

In Table 2.1, the modalities of human linguistic communication are first classified by their relative permanence, and subsequently by their modes of encoding and decoding.³ Speech and signing, which are considered primary modalities (O'Connor, 1996, p. 793), are temporary in nature. Writing, on the other hand, is expected to be, and often is, relatively longer-lasting.⁴ In terms of encoding, speech, comprising sound units or phones, is produced in the human vocal tract. Manual signing, comprising gestures or physical movements, are produced using the hands and, to a lesser extent, the face and upper body. In terms of decoding, speech is most often decoded by the human ear, although it may, in principle, be decoded via the human eye or touch. Manual signing can only be decoded through the eye or touch. Writing, comprising units called graphs (§2.2), is most often encoded with the aim of visual decoding — namely by sight. However, if the graphs are encoded in a tactually perceptible manner — say, by embossing — such writing may also be decoded using human touch. This is the route used most often by Blind and visually impaired persons to read text composed in Braille. In theory, writing intended to be decoded tactually may also be decoded visually, as in a text composed in Braille being decoded visually by a sighted person. By this measure, Braille-based texts meet the definition of writing outlined earlier. This idea was succinctly expressed by Sir Clutha Mackenzie, a Blind activist and administrator who played a leading role in the harmonisation of Braille-based writing systems worldwide (§11.1):

Apart from the facts that seeing [i.e., sighted] people have many scripts and the blind but one, and that the former read by sight and the latter by touch, there is no fundamental difference between a written and an embossed script.⁵

(Mackenzie, 1954, p. 45)

Since most Braille-based writing is intended to be decoded tactually, and targeted primarily at the Blind community, it may be justifiably understood as a distinct sociolinguistic subtype of writing (Bunčić, 2016e, pp. 100–101). Graphematically, however, there is little justification for treating Braille-based writing any differently from prototypical writing intended to be decoded visually. As Meletis (2020, p. 32 footnote 35) asserts, Braille-based writing “is a graphic representation of language and it should be counted as writing”. This justifies the inclusion in this book of Braille-based writing systems for Sindhi.

Building on the above, the subfield of linguistics dealing with “the scientific study of all aspects of written language” (Neef, 2015, p. 711), or, alternatively, “the study of the written modality of language” (Haralambous, 2020, p. 12) is known as grapholinguistics. In English-language scholarship, this discipline has been variously known in the past as grammatology, graphonomy, graphemics and graphematics (Meletis, 2020, pp. 3–8). In German-language academia, though, the field has been known as Schriftlingiustik and Grapholinguistik since at least 1988 (Dürscheid, 2016, p. 2 footnote). The latter term, in the form of grapholinguistics, has found increasing use in English-language academia since the early 2010s, and is slowly but steadily becoming established as the designation for the subfield (Honda, 2021, p. 622 footnote). In this book, grapholinguistics will be understood as encompassing two broad subdomains, graphematics (§2.4) and the sociolinguistics of writing (Bunčić, 2016b). For compactness, the latter domain will be referred to as graphosociolinguistics.

2.2 Graph

Within the paradigm on the modalities of human linguistic communication (Table 2.1), the basic unit of the written modality was termed a graph, and that of the manually-signed modality a sign. However, such usage is by no means universal and well-established. For instance, Daniels and Bright (1996, p. xliv) define sign as “a unit in a communicative system comprising a signifier (what carries the meaning) and a signified (what is meant)”. Unless explicitly clarified, it follows that “communicative system” is susceptible to broad interpretations or misunderstandings. Another term for the basic unit of written language, frequently used in both academic and lay contexts, is symbol. According to Rogers (2005, p. 10), this is “a general term for a graphic mark used in writing”. In the context of written English, the basic unit of writing is popularly known as a letter. As evident, these terms have the potential to be imprecise or even misleading. Meletis (2020, p. 78) opines that terms like sign, symbol and letter “lead double lives as lay terms and quasi-technical terms, and their use is problematic”. These fuzzy terms may have their place in lay discourse, thanks to their familiarity to nonspecialists. In a scholarly grapholinguistic context, though, precise terminology is desirable. Hence, this book will adhere to the paradigm in Table 2.1 and use the terms graph and SIGN to refer to the basic units of the written and manually-signed modalities of human language, respectively. Furthermore, labelling the basic unit of writing as the graph creates a foundation for deriving vocabulary that is harmonious with corresponding vocabulary in other linguistic subdisciplines. With graph as the point of departure, we arrive at grapheme, allograph and graphetics, analogous to phone, phoneme, allophone and phonetics (Haralambous & Dürst, 2019b, p. 129). Terminology derived from graph will be dealt with in the sections that follow.

If multiple individual graphs are used to denote one particular linguistic value, especially a phoneme, the set of graphs may be referred to as a polygraph (Sproat, 2000, p. 136 footnote 2; Osterkamp & Schreiber, 2021) or multigraph (Honda, 2019, p. 203). In the context of writing systems based on the Roman (or Latin) script, the most commonly encountered subcategory of multigraphs are digraphs and trigraphs, exemplified by English-in-Roman ⟨ch⟩ and ⟨tch⟩ for the English phoneme /t͡ʃ/. Less frequently seen are tetragraphs, such as the German-in-Roman ⟨tsch⟩ for the German phoneme /t͡ʃ/. An implied characteristic of multigraphs is that their component graphs remain individually identifiable and segmentable. That is, a multigraph is generally compositionally transparent (Meletis, 2020, p. 254 ff.). If the individual graphs in a sequence are merged such that they become harder to visually segment and lose their individual identities somewhat, the resultant form may well be considered a distinct graph. That said, it is not always clear-cut whether a particular written entity constitutes a distinct and independent graph in its own right, or is actually a multigraph featuring some visual merger between its individual constituent graphs (§2.7).

2.3 Script and Writing System

Despite the potential ambiguity of the term sign as described in Section 2.2, it has been used by several authors to refer to the basic unit of writing. For instance, Coulmas (1996a) defines writing system as:

[a] set of visible or tactile signs used to represent units of language in a systematic way, with the purpose of recording messages which can be retrieved by everyone who knows the language in question and the rules by virtue of which its units are encoded in the writing system.

(Coulmas, 1996a, p. 560)

Notwithstanding its labelling of the basic unit of writing as ‘sign’, Coulmas’ definition reiterates the fundamental idea that written units may be decoded visually or tactually, assuming they are encoded appropriately. That said, the above definition does not mention the term script. Similar to the term writing, the term script, too, has been interpreted variously in the literature (Wang, 2019). Daniels and Bright’s (1996) base their definition of script on the terms signary, writing system and orthography:

signary

a general term for a determined collection of characters (or signs)

orthography

conventional spelling of texts, and the principles therefor⁶

writing system

a signary together with an associated orthography

script

in this book, equivalent to writing system

(Daniels & Bright, 1996, pp. xliii–xlv)

This position is also adopted by Rogers (2005, p. 11). However, Daniels (2018) reinterprets the term script as follows:

script

a particular collection of characters (or signs)

orthography

conventional spelling of texts, and the principles therefor

writing system

a script together with an associated orthography

(Daniels, 2018, p. 155)

That is, Daniels’ (2018) definition of script is identical to what Daniels and Bright (1996) term signary. The equating of script with signary, while decoupling it from writing system, is an epistemological stance adopted by other scholars, including Coulmas (1996a; 2003) and Sproat (2000).⁷ This is also the position of Unicode, whose glossary defines the terms in question as follows:

script

A collection of […] written signs used to represent textual information in one or more writing systems. For example, Russian is written with a subset of the Cyrillic script; Ukrainian is written with a different subset. The Japanese writing system uses several scripts.

writing system

A set of rules for using one or more scripts to write a particular language. Examples include the American English writing system, the British English writing system, the French writing system, and the Japanese writing system.

(Unicode, 2025a)

Unicode’s definitions of script and writing system cited above also succinctly allude to an important distinction. For a writing system to qualify as one, it must comprise a script together with a language (Neef, 2015; Weingarten, 2011). As alluded to at the start of this section, writing as in glottography inherently involves a set of graphs that represent language. Put simply, a writing system requires not just a script, but also a language as an indispensable component. Accordingly, this book will follow Unicode’s definitions of script and writing system as distinct concepts, where the former is a component of the latter. The term signary will be avoided, to avoid confusion stemming from the meaning of sign. Where reference needs to be made to a set of graphs, be it the superset of all graphs in a particular script, or the subset of graphs employed within a particular writing system, the term graph inventory will be used.⁸

Despite the above, there remains the question of how much one can vary the constituent script or language of a writing system without fundamentally changing the identity of the writing system. Consider the examples in (1):

In (1), the language portrayed is the same throughout, namely English. However, the script in each instance is distinct — Roman, Shavian and Deseret, respectively. Compare them with the examples in (2):

In (2), the examples are the inverse of (1) in that the languages are distinct in each case, while the script is the same – Braille. The question is, are examples (1)a, (1)b, (1)c and (2)a each an instance of a different writing system? Or are they collectively representative of an overall ‘English writing system’? It appears scholarly consensus on this issue is yet to emerge. Weingarten (2011, p. 17) considers a writing system to be “an ordered pair of a single language and a single script”. More recently, though, Meletis (2020) opines that:

two different scripts can also materialize one and the same writing system if only the [graphs] are switched out while the linguistic units [e.g., phonemes] they correspond with remain stable (this is more or less the case for biscriptual Serbian, which is written in either Roman or Cyrillic script).

(Meletis, 2020, p. 25 footnote)

Meletis’ statement above implies that the Serbian language written in the Roman script, and the Serbian language written in the Cyrillic script, both “materialize one and the same writing system”. Along similar lines, Honda (2021) suggests that written Japanese is one writing system made up of four different scripts — hiragana, katakana, kanji and rōmaji. While such a position seems reasonable, it also engenders at least two counterarguments. The first is the theoretical consideration that, if the ontogenesis of a writing system lies in a language component and a script component coming together, varying just one of these components should be a necessary and sufficient condition for the writing system to change identity. Such a position also affirms the equal status of language and script in constituting a writing system. While language is an essential component of a writing system, it is not the primary component of a writing system. Put differently, glottography need not, and should not, imply glottocentricity, for this would be a rather self-defeating stance for a grapholinguist to adopt. The second, practical counterargument is that of potential terminological ambiguity. If ‘Serbian writing system’ were understood as comprising both Serbian-in-Cyrillic as well as Serbian-in-Roman, fully specifying which script a Serbian-language text is written in (Bunčić, 2016h) would require additional periphrastic terminology.¹⁰

Consequently, this book adopts the position that every unique combination of language-and-script — that is, every unique language-in-script instantiation — should be understood as constituting a distinct writing system. On this basis, each example in (1) and (2) represents a distinct writing system. This book will also refer to a particular writing system in terms of both its principal constituents. Accordingly, the writing systems in (1) may be termed English-Roman, English-Shavian and English-Deseret, respectively, while those in (2) may be called English-Braille and Tamil-Braille, respectively. Such double-barrelled nomenclature has the advantage of being visually transparent, and has precedent in the works of Weingarten (2011) and Neef (2015). The format has also been adopted by ScriptSource (2022a) for specifying writing systems, albeit with a slight modification. ScriptSource designates writing systems using the format adopted by the Internet Engineering Task Force (IETF) and the Internet Assigned Numbers Authority (IANA). Such a format for depicting a writing system is called a language tag, and is officially defined as follows:

A language tag is composed from a sequence of one or more “subtags”, each of which refines or narrows the range of language identified by the overall tag. Subtags, in turn, are a sequence of [Roman-script] alphanumeric characters (letters and digits), distinguished and separated from other subtags in a tag by a hyphen ("-", [Unicode] U+002D).

(Phillips & Davis, 2009, p. 3)

Per this convention, a language tag for a writing system comprises at least:

• a two or three-letter language subtag in lowercase based on the ISO 639 standard, such as 𝚓𝚊 for the Japanese language;
• a four-letter script subtag in title case based on the ISO 15924 standard, such as 𝙻𝚊𝚝𝚗 for the Latin or Roman script.

Subtags are separated from each other by hyphens. Accordingly, the writing systems seen in (1) and (2) would be 𝚎𝚗-𝙻𝚊𝚝𝚗, 𝚎𝚗-𝚂𝚑𝚊𝚠, 𝚎𝚗-𝙳𝚜𝚛𝚝, 𝚎𝚗-𝙱𝚛𝚊𝚒 and 𝚝𝚊-𝙱𝚛𝚊𝚒, respectively (ScriptSource, 2022a). Optional subtags may be added for greater specificity, such as country codes in uppercase based on ISO 3166-1, or custom subtags in lowercase for particular language varieties or orthographies. For instance, the Serbian language as written in the Roman script in the Republic of Serbia would be specified as 𝚜𝚛-𝙻𝚊𝚝𝚗-𝚁𝚂 (Phillips & Davis, 2009, p. 14) Similarly, English as written in Great Britain in the Roman script and in ‘Oxford spelling’ (Bunčić, Lippert, & Rabus, 2016) may be specified as 𝚎𝚗-𝙻𝚊𝚝𝚗-𝙶𝙱-𝚘𝚡𝚎𝚗𝚍𝚒𝚌𝚝. Evidently, such a granular nomenclature offers clarity and, when presented as a language tag, compactness. Nevertheless, the format may also be seen as redundant when used in the context of languages customarily written in just one script, as in ‘Tamil-Tamil writing system’. Also, one subtag may imply another; the 𝚘𝚡𝚎𝚗𝚍𝚒𝚌𝚝 orthography subtag implies that the script in question is Roman. In such instances, where there is little room for ambiguity, the seemingly redundant component may be dropped, as in ‘Tamil writing system’ or 𝚎𝚗-𝙶𝙱-𝚘𝚡𝚎𝚗𝚍𝚒𝚌𝚝.¹² Accordingly, in this book, these language subtags will be harnessed in various combinations depending on context and need:

• 𝚜𝚍 refers to the Sindhi language, with no mention of script;
• 𝙰𝚛𝚊𝚋 refers only to the Arabic script, independent of language;
• 𝚜𝚍-𝙰𝚛𝚊𝚋 refers specifically to the Sindhi-Arabic writing system;
• 𝚖𝚞𝚕-𝙰𝚛𝚊𝚋 refers to multiple languages written in the Arabic script.

In the context of the Roman script, the fifty-two characters below specified in the ISO 646 standard will be referred to as the basic Roman or basic Latin script:

• ABCDEFGHIJKLMNOPQRSTUVWXYZ
• abcdefghijklmnopqrstuvwxyz

Categorically specifying the constituents of a writing system using a multi-part nomenclature also has another, somewhat inconspicuous, advantage. The format lays bare the somewhat arbitrary and circumstantial nature of the language and script combination comprising the writing system in question. Indeed, (1) and (2) tacitly reveal how English can be reasonably written in various scripts. Thus, a double-barrelled nomenclature for writing systems subconsciously reinforces the absence of any intrinsic link between a language and the script it may be written in. In doing so, it brings to the fore the long-held grapholinguistic maxim that any language may be written in any script (Coulmas, 1996b, p. 1380). Admittedly, academic grapholinguists might find this idea to be a banal truism. To laypersons, however, such an insinuation may well prove dissonant or even offensive from a sociolinguistic perspective, especially to people who view a particular script as being an indispensable part of their language. If conflicting views on this issue exist within the same language community, the result may be pedagogically and sociolinguistically detrimental to the community. In the context of the Sindhi community, this sociolinguistic aspect will be examined further in Chapter 13.

Building on the basic principle of a writing system comprising two equipollent components — a language and a script — it may be possible to further categorise writing systems into homolingual and homoscriptal ones. Homolingual writing systems share a constituent language but differ in terms of their constituent scripts, as seen with 𝚎𝚗-𝙻𝚊𝚝𝚗, 𝚎𝚗-𝚂𝚑𝚊𝚠, 𝚎𝚗-𝙳𝚜𝚛𝚝 and 𝚎𝚗-𝙱𝚛𝚊𝚒 in (1) and (2). Conversely, homoscriptal writing systems differ in their constituent languages but share a constituent script, as in 𝚎𝚗-𝙱𝚛𝚊𝚒 and 𝚝𝚊-𝙱𝚛𝚊𝚒 in (2). To clarify, homolingual or homoscriptal writing systems need not be typologically similar, and two writing systems with a common language or script component might well belong to two different typological categories (§2.10). This book will demonstrate how a set of homolingual writing systems — with Sindhi as their common language — can be typologically quite distinct from one another.

At this juncture, it is worth emphasising that, when designating a writing system in terms of its constituent language and script, there apply the usual caveats of ‘language’ having fuzzy boundaries that may be difficult to determine. By analogy, the boundaries of ‘script’, too, are fuzzy, making the allocation of graphic variants into one or another ‘script’ potentially contentious. For instance, in (1), it seems uncontroversial to state that Roman, Shavian and Deseret are distinct scripts. However, not all instances are as straightforward, as seen in (3):

Based on (3), do Unifon and the Initial Teaching Alphabet constitute distinct scripts? Accordingly, are English-Unifon and English-ITA distinct writing systems? Or should Unifon and ITA simply be considered distinct stylistic variants of the Roman script that just happen to possess certain graphs absent from the conventional Roman-script graph inventory for English? To some extent, the latter proposition has a parallel in spoken language. For instance, individual varieties of spoken English sound different, as they effectively comprise distinct phonological subinventories of the overall English inventory. However, the fact that these varieties are still considered ‘English’, and not distinct languages, suggests that the differences in phonological inventory are not significant enough. As evident, there is an element of subjectivity in such determinations, making them contentious. In the context of this book, the difficulty in deciding how much individual graphs — and, by extension, the scripts they constitute — can vary in visual form before they lose their initial identity and gain a distinct one is aptly illustrated in the context of the Landa inventories used to write Sindhi (Chapter 8).

The emergence of the Unicode standard over the last three decades has provided a useful heuristic for determining whether a particular graph inventory constitutes an independent script. Before a particular set of graphs can be encoded in Unicode as a distinct ‘script’, it usually goes through several rounds of proposals, refinements and deliberation (Unicode, 2025b). Consequently, the outcome may be considered, if not definitive, at least a reasoned one. Despite the rigorous process, the fuzziness involved in such decision-making is aptly illustrated by a proposal to encode Unifon graphs in Unicode, in which the author leaves open the question of whether the graphs should be designated a distinct ‘script’, or simply be included as part of the larger Roman-script inventory (Everson, 2012).¹³

It should be emphasised that the use of a particular writing system is often determined by sociolinguistic factors, and it is not difficult to find instances of full-fledged writing systems going unused, being used only in restricted domains. For instance, the Mandarin language written in the Roman script, known as Hànyǔ Pīnyīn, ticks the boxes for qualifying as a standalone writing system in theory (𝚌𝚖𝚗-𝙻𝚊𝚝𝚗-𝚙𝚒𝚗𝚢𝚒𝚗). In practice, though, Hànyǔ Pīnyīn is typically used as an auxiliary writing system restricted to specific contexts, such as teaching Mandarin to learners or entering Mandarin-language text on electronic devices (Mair, 1996, pp. 204–205; Wiedenhof, 2005). The reasons for the restricted use of 𝚌𝚖𝚗-𝙻𝚊𝚝𝚗-𝚙𝚒𝚗𝚢𝚒𝚗 are primarily sociolinguistic and political, as deployment of the Roman-script-based Hànyǔ Pīnyīn as a primary writing system, on par with or in place of Chinese-origin characters or Hànzì, might be seen as symbolic of Western cultural imperialism. Hence, despite being widely used as an auxiliary writing system, Hànyǔ Pīnyīn has never been officially promoted, nor gained popular acceptance, as a standalone primary writing system (DeFrancis, 2006). The secondary status of 𝚌𝚖𝚗-𝙻𝚊𝚝𝚗-𝚙𝚒𝚗𝚢𝚒𝚗 for writing Mandarin is betrayed by its common description, both in lay and scholarly spheres, as an ‘auxiliary’ writing system or ‘transliteration’ (§2.10).

2.4 Graphematics

In what is termed the Modular Theory (MT) of Writing Systems, Neef (2015) conceives of writing systems being composed of three distinct modules, two mandatory and one optional. The first mandatory module is a script, or, more granularly, a graph inventory, paired with a language. The second module, also mandatory, is graphematics or the graphematic system, defined as the “relation of letters to phonological units” (Neef, 2015, p. 714).¹⁴ This definition has since been extended to mean the correlation of the written units of the graph inventory with the spoken units of the language (Meletis, 2020, p. 26). Here, ‘spoken units’ may refer to phonemes, morphemes, lexemes, suprasegmentals or other orally manifesting linguistic elements. The third and optional module is orthography, taken up in Section 2.9.

Graphematics, thus, constitutes the correspondences between units of the written and spoken modalities of language, the study of which includes (a) what graphical and linguistic units are correlated, and (b) how the units are correlated. Under the what of graphematics, one would describe that, in the 𝚎𝚗-𝙻𝚊𝚝𝚗 writing system, the digraph ⟨sh⟩ typically denotes a voiceless alveolo-palatal fricative /ʃ/. In contrast, the how of graphematics would focus on the fact that ⟨sh⟩ maps onto a phoneme, and not onto a syllable or morpheme. Depending on whether the ‘linguistic unit’ is a phonological or morphological unit, the corresponding graph(s) may be known as phonograms and logograms, respectively.¹⁵ The former category may also encompass syllabograms, and the latter morphograms (Sampson, 2015, p. 24; Sproat & Gutkin, 2021). Section 2.8 deals with these categories in further detail.

Closer scrutiny reveals that the term graphematics may be prone to polysemous use, to denote not only the correlation between graphs and linguistic values, but also the study thereof. Such polysemy is reminiscent of the ambiguity with graphematics’ spoken counterpart, phonology, in that the latter term may denote the sound system of a language, or the study of sound systems of languages. When greater precision and disambiguation is desired, especially in the context of phonograms, the term graph-phone correspondences (or relations or mappings) may be used to denote the typical correlations between written and spoken units in a given writing system.¹⁶

Also provided for in the MT is the concept of graphematic solution space. According to Neef (2015):

[…] graphematics defines what a possible spelling of a word is. […] in natural writing systems, there may be more than one, actually considerably more than one. The set of possible spellings for a word with a specific phonological representation is what I call the ‘graphematic solution space’.

(Neef, 2015, p. 716)

It follows from the above that strict one-to-one correspondences between phones and graphs would result in fewer and more predictable spelling options for a spoken word or utterance. That is, the more transparent and biunique the graph-phone correspondences, the smaller the resultant graphematic solution space. Conversely, if the correspondences between phones and graphs are many-to-one or opaque, the number of possible spellings would be greater, and the size of the graphematic solution space larger (Meletis, 2020, p. 133). In such an instance, an optional orthographic module might be required to constrain the possible spellings of words and utterances (§2.9).

Although Neef defines the graphematic solution space with reference to phonographic writing systems, the concept lends itself to being extended and applied to a variety of writing system types (Honda, 2021). In any event, since the writing systems dealt with in this book are largely phonographic in nature, Neef’s concept may be harnessed without significant modification. The interplay of the graphematic module with the orthographic module will be analysed further in Section 2.9.

The concept of graphematic system allows us to neatly articulate an elusive but crucial sociolinguistic phenomenon. When a script used to write a particular language is adopted to write another language, the process is often termed script adoption. However, as Meletis (2020, p. 341) points out, the process often involves adopting not just the script — that is, the graph inventory — but also the graphematic values associated with the graphs. Accordingly, the process in question may be more accurately characterised as adoption of a graphematic system. The graphematic system may subsequently be modified to suit the target language’s phonology, which may be analysed as adding an orthographic module (§2.9). That said, most scripts used to write multiple languages — that is, used in multiple writing systems — tend to have certain graphs that retain an archetypal linguistic value across writing systems. Thus, the Roman-script graph |a| usually denotes a low vowel (Meletis, 2020, p. 24), while |k| typically denotes a velar consonant. These archetypal linguistic values have been termed prototypical or canonical values (Gnanadesikan, 2017a; 2021; Weingarten, 2011). Besides being useful in analysing graphematic phenomena, the concept of canonical values also permits us to explain why, in an adopted graphematic system, ‘recycling’ a graph and assigning it a linguistic value vastly different from its canonical value may be met with sociolinguistic resistance (Meletis, 2020, pp. 356–357).

2.5 Grapheme

In writing-related contexts, the term grapheme is often defined by analogy to its counterparts in spoken language, such as the phoneme or morpheme (Daniels, 2018, p. 164 ff.). According to an early definition by British linguist Daniel Jones:

[…] phonemes are abstract sounds which one aims at producing, but which emerge in actual speech as a number of differing concrete sounds depending upon the phonetic environment. [Thus,] a phoneme is a family of sounds [called allophones], each of which is appropriate to one or more phonetic contexts […]

(Jones D. , 1944, p. 1)

Broadly, the defining principles of a phoneme as outlined by Jones also hold true for a grapheme. A grapheme may justifiably be considered an abstract written symbol that one aims at producing, but which emerges in actual writing as a number of differing concrete symbols depending upon the context of writing. Simultaneously, a grapheme may also be considered a family of symbols — called allographs (§2.6) — each of which is appropriate to one or more writing contexts. Thus, a grapheme may be understood as the smallest contrastive unit, namely a graph in a writing system that is also linguistically significant.

Despite consensus on the broad concept of a grapheme, a narrow definition of the term has proved elusive. Unlike the concepts of phoneme or morpheme, which are defined based on a single modality of language, the concept of grapheme necessarily involves two modalities, written and spoken. Unsurprisingly, the bimodal implications of a grapheme has led to different authors arriving at slightly different interpretations and definitions of the term, resulting in ambiguity. As a result, scholars have questioned the usefulness of the term (Meletis, 2020, p. 78 ff.), and Unicode avoids using the term in its glossary (Haralambous & Dürst, 2019b, p. 141). Daniels (2018, p. 169) goes so far as to assert that “[t]he term “grapheme” should not be used in the study of writing systems”.

The counterargument to eschewing the term grapheme altogether is that it may amount to throwing the baby out with the bathwater. An alternative pragmatic approach is to salvage the term by refining its meaning and scope. One such attempt at re(de)fining the grapheme is that of Meletis (2020, pp. 94–96), who states that, for a particular graph to qualify as a grapheme within a given writing system, it must be interpreted as being:

1. minimal, in that the graph must not be further divisible into graphic units that are themselves graphemes;
2. contrastive or distinctive within a specific graphematic environment, in that, within a given writing system, substituting the graph in question for another graph of the same status should result in a distinct word or lexeme;
3. linguistically significant, in that the graph in question should have a prototypical or ‘default’ relation to one or more linguistic units, such as phonemes or morphemes.

To qualify as a grapheme within a given writing system, a graph must meet all of the above criteria. That said, a grapheme need not always be unattached to other written units and necessarily occupy its own graphosegmental space; a bound graph or ‘diacritic’ that is graphosubsegmental may well act as a grapheme in a particular writing system. Conversely, within a writing system, a particular multigraph may meet the minimality criterion and qualify as a grapheme, but other multigraphs in the same system may not (Meletis, 2020, pp. 94–96). As evident, such analysis is hardly straightforward. The fraught nature of deciding whether a graph in a writing system qualifies as a grapheme will become evident in Part Two of this book.

Granular qualifying criteria aside, the fact remains that a grapheme is defined at the level of the writing system. That is, a particular graph qualifies as a grapheme of a writing system than of a script. This is evident in the discussion on allographs in the following section. A grapheme or a sequence of graphemes is conventionally enclosed in angular brackets ⟨ ⟩.

2.6 Allograph

Despite rejecting the term grapheme, Daniels (2018, p. 169) concedes that “[t]he one piece of grapheme theory that remains useful is the term allograph”. Within a given writing system, an allograph is understood as a variant form or shape of a grapheme, and is perceived by fluent literates in the language as such. It follows that, regardless of graphemic association, all allographs are necessarily graphs. The adjective graphetic may be used to describe allographic variation in a given context, much like how the adjective phonetic is used to describe allophonic variation. Allographs — and graphs in general — are shown enclosed in pipes | |.

From the definition above, it follows that the existence of allographs in a writing system is ontologically secondary, being predicated on the prior identification of graphemes. Simply put, one can talk about allographs of a particular grapheme only after one has already identified the set of graphemes in that writing system. Accordingly, graphs that belong to the same grapheme in one writing system may be understood as belonging to distinct graphemes in another system. That is, graphs that are considered allographs in complementary distribution or free variation in one writing system, may be considered contrastive graphemes in another writing system. For instance in the 𝚎𝚗-𝙻𝚊𝚝𝚗 writing system, the graphs |e ɛ E Ɛ|, whether printed or handwritten, are normally considered variants of the same grapheme, as seen in (4):

However, in the writing system of the Ewe language in the Roman script (𝚎𝚎-𝙻𝚊𝚝𝚗), |e| and |ɛ| are considered representative of distinct graphemes – namely as ⟨e⟩ and ⟨ɛ⟩. Accordingly, the distribution of the graphs |e ɛ E Ɛ| in the Ewe-Roman writing system would be:

Similarly, in 𝚎𝚗-𝙻𝚊𝚝𝚗, |z| and |ʒ| would be considered allographs in (relatively) free variation.¹⁷ However, in the writing system of the Skolt Sami language in the Roman script (𝚜𝚖𝚜-𝙻𝚊𝚝𝚗), |z| and |ʒ| would be considered representative of distinct graphemes, as ⟨z⟩ and ⟨ʒ⟩, respectively. Similar considerations are seen in the context of graphs shared by the inventories employed by the Arabic-Arabic (𝚊𝚛-𝙰𝚛𝚊𝚋) and Sindhi-Arabic (𝚜𝚍-𝙰𝚛𝚊𝚋) writing systems, in which the shared graphs differ in graphemic status depending on the writing system (Chapter 6).

Broadly speaking, allographic variants of a particular grapheme include graphetic variants. That is, allography may be understood to include not just graphs differing in visual shape, as in |g| and |ɡ|, but also graphs differing in typographic parameters such as line weight (|𝗀| vs. |𝗴|) or calligraphic hand (|ℊ| and |𝖌|). As already outlined, classifying graphs as allographs of a particular grapheme or as representing distinct graphemes may involve grey areas. For instance, if the graphs of a script as used within a given writing system are customarily written in a particular calligraphic or visual style, the visual style may be considered indispensable to the identity of the writing system. This is illustrated by the popular perception of the Fraktur or blackletter style traditionally used to write German in the Roman script (𝖆𝖇𝖈𝖉…𝖝𝖞𝖟), or the Gaelic type used for Irish Gaelic in the Roman script (abcd…xyz). These calligraphic hands have come to be so closely associated with their respective writing systems that each of them is often considered to constitute a distinct script, independent from Roman. This is borne out by the fact that the Fraktur and Gaelic types, despite arguably being calligraphic or stylistic variants of the Roman script, have been allocated their own IETF subtags — 𝙻𝚊𝚝𝚏 and 𝙻𝚊𝚝𝚐, respectively (Phillips & Davis, 2009). Similar is the case with the Nastaliq hand of the Arabic script, which is customarily used to write the Urdu language and, consequently, has gained a status quasi-independent of the Arabic script. As a result, the Nastaliq variant of the Arabic script, too, has been assigned its own subtag — 𝙰𝚛𝚊𝚗 (Phillips & Davis, 2009). Along similar lines, the Sindhi-Arabic writing system (𝚜𝚍-𝙰𝚛𝚊𝚋) almost always appears in the Naskh calligraphic style, whether handwritten or typeset. Written Sindhi appearing in any other calligraphic style, such as Nastaliq, would be seen as typographically and aesthetically anomalous (§6.5). The phenomenon of graphetic variants being popularly perceived as constituting distinct scripts is a good reminder of how sociolinguistic factors may influence graphematic analysis and theory, and how it may well be impossible to mutually insulate graphematic factors from graphosociolinguistic ones.

Despite the fuzzy boundaries separating the graphematic and sociolinguistic dimensions, it is usually possible to distinguish between allographs that are visually distinct variants of a particular grapheme, as in |g| and |ɡ|, and those that are merely stylistic variants or different concrete instantiations of a particular graph(eme), such as |𝗀| and |𝗴|. Meletis (2020, p. 109 ff.) terms the former graphem(at)ic allographs and the latter graphetic allographs. In the context of the latter, Meletis (2020) also states that:

visual resources that are superimposed upon scripts (and thus often considered ‘suprasegmental’), including bold print, italics, etc., […]

(Meletis, 2020, p. 24 footnote)

According to Meletis, variation in the line weight or inclination of graphs may be considered the written counterpart to spoken suprasegmentals. By extension, the “visual resources” of line weight and inclination that are imposed upon graphs may be termed supragraphicals. Most often, supragraphicals such as bold and italic type serve only a graphetic function, and may influence the perception of type design or typographical aesthetics. Occasionally, they may also have a paralinguistic function, in that text printed in bold type may indicate emphasis, while italic text may suggest ‘foreignness’. However, there exist instances where a visual alternation or supragraphical feature in a given writing system may have consistent and identifiable linguistic significance by itself, raising the question of whether such a supragraphical feature is contrastive.¹⁸

Allographs need not be mutually singular, and a single allograph may alternate with a sequence of graphs or graphemes, which is seen in the case of so-called ligatures (§2.7). Finally, and like graphemes, allographs need not always occupy a separate graphosegmental space, and may be graphematically free or bound. As will be seen, several Sindhi writing systems feature free and bound allographs for vowel phones in complementary distribution. In turn, this graphematic feature significantly influences the typology of the writing systems in question (§14.1).

2.7 Additional graphetic concepts

It is noteworthy that the assumption of the graph as the basic unit of writing may not always align with Unicode’s terminology in this regard. Since 1991, the Unicode Consortium has done much of the heavy lifting in coining a technical grapholinguistic vocabulary (Haralambous & Dürst, 2019b, p. 128). However, Unicode’s mandate is to act as a standard for the digital representation of writing, which inevitably gives its terminology a computational bent. In Unicode, the basic or atomic unit (Haralambous & Dürst, 2019b, p. 139) is the character, defined as “[t]he smallest component of written language that has semantic value” (Unicode, 2025a). Significant here is the fact that a particular character is “distinguished from others by its [underlying] meaning, not its specific [superficial] shape” (ScriptSource, 2022a). The latter concept is termed glyph, understood as the “actual, concrete image” depicting a particular character (Unicode, 2025a), or the “shape that is the visual representation of a character” (ScriptSource, 2022a). Thus, the terms character and glyph refer to two sides of the same graphical coin, with the former denoting the underlying, abstract side, and the latter the surface, concrete side. The mapping of glyphs to characters is determined by the writing system and, consequently, the graphosocial context. As Haralambous and Dürst (2019b, p. 145) state, “graphs are defined as units in a graphetic system, while glyphs are defined as socially recognizable renderings of a given character”. For a detailed critique of the terms glyph and character and their homology with graph and grapheme, see Haralambous and Dürst (2019b).

Since this book is oriented towards writing systems per se, and not so much towards their digital representation, it will prefer the term graph and its derivatives, and relegate the terms character and glyph to restricted contexts, such as East Asian writing systems or the sociolinguistic framework of Biscriptality (§2.10).

Related to the concept of glyph is that of basic shape (Haralambous & Dürst, 2019b, p. 129; Meletis, 2020, p. 41), translated from the German original Grundform (Rezec, 2009). In this book, a basic shape will be taken to mean an individual shape, outline, skeleton or template that forms the graphetic basis of one or more graphs in a writing system. For instance, in the context of the Roman script’s graph superset, |A|, |C|, |a| and |ɑ| are graphs that also qualify as basic shapes. In contrast, the graphs |á| and |ç| are not basic shapes; rather, they comprise graphetic enhancements of the basic shapes |a| and |c|, respectively. Thus, a graph in a writing system may comprise just a basic shape, or a basic shape augmented by graphetically distinctive features such as dots or lines. Besides, the graph inventories of homoscriptal writing systems may comprise distinct subsets of basic shapes and their derivatives. Thus, the Roman-script basic shapes |þ ð| are present in the graph inventory of the Icelandic-Roman (𝚒𝚜-𝙻𝚊𝚝𝚗) writing system, but not in the homoscriptal 𝚎𝚗-𝙻𝚊𝚝𝚗 writing system. Furthermore, since basic shapes are graphetic entities (Meletis, 2020, p. 23), their phonological or linguistic values across writing systems are not directly relevant to their analysis. In 𝚎𝚗-𝙻𝚊𝚝𝚗, the basic shapes |a| and |ɑ| would, in most contexts, be considered allographs and, therefore, homophonous. In contrast, |a| and |ɑ| are considered contrastive and heterophonous in the Roman-script-based writing system of the Fe’efe’e language of Cameroon (Priest & Constable, 2005, p. 8). The basic shapes |a| and |ɑ| are also contrastive and heterophonous in transcriptions based on the International Phonetic Alphabet (Haralambous & Dürst, 2019b, p. 129; Neef, 2015). Yet, regardless of context and phonological values, |a| and |ɑ| remain graphetically distinct basic shapes.

Some of the most revealing and illustrative examples of basic shapes and their derivatives are found in Arabic-script-based writing systems. In the graph superset of the Arabic script, |ح| constitutes a basic shape, as well as a standalone graph in the subinventories of the Arabic-Arabic (𝚊𝚛-𝙰𝚛𝚊𝚋), Persian-Arabic (𝚏𝚊-𝙰𝚛𝚊𝚋) and Sindhi-Arabic (𝚜𝚍-𝙰𝚛𝚊𝚋) writing systems. The graph |ج|, formed by adding a dot at a specific location within the graphetic skeleton of |ح|, also acts as a standalone graph in all three writing systems. In contrast, the derivative graph |چ| occurs in 𝚏𝚊-𝙰𝚛𝚊𝚋 and 𝚜𝚍-𝙰𝚛𝚊𝚋 but not in 𝚊𝚛-𝙰𝚛𝚊𝚋. Another derivative graph, |ڄ|, is exclusive to 𝚜𝚍-𝙰𝚛𝚊𝚋. In the context of Arabic-script graphetics, a basic shape such as |ح| is known as a rasm.¹⁹ Originally, the term rasm denoted the entire inventory of basic shapes used in writing Quranic Arabic (Daniels, 2013, pp. 412–413; Gacek, 2009, p. 141). Of late, though, scholars have begun to use rasm to denote an individual basic shape in the Arabic script inventory (Kurzon, 2013), which I adopt in this book. If necessary to distinguish between the two uses of the term, an individual basic shape may be termed a rasmic form. Accordingly, |ح| is a rasm (or rasmic form) that acts as the graphetic skeleton for the derivatives |ج چ ڄ|. That said, the actual occurrence of these derivatives in the graph inventories of various writing systems is a graphematic decision that is somewhat unpredictable and not directly related to their mutual graphetic similarity. Furthermore, in the context of Arabic-script-based writing, each so-called ‘positional variant’ of a graph may also be considered a rasm in itself (Bauer, 1996; see also §6.1).²⁰ That is, each of the graphs |ح ﺣ ﺤ ﺢ| is also a rasm, although the distinctness of |ﺣ| and |ﺤ| is debatable. It follows that considering the rasms |ح ﺣ ﺤ ﺢ| to be allographs is, again, a graphematic decision rather than a graphetic one.

Given the precision and potential utility of the term rasm, this book will extend the term beyond its traditional Arabic-script contexts. In this book, rasm will be considered synonymous with basic shape and used to denote graphetic skeletons or templates in any graph inventory, regardless of graphetic origin. Thus, in the graph superset of the Roman script, |a| and |c| are rasms or rasmic forms, while |á| and |ç| are derived or augmented forms.

Despite the usefulness of the concept, identifying the graphetic boundaries of a rasm is not always straightforward. For instance, in the Roman-script graph |i|, the dot or tittle has traditionally been considered an intrinsic part of the body of the graph, and not a diacritic. According to this view, |i| as a whole may be viewed as a rasm. However, in 1928, a Roman-script-based writing system was introduced for the Turkish language (𝚝𝚛-𝙻𝚊𝚝𝚗), whose graph inventory featured a dotless |ı| distinct from dotted |i|. Graphetically, the introduction of dotless |ı| effectively made dotted |i| into a derivative, in the process depriving the latter of rasmic status. Furthermore, since a rasm is a graphetic entity that exists at the level of script, the creation of dotless |ı| may be argued to impact the Roman-script graph inventory as a whole, irrespective of writing system. As a result, the derived form of dotless |ı| effectively emerges as the rasmic form, and dotted |i| as the derived one. This suggests that graphetic primacy is independent of ontogenetic primacy. Consequently, it may be possible for a rasm to lose rasmic status if a graphetic element were to be hived off from it and spun off into an independent graph. This process may, in turn, affect the graphematic status of graphs within a particular writing system.

A corollary of the above observation is the difficulty in pinpointing what may be called the graphetic baseline — or, more precisely, the graphosegmental baseline — of a rasm. For instance, the Devanagari-script graph inventory features the graphs |ज| and |ज्‍◌|. On purely graphetic grounds, |ज्‍◌| may be considered the rasmic form, and |ज| the derived form. In comparison, when faced with the Devanagari-script graphs |ट| and |ट्ट|, it is much harder to say whether |ट्ट| is simply a derivative of |ट|, or should be considered graphetically rasmic in its own right. Such rasmic fuzziness also exists within the Roman script inventory. Historically, the rasm |C| acted as the graphetic foundation for |G|, as did |V| for |W| (Gnanadesikan, 2009, pp. 231, 248). Hence, should |G| and |W| be considered derivatives of rasmic |C| and |V|, respectively? Or have |G| and |W| undergone sufficient graphetic evolution to be considered rasmic in themselves?

The graph |W| is also described in the literature as a ligature of |V| (Gnanadesikan, 2009, p. 259). A ligature is commonly understood as a graph formed by merging two independent graphs (Haralambous & Dürst, 2019b, p. 151). Yet, the concept of ligature encompasses certain graphetic and phonological presuppositions, which require unpacking. Graphetically, ligatures exist on a spectrum of transparency. For instance, the graph superset of Devanagari contains the graphs |ष्‍◌| and |ट|, which may be graphetically merged or ligated as |ष्‍ट|. However, they may also be represented as |ष्ट|, with a greater degree of graphetic coalescence compared to |ष्‍ट|. Accordingly, at what point does a graphetic merger result in an identifiably independent graph? As expected, this is difficult, if not impossible, to pinpoint. Conversely, within a particular writing system, a particular graph may be considered a ‘ligature’ of two other graphs. Thus, in the Hindi-Devanagari (𝚑𝚒-𝙳𝚎𝚟𝚊) writing system, |क्ष| is considered a ligature of the graph sequence |क्‍ष|, while |क्त| is considered a ligature of the graph sequence |क्‍त|. In the Arabic-Arabic (𝚊𝚛-𝙰𝚛𝚊𝚋) writing system, |لا| is considered a ligature of the right-to-left graph sequence |ل‌ا| (Haralambous & Dürst, 2019b, pp. 151–152). Graphetically, there appears to be little similarity between |क्ष|, |क्त| and |لا| on the one hand, and |क्‍ष|, |क्‍त| and |ل‌ا| on the other. In fact, |क्ष|, |क्त| and |لا| may well qualify as rasms within their respective graph inventories. Yet, |क्ष|, |क्त| and |لا| are commonly considered ‘ligatures’ of |क्‍ष|, |क्‍त| and |ل‌ا|, respectively, based on the fact that they are homophonous. That is, the writing system of 𝚑𝚒-𝙳𝚎𝚟𝚊 assigns both |क्ष| and |क्‍ष| the phonological value of [kɕ(ə)], making them homophonous and, hence, allographs of each other. Likewise, 𝚑𝚒-𝙳𝚎𝚟𝚊 considers |क्त| and |क्‍त| to be allographs representing [kt̪(ə)]. Along similar lines, the writing system of 𝚊𝚛-𝙰𝚛𝚊𝚋 considers |لا| to be homophonous and allographic with |ل‌ا|, with both denoting /laː/.²¹ Taken together, it emerges that, within a particular writing system, a rasm or graphetically distinct form within a graph inventory may be considered a ‘ligature’ of other graphs in that inventory based primarily on mutual homophony and allography. The phonological basis of the label ‘ligature’ is further revealed by the fact that, depending on the writing system, certain ligatures may be considered mandatory and others optional (Haralambous & Dürst, 2019b, pp. 151–152). This corresponds to allographs in complementary distribution and free variation, respectively. For instance, the writing system of 𝚊𝚛-𝙰𝚛𝚊𝚋 requires all instances of |ل‌ا| to be mandatorily substituted with |لا|, making them allographs in complementary distribution. Along similar lines, the writing system of 𝚑𝚒-𝙳𝚎𝚟𝚊 considers |क्ष| a mandatory ‘ligature’ or allograph of |क्‍ष|, putting the two forms in complementary distribution. In contrast, 𝚑𝚒-𝙳𝚎𝚟𝚊 licenses |क्‍त| as well as |क्त|, making them free variants of each other. These examples suggest that the mandatory or optional nature of a so-called ligature — that is, whether it is in complementary distribution or free variation with its allographs — is decided at the level of writing system, particularly its orthography (§2.9), and not at the level of graph inventory. As seen, written elements commonly described as ligatures may justifiably be considered rasms in their respective graph inventories.

Thus, it emerges that the use of the term ligature to refer to graphetic fusion as well as phonological equivalence renders it polysemous. Consequently, the term will be judiciously used in this book, primarily to denote graphetically transparent forms like |ष्ट| that lie on the boundary between qualifying as a sequence of a graphs or as an independent graph by itself. Although the term conjunct may be suggested as an alternative, this term may denote either a sequence of graphs or a sequence of phones (Unicode, 2025a). Since conjunct may refer either to a graphical entity or to a phonological one, it, too, has the potential to prove ambiguous and will also be avoided to the extent possible in this book.

In the context of Indic writing systems, a concept closely related to those of ligature and conjunct is that of akshara. This term has been variously understood and depicted in the academic literature, due to which an unambiguous definition of this term is lacking. Consequently, the term is avoided in this book. For a discussion on the concept of akshara and a proposed definition, see Iyengar (2024).

Finally, when it comes to digital encoding, an individual character or glyph in Unicode may represent graphs that are rasmic, graphetically augmented or of graphetically ambiguous status (Unicode, 2025a). Where Unicode assumes greater importance is in the availability of graphetic forms for enabling electronic text. Since grapholinguistic publishing in the digital era presupposes the availability of necessary graphs or glyphs — by way of fonts containing such glyphs — it follows that the lack of suitable glyphs and fonts represents a significant hurdle for authors in the field. As Haralambous & Dürst (2019b, p. 144 & footnote) observe, typographers and font designers effectively become the “gatekeeper[s]” of glyphs, especially in rare or understudied scripts and writing systems. As a result, authors without the necessary design competencies to design the requisite glyphs and fonts themselves may be forced to use suboptimal existing glyphs, in the process potentially undermining the perceived authenticity and credibility of their work. While much care has been taken to custom-design any necessary glyphs and fonts for this book, certain typographical and aesthetic shortcomings may well remain.

2.8 Phonography and Logography

Most writing systems tend to contain a mix of phonographic and logographic elements (Sampson, 2015, p. 24; Sproat & Gutkin, 2021). For instance, the graph inventory of the predominantly phonographic English-Roman (𝚎𝚗-𝙻𝚊𝚝𝚗) writing system also includes the ampersand, currency symbols and numerals, the last of which are also termed digits or ciphers (Honda, 2021). Logography in a writing system may also manifest at a broader level, with various functions. According to Gnanadesikan (2017a, p. 15), the distinction between the English-language homophones spelt ⟨deer⟩ and ⟨dear⟩ in 𝚎𝚗-𝙻𝚊𝚝𝚗 is essentially logographic in nature. A more illustrative example from French is that of the utterance /aʁive/ corresponding to at least four distinct forms in the French-Roman (𝚏𝚛-𝙻𝚊𝚝𝚗) writing system: ⟨arrivé⟩, ⟨arrivée⟩, ⟨arrivés⟩, ⟨arrivées⟩, all of which mean ‘arrived’ but differ in the referent number and gender they denote. Logograms may also occur as allographs of phonograms, as attested by the Marathi-Devanagari (𝚖𝚛-𝙳𝚎𝚟𝚊) spellings ⟨ॐकार⟩ and ⟨ओमकार⟩. Both written forms reflect the pronunciation /omkaɾ/ and mean ‘the symbol Om’. Here, the benedictory logogram |ॐ| [om] occurs as an allograph in free variation with the homophonous phonogram sequence |ओम|. Whereas these examples illustrate how predominantly phonographic writing systems also comprise logographic elements, the converse is also true. In fact, Daniels (1996a, p. 4) states that writing that is purely logographic may not exist.

Just as the line between phonograms and logograms may be blurry, the line where logography crosses the line into semasiography may also be difficult to determine. Aptly exemplifying the fuzzy boundary in question is the status of punctuation marks across writing systems. On the one hand, punctuation marks may be interpreted as minimal, contrastive and linguistically significant and, hence, may even qualify as graphemes (§2.5). On the other hand, it may be argued that punctuation marks are not so much linguistically meaningful as paralinguistically, in that they graphically indicate pauses, emphasis or intonation patterns. On this basis, punctuation marks may be considered less glottographic and more semasiographic. Scholarly opinion on the status of punctuation marks is equivocal. Neef (2015, pp. 711–713) considers logograms, ciphers and punctuation marks as “belong[ing] to the script but not to the writing system”. Meletis (2020, p. 95) admits that the status of punctuation marks in a writing system is yet to be suitably resolved. In the absence of epistemological and terminological consensus, this book will tentatively categorise and analyse numerals, punctuation marks and certain benedictory symbols within a given writing system as logograms.

Also worth remembering is that a graph is a phonogram or logogram within the context of a particular writing system. Among homoscriptal writing systems, certain graphs used as logograms in one system may be used as phonograms in another, and vice versa. For instance, the Squamish language when written in the Roman script (𝚜𝚚𝚞-𝙻𝚊𝚝𝚗) uses the graph |7| to denote /ʔ/ (Dyck, 2004). The Roman-script writing systems of several Khoisan languages employ the graph |ǃ| to denote an alveolar click, coinciding with IPA usage (DOBES, 2017). A 1993 proposal for writing the Turkmen language in the Roman script (𝚝𝚔-𝙻𝚊𝚝𝚗) listed |$| and |¢| as uppercase and lowercase allographs for the phoneme /ʃ/ (Clement, 2008; Pedersen, 2003).

Graphematically, utilising a graph as a logogram in one context and as a logogram in another is completely plausible. Sociolinguistically, and especially within a writing community or graphosphere (§2.9) that is largely monoscriptal, the use of logograms in a phonographic function may be implicitly seen as transgressing established conventions and, consequently, viscerally resisted. At best, such usage may be looked upon as a grapholinguistic curiosity. At worst, they may cause heated debate, and have ramifications for the perceived violators of graphematic and graphosociolinguistic norms.²² Conversely, the absence of such debate in a writing community and a diversity of scribal practices may point to the lack of importance given to graphematic standardisation in that community. This, in turn, raises critical questions for our conception of standardisation, spelling and orthography.

2.9 Spelling and Orthography

Just as phonotactics is understood as the restrictions on how phonological elements within a language can combine with each other, graphotactics refers to the “restrictions on ways in which the elements of a writing system may combine with each other” (Meletis, 2020, p. 131). As with allography, two subtypes of graphotactic rules may be identified depending on their level of applicability. Rules that govern the combination and manifestation of concrete graph shapes fall under graphetic graphotactics, whereas those governing the co-occurrence of graphs are covered by graphematic graphotactics. As the names suggest, the former merges into the domains of calligraphy and typography, while the latter overlaps with rules and conventions regarding spelling and orthography. Evidently, a writing system comprising a graphetically simple inventory may not require a separate graphetic graphotactics, whereas one with a (nearly) biunique set of correspondences between spoken and written units may render a distinct graphematic graphotactics redundant (Meletis, 2020, p. 133). Consequently, and depending on the characteristics of the writing system being studied, it may be possible to subsume graphotactic considerations under other disciplines and dispense with a standalone category of graphotactics.

Most scholarly works on writing systems do not provide an explicit definition of the concept of spelling. An exception is Coulmas (1996a, p. 477), who defines spelling as the conventions for using the graphs of a writing system to write the language in question. In a sense, spelling bears similarities to pronunciation, as a popular if imprecise term for how smaller linguistic units — be they written or spoken — are generally combined and collectively realised. Coulmas considers orthography to be a codified set of spellings or spellings rules, which aligns with Daniels and Bright’s (1996) definition of the term (§2.3).²³ According to Coulmas, the distinction between a spelling system and orthography lies in the presence of codification, be it implicitly by longstanding tradition or explicitly by official decree. In the absence of any discernible codification, it becomes problematic to assert that the writing system possesses an orthography (Coulmas, 1996b, p. 1380). Instead, such a writing system may be considered a ‘bare’ graphematic system with generally accepted graph-phone correspondences but an unconstrained graphematic solution space.

Coulmas’ characterisation of orthography as an optional add-on that codifies and constrains graphematic manifestations is reflected in the structure of the Modular Theory. The MT posits orthography as a third module superimposed over the graphematic module. Also, the orthographic module is considered optional, and a writing system may, in principle, comprise just a graph inventory and a graphematic system (but see Honda (2021)). As alluded to in Section 2.4, a graphematic system featuring biunique correspondences between spoken and written units and, consequently, a predictable set of licensed spellings, may have limited use for an orthographic module (Meletis, 2020, p. 28). However, if a graphematic system features nonbiunique correspondences between spoken and written forms, it may require an orthographic module to constrain its graphematic solution space. Thus, an overlaid orthographic module:

prescribes how to write correctly within the limits of the graphematic solution space. This module particularly aims at constant spellings of words.

(Neef, 2015, p. 716)

As Neef’s statement implies, underpinning the concept of orthography is the notion of ‘correctness’ and, by extension, prescriptivism. Thus, the graphematic module simply generates a set of possible spellings without prejudice, so to speak. In contrast, the orthographic module restricts these possibilities, and effectively prescribes the ‘correct’ spellings of words.²⁴ In terms of orthography subtypes, the MT aligns with Coulmas’ (1996a, p. 477) observations mentioned above. Thus, an orthography may develop implicitly from within the graphematic system as a set of practices emerging organically from the collective practice of reader-writers and stabilising over time. This is an internal, implicit or a bottom-up orthography. Alternatively, an orthography may comprise a set of rules decreed by a language body, typically one with governmental or official authority. This counts as an external, explicit or top-down orthography. Neef (2015) labels the former type a systematic orthography and the latter a conventional one.²⁵ According to this model, the English-Roman writing system would be considered to have a systematic orthography codified not by a quasi-official linguistic authority but by multiple private lexicographic institutions worldwide. In contrast, the German-Roman (𝚍𝚎-𝙻𝚊𝚝𝚗) writing system possesses a conventional orthography, given the greater role played by official bodies in its codification (Meletis, 2020, p. 155 footnote 172). That said, distinguishing the two orthography types is not always straightforward. For instance, an externally-imposed orthography is not conjured ex nihilo; rather, it typically comprises a regularisation of existing system-internal practices. Similarly, an orthography that emerges from within the system becomes externally imposed when promulgated by a lexicographic institution. As a result, it may be difficult to definitively state what subtype of orthography a particular writing system possesses, or whether it possesses an orthography at all (Meletis, 2020, p. 155). As demonstrated in this book, this dilemma is frequently encountered across most, if not all, of Sindhi’s writing systems.

In the context of phonographic writing systems, a graphematic system with nonbiunique correspondences between phones and graphs may be of two types (Osterkamp & Schreiber, 2021, p. 193). If a particular phone may be represented by more than one graph — that is, if there is homophonous heterography — the number of spelling variants licensed by the graphematic solution space will increase. Conversely, if several phones map onto the same graph — namely in the event of heterophonous homography — the number of spelling variants may be fewer, but the difficulty of decoding spellings increases. Both situations may be addressed by an orthography. In cases of homophonous heterography, the orthography may pick a particular spelling for a particular lexico-semantic entity. As alluded to in Section 2.8, the English utterance /dɪə(ɹ)/ may be spelt ⟨deer⟩ or ⟨dear⟩ in 𝚎𝚗-𝙻𝚊𝚝𝚗, as both spellings are licensed by its graphematic solution space. It is the orthographic module that mandates the spelling ⟨deer⟩ or ⟨dear⟩ depending on the meaning of /dɪə(ɹ)/ in a given context. In cases of heterophonous homography, where the choice of written forms is limited, the orthography may prescribe — or seek to prescribe — spellings not licensed by the graphematic solution space (Meletis, 2020, p. 158). An orthography may also legitimise extrasystemic spellings for reasons of etymology or prevalence, as in 𝚎𝚗-𝙻𝚊𝚝𝚗 ⟨tsunami⟩ or 𝚍𝚎-𝙻𝚊𝚝𝚗 ⟨Computer⟩. Within Sindhi’s writing systems, both homophonous heterography and heterophonous homography are attested, and offer valuable insights to further our understanding of these phenomena.

Within the MT, the relation between the orthographic module and the graph inventory module remains underresearched. In particular, the influence orthography has on the graph inventory of a writing systems is worthy of further investigation. For instance, the German-Roman writing system as used in Germany (𝚍𝚎-𝙻𝚊𝚝𝚗-𝙳𝙴) makes use of the grapheme ⟨ß⟩ /s/, while the German-Roman writing system as used in Switzerland (𝚍𝚎-𝙻𝚊𝚝𝚗-𝙲𝙷) does not. Words featuring ⟨ß⟩ in 𝚍𝚎-𝙻𝚊𝚝𝚗-𝙳𝙴 are spelt in 𝚍𝚎-𝙻𝚊𝚝𝚗-𝙲𝙷 with ⟨ss⟩ instead. Thus, 𝚍𝚎-𝙻𝚊𝚝𝚗-𝙳𝙴 and 𝚍𝚎-𝙻𝚊𝚝𝚗-𝙲𝙷 differ in their graph inventories. Yet, scholarly and popular opinion generally describes this phenomenon as a difference in the orthographies of 𝚍𝚎-𝙻𝚊𝚝𝚗-𝙳𝙴 and 𝚍𝚎-𝙻𝚊𝚝𝚗-𝙲𝙷 (Bunčić, 2016c, p. 66; Meletis, 2020, p. 373). Similarly, the Russian-Cyrillic (𝚛𝚞-𝙲𝚢𝚛𝚕) writing system underwent significant modifications in 1708 and 1917, which impacted not just the spellings of words but also graph-phone correspondences, the graph inventory and the calligraphic style of the graphs (Coulmas, 1996b, pp. 107, 447). Despite affecting the size and composition of the 𝚛𝚞-𝙲𝚢𝚛𝚕 graph inventory, these reforms are commonly understood as orthographic reforms (Bunčić, Kislova, & Rabus, 2016). The distinct graph inventories used to write Mandarin, commonly known as Traditional and Simplified characters, have also been characterised as a distinction in orthography (Klöter & Bunčić, 2016). A particularly intriguing and illustrative case in this context is that of the Azeri language in the Roman script (𝚊𝚣-𝙻𝚊𝚝𝚗). Over the twentieth century, Azeri has been written in the Roman script but at least three distinct graph inventories and, consequently, with differing graph-phone correspondences. Again, this situation is commonly portrayed as Azeri having been written in different Roman-script orthographies (Coulmas, 1996b, pp. 29–30). In the context of scripts featured in this book, modifications to the Uyghur-Arabic (𝚞𝚐-𝙰𝚛𝚊𝚋) writing system in the late twentieth century (Coulmas, 1996b, p. 525), and to the Kashmiri-Devanagari (𝚔𝚜-𝙳𝚎𝚟𝚊) writing system in the early twenty-first century (Ishida, 2021) have been characterised in terms of orthographic reform, despite the modifications entailing the addition of new graphs and the recalibration of graph-phone correspondences.

Thus, in terms of the MT, it needs to be recognised that a change in the orthographic module may impact not just the graphematic module, but also the graph inventory. Such an approach also aligns somewhat with the format of IETF language tags. Within the language tag 𝚍𝚎-𝙻𝚊𝚝𝚗-𝙳𝙴, the subtag DE denotes the presence of the Germany-specific orthographic module and, by extension, the Germany-specific graph inventory that includes ⟨ß⟩. Likewise, the 1708 and 1917 subvariants of the Russian-Cyrillic writing system are assigned the IETF language tags 𝚛𝚞-𝙲𝚢𝚛𝚕-𝚙𝚎𝚝𝚛𝟷𝟽𝟶𝟾 and 𝚛𝚞-𝙲𝚢𝚛𝚕-𝚕𝚞𝚗𝚊𝟷𝟿𝟷𝟽, respectively. However, this is not always the case with IETF subtags. Although academics have described the Traditional and Simplified Hànzì graph inventories for Mandarin as distinct orthographies, the IETF considers them to be distinct scripts and assigns each of them a separate script subtag. Thus, Mandarin in Traditional and Simplified Hànzì bear the language tags 𝚌𝚖𝚗-𝙷𝚊𝚗𝚝 and 𝚌𝚖𝚗-𝙷𝚊𝚗𝚜, respectively. The vacillation in denoting distinct graph inventories, by using orthography subtags on the one hand and script subtags on the other, reiterates how writing system taxonomy remains a fraught area (§2.3).

It follows from the above that most writing systems that are or have been in active use can be understood to possess an orthographic module. In Neef’s (2015, p. 719) words, the “ideal” writing system comprising just a bare graphematic system with biunique correspondences between spoken and written units is unlikely to exist in practice. Even if one did exist, asynchronous phonological and graphematic evolution would eventually disrupt the one-to-one mapping between its spoken and written units. Consequently, Neef considers the presence of an orthographic module to be the “normal case”.

The MT also has the potential to act as a graphematic mirror to certain (grapho)sociolinguistic phenomena. Analogous to a speech community that shares a common language, it is possible to conceive of a “writing community” (Coulmas, 1996a, p. 556) or graphosphere (Franklin, 2011). Depending on the strictness of interpretation, a graphosphere may indicate a group of people with a shared orthography (𝚎𝚗-𝙻𝚊𝚝𝚗-𝚄𝚂 versus 𝚎𝚗-𝙻𝚊𝚝𝚗-𝙶𝙱), a shared writing system (𝚎𝚗-𝙻𝚊𝚝𝚗) or simply a shared script without a shared language (𝙻𝚊𝚝𝚗). In Coulmas’ (1996a, p. 556) opinion, once writing systems or parts thereof become established, “they define cultural and religious spheres and thus communities”. This sociolinguistic interpretation may be neatly expressed in graphematic terms using the MT. The greater the number of shared modules (script, language, orthography…) among communities, the greater the emic and etic feeling that those communities belong to the same graphosphere.

Moving further into the graphosociolinguistic realm, the process of writing down a previously spoken-only language for the first time — sometimes unfortunately characterised as ‘reducing’ a language to writing — may be termed graphisation (Ferguson, 1996 [1968], p. 41). Here, ‘reduction to writing’ should be understood as “creation or adoption of a writing system” (Gold, 1982). With time, a pan-regional and pan-social norm may develop for the language in question, both in speech and in writing. The development of such norms usually amounts to standardisation (Ferguson, 1996 [1968], p. 41). Within the written domain, standardisation may include the development of a normative graph inventory and spelling rules, namely an orthography. The process of developing an orthography is orthographisation, by analogy with graphisation. Sometimes, the process of graphisation and orthographisation may coincide (Lüpke, 2011). However, in the context of Sindhi, the two are unarguably distinct. Hence, this book will maintain a strict distinction between these two concepts and terms, and use them accordingly.

Near-synonymous with graphisation is literisation, which Siegel (1981, p. 20) uses to mean “development, both planned and unplanned, of a previously unwritten language into a written one”. Evidently, the process of literisation would involve at least some degree of standardisation and orthographisation. Pollock (1998, p. 41) draws a further distinction between literisation and literarisation, where the former term indicates the act of committing a spoken-only language into writing, while the latter refers to the ongoing deployment of a now-literised or graphised language for literary purposes. In turn, these processes may overlap with those of transliteration and transcription (§2.10), but remain distinct from them.

In this book, where no official IETF language subtag exists for a specific orthographic module within a writing system, a private use subtag will be assigned. The term ‘private use’ indicates that the subtag is not approved or specified by the IETF, and is employed purely for nomenclatural expedience. All private-use subtags in this book will have a preceding 𝚡-, in line with the IETF-prescribed format.

2.10 Writing system typologies

Writing systems may be divided into subcategories based on their internal structure, or on their use in everyday settings. In simple terms, we can classify writing systems based on how they are, or what they are used for. The former may be considered a graphematic typology of writing systems, while the latter constitutes a sociolinguistic typology.

2.10.1 Graphematic or structural typology

A graphematic or structure-based typology of writing systems focuses on the nature of mapping or correspondence between a writing system’s graphs and their linguistic values. As outlined in Section 2.8, writing systems may be graphematically subcategorised into logographies and phonographies. In turn, phonographies — the focus of this book — may be further classified into syllabaries and segmentaries (Gnanadesikan, 2017a; Sampson, 2015). In a syllabary, an individual graph corresponds to a phonological syllable, whereas, in a segmentary, the individual graph corresponds to a phonological segment, namely a consonant or a vowel. Thus, syllabaries and segmentaries differ in terms of the basic grain size represented by their graphs — the phonological syllable in the former, and the phonological segment in the latter (Gnanadesikan, 2017a). Segmentaries themselves may be further divided into the kinds of segments they represents. One that comprises graphs for consonantal segments only, and leaves vowel segments unrepresented in writing, is termed a consonantary or abjad (Daniels & Bright, 1996, p. xxxix). In contrast, vowelled segmentaries that contain graphs for consonant as well as vowel segments may be of various subtypes. Figure 2.2 summarises the graphematic typology of writing systems described thus far.

The prototypical vowelled segmentary is an alphabet, in which phonological consonants and vowels are denoted by distinct graphs, usually of graphetically comparable size (Daniels & Bright, 1996, p. xxxix). However, other recognised categories of writing systems, such as alphasyllabaries and abugidas, are also graphematically capable of denoting phonological consonants and vowels with distinct graphs. In an alphasyllabary, graphs denoting phonological vowels generally have two allographs, free and bound (Bright, 1999), whose distribution is determined by phonological and graphotactic-orthographic considerations. In general, if a phonological consonant is followed by a phonological vowel, the latter is shown with the appropriate bound allograph.²⁶ The distinctive feature of an abugida is that a basic graph in the inventory corresponds to a sequence of a distinct phonological consonant and a common phonological vowel. The systematically common phonological vowel is known as the ‘inherent’ or ‘default’ vowel. Its quality depends on the language represented. Against the above, Daniels (1996a, p. 4) and Bright (1999) characterise the graphematic distinction among alphabets, alphasyllabaries and abugidas as a ‘formal’ versus ‘functional’ one. Gnanadesikan (2017a) refines this distinction into one of which vowel segments are overtly represented in the graphematic system, versus how they are graphematically realised. Yet, there remain areas of overlap among these definitions. For instance, in Hindi-Devanagari (𝚑𝚒-𝙳𝚎𝚟𝚊), postconsonantal vowels are shown using the appropriate bound allograph for the vowel, which makes the writing system an alphasyllabary. Concurrently, 𝚑𝚒-𝙳𝚎𝚟𝚊 denotes all phonological vowels overtly except one, with the vowel /ə/ considered ‘inherent’ to basic graphs denoting phonological consonants. This makes the writing system an abugida. Thus, it is possible for a particular writing system to simultaneously fall into multiple categories. On this basis, harnessing Gnanadesikan’s (2017a) dimensions of which and how along different axes, Iyengar (2023) argues that the key distinguishing factor among various types of vowelled segmentaries is the graphematic (γ) representation of a phonological (φ) vowel after a phonological consonant. Based on the graphematic and phonological manifestation of a postconsonantal vowel, Iyengar proposes a four-way classification of vowelled segmentaries, shown in Figure 2.3.

As seen in Figure 2.3, Iyengar introduces the term plenary to denote a writing system in which all phonological vowels are overtly written, with none considered ‘inherent’ to a graph.²⁷ This property distinguishes a plenary from an abugida, the latter of which requires one phonological vowel to be systematically ‘inherent’ in graphs representing consonants. Based on the paradigm outlined in Figure 2.3, Hindi-Devanagari (𝚑𝚒-𝙳𝚎𝚟𝚊) is an abugidic alphasyllabary, which is harmonious with its graphematic properties described earlier. In contrast, 𝚎𝚗-𝙻𝚊𝚝𝚗 is a plenar alphabet, since it represents phonological vowels using free allographs at all times, and represents them overtly. Sharing graphematic similarities with both 𝚑𝚒-𝙳𝚎𝚟𝚊 as well as 𝚎𝚗-𝙻𝚊𝚝𝚗 is the writing system for the Divehi or Maldivian language based on the Thaana script (𝚍𝚟-𝚃𝚑𝚊𝚊). In 𝚍𝚟-𝚃𝚑𝚊𝚊, postconsonantal vowel phones are shown as bound allographs, making it an alphasyllabary like 𝚑𝚒-𝙳𝚎𝚟𝚊. However, all vowel phones are represented overtly, making it a plenary like 𝚎𝚗-𝙻𝚊𝚝𝚗 (Gnanadesikan, 2017b). Finally, the Mongolian language written in the ’Phags-pa script (𝚖𝚗-𝙿𝚑𝚊𝚐) denotes vowel phones only as free allographs, making it an alphabet, but also assigns an inherent vowel phone to each of its consonant graphs, making it an alphasyllabary (Bright, 1999; Gnanadesikan, 2017a).

Coupled with the double-barrelled nomenclature of writing systems adopted in this book, the typological framework proposed in Figure 2.3 helps prevent conflating homoscriptal writing system types with each other. Thus, the Hebrew language written in the Hebrew script (𝚑𝚎-𝙷𝚎𝚋𝚛), with all vowels overtly represented, constitutes a plenar alphasyllabary (Bright, 1999, p. 50 footnote 2). In contrast, Yiddish in the Hebrew script (𝚢𝚒-𝙷𝚎𝚋𝚛) shares graphematic similarities with 𝚎𝚗-𝙻𝚊𝚝𝚗 in functioning as a plenar alphabet (Aronson H. I., 1996). This observation, in turn, reinforces the idea that writing system typology is actually graphematic typology (Meletis, 2020, p. 393) rather than script typology. In other words, writing systems fall into a particular category based on their graphematic makeup, and not because of some inherent property within the script itself. Additionally, and as alluded to in Section 2.9, the framework in Figure 2.3 is consistent with the notion that an orthography-induced modification to the graph inventory and/or graph-phone correspondences within a writing system may cause the system to deviate from its original type. Such a position also reaffirms the interdependent nature of a writing system’s modules.

Since this book deals only with segmentaries, speaking of ‘phonological vowels’ and ‘phonological consonants’ on the one hand, and ‘graphematic vowels’ and ‘graphematic consonants’ on the other, is theoretically justified and consistent. To allow for brevity and compactness, the Greek graphs φ and γ will be used as notations to indicate ‘phonological’ and ‘graphematic’, respectively (Haralambous, 2019a). Thus, a graphematic consonant — namely a consonant graph — will be represented as γ-consonant. Likewise, a phonological vowel or vowel sound will be denoted by φ-vowel. The notation also lends itself well to denoting non-segmental grapholinguistic units, such as φ-syllable for ‘phonological syllable’.

The study of written Sindhi’s graphematic and sociolinguistic history reveals how certain graphematic labels used to describe the language’s writing systems — such as abjad or alphasyllabary — were actually based on sociolinguistic practices. The following section sheds more light on the issue.

2.10.2 Sociolinguistic or contextual typology

In contrast to a graphematic typology, a sociolinguistic or context-based typology of writing systems derives from the employment of a writing system in a particular context or for a particular purpose. According to Mountford (1996, p. 627), a sociolinguistic typology of writing system seeks to answer the question “who uses which writing system to whom, in what context, and for what purpose”.²⁸ Broadly speaking, Mountford distinguishes between general-purpose and special-purpose writing systems. The former comprises what Sampson (2015, p. 146) describes as “the ordinary writing system of a society” used for everyday purposes. The latter encompasses writing systems employed in specialised situations or for special purposes. Such specialised writing systems include shorthand, cryptic codes and systems used primarily in pedagogical or professional settings, regardless of their graphematic type. The use of a writing system in a restricted situation or context may influence its popular perception, and prevent the system from being used for everyday purposes, regardless of its graphematic properties. This explains why Hànyǔ Pīnyīn (𝚌𝚖𝚗-𝙻𝚊𝚝𝚗-𝚙𝚒𝚗𝚢𝚒𝚗; §2.3), English in the Initial Teaching Alphabet (see (3)) or English transcribed in the IPA (𝚎𝚗-𝙻𝚊𝚝𝚗-𝚏𝚘𝚗𝚒𝚙𝚊) are not used in quotidian settings, despite being graphematically comprehensive systems. Hence, the use of a writing system for a specific purpose or by a specific user group may well create a semiotic association between the system and its use or users. If such a semiotic association becomes entrenched over time, attempting to use the writing system outside of its perceived domain of use may be sociolinguistically resisted. Such semiotic associations, and resistance to modify them, are especially evident in the context of written Sindhi in contemporary India (§13.3).

Transliteration, Transcription and Shorthand

Within the context of a particular language — that is, in a homolingual context — three noteworthy special-purpose types of writing system are transliteration, transcription and shorthand. According to Daniels (2018, p. 15), transliteration involves replacing one script with another, ensuring that (nearly) every graph in the source inventory has an equivalent in the target inventory. If the target script is Roman or Latin, the transliteration is called a romanisation or latinisation.²⁹ In contrast, a transcription involves equating the phones of a language with the graphs of a particular script (Daniels, 2018, p. 15). Graphematically, the phenomena differ in terms of focus and point of departure. Transliteration focuses on graph-graph correspondence between two different scripts or inventories, while transcription focuses on phone-graph correspondences between a language and a script. Also, transliteration in the strict sense presupposes that the language in question already has a writing system, thereby being applicable only to graphised languages. Transcription, on the other hand, can apply to any language irrespective of graphisation status. Both processes, however, lead to the establishment of a new language-script pair, thereby resulting in a new writing system. Consequently, the terms are often used interchangeably. Besides, they are also used as near-synonyms for a writing system perceived as secondary or auxiliary in nature. Such a characterisation, however, derives from the sociolinguistic use and perception of the writing system in question, and not from its graphematic properties per se. As described earlier in this section, Hànyǔ Pīnyīn (𝚌𝚖𝚗-𝙻𝚊𝚝𝚗-𝚙𝚒𝚗𝚢𝚒𝚗) is sociolinguistically perceived as a secondary or auxiliary system, despite being a graphematically full-fledged writing system in its own right. Hànyǔ Pīnyīn does, however, meet the definition of transliteration — specifically, romanisation — as outlined above, since there exists a consistent correspondence between the graphs of 𝚌𝚖𝚗-𝙻𝚊𝚝𝚗-𝚙𝚒𝚗𝚢𝚒𝚗 and those of Mandarin in Simplified Hànzì (𝚌𝚖𝚗-𝙷𝚊𝚗𝚜).

In the context of this book, there arise two questions regarding transliteration. First, since a transliteration is intended to serve as an auxiliary writing system, should one be used in this book, especially given the large number of writing systems in focus? Second, if a transliteration is essentially a full-fledged writing system, should this book also include past and present transliterations of Sindhi in its purview? As defined above, a transliteration reflects graph-graph correspondences between two homolingual writing systems. This implies that, for a transliteration to be effective, the number of pre-existing writing systems should be as few as possible — ideally just one. In this book, however, no less than seven distinct writing systems are described and analysed, with each of them having numerous subvariants featuring distinct graph inventories, graph-phone correspondences and orthographies. In such a scenario, an effective transliteration would, on the one hand, need to accurately capture and portray a vast amount of graphemic, graphetic, phonemic, phonetic and orthographic information. On the other hand, it would also need to be as economical and legible as possible. Clearly, these criteria involving simultaneous comprehensiveness and compactness are difficult to reconcile. For this reason, this book will not employ a transliteration to depict Sindhi’s writing systems. Instead, it will make use of transcriptions of Sindhi words into the IPA, both phonemic and phonetic, as needed. In the absence of a specific transliteration, Sindhi and South Asian names and terms will be spelt according to their generally accepted 𝚎𝚗-𝙻𝚊𝚝𝚗 spellings, as in ⟨Sindh⟩, ⟨Hyderabad⟩ and ⟨Kutch⟩. If a word has more than one commonly attested 𝚎𝚗-𝙻𝚊𝚝𝚗 spelling — as in ⟨Devanagari⟩ and ⟨Devnagri⟩ — one of them will be adopted on a discretionary basis and, if required, be accompanied by an IPA transcription. In terms of transliterations as objects of study, the book will make one borderline exception — the Sindhi-Roman writing system as used in the Linguistic Survey of India (𝚜𝚍-𝙻𝚊𝚝𝚗-𝚡-𝚕𝚜𝚒; §12.2). Despite being explicitly designed and used as a transliteration for auxiliary purposes, 𝚜𝚍-𝙻𝚊𝚝𝚗-𝚡-𝚕𝚜𝚒 is included as a subvariant of Sindhi-Roman due to its graphematic insights and historical-sociolinguistic significance. The question of sociolinguistically classifying Sindhi-Braille (𝚜𝚍-𝙱𝚛𝚊𝚒) is taken up in detail in Chapter 11.

Aside from transliteration and transcription, a third type of sociolinguistically defined homolingual writing system is shorthand. According to Daniels (1996b, p. 807), shorthand is a “notation system” for recording words quickly, and designed to prioritise speed over phonological accuracy. The sociolinguistic aim and graphematic design of a shorthand system result in two distinctive characteristics. First, since a text written in shorthand may gloss over certain phonological aspects of the spoken utterance it records, it usually needs to be rewritten in another, more widely-known homolingual writing system — or orthography — for the recorded phonological information to be made transparent. Else, deciphering the shorthand text may prove difficult, if not impossible, for anyone other than the writer. Second, since shorthand notations are usually known to and utilised by specialised user groups, a shorthand system may serendipitously serve as a ‘secret’ writing system restricted to its user base.

Despite these properties, there remains a third requirement for a writing system to be termed shorthand — the existence of a homolingual ‘full’ writing system, or longhand. Hence, a shorthand can only be considered to exist when there also exists a corresponding graphematic longhand. Sociolinguistically, the longhand serves as the general-purpose writing system, while the shorthand acts as a special-purpose system restricted to specific uses or users.

As alluded to at the end of the previous section, Sindhi’s writing systems have been classified in academic and popular discourse under a variety of descriptors. Just as the graphematic labels abjad and alphasyllabary were applied based on sociolinguistic practices, so too were labels like auxiliary system and shorthand. In the absence of grapholinguistic tools in the past, terminological equivocation was understandable. However, in view of recent scholarly advancements and the introduction of sophisticated grapholinguistic models, a fresh approach to analysing Sindhi’s writing systems from graphematic and sociolinguistic perspectives is now necessary. Such an endeavour includes reclassifying Sindhi’s writing systems under appropriate graphematic and sociolinguistic labels, based on their synchronic use.

Biscriptality and multiscriptality

The specific phenomenon investigated in this book is that of a particular language being written in multiple ways, be it in more than one script, orthography or graphetic style. This phenomenon has been recognised in the literature for some time now, typically under the label digraphia (Grivelet, 2001) and characterised as the written counterpart to spoken-language diglossia (Ferguson, 1959). It has also been referred to by a variety of other appellations, including biscriptalism and multialphabetism, among others (Bunčić, 2016a, p. 16). Inconsistent terminology, coupled with a lack of robust methodologies, have hindered rigorous analyses — especially comparative analyses — of such sociolinguistic situations. Aiming to plug this epistemological gap, Bunčić, Lippert and Rabus (2016) propose a typological framework for analysing what they term biscriptality. According to the authors, biscriptality is the simultaneous use of more than one script, orthography or graphetic style in writing a given language (Bunčić, 2016c, pp. 51, 54). It follows that all instances of biscriptality are, by definition, homolingual in nature. The proposed analytical framework categorises biscriptal writing along two principal dimensions of variation – graphematic and sociolinguistic.³⁰ The graphematic dimension or axis further distinguishes among biscriptal writing in three ways, depending on whether they differ in the scripts, graphetic styles or orthographies they employ. Along the sociolinguistic axis, biscriptal writing is again categorised into three subtypes, based on the conditioning factors at play — the situation of use, the user group, or none. When the choice of biscriptal variant is decided by the purpose of writing, the result is complementary distribution between the variants, often accompanied by an implied difference in status. One may say the variants exist in ‘vertical’ opposition. When the choice of variant depends less on what the writing is about and more on who the writers (and intended readers) are, the outcome is complementary distribution based on user group – a ‘horizontal’ opposition. When there is no evident conditioning factor behind the choice of biscriptal variant, the choice may be considered ‘free’ in principle — although this is rarely the case in practice. The resultant 3 × 3 matrix allows for precisely and transparently situating attested instances of biscriptality based on their characteristics, as shown in a simplified format in Table 2.2.³¹Additionally, some of the nine categories of biscriptality may feature further subcategories, not shown here for brevity.

Despite the name, the Biscriptality framework also overs homolingual writing practices that differ in not just in the script used, but also the graphetic style or orthography employed. Consequently, going by the definitions and approaches used in this book, only those biscriptal variants that differ in their script component would be considered distinct writing systems (§2.3). Biscriptal variants that differ only in graphetic style or in orthography are usually considered subvariants of the same writing system. Also, graphetic style here refers to the specific typeface or calligraphic hand in use, exemplified by German in the Roman script appearing in the Fraktur style (𝖆𝖇𝖈𝖉…𝖝𝖞𝖟) on the one hand, and in the Roman or ‘Antiqua’ style (abcd…xyz) on the other (Spitzmüller & Bunčić, 2016). As alluded to in Section 2.2, the Biscriptality framework uses the term glyph to describe the surface typographic or calligraphic allograph of an underlying graph.

With regard to scope, the authors intend the term biscriptality as a cover term for a language written in multiple ways, although they make a provision for the term multiscriptality to be used where warranted:

[We] treat biscriptality as a cover term encompassing multiscriptality as a special case where more than two [scripts, graphetic styles or orthographies] are involved. Mathematically one might of course argue that multi ‘many’ means more than one, so that biscriptality should be a special case of multiscriptality. However, the analysis of concrete examples of multiscriptality with three (or even more) [biscriptal variants] shows that these are often complex cases of double (or triple) biscriptality nested into each other. […] Therefore biscriptality seems to be the term with the greater analytic power (although multiscriptality can of course be used to describe a situation of complex biscriptality).

(Bunčić, 2016c, p. 52; emphasis in original)

Given that the case of Sindhi seems to qualify as “complex biscriptality”, the term multiscriptality will be preferred in this book.

A cornerstone of the Biscriptality framework is its restriction to instances of simultaneous, concurrent or synchronic biscriptality (Bunčić, 2016c, p. 54). In simple terms, a language must be written in two different ways during the same time period or era. The framework does not intend to analyse diachronic changeovers in script, graphetic style or orthography. That said, it is capable of analysing multiple instances of (synchronic) biscriptality within the same language but at different points in time. A language that undeniably fits this description is Sindhi, which not only features multiple instances of biscriptality across the centuries, but also exemplifies different biscriptal subtypes in different eras. This makes written Sindhi a particularly rich sociolinguistic site for harnessing the full potential of the Biscriptality paradigm.

In terms of its relation with graphematic typologies of writing systems, the Biscriptality framework appears, especially at first glance, to align somewhat with the concept of writing system modules as envisaged by the Modular Theory (MT). Despite superficial similarities, the overlap in scope is limited. The neatest overlap of the sociolinguistically oriented Biscriptality framework with the graphematically oriented MT is in terms of script, which, in both frameworks, refers to a graph superset underspecified for typographical or calligraphic style. When it comes to orthography, the Biscriptality framework aligns with the MT in certain cases such as American English and British English. Here, the sociolinguistic instance of orthographic pluricentricity represents the graphematic phenomenon of different orthographic modules being superimposed on a shared graph inventory and graphematic system. These sociolinguistic and graphematic interpretations are also conveniently captured by the IETF language tags applicable in this case — 𝚎𝚗-𝙻𝚊𝚝𝚗-𝚄𝚂 and 𝚎𝚗-𝙻𝚊𝚝𝚗-𝙶𝙱, which share a common language and script subtag but differ in their country subtags. In contrast, the Biscriptality framework analyses the Traditional and Simplified characters used in writing Mandarin as a difference in orthography (Klöter & Bunčić, 2016). However, the Traditional and Simplified characters may also be interpreted as different graph subsets of the same script. Yet another interpretation is that the two inventories represent two distinct scripts, suggested by the IETF language tags 𝚌𝚖𝚗-𝙷𝚊𝚗𝚝 and 𝚌𝚖𝚗-𝙷𝚊𝚗𝚜, respectively (§2.9). Finally, graphetic or glyph style as envisaged by the Biscriptality framework is not explicitly provided for under the MT, although it may be argued that a particular graphetic style may well be instituted by an orthographic module. Muddying the waters yet again is the fact that the IETF assigns script subtags to certain stylistic or glyph variants of the same script (§2.6). As a result, German-Roman printed in Antiqua and Fraktur styles may be indicated with the language subtags 𝚍𝚎-𝙻𝚊𝚝𝚗 and 𝚍𝚎-𝙻𝚊𝚝𝚏, respectively.

Even so, the limited concurrency between the Biscriptality framework and the MT should not impact on the perceived versatility and utility of each. Both frameworks carry potential for plausible and revealing analyses in their respective domains of specialisation, while also allowing for modifications and ameliorations as needed. For this reason, the MT and the Biscriptality framework will form the foundation for analysing and interpreting the graphematic and sociolinguistic properties exhibited by Sindhi’s writing systems, past and present.

☙ ❧

3 Sindh and the Sindhi people

This chapter provides a brief history of Sindh and its inhabitants, from prehistoric times to the modern day. In doing so, it recounts the twentieth-century split of Sindhi society into its present-day Pakistani, Indian and diasporic communities. This sociohistorical overview will contextualise the chapters that follow on the Sindhi language and its writing systems.

3.1 Ancient and pre-Islamic times

The Sindh region is located in the north-west of South Asia. Geopolitically, contemporary Sindh is a province of Pakistan. Geographically, it is flanked by the Pakistani provinces of Balochistan and Punjab to the west and north, respectively, and by the Indian states of Rajasthan and Gujarat to the east and south, respectively. To the south-west of the Sindh province lies the Arabian Sea and the delta of the Indus, an economically, culturally and historically significant river for much of South Asia.

Historically, the Sindh region finds mention as sindhu in the Mahabharata, an epic poem in Sanskrit about events supposed to have taken place around 3102 BC (Buck, 2000, p. xiv; Winternitz, 1981, p. 453).³² Up until the early twentieth century, these mythological allusions were the earliest recorded references to the region. In fact, the history of South Asia as a whole had until then been attested with some certainty only up to 326 BC, the year Alexander of Macedonia had invaded the north-west of the Subcontinent (Possehl, 2002, p. 3). However, the 1920s represented a watershed moment for our knowledge of Sindh’s history. Following the discovery of numerous large mounds along the valley of the Indus river, the Archaeological Survey of India (ASI) carried out a series of excavations at these mounds. Buried beneath were hitherto unknown sites of human settlement dating back to the Bronze Age (3300–1100 BC) (Habib, 2002; McIntosh, 2008). The discovery of these settlements pushed back South Asia’s recorded history to at least 2500 BC. Since the first settlement discovered was near the village of Harappa (in present-day Punjab province, Pakistan), the civilisation was termed the Harappan Civilisation (McIntosh, 2008, pp. 3–4).³³ Eventually, the sites as a whole became commonly known as the Indus Valley Civilisation (Marshall, 2004 [1931]). The civilisation reached its pinnacle between 2600 and 1900 BC (‘Mature Harappan Phase’; see Figure 3.1), and was highly advanced for its time.

After Harappa, another well-known site of the Indus Valley Civilisation was unearthed under a mound near the town of Larkano, in the north-west of present-day Sindh province. Estimates suggested that this settlement was one of the largest cities in the world for its time (Petrie, 2013, p. 88). Also found at the site were several skeletal remains of human bodies in unusual positions, apparently abandoned on the city’s streets. Theories have been put forward to explain the presence and positions of these skeletons (Wheeler, 1953, pp. 91–93), but none has been widely accepted (Habib, 2002, p. 64). Nevertheless, the skeletons have likely served as the inspiration for the Sindhi name of the site — [mʊənᶦ d͡ʑo d̪əɽo], meaning ‘mound of the dead’. The name has conventionally been spelt Mohenjo-daro in 𝚎𝚗-𝙻𝚊𝚝𝚗 since the publication of Marshall’s (2004 [1931]) work, with the alternative Moenjo-daro appearing occasionally (Kenoyer, 2016; Possehl, 2002, p. 3). Both spellings may also occur unhyphenated (Encyclopedia Britannica, 2021). Of late, Mohenjo-daro has been increasingly harnessed by Sindhi intellectuals as a symbol of the supposed antiquity of Sindhi culture (Falzon, 2004, p. 78; Kothari, 2009, pp. 1–2). This is notwithstanding a dearth of substantiated facts on the culture, religion or language of the Indus Valley civilisation.

The Indus Valley Civilisation started declining with the abandonment of its cities from 1900 BC onwards, the reasons for which are again not clearly known. Prominent hypotheses that have been advanced include a decline in trade, a change in course of the river Indus and destruction by invaders (Habib, 2002, pp. 61–66; Possehl, 2002, pp. 237–246). Albeit yet unproven, the last of these hypotheses has received some attention, for Sindh’s geographical location on the north-western border of South Asia has historically rendered it vulnerable to invasions from Central and Western Asia. The area was conquered by the Persian Achaemenid Empire in the sixth century BC, and by the Macedonian-Greek army of Alexander in 326 BC (Kulke & Rothermund, 2004, pp. 60–61). Subsequently, the Sindh region came under the rule of the Mauryans, Graeco-Bactrians, Scythians, Kushans, Sassanids and Huns, who brought with them and pursued various cultural and religious practices including Greek, Vedic and Buddhist (Bowersock, Brown, & Grabar, 1999).

3.2 711 to 1843: Arab and Islamic influence

In 711 AD, the seventeen-year old Arab general Muhammad bin Qasim invaded Sindh and defeated the local king Dahir.³⁴ This made Sindh the easternmost province of the Umayyad Caliphate (Campo, 2009). In 1025 AD, Sindh was seized from the Arabs by the Afghan king Mahmud of Ghazni. From 1050 AD onwards, Sindh was ruled by the native Sumro and Samo warrior clans, and then by the Turco-Mongol Arghun dynasty, all of whom had adopted Islam (Qalichbeg, 1902).³⁵ In 1593, Sindh was taken over by the emperor Akbar (Richards, 1995, p. 51) and annexed to the Mughal empire, which ruled much of the Subcontinent until British colonisation. From the seventeenth century onwards, Sindh was ruled by the local Kalhoro clan and then the Talpur clan as vassals of the Mughals (Qalichbeg, 1902).³⁶ Thus, despite not remaining under direct Arab rule for long, Sindh did remain under local Muslim rulers, predominantly Sunni, for more than a millennium. This rule left a profound Islamic influence on Sindh’s culture and language.

Yet, throughout the period of Muslim rule, Sindh continued to have significant communities of minority faiths (Boivin, 2008a; Boivin & Cook, 2010). Prominent among them were Shia Muslims of the Nizari Ismaili community (§5.1.4; Chapter 10), whose faith traditions incorporated numerous Indic elements (Asani, 1991; Gillani, 2004; Khakee, 1972). The non-Muslim minority of Sindh’s population — commonly, if anachronistically, labelled ‘Hindu’ — historically worshipped the water deity Jhulelal (Parwani, 2010). Later, they also began following the teachings of Guru Nanak (1469–1539), the founder of the Sikh faith (Khan, 2008).³⁷ Consequently, the belief system of non-Muslims in Sindh came to be “an easy blend of Sikhism and Hinduism” (Daswani & Parchani, 1978, p. 21). Major life events were marked by rites and rituals at Sikh temples, alongside Vedic and folk practices on an everyday level. Pending the emergence of a widely accepted umbrella term for Sindhis who observed a mélange of Sikh, Vedic and folk traditions, this socioreligious group will be referred to in this book as Hindu-Sikh. Contributing to the syncretism of Sindhi faith practices was a significant Sufi mystic influence (Ramey, 2008, p. 178). The influence of Sufism on Sindhi spiritual life persisted even after the British conquest of Sindh, leading Boivin (2020, p. 298) to claim that, up until the early twentieth century, adherence to the Sufi paradigm was “the best way to claim to be Sindhi, whatever the religion at hand”.

The intertwined and nonsegregated nature of belief systems meant that, in pre-British Sindh, “religious distinctions were not particularly important political categories of identification” (Boivin, Cook, & Levesque, 2017, p. 3). However, these fluid traditions were sometimes met with disapproval by orthodox observers, leading to discrimination. Despite negative reception in some quarters, the syncretic nature of Sindhi expressions of faith significantly contributed to sociocultural harmony in pre-British Sindh. It also left a lasting impact on the language’s writing systems (Chapter 5).

3.3 1843 to 1947: British rule

In 1843, the ruling Talpur chieftain of Sindh was defeated by the forces of Charles James Napier, a British general in the private army of the London-based East India Company (Cook M. A., 2016a). In 1857, the British government dissolved the East India Company and took over its possessions in South Asia, including Sindh. Subsequently, Sindh was amalgamated into the neighbouring Bombay Presidency. In 1936, it was made a separate province with its own Assembly (Boivin, Cook, & Levesque, 2017).

The early twentieth century saw the struggle for independence gaining momentum in British-ruled South Asia, or British India as it was then known. This period also saw a rise in Hindu-Muslim tensions, as the Muslim elite in certain parts of British India began to fear being dominated by the majority Hindus in an independent India. In 1940, the All-India Muslim League passed the Lahore Resolution — today known as the Pakistan Resolution (Swaab, 1973). The Resolution called for the creation of “independent states” for Muslims in British India, and stated, among other things, that:

no constitutional plan would be workable in this country or acceptable to the Muslims unless it is designed on the following basic principle, viz., that geographically contiguous units are demarcated into regions which should be so constituted with such territorial adjustments as may be necessary that the areas in which the Muslims are numerically in a majority as in the North-Western and Eastern zones of India should be grouped to constitute “Independent States” in which the constituent units shall be autonomous and sovereign.

(Swaab, 1973, p. 44)

This Resolution was passed in the Sindh Assembly in June 1947 (Jalal, 1994, p. 290), which meant that Sindh would become part of the new nation of Pakistan. Pakistan and independent India, comprising the Muslim and Hindu-majority areas of British India, respectively, came into being simultaneously at midnight between 14 and 15 August 1947. The port city of Karachi, which was the capital of Sindh and now Pakistan’s largest city, was designated the country’s interim capital (Khan Y. , 2017).

The Partition of British India triggered massive bloodshed and resulted in one of the largest mass migrations in recorded history. Between twelve and seventeen million people were displaced (Butalia, 1998, p. 3; Jalal, 1994, p. 1), and between two hundred thousand and two million were left dead (Butalia, 1998, p. 3). On both sides of the border, religious minorities began to fear for their lives. Most Hindus and Sikhs in what was now Pakistan, as well as a large number of Muslims from independent India, fled their ancestral homes and sought refuge on the other side of the border (Khan Y. , 2017). Initially, Sindh initially remained relatively calm, especially when compared to the level of violence seen in neighbouring Punjab. Hence, in the immediate aftermath of Partition, much of Sindh’s non-Muslim population stayed in place, albeit tentatively. However, in the months that followed, Karachi began to receive a large influx of Muslim refugees from independent India. By January 1948, religious riots began erupting in Sindh, resulting in an exodus of the province’s non-Muslim population (Zamindar, 2010). Although a section of non-Muslims of socioeconomically disadvantaged background remained in Sindh (Boivin, Cook, & Levesque, 2017, p. 7), most Sindhis of Hindu and Sikh background left their homeland to make a new beginning in independent India and elsewhere around the world.

3.4 Present-day

Pakistan

In terms of ethnolinguistic identity, the Pakistani census only reports figures based on self-reported ‘mother tongue’. Consequently, these figures need to be used as a statistical proxy for gauging ethnolinguistic self-identification. According to the 2017 Pakistani census, Sindh (Figure 3.2) had a population of 47.8 million (Pakistan Bureau of Statistics, 2017a), of whom 29.4 million were reported Sindhi as their mother tongue. Fewer than a million native Sindhi speakers were based in other parts of Pakistan, with the total number of Sindhi speakers in the country numbering 30.2 million (Pakistan Bureau of Statistics, 2017b; 2017c). While data on speakers of Sindhi as a second language are absent from the census figures, the number of such speakers would likely be statistically insignificant.

The ethnolinguistic composition of Sindh today represents a significant transformation from pre-Partition times, when native Sindhi speakers were in a much greater majority. In 1941, Sindh had a population of just over 4.5 million (Census of India, 1941), of whom the overwhelming majority were native Sindhi speakers. Partition led to around 800,000 Sindhi speakers fleeing Sindh, whose place was taken up primarily by Urdu speakers arriving from what was now independent India (Boivin, Cook, & Levesque, 2017, p. 6). These refugees came to be known as muhajir, Arabic for ‘refugee’ (Platts, 1884, p. 1098).³⁸ Although disliked by some in-group members due to its connotations of nonindigeneity, the label muhajir has today come to be somewhat coterminous with ‘Urdu speaker’ in Pakistan (Ayres, 2009; Siddiqi, 2012). Many Muhajirs settled in Karachi, while others ended up in regional cities of Sindh such as Hyderabad and Shikarpur.³⁹ As a result, the populations of these cities have gone from majority Sindhi-speaking before Partition to majority Urdu-speaking in present times (Khubchandani, 1998, p. 12; Rahman, 1995, p. 1008).

Today, Sindhi remains the official language of Sindh and is used in education and mass media in the province. In rural areas of Sindh, where the population numbers 23 million, the overwhelming majority — 21 million — are native speakers of Sindhi. In urban Sindh, however, the language has been reduced to a minority. Of an urban population of 24.8 million, only 8.3 million speak Sindhi as a mother tongue, compared to almost 8.5 million Urdu speakers. The linguistic divide is particularly stark in Karachi, where the population of 14.9 million people comprises only 1.2 million native Sindhi speakers but 6.7 million Urdu speakers. (Pakistan Bureau of Statistics, 2017b; 2017c). Thus, in the capital city of Sindh, Sindhi speakers are outnumbered by Urdu speakers by more than five-to-one. Consequently, Sindh has seen several instances of Sindhi-Muhajir tensions erupting into violence, not least over the issue of language (Shackle, 2014a). For an overview of ethnolinguistic conflicts in the recent past in Sindh and Pakistan, see Rahman (1995; 1999), Ayres (2009) and Siddiqi (2012).

Present-day Sindh also has a minority of just over 3.3 million Hindus, although it is unclear how many of them are native speakers of Sindhi (Pakistan Bureau of Statistics, 1998c). At any rate, since Hindus comprise less than one per cent of Sindh’s 47.8 million residents, they do not significantly impact the province’s language demographics.

India

Sindhis in India are dispersed all over the country, although the vast majority are settled in the western states of Maharashtra and Gujarat. Of the Hindu-Sikh Sindhis that migrated to independent India following Partition, most settled in metropolitan areas such as Bombay (now Mumbai), Poona (now Pune), Delhi and Ahmedabad. Since Hindu-Sikh Sindhis have historically been a mercantile community, Bombay as the commercial capital of India was a natural destination for them (Anand, 1996, p. 52; Tan & Kudaisya, 2000, p. 233). Hence, Bombay received the main wave of migration. Cities further away from Sindh, such as Bangalore (now Bengaluru) and Madras (now Chennai), and smaller towns such as Jaipur and Bhopal, received a smaller influx of Sindhi settlers. Those unable to find shelter in the cities proper ended up in refugee camps on their outskirts, often in abandoned British military barracks. Although the Indian government provided some help, the refugees in these camps were faced with lack of housing and the travails of starting life anew in a foreign land (Falzon, 2004, p. 41). Over time, several of these refugee camps grew into towns in their own right, prominent among them being Ulhasnagar and Pimpri in present-day Maharashtra state, Kubernagar and Gandhidham in Gujarat state, and Bairagarh in Madhya Pradesh state.

The total number of Sindhi refugees in India just after Partition was estimated at 800,000 (Daswani & Parchani, 1978, p. 7). According to the 1951 Indian census, 337,000 Sindhi refugees arrived in western India, of whom almost 88 percent had settled in urban districts due to their mercantile pursuits (Barnouw, 1966). Consequently, the number of Sindhis listed as tradespersons was 41 percent, compared to only eight percent for the overall Indian population. More significantly, the Sindhis had a relatively high literacy rate of 53 percent, compared to 24 percent for the overall population at the time (Falzon, 2004, p. 41).

As with the Pakistani census, the Indian census, too, only collects data on self-reported ‘mother tongue’ and not on ethnicity per se. This effectively makes the terms ‘Sindhi’ and ‘Sindhi speaker’ synonymous. According to the 2011 Indian census, (Iyengar & Parchani, 2021, p. 4), there were slightly fewer than 2.8 million Sindhi speakers in India, although this figure includes a million speakers of the Kutchi variety (Census of India, 2011a). In direct contrast to the Pakistani situation, the bulk of Indian Sindhis are based in urban areas of the country. Those who arrived from large urban centres of Sindh such as Karachi, Hyderabad and Shikarpur are often settled in correspondingly large Indian cities such as Mumbai, Pune, Delhi and Ahmedabad. On the other hand, persons hailing from the villages and smaller towns of Sindh typically reside in smaller Indian towns such as Ulhasnagar (Falzon, 2004, pp. 41–42), Pimpri, Bhopal and Adipur-Gandhidham. Currently, the largest concentration of Sindhi speakers in India is at Ulhasnagar, at an estimated 400,000 (Tare, 2010).

Sindhi diaspora

Although receiving a significant numerical boost thanks to the post-Partition exodus of non-Muslim Sindhis, the Sindhi diaspora around the world had been in existence for at least one hundred years prior. In fact, Falzon (2004, pp. 5–6) characterises the emergence of the worldwide Sindhi diaspora in terms of three distinct waves of emigration from Sindh. The first occurred with the British annexation of Sindh in 1843. Faced with uncertain business prospects following the takeover, Sindhi traders, primarily from Hyderabad and Shikarpur, started venturing into new lands to try their luck (Markovits, 2008). The second migration was an outcome of Partition, with those fleeing Sindh heading to independent India or to other parts of the world where they had family or business links. For some, India served only as a temporary post-Partition base before they moved on to other countries, both in the East and West. The third migration according to Falzon (2004) coincides with modern-day emigration from India for economic reasons. These waves of migrations have resulted in a present-day Sindhi diaspora spread across more than a hundred countries worldwide, from Japan to Malta to Panama.

3.5 Summary

This chapter has traced the rich and diverse history of Sindh and its inhabitants through the ages. More than a millennium of Islamic influence, coupled with frequent changes in rulers, have left a lasting impression on the people, culture and language of Sindh. In particular, it has made the people accustomed to uncertainty, and instilled in them an ethos of adaptability and pragmatism (Anand, 1996; Falzon, 2004). How these historical events have impacted the Sindhi language and its writing systems over the centuries and into the present day is explored and analysed in detail in the remainder of this book.

☙ ❧

4 The Sindhi language

The Sindhi language is an Indo-Aryan language (Khubchandani, 2007) native to Sindh. Varieties of the language are also indigenous to regions bordering Sindh (see Figure 4.2). As outlined in Section 3.4, Sindhi is spoken by around 30.2 million people in Pakistan, 2.8 million in India, and an unknown, albeit small, number in the 500,000-strong Sindhi diaspora worldwide (Falzon, 2004, p. 6), a speaker base of 2.8 million amounts to only 0.2% of India’s population, which, as of 2022, stood at 1.41 billion (World Bank, 2022). Consequently, the Sindhi community in India is, in Khubchandani’s words, a “microscopic minority” in the country (1995, p. 309).

This chapter describes the key elements of the language, including its phonology and morphosyntax, and concludes with a sketch of the language’s sociolinguistic development in modern times. In doing so, this chapter acts as a foundation for Part Two, which forms the descriptive core of this book.

4.1 Linguistic affiliation

The Sindhi language is grouped under the north-western branch of the Indo-Aryan subfamily of the Indo-European language family (see Figure 4.1). However, this classification has occasionally been contested. In the popular literature, one frequently encounters unproven assertions that the language of Mohenjo-daro, or of the Indus Valley Civilisation in general, was an ancient form of Sindhi. Less frequently, such assertions may surface in academic works (Allana, 1991, p. 1; Asani, 2003, p. 613). Khubchandani summarises prominent claims linking the modern Sindhi language to the yet-unknown language of the Indus Valley Civilisation, and characterises them as made “[u]nder the spell of language chauvinism” (Khubchandani, 2007, p. 687). There also exists scholarly conjecture that Sindhi might well contain a non-Aryan substratum. The author of one of the first major Sindhi grammars, Ernest Trumpp (1872), states on the one hand that “Sindhi is a pure Sanskritical language, more free from foreign elements than any other of the North Indian vernaculars” (p. i).⁴⁰ On the other hand, he also claims that Sindhi possesses “a certain residuum of vocables, which we must allot to an old aboriginal language, of which neither name nor extent is now known to us” (p. iii). Sindhi lexicographer George Shirt (1878) asserts that this underlying language is Dravidian in nature, and attempts to demonstrate the link between Sindhi and Dravidian languages by means of lexical-etymological comparison. In making his claim, Shirt alludes to the modern-day existence of a Dravidian language called Brahui in parts of Sindh and its environs. Shirt’s argument is reasserted by Sindhi pedagogue Parso Gidwani (2007) in his revised and enlarged comparative list of phonologically similar lexical roots in Sindhi and Dravidian languages. To bolster his argument, Gidwani reiterates the presence of Brahui in Sindh. The speculation of Sindhi having a Dravidian substratum is in a mutually reinforcing relationship with the hypothesis that the yet-unknown language of the Indus Valley Civilisation was of Dravidian stock (Habib, 2002, pp. 50–51; Mahadevan, 1977; Parpola, 1996; 2009). Still others have attempted to present a Semitic origin for Sindhi (Baloch, 1962). Nonetheless, based on available linguistic evidence, the majority of contemporary scholarly works classify Sindhi as Indo-Aryan (Cole, 2006).

To be precise, Sindhi is considered to have descended from a certain form of Prakrit, a group of Middle Indo-Aryan vernaculars (Bubenik, 2007), spoken in what is today Sindh (Grierson, 1919, p. 4; Jetley, 2000; Khubchandani, 2007, pp. 686–687). However, due to the region having been under Islamic influence for more than eleven hundred years, Arabic and Persian lexical items, phonemes and, to a lesser extent, morphosyntactic constructions have found their way into the Sindhi language (Cole, 2006; Jetley, 2000). Cole (2006, p. 384) states that Sindhi “undeniably reveals the impact of its long history of contact with speakers of other languages”. George Stack, a pioneering British author of Sindhi dictionaries and a grammar, has succinctly summed up the syncretic nature of modern Sindhi in stating that “[t]he Sindhi […] borrows from the Arabic, the Persian, and the Sanscrit [sic], to an extent only limited by the learning and fancy of the writer” (1849a, p. iii). Despite these influences, Jetley (2000, p. 40) claims that the basic Indo-Aryan structure of the language has “remained mostly unchanged”.

4.2 Spoken varieties

Standard Sindhi is based on the Vicholi variety (Khubchandani, 2007, p. 683; Nihalani, 1999, p. 131) spoken in Vicholo, central Sindh. Since the main city in the region is Hyderabad, the variety is occasionally referred to as Hyderabadi (Covernton, 1906, p. 10). In Sindh, the Vicholi variety is considered the standard for administration, literature and education. Other principal dialects of Sindhi listed by Grierson (1919) and Khubchandani (2007) include:

• Siroli, spoken in Siro, upper Sindh;
• Lari, in Lar, lower Sindh;
• Lasi, in western Sindh and the Lasbelo region of neighbouring Balochistan province in Pakistan;
• Thari or Thareli, in the Thar region of south-east Sindh.⁴¹

The Siroli dialect is also cited in the literature as Siraiki (Nihalani, 1978, p. 8). However, this is also the name given to a variety of southern Punjabi (Grierson, 1919, p. 140), and is being increasingly used in this sense (Masica, 1991, p. 18; Shackle, 2014b). Meanwhile, the northern Sindhi variety has come to be known as Siroli (Bughio, 2006; 2009) or Utaradi (Bughio, 2009, p. 30), meaning “of the north” (Mewaram, 1910, p. 3).⁴² Other speech varieties usually classified as Sindhi dialects include Kutchi (or Kachchhi) in the Kutch region of Gujarat state, and Jaisalmeri in the Jaisalmer region of Rajasthan state, both in present-day India.⁴³ Although the Indian census figures on the number of Sindhi speakers in the country includes speakers of Kutchi and Jaisalmeri, native speakers of these varieties may consider their mother tongues to be distinct from Sindhi proper (Khubchandani, 2007, p. 683). Such self-distinction is particularly noticeable with Kutchi speakers. Over the last several centuries, Sindh and Kutch have developed as separate political entities (Mallison, 2008, p. 67). Since 1947, the two regions have also been separated by the international border between Pakistan and India. Hence, Kutchi speakers may prefer to identify with the neighbouring Gujarati-speaking population, or maintain a distinct sociocultural identity. For an overview of the phonological differences between the Kutchi and Vicholi varieties, see Khubchandani (2007, p. 690). Regarding Jaisalmeri, Grierson (1919, p. 143) considers it a subvariant of Thareli, opining that “all these specimens are the same mixed forms of speech”, between Sindhi on the one hand and the Marwari speech variety in Rajasthan on the other.

Apart from Kutchi and Jaisalmeri, there are no Sindhi-related varieties native to present-day India. For a comprehensive treatment of Sindhi dialect features, see Grierson (1919). Grierson’s work is invaluable both for the wealth of information it contains, as well as for its historical significance.

4.3 Phonology

The phonemic inventory and phonological makeup of Sindhi is quite similar to that of most north-western and western Indo-Aryan languages. That said, scholars are not unanimous on the exact number and nature of phonemes in Sindhi. This section provides an overview of several areas of Sindhi phonology and highlights certain features that are significant from a grapholinguistic perspective. These features include implosive stops, consonant gemination, reduced vowels and age-based phonological variation. Scholarly disagreement in these areas is also discussed.

4.3.1 Consonants

Table 4.1 shows the phonemic inventory of consonants in standard Sindhi, based on Nihalani (1999), Cole (2006), Khubchandani (2007) and Iyengar and Parchani (2021). Phonemes of marginal or undecided status are shown in parentheses, while allophones are shown in square brackets.

Sindhi has the most comprehensive stop system of all Indo-Aryan languages (Nihalani, 1974). Traditionally, Sindhi consonants are classified into five places of articulation — labial, dental, retroflex, palatal and velar. Based on the Sanskritic model, phonetically alveolar oral stops are considered phonemically retroflex. Similarly, alveolar sibilants and liquids are classified as dental. Alveolo-palatal affricates are traditionally grouped under palatal stops (Masica, 1991, p. 94). Sindhi thus shares a large part of its consonantal inventory with neighbouring Indo-Aryan languages such as Punjabi, Hindi-Urdu and Gujarati. However, the Sindhi consonantal inventory transcends those of its neighbours in terms of two major features. First, Sindhi has a full set of five nasal consonants. This contrasts with the neighbouring Indo-Aryan languages, where velar [ŋ] and palatal [ɲ] do not for the most part manifest as independent phonemes. In fact, [ɲ] may be altogether absent in these languages, while [ŋ] may only appear as an allophone of /n/ when preceding a velar stop in a consonant cluster. That said, in Sindhi, only /n/ and /m/ among these five nasal phonemes appear word-initially. Second, Sindhi features a series of voiced implosive stops /ɠ ʄ ɗ ɓ/, which occur word-initially and medially (Table 4.2). Of these, /ʄ/ is traditionally considered palatal and /ɗ/ retroflex. This classification agrees with the etymological origin of the implosives from the gemination of the corresponding plosives /ɡ d͡ʑ ɖ b/, respectively (Trumpp, 1872, pp. 13–19). A dental implosive corresponding to the dental plosive /d̪/ is absent from the phonemic inventory. For a detailed articulatory analysis of implosives in Sindhi, see Nihalani (1986).

The realisation of the velar and palatal nasals, as well as of the implosives, becomes increasingly unstable as one moves away from the Vicholi region. In Thareli-Jaisalmeri towards the east and Kutchi towards the south, the implosives are reported to be poorly established and highly variable (Grierson, 1919, p. 147; Khubchandani, Sindhi, 2007, p. 690), or altogether absent (Grierson, 1919, p. 185). In recent years, the implosives have also been disappearing from standard Sindhi varieties in India, particularly in the speech of younger speakers who have limited exposure to spoken Sindhi. In the speech of such speakers, the implosives are typically realised as the corresponding plosives /ɡ d͡ʑ ɖ b/, respectively (Lekhwani, 2011, p. 34; Parchani, 1998, p. 18; Iyengar & Parchani, 2021).

Consonant allophony

Table 4.1 shows allophone symbols enclosed in square brackets. [n̪] is an allophone of /n/ preceding dental stops in consonant clusters, as in /ɗənd̪ʊ/ [ɗən̪d̪ᶷ] ‘tooth’. [ɽ̃] is an intervocalic allophone of /ɳ/, as in /maɳʱu/ [maɽ̃ʱu] ‘man’. This is reflected in the observation by early grammarians that /ɳ/ was occasionally interchangeable with /ɽ/ (Trumpp, 1872, p. 16). [w] and [wʱ] tend to be the realisations of /ʋ/ and /ʋʱ/ before back vowels, although this is subject to high idiolectal variation. In addition, [ɽ] and [ɽʱ] freely alternate with [ɾ] and [ɾʱ] in certain speakers’ idiolects (Nihalani, 1978, p. 103).

Aspirated sonorants

The symbols enclosed in parentheses in Table 4.1 are described by Nihalani (1999) as representing independent phonemes in Sindhi, but by Cole (2006) as phonologically indeterminate. Cole notes that it is difficult to assert independent phoneme status for [mʱ nʱ ɳʱ lʱ ɽʱ ʋʱ] in Sindhi, since they do not contrast with a phonological cluster of a sonorant and [ɦ] in the language. Khubchandani (2007), too, does not characterise the parenthesised symbols as representing independent phonemes in Sindhi. This perspective is graphematically reflected in the current Sindhi-Arabic (𝚜𝚍-𝙰𝚛𝚊𝚋) writing system (Chapter 6) in that each of [mʱ nʱ ɳʱ lʱ ɽʱ ʋʱ] is written as a multigraph.

Consonant clusters

Consensus is also lacking on the phonemic status of word-medial consonant clusters in Sindhi. Trumpp (1872, p. xxxiii) observes that there is great idiosyncratic variation in the pronunciation of a “compound consonant”, ranging from clear articulation as a cluster to phonetically rendering them as separate consonants with an intervening epenthetic [ᶦ]. He notes that this epenthetic vowel is “scarcely perceptible”. This phenomenon has been attested by several scholars over the years (Grierson, 1919, p. 23; Khubchandani, 2007, p. 691; Lekhwani, 1996, p. iv). As will be seen in Part Two of this book, idiosyncratic variation in the pronunciation of medial clusters results in corresponding variation in their graphematic representation, regardless of the writing system used. That said, certain varieties of Sindhi, especially Siroli in the north, feature the consonant clusters [ʈɾ] and [ɖɾ], including word-initially (Bughio, 2009). In the Vicholi variety of south-central Sindh, these clusters are realised as simple [ʈ] and [ɖ], respectively. Present-day 𝚜𝚍-𝙰𝚛𝚊𝚋 spelling (§6.5) reflects the Vicholi pronunciation (Grierson, 1919, p. 23). Thus, the graphematic representation may not be reflective of the pronunciations of speakers of northern varieties. Attempts have been made in the past to represent these phones with distinct graphs (Stack, 1849a, p. 9), but this practice has not caught on, presumably as it amounts to graphematic overrepresentation for the standard Vicholi variety. Cognates in southern varieties, especially Kutchi, are realised with [ʈɾ] replaced by [t̪ɾ]. Thus, the word for ‘three’ is realised in Siroli as [ʈɾe], Vicholi as [ʈe] and Kutchi as [t̪ɾe] (Grierson, 1919, pp. 140, 185).

Consonant gemination

Scholarly consensus is also lacking on whether consonant gemination in Sindhi is contrastive and, therefore, phonemic. Grierson characterises the Sindhi implosive stops as “double consonants”, but states that “[t]hese are really the only double letters [sic] in Sindhī” (1919, p. 22). Nihalani (1978; 1999) only states that intervocalic stops tend to be longer when they follow a lax vowel. Allana (1993 [1964], p. 151) states that phonological gemination does occur in Sindhi in certain cases, but does not elaborate further. Khubchandani (2007, p. 691) affirms that “doubling of consonants is not significant in Sindhi”, but then goes on to refer to the existence of a diacritic marker in the 𝚜𝚍-𝙰𝚛𝚊𝚋 graph inventory sometimes used to represent consonant gemination. He also offers the example of a geminate consonant in the word [ɪzzət̪] ‘respect’. In contrast, Mewaram’s (1910, p. 373) and Lekhwani’s (1996, p. 12) spelling for the same word reflects the pronunciation [ɪzət̪ᵊ]. Lekhwani is likely the most explicit on the issue of gemination. He states that phonemic gemination is not semantically significant in Sindhi, and on this basis, recommends against representing gemination in writing (Lekhwani, 1996, p. iii; 1997, p. xv).

Loan consonants and mergers

Due to a large number of lexical borrowings or loanwords from other languages, standard Sindhi has absorbed a number of nonindigenous φ-consonants. Among these, Khubchandani (2007, p. 689) considers the voiceless uvular stop [q] marginal, mentioning that it is only found in the “formal speech of Persian-oriented speakers”. Masica (1991, p. 105) echoes this sentiment in claiming that phonemic /q/ is either absent or not well established in Sindhi. Of late, Cole (2006) observes that Urdu influence is strengthening the phonemic status of [q] in urban Pakistani Sindhi. In contrast, Iyengar and Parchani (2021) note that [q] is essentially absent from modern Indian Sindhi. From a graphematic perspective, Varyani and Thakwani (2003, p. x) explicitly state that, 𝚜𝚍-𝙳𝚎𝚟𝚊 |क़्| is pronounced [k] in Indian Sindhi, making it homophonous with |क्| [k] (see Chapter 7). For a comparative overview of post-Partition phonological developments in Pakistani and Indian Sindhi, see Iyengar and Parchani (2021).

Historically, Sindhi phonology has absorbed [f z x ɣ] by way of loanwords. While [x ɣ] are found only in Arabic and Persian loans, [f z] occur in Arabic, Persian and English loans. Among Sindhis in India, [f z] seem to be relatively stable in the speech of younger speakers due to their proficiency in English and the reinforcing presence of [f z] in English. On the other hand, [x ɣ] are disappearing in the speech of younger Indian Sindhis, merging with [kʰ ɡ], respectively (Lekhwani, 2011, p. 34; Nihalani, 1978, pp. 2–3; Parchani, 1998, p. 19; Iyengar & Parchani, 2021). This merger is leading to the emergence of new homophones. For instance, both /səkʰi/ ‘female friend’ and /səxi/ ‘generous’ are realised as [səkʰi] in the speech of younger Indian Sindhis. An emerging phenomenon in Indian Sindhi is that of /pʰ f/ both collapsing into [f], even in native Sindhi words, again resulting in new homophones. An example is that of /pʰoʈo/ ‘cardamom’ merging with /foʈo/ ‘photograph’, with both being realised as [foʈo]. A similar phenomenon is attested in modern Hindi by Shapiro (2007, p. 286). Compared to standard Sindhi, Kutchi is reported as lacking not only phonemic /f z x ɣ/, but also /ɕ/, with the last one surfacing only as an allophone of /s/ (Acharya, 1966).

Other phones in the source forms of loanwords are usually approximated to the closest native Sindhi phoneme; for instance, the English alveolar stops /t d/ are usually realised in Sindhi as the corresponding retroflex stops /ʈ/ and /ɖ/, respectively, and represented as such in all its writing systems.

4.3.2 Vowels

Standard Sindhi has a system of ten phonemic vowels, namely /ə a ɪ i ʊ u e ɛ o ɔ/ (Khubchandani, 2007; Nihalani, 1999), shown in Table 4.3. Sindhi also features phonemic vowel nasalisation, with all ten oral vowels possessing nasalised counterparts.

Based on the Sanskritic model, the vowels of modern Indo-Aryan languages, including that of Sindhi, have been conventionally grouped into short and long categories (Masica, 1991, p. 111). In Sindhi, the former comprises the monophthongs /ə ɪ ʊ/, while the latter includes the monophthongs /a i u e o/ and so-called diphthongs /ɛ ɔ/. Notwithstanding the Sanskritic model, vowel length is not phonemically significant in modern Sindhi, and distinctions between phonological vowels are primarily qualitative (Nihalani, 1978). Therefore, this book categorises standard Sindhi’s vowel phonemes as lax /ə ɪ ʊ/ and tense /a i u e o ɛ ɔ/, respectively. In general, lax vowels tend to be phonetically short, and tense vowels phonetically long (Keerio, Channa, Mitra, Young, & Chatwin, 2014), although this depends greatly on word environment (Nihalani, 1978). Hence, the terms lax and tense are used in this book primarily for nomenclatural convenience rather than as accurate descriptors of phonetic quality and quantity. The so-called diphthongs /ɛ ɔ/ show variable realisation, ranging from [əɪ̯ əʊ̯] to [e o], respectively.

The distribution of pronunciation is described variously by different authors. Nihalani (1999, p. 133) only states that /ɛ ɔ/ “tend to be diphthongized”. Khubchandani (2007, p. 693) claims that /ɛ ɔ/ occur mostly in loanwords, and are often replaced by /e o/. The most comprehensive sociolinguistic treatment of these vowels is likely that of Bughio (2001). Bughio attributes the varying realisation of /ɛ ɔ/ to a nominal diachronic and synchronic stratification in Sindhi pronunciation, albeit only in the context of Pakistan. First, he draws a distinction between the speech of older, rural Sindhi speakers and younger, urban Sindhi speakers. These are termed the Old Variety and New Variety, respectively. The New Variety is further divided into so-called Hindu and Muslim varieties. Sindhi sociolects are, thus, classified into religiolects and chronolects (Adamson, 1998; Frellesvig, 1996). According to Bughio, Hindu New Variety speakers tend to realise these vowels as the mid-low monophthongs [ɛ ɔ], while Muslim New Variety speakers usually featuring the diphthongs [əɪ̯ əʊ̯] in their phonologies. In contrast, Old Variety speakers, both Hindu and Muslim, tend to realise these vowels as [e o], respectively. Bughio’s characterisation of the Hindu New Variety of Pakistani Sindhi is reasonably applicable to Indian Sindhi pronunciation as well.

Notwithstanding Bughio’s classification, there remains significant idiolectal variation in the realisation of /ɛ ɔ/. Unsurprisingly, scholars disagree on the quality of not just these two phones, but on the very nature of diphthongs in Sindhi (Keerio, 2011, p. 62). Regardless, Bughio’s classification of Sindhi dialectal variation into religiolects and chronolects is a compact and convenient one.

Reduced vowels

An oft-cited feature of Sindhi phonology is its vowel-finality. Some authors assert that all Sindhi words are vowel-final (Bughio, 2006, p. 98; Grierson, 1919, p. 22; Hardwani, 1991, p. iii). Others state that indigenous Sindhi words are vowel-final, while certain consonant-final loanwords that have not yet been phonologically indigenised remain as they are (Khubchandani, 2007, pp. 691, 701). Conversely, if a loanword that is consonant-final in the source language is considered to have been assimilated into Sindhi, it may have a lax vowel suffixed. This word-final lax vowel manifests in a ‘reduced’ form (see Table 4.4).

The reduced vowels [ᵊ ᶦ ᶷ] are essentially unstressed phonetic realisations of the lax vowels /ə ɪ ʊ/, respectively, in morpheme-medial or morpheme-final positions in polysyllabic words (Addleton & Brown, 2010, p. 15).⁴⁴ The distribution of reduced lax vowels in Sindhi aligns with the phenomenon known as the Indo-Aryan schwa syncope or schwa deletion rule (Bhatia, 1987; Choudhury, Basu, & Sarkar, 2004). Compared to their Sanskrit etyma, polysyllabic lexical items in modern Indo-Aryan languages often feature a deleted or elided oral schwa in certain phonological environments.⁴⁵ The rule may also apply to non-Sanskritic lexical items, especially when a root and affix(es) combine to form a stem (Bhatia, 1987, pp. 39–40). However, the schwa deletion rule applies to different extents to different modern Indo-Aryan languages. The environments where schwa may be potentially deleted in polysyllabic words, and their applicability to various modern Indo-Aryan languages are detailed in Table 4.5.

As evident from Table 4.5, the schwa deletion rule is not universally applicable to modern Indo-Aryan languages. In general, the applicability of the rule gets weaker as one proceeds to the peripheries of the Indo-Aryan-speaking region (§4.1). Sinhala does not exhibit schwa deletion at all, while Bengali and Marathi feature word-final deletion with exceptions (Masica, 1991). In Sindhi, the rule involves vowel reduction, and applies to all the lax vowels /ə ɪ ʊ/. Thus, applying the rule of lax vowel reduction in Sindhi results in the reduced forms in [ᵊ ᶦ ᶷ] in certain word positions. Furthermore, the rules in Table 4.5 may apply multiple times within the same word. Thus, if a Sindhi word contains multiple /VCVCV/ sequences where the medial /V/ is any of lax /ə ɪ ʊ/, this may result in more than one theoretical surface realisation, as shown in Table 4.6:

Arriving at the correct pronunciations for the examples in Table 4.6 requires applying the rules from Table 4.5 at the morpheme level before applying them at the lexical or word level. Clearly, this requires knowledge of morpheme boundaries and, consequently, a certain level of linguistic proficiency in Sindhi.

Reduced [ᵊ ᶦ ᶷ] in word-final position are grammatically significant in Sindhi, since the quality of the vowel often encodes person-​number-​gender (PNG) information (Grierson, 1919, p. 22). Compare, for instance, /kɪt̪abʊ/ [kɪt̪abᶷ] ‘book’ and /kɪt̪abə/ [kɪt̪abᵊ] ‘books’, or /masat̪ʊ/ [masat̪ᶷ] ‘mother’s sister’s son’ and /masat̪ɪ/ [masat̪ᶦ] ‘mother’s sister’s daughter’. Final [ᵊ ᶦ ᶷ] are also semantically significant, as illustrated by the words /bãsə/ [bãsᵊ] ‘smell’ and /bãsʊ/ [bãsᶷ] ‘bamboo’. Furthermore, the quality of a final phonological vowel is lexically conditioned, and not predictable from the phonological structure of the word. Also, consonant-final loanwords in Sindhi may be assigned a final reduced vowel in an arbitrary manner, much like the arbitrary assignment of grammatical gender in Sindhi to an inanimate loanword. That said, final vowels and grammatical gender often align; certain final vowels are characteristic of masculine words and others of feminine words. This is treated in further detail in Section 4.4.

The acoustic indistinctness of reduced [ᵊ ᶦ ᶷ] may lead to free variation among them in pronunciation, resulting in Sindhi [ɾət͡ɕᵊna] (see Table 4.5) also being realised as [ɾət͡ɕᶦna]. Besides being imperceptible to a nonfluent listener (Cole, 2001; Stack, 1849a, p. 10), reduced [ᵊ ᶦ ᶷ] are often dropped in rapid speech (Khubchandani, 2007, p. 692). Cole (2001) notes that reduced vowels in final position may also be devoiced following voiceless stops, adding to their imperceptibility. These reduced vowels are being lost in the speech of Sindhi youth who do not live in a comprehensive Sindhi-speaking environment. Remarkably, this has been attested not just in India, but also in Pakistan (Cole, 2006). More accurately, reduced [ᵊ ᶦ ᶷ] tend to feature only epenthetically in the New Variety phonologies, rather than contrastively. This has resulted in the emergence of new homophones, and consequently impinged upon the morphology of the language. For instance, both [kɪt̪abᶷ] and [kɪt̪abᵊ] may be pronounced [kɪt̪ab] by New Variety speakers. This is often frowned upon by older speakers (Iyengar & Parchani, 2021).

In Indian Sindhi, the merger or loss of various consonants and vowels as described above has shaped the New Variety of the language in India. In this chronolect, reduced vowels have largely been lost (Bughio, 2001), and implosives and velar fricatives have merged with their corresponding stops. The New Variety coexists with the Old Variety spoken by the older generation of Indian Sindhis, which remains phonologically more conservative, preserves reduced vowels, and also retains the distinctiveness of implosives and velar fricatives. The intergenerational difference in phonology has implications for Sindhi pedagogy and orthography design and forms a significant thread in Part Two of this book.

The phonologies of the old and new chronolects, therefore, form a spectrum. In such a scenario, speaking of a supposedly authentic Sindhi phonology becomes problematic. For this reason, using either extreme as the ‘standard’ phonology will be suboptimal as a reasonably accurate representation of the other. That is, both /kɪt̪abʊ/ and /kɪt̪abə/ on the one hand, and [kɪt̪ab] on the other, would be found awkward by speakers of the other chronolect. To overcome this dichotomy, IPA transcriptions of Sindhi in this book will depict a mid-point pronunciation that explicitly show lax vowels as reduced. In other words, phonemic /ə ɪ ʊ/ in unstressed or final position will always be shown as phonetic [ᵊ ᶦ ᶷ], respectively. This practice has some precedence in Sindhi grapholinguistics. In writing Sindhi in the Roman script (𝚜𝚍-𝙻𝚊𝚝𝚗) Grierson (1919) explicitly represented a reduced φ-vowel by a superscript version of the corresponding graph (see Chapter 12).

Reduced φ-vowels are also significant from a graphematic point of view. In present-day 𝚜𝚍-𝙰𝚛𝚊𝚋 writing, lax and reduced φ-vowels are conventionally unrepresented. In contrast, Sindhi-Devanagari (𝚜𝚍-𝙳𝚎𝚟𝚊) convention requires that all φ-vowels, whether tense, lax or reduced, be explicitly represented in writing. This results in disagreement on how to spell certain words in 𝚜𝚍-𝙳𝚎𝚟𝚊, since there is high intergenerational as well as idiosyncratic variation in the pronunciation of reduced φ-vowels. The fraught nature of representing reduced φ-vowels in contemporary written Sindhi is described and analysed in Sections 6.5, 7.5 and 13.2.

Speech varieties geographically distant from Vicholi, such as Thareli-Jaisalmeri and Kutchi, are generally understood as permitting φ-consonants to occur in clusters and in word-final position (Masica, 1991, p. 196). As a result, reduced φ-vowels in such varieties, to the extent they exist, may be considered epenthetic rather than phonemic (Grierson, 1919, pp. 147, 184–185). In the limited instances that such varieties are written, the above phonological features may be reflected in the graphematic practices adopted, with φ-consonant clusters written as such, and word-final reduced φ-vowels left unwritten.

Vowel allophony

In standard Sindhi, certain environments can cause φ-vowels to change their quality; for instance, the sequence /əɦɪ/ is realised allophonically as [ɛɦᵋ], (Grierson, 1919, p. 22; Trumpp, 1872, p. x). Similarly, /əɦʊ/ often surfaces as [ɔɦᵓ] in rapid speech. Thus, standard Sindhi /ɡəɗəɦɪ/ ‘female donkey’ is realised as [ɡəɗɛɦᵋ], while /ɡəɗəɦʊ/ ‘male donkey’ may be realised as [ɡəɗɔɦᵓ]. Metathesis may also be involved, resulting in [ɡəɗʱɛ] and [ɡəɗʱɔ], respectively. A similar — although not identical — phenomenon is attested in the neighbouring and phonologically similar Punjabi language (Gill, 1996, p. 397). An exception to this rule among Sindhi varieties occurs in Kutchi, in whose phonology surface [ɛ ɔ] tend to be analysed simply as underlying /ɛ ɔ/. This is reflected in the fact that, in written Kutchi, these φ-vowels are usually represented with distinct graphs of their own (Jobanpotra, 1922).

4.4 Morphology and syntax

This section provides a brief overview of Sindhi syntax and morphology for the uninitiated reader. Since sections of Part Two touch upon aspects of Sindhi noun and adjectival morphology, the scope of this section is restricted to these aspects. For comprehensive treatments of Sindhi verbal morphology, see Grierson (1919) and Cole (2001).

In terms of syntax, Sindhi has features similar to neighbouring Indo-Aryan languages. Sindhi is a head-final language; word order is nominally subject-object-verb (SOV), but is flexible to allow for topicalisation. In terms of morphology, Sindhi has a rich system of noun declensions and verb conjugations based on case, number and gender (Cole, 2001). Nouns usually end in one of [ᵊ a ᶦ i ᶷ u o], with those ending in [e ɛ ɔ] being uncommon. All nouns are classified into masculine or feminine genders, and are often distinguishable based on their final vowel. Nouns ending in [ᵊ a ᶦ i] are usually feminine, whereas those ending in [ᶷ u o] are usually masculine (see Table 4.7). Exceptions typically comprise animate nouns, such as [ɾad͡ʑa] ‘king’ and [maᶷ] ‘mother’. However, a few inanimate nouns are also irregular in this regard; [mot̪i] ‘pearl’ is masculine, and [kʰəɳɖᶷ] ‘sugar’ is feminine (Grierson, 1919, pp. 23–24; Shahaney, 1967 [1906], pp. 41–48).

Sindhi makes use of postpositions, which is typical of Indo-Aryan languages. Most grammatical cases in Sindhi are formed by a noun followed by a postposition, wherein the noun appears in a declined form known as the oblique (see Table 4.8).

Verbs are marked for tense, aspect, mood and agreement, with agreement occurring in number and gender with the head noun (Cole, 2001; Grierson, 1919). Adjectives in Sindhi are classified as declinable or indeclinable. Declinable adjectives agree in case, number and gender with the noun they qualify, whereas indeclinable adjectives are invariant (see Table 4.9).

Of late, a few morphological changes have emerged in Sindhi usage in India. In the literary variety, this primarily involves replacing vernacular Sindhi adjectival declensions with Sanskritic adjectival suffixes. Lekhwani (2011, p. 34) illustrates this phenomenon with the example of the Sindhi word [səmad͡ʑᶷ] ‘society, community’. While this word has traditionally been adjectivised as [səmad͡ʑi] ‘societal’, recent Indian practice is to use the Sanskritised adjectival form [səmad͡ʑɪkᶷ]. In the spoken variety, morphological changes are most evident in the speech of the younger generation. For instance, Parchani (1998, p. 20) observes that the negative copula [naɦe] ‘is not’ may be realised by younger Sindhi speakers as [nə aɦe], where the negative particle [nə] ‘not’ is simply juxtaposed with the copula [aɦe] ‘is’, rather than phonologically merged with it. Such changes may be considered the morphological counterparts to the intergenerational phonological changes described in Section 4.3. For a deeper examination of morphophonological and morpholexical changes in Indian Sindhi since Partition, see Iyengar and Parchani (2021).

4.5 Sociolinguistic overview

4.5.1 Linguistic status

Following Partition and the arrival of non-Sindhi Muslim refugees from independent India in Karachi, Sindhis became a minority in their province’s capital city (§3.4). Consequently, Urdu emerged as the lingua franca among Karachi’s multilingual population, which effectively eliminated the need for non-Sindhis in the city to learn the Sindhi language (Rahman, 1999, p. 27). A government proposal to have non-Sindhi children in Sindh’s schools learn both Urdu and Sindhi as part of the curriculum was never implemented (Bughio, 2006, p. 99). Instead, in 1954, Sindh and its neighbouring regions were assimilated into the territory of West Pakistan, as part of the government’s ‘One Unit’ policy (Rahman, 1995, p. 1010). As a result, Sindh ceased to exist as a distinct geopolitical entity. In 1958, Pakistan’s democratically elected government was overthrown in a military coup, resulting in further centralisation of language policy and the promotion of Urdu at the expense of the country’s regional languages (Bughio, 2006, p. 100). In 1970, the One Unit policy came to an end. With it, Sindh was restored as a province of Pakistan, and Sindhi was once again declared the province’s official language. However, this declaration was strongly opposed by the province’s now-sizable non-Sindhi population, who feared socioeconomic and political sidelining. Simmering linguistic tensions between Sindhis and non-Sindhis culminated in riots in several of the province’s cities (Bughio, 2006; Rahman, 1999). Although a compromise was negotiated, tensions between Sindhis and non-Sindhis persisted. Another period of military rule from 1977 to 1988 pushed the language controversy into the background, following which an uneasy linguistic status quo has prevailed in Sindh up to the present day. Recent qualitative research in Sindh reveals that, despite a high level of bi- and trilingualism and increasing use of Urdu and English in their daily lives, both Muslim and Hindu Sindhis in the province continue to retain a strong sense of ethnolinguistic identity (Abbasi, Khemlani David, & Ali, 2021; Khemlani David & Ali, 2021). For a history of language learning, teaching and policy in Sindh and Pakistan, see Rahman (1999; 2002).

In India, the post-Partition settlement of Sindhi refugees all across the country, primarily in large cities, resulted in a geographically scattered community. Consequently, the urban, cosmopolitan and multilingual environment that the Indian Sindhi community has found itself in since Partition has triggered linguistic changes. According to Lekhwani (2011, p. 33), Sindhi dialectal variation is disappearing in India, with dialects being levelled and the Vicholi variety tending to predominate. In contrast, Daswani (1989, p. 59) states that there is no accepted standard dialect of Sindhi in India, and that every speaker considers their own familial variety to be the reference. Khubchandani (2007, p. 684) notes that the Sindhi language has developed in different directions in Pakistan and in India in the post-Partition era. In Pakistan, the language is undergoing progressive Arabicisation and Urdu-isation, while in India, it is being infused with Sanskrit and Hindi elements, accompanied by a simultaneous purging of unassimilated Perso-Arabic elements. A common element, though, is that urban varieties of the language in both countries are also experiencing increasing Anglicisation. This is corroborated by Nihalani (1978, pp. 3–4) when he points out the reduction of Perso-Arabic elements in colloquial Indian Sindhi and an increase in Hindi and English influence. Sociolinguistically, Sindhi in India is typically used only in familial and cultural environments, with formal education in the language being minimal. Despite a steady increase in speaker numbers over the last seven decades, Sindhi in India has undergone a considerable reduction in functional load since Partition (Iyengar & Parchani, 2021). In large Indian cities where English, Hindi and one or more regional languages dominate, language shift away from Sindhi has also been attested (Iyengar, 2013). For analyses and critiques of phonological, morphological, lexical and sociolinguistic changes in Indian Sindhi since Partition, see Khubchandani (1963; 1998; 2007), Daswani (1979; 1985; 1989), Daswani and Parchani (1978), Lekhwani (2011, pp. 33–35) and Iyengar and Parchani (2021).

Outside of Pakistan and India, the Sindhi language continues to be spoken in the diaspora, albeit primarily by the older generation. Although reliable statistics are lacking, sociolinguistic studies carried out at regular intervals among sections of the worldwide Sindhi diaspora attest to language shift away from the language. For analyses of Sindhi-language proficiency and the sociolinguistics thereof in the worldwide diaspora, see Dewan (1989), Detaramani and Lock (2003), Khemlani David (1991; 1998; 2001) and Raina Thapan (2002).

4.5.2 Education and media

Although not compulsory anywhere in Pakistan or India, state-funded Sindhi-language education is available in both countries. In Sindh, education in the language exists at all levels, from primary to tertiary. Prominent public universities in the province offering doctoral research opportunities in Sindhi include the University of Sindh and University of Karachi.

In India, the initial years of the post-Partition era saw several Sindhi philanthropists endeavouring to set up Sindhi-language schools in areas where the community had settled in considerable numbers. Consequently, a number of schools providing instruction in Sindhi emerged in western Maharashtra and Kutch. However, Sindhis who had settled in other parts of India found it difficult to provide their children with education in their home language. In the first fifteen years after Partition, Khubchandani (1963, p. 29) notes that less than half of the total Sindhi population of India at the time had the opportunity of sending their children to Sindhi-medium schools. That said, he observes that a few such schools had introduced Sindhi instruction in 𝚜𝚍-𝙳𝚎𝚟𝚊, although the majority of them, especially in Maharashtra and Gujarat, taught in 𝚜𝚍-𝙰𝚛𝚊𝚋. While 𝚜𝚍-𝙳𝚎𝚟𝚊 has since become more widely available, the overall demand for Sindhi-language education has fallen drastically (Daswani, 1989, p. 59).

Since 1967, Sindhi has been listed in the Eighth Schedule of the Indian constitution, which makes it an official language of the country and eligible for government support (Daswani, 1979; Vaish, 2008). Apart from being the primary medium of instruction, Sindhi is also eligible to be taught as a language subject in schools under the Three-Language Formula mandated by the Indian government. This formula recommends that children be taught their “mother tongue”, Hindi and English in school, with one of the three languages being the primary medium of instruction. The choice of the “mother tongue” or third language is at the discretion of individual schools (Benedikter, 1999). However, elite schools in urban areas typically employ English as the primary medium of instruction and offer the state language as the third language. These are the schools often preferred by the Sindhi community (Daswani & Parchani, 1978, pp. 88–89). Such a schooling choice often precludes any presence of the Sindhi language in Sindhi children’s education.

As a result of diminishing community demand for education in the language, most Sindhi-medium schools in India have either closed down (Sharma, 2016; Vora, 2016; Wajihuddin, 2010), or are changing their medium of instruction to English, Hindi or a regional language (Sindhi Sangat, 2016). They retain Sindhi only as a language subject, if at all (Anand, 1996, pp. 114, 127). Unfortunately, reliable figures regarding the number and spread of Sindhi schools in India are scarce. The absence of up-to-date and openly accessible government figures means that statistics on Sindhi education in India need to be pieced together from independent academic sources. In western India, Anand (1996, p. 168) provides a figure of 28 schools in Ulhasnagar town that taught Sindhi at the time, at least as a subject. In the metropolises of Mumbai and Pune there are likely no Sindhi-medium schools remaining. A few schools and colleges in these cities offer the language as a subject (Jai Hind College, 2019; MUCC, 2016; St Mira’s, 2021). When taught as a subject, the written form is usually 𝚜𝚍-𝙳𝚎𝚟𝚊. In northern India, Lekhwani (2015) opines that opportunities for formal education in Sindhi are almost nil, save for Ajmer town. That said, language courses at certificate and diploma levels are run throughout the country with the help of central government funding (NCPSL, 2005; 2014; 2015). Options for further study in the language up to the doctorate level are also available at a few major universities, particularly the University of Mumbai.

In Pakistan, Sindhi enjoys a vibrant and vigorous literary output and media presence in its home province, with a significant number of books, magazines and television channels churning out Sindhi-language content in various genres. Lying in stark contrast to the situation in Pakistan is the minimal media presence of Sindhi in India. In terms of audio-visual content, the state-owned All India Radio (AIR) allocates 16 hours per week for Sindhi programmes, including news (Central Institute of Indian Languages, n.d.). Certain channels run by the Indian state-owned television broadcaster, Doordarshan, also have slots for Sindhi-language programming (SABSindhi, 2016). A private Sindhi television channel called ‘Sindhi Kutchi TV’, relaying from Adipur-Gandhidham in Kutch, was briefly available between 2013 and 2016 ("Encyclopedia of Sindhi", 2016). However, at the time of writing this book, there was no widely available Sindhi-language television channel, whether state-funded or private. A campaign has been underway to have the Indian government start a Sindhi-language channel (Live Law, 2015; Punjabi, n.d.), with prominent Indian Sindhi politicians lending their support to the cause (Sindhi Sangat, 2008).

In the meantime, technological advances the twenty-first century have enabled several Pakistan-based Sindhi-language media outlets, as well as smaller Indian ones, to make their audio-visual content available online on video-sharing websites. Being a fairly recent phenomenon, the rise and spread of online Sindhi-language content remains largely unexplored from a scholarly perspective. Although beyond the scope of this book, the linguistic and pedagogical impacts of online Sindhi-language content on the Sindhi community worldwide is a promising avenue for future research.

4.5.3 Literary development

While noting that Sindhi literature has often experienced interruptions due to social turmoil, Kloss and McConnell (1978, p. 466) reiterate that fresh literary activities in the Sindhi language have not been found wanting both in Pakistan and in India. Overall, post-Partition literary productions in Sindhi, both in Pakistan and India, encompass drama, essays, critical works, short stories and novels, biographies, folklore, translations and works in linguistics. A list of prominent post-Partition linguistic works on Sindhi can be found in Khubchandani (2007, pp. 685–686). An overview of lexicographic works in the language in modern times is given in Khubchandani (1988). For comprehensive overviews of Sindhi literature from medieval to post-Partition times, see Allana (1991; 2009), Jetley (1992), Jotwani (1992; 1996), Lekhwani (2011) and Schimmel (1963; 1974).

Following Partition and arrival in India, Sindhis in the country attempted to maintain their linguistic identity by publishing Sindhi-language newspapers, magazines and books, primarily in 𝚜𝚍-𝙰𝚛𝚊𝚋 and 𝚜𝚍-𝙳𝚎𝚟𝚊. A variety of material, including poetry and prose, was thus produced. At the time of writing this book, the Registrar of Newspapers for India (2022) listed the number of active registered newspapers and periodicals in 𝚜𝚍-𝙰𝚛𝚊𝚋 at twenty-one and 𝚜𝚍-𝙳𝚎𝚟𝚊 at two. Most of these periodicals are published from areas of high Sindhi concentration such as Ajmer, Ahmedabad and Mumbai-Ulhasnagar. In terms of literary output, Khubchandani (personal communication, August 2, 2012) notes that Sindhi-language output and writers, although not thriving, are nevertheless on the rise. This can be attributed in part to government grants and recognition such as the annual awards by the Sahitya Akademi, or the National Academy of Letters in India. Similarly, Rohra (2015) highlights the positive impact that governmental and institutional aid and awards given to Sindhi writers have had in increasing the amount of Sindhi literature published. Rohra, however, characterises the situation of written Sindhi in India as “strange”, considering that the number of publications in the language is increasing despite the number of readers decreasing. In contrast, Daswani (1989, p. 59) opines that literary activity in the language is very limited, as is the number of emerging writers.

Both Pakistan and India have state-funded institutions to ensure the Sindhi language’s development and promotion. In Pakistan, the Hyderabad-based Sindhi Language Authority (SLA) has consistently produced a wealth of monographs and pedagogical resources since its establishment in 1990, while also maintaining a healthy output of Sindhi-language academic journals. In India, the state-aided Indian Institute of Sindhology (IIS) and the National Council for Promotion of Sindhi Language (NCPSL) were established contemporaneously with the SLA. The IIS, founded in 1989, claims to be:

[A] centre for advanced studies and research in the fields related to Sindhi Language, Literature, Education, Art and Culture. Its primary aim is to preserve and promote the Cultural heritage of [the] Sindhi Community and ensure its continuity by disseminating it in the younger generation.

(IIS, n.d.)

The NCPSL, formed in 1994, has similar aims, undertaking:

action for making available in Sindhi language the knowledge of scientific and technological development as well as knowledge of ideas evolved in the modern context, […] [and undertaking] any other activity for the promotion of Sindhi language as may be deemed fit by the Council.

(NCPSL, n.d.)

Like the SLA, the IIS and NCPSL are both engaged in publishing Sindhi literary and educational material, as well as conducting language courses. The IIS also runs a state-recognised Sindhi-language school in Adipur, Kutch.

4.6 Summary

This chapter has outlined the linguistic features of the Sindhi language relevant to this study, especially its phonology. It has also shown how the transplantation (Khubchandani, 1995) of the community in independent India has brought about significant changes in the language’s present-day status, role in education, and consequently, everyday use. Evidently, the changes in domains and patterns of language use have brought with them corresponding changes in writing system use.

Changes in use, however, are not new in the context of the Sindhi language, be it in its spoken or written forms. As evinced by the phonological, morphosyntactic and sociolinguistic descriptions in this chapter, the language has absorbed a variety of influence over the centuries. Accordingly, its spoken form has undergone several metamorphoses, as have its written forms. Part Two of this book is devoted to tracing and examining the evolution of Sindhi’s various written forms over the centuries, and to unearthing graphematic and sociolinguistic insights from them.

☙ ❧

5 Written Sindhi: A chronology

The history of Sindhi’s writing systems goes hand in hand with the history of written literature in the language. Yet, scholarly works focusing primarily on Sindhi writing systems per se are few and far between. Hence, works on Sindhi language and literature often serve as useful secondary sources for information on historical script trends. Harnessing these secondary sources to enrich and contextualise original research, this chapter presents a comprehensive chronology of the writing systems used for the Sindhi language, from the tenth century AD to the present day.⁴⁶

5.1 Pre-1843

As described in Sections 3.1 and 4.1, neither do we have much substantiated information on the language of the Indus Valley Civilisation, nor have the graphical marks and symbols on the seals excavated at Indus Valley sites been convincingly deciphered. This is despite more than a hundred attempts at decipherment having been made since the 1920s (Robinson, 2015, p. 500). In terms of linguistic decipherment, two of the best-known recent attempts are those of Mahadevan (1977) and Parpola (1996; 2009), both of whom posit that the language of the Indus Valley Civilisation is of Dravidian stock. This hypothesis remains to be universally accepted (Possehl, 2002, pp. 127–140). Alternative claims are even more contested, such as that of the language being related to Vedic Sanskrit (Bright, 1990a; Mitchiner, 1978; Rao S. R., 1982; 1994). With regard to decoding the symbols on Indus Valley seals, researchers disagree on whether these symbols represent a linguistic writing system at all (Rao, et al., 2009; Sproat, 2010b). Yet, claims of graphematic decipherment are common. In his initial work on the Indus Valley Civilisation excavations, Marshall (2004 [1931], p. 423) claims that these symbols were the precursor to the Brahmi script of ancient South Asia. He concedes, though, that the language represented by these symbols is unknown. Certain scholars have attempted to draw a distinction between linguistic and graphematic decipherment, especially in the absence of any evident link between the Indus Valley symbols and subsequent glottography in South Asia. Notwithstanding his position of the Indus Valley language being of Dravidian origin, Parpola (1996, p. 165) emphasises that “[t]here is no connection whatsoever [of the symbols found on Indus Valley seals] with the earliest scripts of South Asia, Brahmi and Kharoshthi”. In this regard, Masica notes that:

[A] serious obstacle to any attempt to link the Brahmi script with the Harappan is the time-gap between the two, which may amount to 1,500 years. […] Despite these difficulties, it may be said that derivation of Brahmi from the Harappan script — somehow — remains the most popular view in India.

(Masica, 1991, p. 134)

Reflecting Masica’s sentiment is a cohort largely if not entirely comprising Sindhi primordialists, whose membership cuts across national and religious lines. This cohort claims that the language of the Indus Valley represents an ancestor of modern Sindhi, predating other linguistic influences (Asani, 2003, p. 613). By extension, the Indus Valley symbols are depicted as a precursor to the present-day writing systems of Sindhi. The intention behind such claims is to confer antiquity to the Sindhi language and its written forms. Khubchandani (2007, p. 687) lists a few authors who have attempted to link the Indus Valley symbols with a proto-Sindhi language and writing system. For reasons evident, such claims remain popular among the Sindhi intelligentsia, whether in Pakistan, India or in the diaspora. These claims may also have adherents in the academy, although scholars of repute tend to express their stance in a measured and nuanced manner, as exemplified by Allana (1993 [1964]):

هت موهن جي دڙي واري لکت کي نظرانداز ڪيو ويو آهي، ڇاڪاڻ تہ جيستائين اها لکت پڙھي نہ ويئي آهي، ۽ اهي ثابت نہ ڪيو وئو آهي تہ اُن لکت واري ٻوليءَ جي ساخت ۽ سٽاءُ ڪھڙي آهي، تيستائين موهن جي دڙي واري ٻوليءَ کي ڊاڪٽر بلوچ صاحب جي لفظن ۾ ”قديم سنڌ جي ٻولي“ چئبو، ۽ موهن جي دڙي واريءَ لکت کي ”قديم سنڌ جي ٻوليءَ جي لکت“ چئبو، نہ ڪ موجوده سنڌي ٻوليءَ جي لکت […]

I have not taken into account here the writing found at Mohenjo-daro. Until this writing has been deciphered and its underlying system and structure established, the language of Mohenjo-daro, in the words of Dr [Nabi Bakhsh Khan] Baloch, should be termed “the language of ancient Sindh”, and its writing “the written form of the language of ancient Sindh”, and not that of the present-day Sindhi language […]

(Allana, 1993 [1964], p. 34 footnote 1)

In this sense, the topic of the Indus Valley symbols is one that arises frequently in present-day intellectual opinion on Sindhi writing systems, particularly in India (§13.3).

In terms of attested folklore, oral poetry in what may be reasonably considered ‘Sindhi’ has been dated to the ninth century AD (Asani, 2003, p. 615). Over the years, a rich tradition of poetry, religious hymns and folk ballads emerged in the language. Indeed, Sindhi has been characterised as having one of the most extensive folkloric traditions among Indo-Aryan languages (Schimmel, 1974; Baloch, 1993). That said, several of these compositions were performative in nature, and tended to be propagated through oral means (Boivin, 2020). Therefore, early Sindhi ‘literature’ may be more accurately described as oral literature or orature. This was a trend common to folkloric traditions throughout South Asia, including in Sindh’s immediate vicinity. In the context of the Punjabi language, Grierson (1919) states that:

[i]t contains no prose literature, and the poetry written in it is of the nature of folk-poetry, stored in the memory, and seldom committed to writing.

(Grierson, 1919, p. 247)

Yet, for nineteenth-century British and European scholars and administrators, oral traditions, however rich, did not merit the same prestige as a written body of literature. As a result, Sindhi was often depicted by colonial-era writers as a language poor in literature (Grierson, 1919, pp. 12–13; Imperial Gazetteer of India, 1908), with the status of ‘literature’ being reserved mainly for written works composed under the aegis of the British (Aitken, 1907, pp. 481–485). Even those Europeans who devoted their lives to studying the Sindhi language often painted its associated folklore in unsympathetic tones, such as the German-born Christian missionary and Sindhologist Ernest Trumpp. In his pioneering publication of the verses of celebrated eighteenth-century Sindhi poet Shah Abdul Latif, Trumpp (1866) describes Sindhi folklore thus:

I endeavoured to collect, whatever I could, of the old literature of the country, consisting, as I soon found out, of an inexhaustible store of legends, ballads and songs, which had been handed down orally and committed to the memory of the travelling bards and singers; for written books are scarce in Sindh […] unfortunately I could hit upon no native work of any merit, prose being peculiarly disliked by the Sindhis, who have only an ear for jingling verses, like most semi-barbarous nations.

(Trumpp, 1866, pp. v–vi)

Trumpp’s unflattering portrayal of Sindhi orature and implicit preferencing of written literature is illustrative of colonial-era attitudes of the time. A pithy encapsulation of these colonial attitudes is provided by Boivin (2020, p. 26), who notes that “literacy and illiteracy were a primary cleavage for identifying a civilized and uncivilized people”.

In contrast to British-era perspectives, present-day scholarship tends to agree that written material in Sindhi existed as far back as the ninth century AD, making it more or less contemporaneous with the oldest attested orature in the language. Unfortunately, no specimens from this era have survived, partly due to the tropical climate of South Asia and the perishability of commonly-used manuscript materials such as palm leaf and birch bark (Salomon, 2007, p. 82). Consequently, the existence of the earliest written texts in Sindhi is inferable only through secondary references. Khubchandani (2007, p. 688) and Lekhwani (2011, pp. 26–27) state that the Quran was translated into Sindhi in the ninth century AD, presumably using an Arabic-based script. However, Schimmel (1963, p. 224) adds a layer of nuance to this claim, noting the lack of clarity on whether the translation was into an earlier stage of Sindhi or of a related north-western Indo-Aryan language. Chatterji (1958) refers to the supposed existence of a Sindhi version of the Sanskrit-language epic Mahabharata in the eleventh century AD, known only through its subsequent Arabic and Persian translations. The script used in the Sindhi composition, however, is not known. In discussing specimens of written Sindhi, Allana (1993 [1964]) lists several toponyms of Sindh mentioned in the Chachnama, a thirteenth-century Persian translation of a now-lost Arabic-language chronicle of the Arab conquest of Sindh (Qalichbeg, 1900).⁴⁷ Strictly speaking, the Sindh-related toponyms in the Chachnama are transcribed in accordance with the Persian-Arabic (𝚏𝚊-𝙰𝚛𝚊𝚋) writing system. For Allana, though, they count as early written specimens of Sindhi.

Nevertheless, there exists evidence that earlier stages of what might be reasonably considered Sindhi were written using multiple scripts even a thousand years ago. The tenth-century Arab traveller Al-Nadim cites accounts of there being two hundred scripts being used in Sindh at the time (Dodge, 1970, p. 34). Al-Nadim’s contemporary, the Persian scholar Al-Biruni, reports a more plausible three scripts in vogue in the region at the time. He gives their names as Ardhanagari (half-Nagari), Malwari and Saindhava (Sachau, 1910, p. 173).⁴⁸ Considering that Sindh was under Arab rule at the time (§3.2), there exists scholarly conjecture (Allana, 1993 [1964], p. 34; Baloch, 1962; Rahman, 1999, p. 22) that at least one of these scripts would have been an offshoot of the Arabic script. This speculation aligns with the evidence that, in pre-British times, the graph inventories used for writing Sindhi essentially fell into two broad categories — those of Semitic origin, namely the Arabic script, and those derived from the ancient Brahmi script of South Asia, known as Brahmic or Indic.

5.1.1 Unstandardised Landa

The Indic inventories used for Sindhi in pre-British times were part of a macrofamily of related but unstandardised inventories known collectively as Landa.⁴⁹ Being an unstandardised macrofamily, there is no IETF script subtag for the Landa inventories. Their genealogical relationship to Brahmi is shown in Figure 5.1.

The Landa inventories were well established in north-western South Asia by the sixteenth century AD (Grierson, 1916, p. 624), and were prevalent from Kashmir in the north to Gujarat in the south.⁵⁰ Besides sharing a common genealogy, these inventories were also visually similar. They existed along a graphetic continuum, varying gradually based on geographical location and user group (Stack, 1849a, p. 1; Trumpp, 1872, p. 1). Sociolinguistically, these unstandardised inventories were primarily used by traders for private bookkeeping purposes. Consequently, the Landa inventories came to be identified with their mercantile user group, and derived their local names accordingly. In Sindhi, the Landa inventories were known as Vanika (Asani, 2003, p. 623; Grierson, 1919, p. 14; Pandey, 2010a, p. 1) or Hatavanika (Lekhwani, 2011, p. 36), both being adjectival forms of vanio ‘trader’ (pl. vania). In later years, the synonymous Hindi-origin designation bania ‘trader’ was also applied to these inventories, particularly by European colonial authors.⁵¹ Graphematically, the unstandardised Landa inventories comprised a sufficient number of discrete graphs to permit distinct representation of most — although not all — phonological consonants present in north-western Indo-Aryan speech varieties (§4.3.1). Where they fell particularly short was in the available number of discrete graphs to distinctly represent φ-vowels. Typically, the Landa inventories possessed only three basic shapes or rasms for representing the ten or so φ-vowels typical of north-western Indo-Aryan speech varieties (§4.3.2). In the absence of graphetic or graphematic augmentation, the three available rasms were generally used to denote three underspecified vowel archiphonemes: low, high front and high back. Graphematically unencoded phonological information on vowel quality, including nasalisation, had to be mentally filled in by the reader from context. Also, the three rasms for vowel archiphonemes were typically restricted to word-initial or syllable-initial position, although they were occasionally, albeit unpredictably, used in final position in monosyllabic words. Unstandardised writing systems based on the Landa inventories were notable for having no generally accepted provision to represent word-medial φ-vowels (Shackle & Moir, 1992, pp. 34–35). It was up to individual writers to idiosyncratically insert one of the three rasms as matres lectionis if desired (Shackle & Skjærvø, 2006, p. 547). Unsurprisingly, graphematic practices commonplace today, such as word spacing and punctuation, were rarely observed consistently in the past (Pandey, 2010a, p. 1). It was the perceived graphematic deficiency inherent in such writing practices that resulted in the name Landa, meaning ‘clipped’ or ‘tailless’ (Grierson, 1904, p. 68; Pandey, 2010a, p. 1; Pandey, 2010d, p. 2).

Colonial-era British and European authors who encountered the unstandardised Landa-based systems in the nineteenth century were generally contemptuous of them. In his Sindhi grammar, Trumpp (1872, p. 1) dismisses the Landa-based systems as “utterly unfit for literary purposes”, claiming that “merchants themselves, after a lapse of time, are hardly able to reproduce with accuracy what they have entered in the ledgers”. Sharing Trumpp’s sentiment, Grierson (1919, p. 14 footnote) notes that:

[a]s regards the illegibility of this Wāṇikō character, there is a proverb, Wāṇikā akharᵃ bbuṭā, sukā paṛhaṇᵃ-khā̃ chhuṭā, the Wāṇikō letters are vowelless, (as soon as the ink is) dry, they are […] illegible.⁵²

(Grierson, 1919, p. 14 footnote)

Grierson (1919, p. 14 footnote) also reports a story on the supposed perils of employing Landa-based systems in written communication. A merchant once wrote to his son in Sindhi-Landa, asking to send [nəɳɖʱ(ɾ)i ʋəɦi puʈʰe sud̪ʱi] ‘the small accounts book with the cover’. The son interpreted the message as being asked to send [nəɳɖʱ(ɾ)i ʋəɦu puʈ(ɾ)ə sud̪ʱi] ‘the younger daughter-in-law with (her) son’. Similar apocryphal tales on the supposed misinterpretations arising from quick-and-dirty Landa-based writing continue to be popular among Sindhis even today (Falzon, 2004, p. 272). In a similar anecdote related to me by a contemporary Sindhi litterateur (interviewee 26M; see Chapter 13), someone wrote the following message to an acquaintance in Sindhi-Landa: [ɦu əd͡ʑᶦmeɾᶦ ʋᶦjo] ‘He went to Ajmer town’. The recipient read it as [ɦu əʄᶷ məɾi ʋᶦjo] ‘He died today’. Whether the interpretations cited in these jokes are graphematically plausible is a matter of debate, and will be examined further in Section 8.1 (Example (20)). Regardless, the persistence of humorous tales arising from the supposed illegibility of Sindhi-Landa implicitly indexes a more sober reality — that present-day literacy in Sindhi-Landa is practically nil.

Despite the now-legendary ambiguity of Landa-based writing and its consequent exoticisation, the practices observed seem graphematically plausible and consistent with their sociolinguistic functions. Graphematically, representing nearly ten φ-vowels in word-initial position using just three rasms has been attributed to influence from Semitic writing practices (Masica, 1991, pp. 150–151; Salomon, 2007, p. 93; Shackle, 2007, p. 651). Lekhwani (personal communication, December 8, 2014) also surmises that the Arabic-Arabic (𝚊𝚛-𝙰𝚛𝚊𝚋) convention of leaving word-medial φ-vowels unmarked, especially in quotidian writing, may have influenced the avoidance of medial vowel graphs in Landa-based writing. In other words, the possibility of a spillover effect from 𝚊𝚛-𝙰𝚛𝚊𝚋 practices onto Sindhi-Landa cannot be ruled out. Lekhwani’s conjecture is echoed by a major nineteenth-century work on world scripts (Faulmann, 1880a, p. 121), which ascribes Sindhi-Landa’s focus on consonant graphs to Semitic abjadic practice. To illustrate the nature of such abjadic writing, Sproat (2010a) provides the following example based on 𝚎𝚗-𝙻𝚊𝚝𝚗:

f u cn rd ths u cn b trnd as a scrtry nd gt a gd jb

(Sproat, 2010a, p. 53)

According to Sproat, a competent speaker-reader of English and 𝚎𝚗-𝙻𝚊𝚝𝚗 should be able to fill in the missing graphematic — and, in turn, phonological — information to accurately decode the message as ⟨if you can read this, you can be trained as a secretary and get a good job⟩. It is also evident that such writing, despite posing an increased cognitive load for the reader, is extremely convenient for the writer in terms of speed. Sproat’s example, thus, serves as a convenient demonstration of how, and why, the supposedly ‘deficient’ Landa-based systems managed to function effectively. For starters, the Landa-based systems were primarily used to maintain mercantile records and account books, which remained largely restricted to their authors and close associates. Consequently, readability by a wider audience was not a priority. Even when the intended reader was a third party, the writing did not pose insurmountable hurdles as long as the reader was reasonably proficient in the language represented. Moreover, as noted by Prinsep (1837, p. 352), “the inconvenience of this omission [of vowel graphs] is not much felt in the limited scope of mercantile correspondence, […] where the same sentences are constantly repeated”. Above all, the near-exclusion of vowel graphs allowed for quick writing and note-taking. Sociolinguistically, this proved to be a particularly valuable convenience to its primary user group of traders and merchants (Shackle & Moir, 1992, p. 35).

It is often mentioned in the literature that the unstandardised Landa inventories were a secret traders’ script (Anand, 1996, p. 8; Boivin, 2015; Falzon, 2004). Such depictions may be understood to imply that the graphematically opaque practices in question were deliberate, allowing traders to maintain their accounts in a cryptic form and conceal details of their dealings from the authorities. However, and as mentioned, it seems more plausible that these seemingly opaque practices arose simply from traders’ need for expedience and compactness rather than phonetic fidelity. Indeed, the two functions of Sindhi-Landa as an expedient writing system, as well as a secret code, go hand in hand. In other words, to the extent Sindhi-Landa was used as a secret code, it was likely a convenient extension of it being known only to a restricted user group (Falzon, 2004, p. 271). For these reasons, Landa-based writing systems remained in vogue for centuries across north-western South Asia, including in Sindh (Shackle & Skjærvø, 2006, p. 547).

Irrespective of their popularity for bookkeeping and private note-taking, the unstandardised Landa-based systems exhibited high levels of idiosyncratic variation and low levels of graphetic and graphematic normativity. These features prevented the systems as such from being extensively used for literary purposes (Hotchand & Bhojwani, 1982 [1915], p. 165 footnote). This was unsurprising, since literary writing is a genre intended to be read and understood by a wide audience, thereby necessitating consistency in graphetic form and graph-phone correspondences. Moreover, in pre-British Sindh, propagation by oral means remained the most common mode of transmitting folklore, with written compositions limited in number. Those Sindhi-language compositions that were written down were often intended for a specific audience, which, in turn, dictated the choice of script.

5.1.2 Arabic

In addition to the unstandardised Landa scripts, Sindhi in pre-British times was written in the Arabic script, not least because educated persons in Muslim-majority Sindh often acquired basic competence in 𝚊𝚛-𝙰𝚛𝚊𝚋 and 𝚏𝚊-𝙰𝚛𝚊𝚋 (Lekhwani, 2011, p. 38). As described earlier in this section, it is plausible that Sindhi was written in the Arabic script as far back as a thousand years ago, although first-hand evidence of such writing has not survived. Hence, the graph inventory and graph-phone correspondences of early Sindhi-Arabic (𝚜𝚍-𝙰𝚛𝚊𝚋) writing is not known with certainty. According to Jetley (1987b, pp. 196–197), a collection of Persian and Sindhi verses attributed to Sufi preacher Abdul Karim of Bulri (1537?–1622) were written down by his disciple Muhammad Shah in the year 1628. Jetley states that the Sindhi-language verses were written down in “Persian script”, although details of the graph inventory and graph-phone correspondences used are not stated. Also present in the same collection are Sindhi-language verses attributed to Qazi Qadan, whose work has also been found transcribed in Devanagari (§5.1.3).

In the seventeenth century, the 𝚜𝚍-𝙰𝚛𝚊𝚋 graph inventory and graph-phone correspondences were substantially refined by Abul Hasan, a theologian from the politically significant town of Thatta in south-central Sindh.⁵³ The refinements primarily comprised adding dots to existing graphs or rasms in the inventory to create distinct graphs for Sindhi-specific phonemes absent in the Arabic or Persian languages. Graphematically noteworthy among these modifications was the tendency to depict of aspirate φ-stops using only one rasm, modified by dots as required. This was illustrative of the emic view that the aspirated φ-stops were single phonemes, and not clusters of a φ-stop and /h/ (Trumpp, 1872, p. 3). This augmented version of 𝚜𝚍-𝙰𝚛𝚊𝚋 came to be known as “Abul Hasan Sindhi” (§6.2). In spite of this attempt at standardisation, ambiguities persisted in the 𝚜𝚍-𝙰𝚛𝚊𝚋 graph inventory and graph-phone correspondences in pre-British times. These mainly concerned inconsistencies in the application of dots for creating new graphs, and the graphematic underdifferentiation of certain phonemes (Trumpp, 1872, pp. 2–3). These matters would not be addressed until after the British conquest of Sindh.

5.1.3 Devanagari

In addition to the unstandardised and proto-standardised inventories from the Landa macrofamily, a non-Landa Indic script said to be occasionally used for the Sindhi language in pre-British times was Devanagari. That said, the name of the script as used in the context of pre-British era Sindhi writing requires further qualification. Strictly speaking, Devanagari (from Sanskrit deva- ‘divine’ and -nagari ‘of the city, urban’) is a specific graph inventory from the Nagari macrofamily (Maurer, 1976; Salomon, 1998, pp. 40–41). However, the term is sometimes used by present-day authors to refer to any Brahmic or Indic graph inventory, including Landa-based ones. According to Taylor (1883), such broad-based use of the term ‘Devanagari’ likely became entrenched thanks to its use in British-authored works of the late eighteenth century:

The term Devanagari, which would mean the divine or sacred Nagari, is not used by the natives of India, and seems to have been invented by some ingenious Anglo-Indian [i.e., English person in India] about the end of the [eighteenth] century. It has, however, established itself in works on Indian Palæography, and may be conveniently retained to denote that particular type of the Nagari character employed in printed books for the sacred Sanskrit literature, while the generic term Nagari may serve as the designation of the whole class of vernacular alphabets of which the Devanagari is the literary type.

(Taylor, 1883, p. 349 footnote 4)

In line with Taylor’s observations, specimens of written Sindhi composed in a Landa inventory may be described by certain authors as ‘Sindhi in Devanagari’. It is with these provisos in mind that we need approach Sindhi-language works written in graph inventories collectively identified as ‘Devanagari’. Most works in this category comprise Sindhi-language poetry on Sufi, Vedantic and spiritual themes. Prominent among them are the Sufi verses of Qazi Qadan (d. 1551), hailed as one of the earliest Sindhi poets (Asani, 2003, p. 616 ff.; Lekhwani, 2011, p. 37; Schimmel, 1974, p. 11), as well as the Sindhi-language verses of sixteenth and seventeenth-century mystics Dadu Dayal and Pran Nath (Khubchandani, 2007, p. 693; Lekhwani, 2011, p. 37).⁵⁴

As mentioned, the blanket characterisation of the graph inventories in question as ‘Devanagari’ may inadvertently gloss over underlying graphetic (and graphematic) differences among them. Moreover, the gradual variation in graphetic appearance among Indic-origin graph inventories used to write Sindhi has resulted in disparate interpretations and descriptions. That said, certain authors are particular about distinguishing between the graphetic forms used in the pre-British Sindhi manuscripts and in contemporary Devanagari. In analysing and interpreting these poems, Sindhi litterateur Hiro Thakur characterises the manuscript containing Qazi Qadan’s poems as being written in “پراڻي ديوناگري” [pʊɾaɳi d̪eʋᵊnaɡᵊɾi] ‘old Devanagari’ (Thakur, 1996, pp. 35–36). He also notes that several graphs in the inventory used were graphetically distinct from their present-day equivalents. Likewise, a Gujarati-language collection of Dadu Dayal’s works (Bhavnani, 2000) characterises their written form as follows:

એમનું સમગ્ર સર્જન એક જ લિપિમાં લખાયું છે. આજની દેવનાગરીના જેવી જ એ લિપિ છે.

His entire body of work has been written down in the same script, which is very similar to present-day Devanagari.

(Bhavnani, 2000, p. 11)

Along similar lines, the graphetic differences between modern Devanagari and the script in which Pran Nath’s poetry has been found written are evident in a facsimile on display at the Sindhi Language Authority in Hyderabad, Sindh (Hauze, 2016a; 2016e). Also relevant in this regard is the fact that some of these written specimens were discovered much after their contents were initially composed in oral form. For instance, a substantial portion of Qazi Qadan’s Devanagari-script works were discovered only in the 1970s, among a multilingual repository of manuscripts found in India’s Haryana state (Thakur, 1996).

Against this background, one needs to bear in mind at least two caveats when analysing works considered to be in Sindhi-Devanagari (𝚜𝚍-𝙳𝚎𝚟𝚊). First, the term ‘Devanagari’ may simply mean a graph inventory that is graphetically similar to present-day Devanagari, while also visually resembling other historical and contemporary scripts and inventories. Second, the 𝚜𝚍-𝙳𝚎𝚟𝚊 works in question may have been written down by other authors, sometimes much after they were initially composed. In such a case, 𝚜𝚍-𝙳𝚎𝚟𝚊 may simply represent a transcription of the original oral composition, with the choice of script being determined by the writer’s or intended audience’s proficiency. In other words, the choice of script may not have been the original composer’s choice, but decided based on user-oriented or use-oriented factors (§5.1.6).

Notwithstanding the above, there exists evidence of the Gospel of Matthew from the Bible being translated into Kutchi and printed in the Devanagari script in 1834. For specimens and an analysis of this text, see Section 7.2.

5.1.4 Khudawadi, Gurmukhi and Khojki

Although most Landa-based writing was unstandardised and idiosyncratic, there existed at least three inventories from the Landa macrofamily that exhibited evidence of proto-standardisation. During the eighteenth-century rule of the Kalhoro clan in Sindh (§3.2), the Landa inventory prevalent in the capital city at the time, Khudabad, underwent a modicum of graphetic standardisation. Consequently, the Landa-based graph inventory characteristic of Khudabad came to be known as Khudabadi or Khudawadi (Lekhwani, 2011, p. 36; Pandey, 2010a; 2010d).⁵⁵ Despite basic emic standardisation, Khudawadi was not used for writing any substantial amounts of Sindhi-language literature. In contrast, two Landa-derived graph inventories that did go on to be used in literary contexts were Gurmukhi and Khojki. Given the organic standardisation they went through, both Gurmukhi and Khojki are commonly understood as full-fledged ‘scripts’. Despite being associated with different faith groups in the popular imagination, the two scripts bear striking similarities to each other in terms of the time period in which they emerged, and the sociolinguistic contexts in which they were used.

Gurmukhi is a script whose Landa-based graphetic predecessor was geographically located in Punjab, and not in Sindh (Figure 5.1).⁵⁶ Its traditional inventory of thirty-five graphetic rasms led to the alternative appellation Paintih ‘the thirty-five’ (§9.1). According to legend, Gurmukhi in its present form was created, or least standardised, by the Sikh spiritual leader Guru Angad (1504–1552). The motivation behind the script’s standardisation was to accurately transcribe the Sikh community’s religious hymns, composed in Punjabi, Sindhi and other north-western Indo-Aryan varieties. Initially compiled in the early seventeenth century into a collection known as the Adi Granth (‘First Book’), the final, enlarged collection transcribed came to be known as the Guru Granth Sahib.⁵⁷ Accordingly, the Sindhi-language hymns appearing in the Adi Granth in Gurmukhi script form the earliest known examples of the Sindhi-Gurmukhi writing system (𝚜𝚍-𝙶𝚞𝚛𝚞). As Sikhism spread to Sindh and gained popularity, a section of the faith’s adherents in the region learnt Gurmukhi to be able to read Sikh scriptures. Predominant among the Gurmukhi-literate Sindhi population were women, as they tended to devote more time to religious rituals than men (Falzon, 2004, p. 54; Khemlani David, 2001, p. 231).

Aside from Sikh scripture, probably the most noteworthy examples of written literature in 𝚜𝚍-𝙶𝚞𝚛𝚞 dated to the pre-British era are the couplets of Sindhi poet Sami (1743?–1850) (Hauze, 2016f). Overall, though, Gurmukhi remained restricted to Sikh temples and books, and did not find wide use as an everyday script for the Sindhi language (Lekhwani, 2011, pp. 37–38). An in-depth description and graphematic analysis of 𝚜𝚍-𝙶𝚞𝚛𝚞 follows in Section 8.3.

Besides Gurmukhi, another Landa-based graph inventory that underwent a degree of emic normativisation in pre-British times was Khojki. The script originated within the Nizari Ismaili branch of Shia Muslims in Sindh and Kutch, and was used to transcribe the community’s canon of ceremonial hymns or ginans composed in Sindhi, Kutchi and other north-western Indo-Aryan speech varieties (Asani, 1991, pp. 4–5).⁵⁸ The name Khojki was first attested only in the mid-twentieth century (Virani, 2022), and is a neologised adjectival form of Khoja, a commonly used emic and etic ethnonym for the Nizari Ismaili community of South Asia (Chapter 10).⁵⁹ The term Khojki has since gained some currency in scholarly circles, but remains far from ubiquitous within the lay community (Virani, 2022). Based on its traditional forty-graph inventory, Khojki was also known as chaliha akhari or chari akhari ‘forty-letter script’ (§10.2).

Much like Gurmukhi, the origins of Khojki are folklorically traced to a community spiritual figure. Legend has it that Khojki was created, or at least polished, by Pir Sadruddin (also spelt |Sadardin|; d. 1416?), a fifteenth-century Ismaili missionary or pir active in the Sindh region (Allana, 1993 [1964], p. 37; Asani, 1987, p. 439).⁶⁰ Notwithstanding creation legends, the oldest extant Khojki-script manuscript is dated 1737 AD (Khakee, 1972). Yet, it is possible that some of the content of later Khojki-script texts was originally composed as far back as the fourteenth century (Shackle & Moir, 1992, p. 10). Given the potential presence of fourteenth-century elements, Schimmel (1963, p. 224) opines that the oldest surviving Khojki-script texts may well represent some of the earliest extant evidence of Sindhi religious literature. However, Allana (1993 [1964], pp. 32–34) posits a much older timeframe for the oldest surviving Khojki-script specimens. In the early 1960s, the Department of Archaeology of the Government of Pakistan excavated several potsherds around the town of Bhambhore in southern Sindh. The potsherds were dated to the eighth century AD and featured certain symbols on them. Allana argues that the symbols are glottographic in nature, and, based on graphetic comparison, equates individual symbols with specific graphs of the Khojki inventory. In so doing, he also attempts to reconstruct a Sindhi-language message encoded in the graphs, thereby implying that the inscriptions on the potsherds are an instance of the Sindhi-Khojki (𝚜𝚍-𝙺𝚑𝚘𝚓) writing system. Also, since the potsherds were dated to the pre-Islamic and early Islamic period in Sindh (§3.2), Allana contends that the symbols on them affirm that Sindhi was written in Khojki even in pre-Islamic times. In view of the unstandardised nature of the Landa inventories and the attested graphetic variation therein, Boivin (2015) notes that the symbols on the potsherds may well be compared with another proto-standardised daughter inventory of Landa, namely Khudawadi. Boivin does, however, concede that the symbols on the potsherds, if glottographic, may well be considered early graphetic variants of Landa-based writing. For a detailed graphematic analysis of the Sindhi-Khojki and Kutchi-Khojki (𝚔𝚏𝚛-𝙺𝚑𝚘𝚓) writing systems, see Chapter 10.

Khojki, Gurmukhi and, to a lesser extent, Khudawadi, represent the most significant and successful instances of proto-standardised Landa-based inventories emerging in pre-British times. Overall, the Landa inventories may be considered the indigenous scripts of north-western South Asia, particularly in Sindh and Punjab. Colonial-era authors who were often disdainful of the variable graphematics of Landa-based systems were nevertheless obliged to recognise their sociolinguistic significance. In the context of Punjab, Grierson (1916, p. 624) describes the regional Landa forms as the “true alphabet” of the region. Similarly, Stack (1849a) designates the local Landa inventories of Sindh — including Khojki and Khudawadi — simply as “Sindhi” to mirror community practice. Indeed, despite Khojki and Khudawadi undergoing considerable graphetic standardisation in British times, their respective user groups continued to refer to them homonymously as ‘Sindhi’ well into the twentieth century (Virani, 2022). An intriguing, if coincidental, outcome of this nomenclatural legacy is the four-letter IETF script subtag assigned to Khudawadi, which is 𝚂𝚒𝚗𝚍 (Pandey, 2010f).

5.1.5 Roman

By the early 1800s, the growth of British control over parts of South Asia resulted in increasing geopolitical focus on Sindh (Cook M. A., 2013; 2016a). Consequently, British and European interest in the people, culture and language of the region began to grow. The earliest European-authored works on Sindhi varieties were translations of the Bible in the 1820s and early 1830s, printed in the Arabic and Devanagari scripts (§6.2, §7.2). The choice of script was most often dictated by the availability of metal types.

European-authored works on Sindhi grammar and lexicography began to appear by the mid-1830s (Hoernle, 1885, p. 162). Significantly, most of these works featured Sindhi transcribed in Roman, resulting in a distinct writing system (𝚜𝚍-𝙻𝚊𝚝𝚗). Again, the choice of script was dictated primarily by typographical constraints, although the intended readership — people of European and British origin — also played a part. Probably the earliest such work is the 1836 Sindhi grammar of British colonial administrator and epigraphist William Henry Wathen, reviewed by famed Indologist James Prinsep in the Journal of the Asiatic Society (1837). Through Prinsep (1837, p. 351) and Stack (1849a, p. v), we know that Wathen’s work featured Sindhi transcribed in 𝚜𝚍-𝙰𝚛𝚊𝚋 and 𝚜𝚍-𝙻𝚊𝚝𝚗, although, in terms of function, the latter was intended only as a transliteration to the former. Unfortunately, I was unable to obtain a physical or electronic copy of Wathen’s work, because of which the details of the 𝚜𝚍-𝙰𝚛𝚊𝚋 and 𝚜𝚍-𝙻𝚊𝚝𝚗 graph-phone correspondences as used by Wathen remain unclear. Significantly, Wathen’s work contains a sample of the Khudawadi graph inventory, cited by Prinsep (1837) as “Sindhi”. This makes Wathen’s work likely the first instance of Sindhi-Khudawadi (𝚜𝚍-𝚂𝚒𝚗𝚍) appearing in print. This was followed by another instance of 𝚜𝚍-𝚂𝚒𝚗𝚍 in print, in the form of a lithographed handwritten English-Sindhi wordlist by British civil servant Edward Backhouse Eastwick (1843a). Eastwick’s work features Sindhi-language entries alongside their English glosses, with the Sindhi entries listed in both 𝚜𝚍-𝚂𝚒𝚗𝚍 and 𝚜𝚍-𝙻𝚊𝚝𝚗. The 𝚜𝚍-𝚂𝚒𝚗𝚍 writing featured in Eastwick’s work was illustrative of the graphematically unstandardised and phonologically underdifferentiated practice in vogue at the time. The 𝚜𝚍-𝙻𝚊𝚝𝚗 system essentially comprised applying 𝚎𝚗-𝙻𝚊𝚝𝚗 graphematic principles to transcribe Sindhi-language lexicon, augmented only by the inconsistent use of the graphs |á í ú| to denote the Sindhi φ-vowels [a i u]. Overall, neither the 𝚜𝚍-𝙻𝚊𝚝𝚗 nor the 𝚜𝚍-𝚂𝚒𝚗𝚍 systems seen in Eastwick’s work consistently distinguish between lax and tense φ-vowels, or between dental and retroflex φ-consonants.⁶¹ A subsequent version of this work (Eastwick, 1843b) was printed using movable type, albeit only featuring entries in 𝚜𝚍-𝙻𝚊𝚝𝚗 since Khudawadi-script types had not yet been developed at the time.

5.1.6 User-oriented and use-oriented writing

In the centuries immediately preceding British colonisation, Sindh remained under the sphere of influence of the Persianate Mughal dynasty (§3.2). Consequently, Persian served as the court and prestige language in Sindh for much of this period (Burton, 1851, pp. 63–69), and was used to maintain government records (UK Parliament, 1854). It followed that official scribes, regardless of socioreligious background, needed to acquire reasonable proficiency in 𝚏𝚊-𝙰𝚛𝚊𝚋. As it turned out, such Persian-literate scribes were often of Hindu-Sikh background, and came to be identified as a distinct social class called Amil (Burton, 1851).⁶² Significantly, even 𝚏𝚊-𝙰𝚛𝚊𝚋 writing existed in a complementary distribution similar to that observed in Sindhi-Landa. While official documents were written in ‘full’ 𝚏𝚊-𝙰𝚛𝚊𝚋, everyday or personal writing in 𝚏𝚊-𝙰𝚛𝚊𝚋 was often written using the shikasta or ‘broken’ calligraphic style — a handwriting style that enabled quick writing but was often difficult to read (Meidani, 2020).⁶³ In this regard, the graphematic practice in 𝚏𝚊-𝙰𝚛𝚊𝚋 attested in Sindh was consistent with practice in other systems, particularly the Landa-based ones, and similarly criticised by British observers (Burton, 1851, p. 69; UK Parliament, 1854, p. 347).

In contrast to educational and writing practices for lay purposes, education and writing practices in religious contexts was primarily in liturgical languages, namely Arabic for Muslim clergy and Sanskrit for Hindu priests. This also necessitated becoming functionally literate in the corresponding writing systems — 𝚊𝚛-𝙰𝚛𝚊𝚋 and Sanskrit-Devanagari (𝚜𝚊-𝙳𝚎𝚟𝚊), with the label ‘Devanagari’ subject to the caveats outlined in Section 5.1.3.⁶⁴ The mercantile classes, as already described, were literate in one of the various Landa inventories, the appearance of which was nominally conditioned by geography and socioreligious background. As a result, distinct graphetic styles of the Landa inventories became associated with specific areas of Sindh, as well as with socioreligious groups. Geographically identifiable graphetic styles were named after the principal city of the region, resulting in names like Khudawadi, Shikarpuri and Thattai. Styles typical of socioreligious groups took on the demonyms of those groups, such as those used by the Shia Ismailis and Sunni Memons (Stack, 1849a, pp. 3–8).⁶⁴ Although reliable statistics are unavailable, it is conceivable that a minority of merchants would have been conversant in their group-specific Landa inventory as well as in the script used for their community’s liturgical language (Allana, 1993 [1964], p. 36). Moreover, as outlined in Section 5.1.4, certain Hindu-Sikh women acquired proficiency in Gurmukhi, to read Sikh scriptures. Notably, Sindhi was transcribed in Roman by European and British authors just prior to British colonisation, although such use stemmed primarily from typographical limitations. Overall, though, education and writing practices in Sindh prior to the British conquest of 1843 was stratified by “vocational relevance” (Khubchandani, 1977, p. 34), further categorisable into a user-oriented and use-oriented stratification. Such stratification of graphematic adoption based on vocation, socioreligious background and even gender meant that different user groups tended to be conversant with different scripts. Regardless, spoken Sindhi, in its regional varieties, remained common to all sections of society (Aitken, 1907, p. 473).

Despite Sindhi varieties being the predominant medium of oral communication in the region, none of these local varieties had any official status up until the mid-nineteenth century (Boivin, 2020). The lack of administrative or societal recognition for Sindhi led Stack (1849a) to write:

The study of [Sindhi] has been always despised — the most by those who had pretensions to education. Learned Musalmans read Arabic and Persian; Hindoos the latter, or Punjabi and Hindi. The speech they learnt at their mothers’ breasts, was thought only fit for clowns.

(Stack, 1849a, p. vii)

It is debatable whether Sindhi was “despised” and “thought only fit for clowns” by its speakers at the time. What is less debatable is that the language was subordinate in social prestige to Persian, Arabic and Sanskrit. In part, the lack of official status for Sindhi explains why, prior to the British era, the language was written primarily in informal contexts and in a plethora of graphetically unstandardised graph inventories (Asani, 2003, p. 622).

5.2 1843–1947

By 1843, Sindh had fallen into British hands. The new rulers had a policy of running lower-level administration in the local language (Khubchandani, 2007, p. 696). However, Bartle Frere, the British-appointed Commissioner in Sindh, noted that knowledge of the Sindhi language was low among British officers stationed in the region, with only two officers reported to be conversant as of the late 1840s (Martineau, 1895, p. 122). Sindhi scribes and interpreters employed in the administration spoke Sindhi natively, but read and wrote primarily in 𝚏𝚊-𝙰𝚛𝚊𝚋 based on practice in the pre-British era and the absence of a generally accepted writing system for their native language (Aitken, 1907, p. 474). Proceedings in British-instituted courts in Sindh were described by Frere’s biographer thus:

The usual process in all official proceedings before Europeans was that the Sindi [sic] parties and witnesses spoke Sindi, a Moonshee interpreted between them and the European officer in Hindustani [Hindi-Urdu], and all was written down in bastard Persian.⁶⁶

(Martineau, 1895, p. 122)

For the colonisers, oral and grapholinguistic fluidity in official contexts evidently came across as aberrant practice, and needed to be replaced by a uniform spoken and written language. Consequently, Frere decided that any use of Persian in administrative contexts would be done away with, and would be replaced by Sindhi as the official language of the colonial administration of Sindh. In 1851, Frere decreed that administrative officers would need to pass an examination in spoken Sindhi (Aitken, 1907, p. 474). Additionally, he proposed a higher level of examination in which knowledge of spoken and written Sindhi would be tested, and recommended that the government bear the cost of printing Sindhi-language books in any script (Martineau, 1895, p. 123). However, employing written Sindhi in government records or in book printing necessitated a generally accepted writing system for it, comprising at least a standardised graph inventory and reasonably consistent graph-phone correspondences (Allana, 1993 [1964], p. 96). Under the chairmanship of Barrow Ellis, Assistant Commissioner in Sindh, a committee was established to choose a script and decide on the graph inventory and graphematic fundamentals of an official writing system for Sindhi. In addition to Ellis, the committee comprised eight local and two European members (Allana, 1993 [1964], p. 96; Jetley, 2007; Matlani, 2008). The local members of the committee were:

• Rai Bahadur Narayan Jagannath Vaidya, a Marathi Brahmin originally from Ratnagiri (today in Maharashtra state, India);
• Khan Bahadur Mirza Sadiq Ali Beg of Hyderabad;
• Diwan Pribhdas Anandram Ramchandani of Hyderabad;
• Diwan Udharam Thanvardas Mirchandani of Hyderabad;
• Diwan Nandiram Navani of Sehwan;
• Mian Mahomed Shah of Hyderabad;
• Qazi Ghulam Ali of Thatta;
• Mian Ghulam Husain of Thatta.

The two European members were George Stack and Richard Francis Burton, both civil servants with the colonial administration.

5.2.1 Arabic

Stack was the author of the first comprehensive dictionaries of Sindhi (Stack, 1849b; 1855) as well as a detailed grammar of the language (1849a). Despite favouring a standardised Landa-based inventory as the official script for Sindhi (Government of Bombay, 1857, p. 105), Stack used Devanagari to print Sindhi-language content in all his works. In the introduction to his grammar (1849a, pp. v–vi), he decides against any of the Landa inventories — which he collectively terms “Sindhi” — due to their supposed “scanty use of vowels”. He rejects Roman since its inventory would supposedly require considerable augmentation if a transparent writing system is desired. According to Stack, acquainting oneself with this augmented graph inventory would allegedly involve as much effort as learning a new script altogether. Yet, he acknowledges that the existing inventory of Devanagari, too, would require augmentation to align it with Sindhi phonology, albeit justifying such augmentation by claiming it would be minimal in nature. Notably, Stack admits that Gurmukhi had an advantage in “being […] more known to Hindoos in Sindh” than Devanagari, and concedes that the latter was largely unfamiliar to Sindhis themselves. Eventually, he selects Devanagari for his grammar and dictionaries on the grounds that it was better known to Europeans at the time (1849a, p. vi), since his works were primarily aimed at British officers in Sindh. For details of Stack’s version of 𝚜𝚍-𝙳𝚎𝚟𝚊, see Section 7.3. Regardless of his script preferences, Stack’s grammar remains invaluable for its comparative handwritten chart of the various graph inventories in vogue at the time for writing Sindhi (1849a, pp. 3–8).

Along similar lines, Burton (1851, pp. 152–157) outlines what he feels are the relative advantages and disadvantages of the various candidate scripts for an official writing system for Sindhi — proto-standardised Khudawadi, Gurmukhi, Devanagari and Arabic. Burton rejects both Khudawadi and Gurmukhi on the grounds of them being known only to small sections of the population. He considers Devanagari a strong candidate based on its familiarity to Europeans in the colonial administration but, like Stack, notes that the script was unfamiliar to most Sindhis themselves. Furthermore, he claims that the existing Devanagari inventory would need to be suitably augmented to create a 𝚜𝚍-𝙳𝚎𝚟𝚊 system that is suitably transparent and as biunique as possible. Ultimately, Burton advocates using Persian-Arabic (𝚏𝚊-𝙰𝚛𝚊𝚋) as the inventorial foundation for an official Sindhi writing system, since this inventory had already been employed in some 𝚜𝚍-𝙰𝚛𝚊𝚋 works by that time.⁶⁷ Burton does admit, though, that the 𝚏𝚊-𝙰𝚛𝚊𝚋 inventory had been “carelessly adapted to the language of Sindh, and by the confusion of points and the multitude of different sounds expressed by one letter appears difficult and discouraging” (1851, p. 155). That said, he asserts that the script would be familiar to most educated Muslims and Persian-educated Amils (§5.1.6). He also hopes that mass education would result in further standardisation of the writing system and elimination of its shortcomings.

Both Stack (1849a) and Burton (1851) tend to wax eloquent about their preferred scripts and gloss over their flaws, and focus instead on the purported drawbacks of the other scripts in the fray. For instance, Burton (1851) ignores the fact that the 𝚏𝚊-𝙰𝚛𝚊𝚋 inventory comprises distinct graphs for φ-consonants present in Persian and Arabic but absent in Sindhi. In fact, Burton seems to suggest that such graphs are necessary in any 𝚜𝚍-𝙰𝚛𝚊𝚋 system simply because Sindhi lexicon comprises a much greater number of Persian and Arabic loanwords than do other South Asian languages (p. 400). In doing so, he glosses over the fact that these loanwords have been assimilated into Sindhi’s phonology. Burton claims that the Landa inventories are unsuitable due to their lack of distinct graphs for φ-vowels, but underplays the fact that, in 𝚊𝚛-𝙰𝚛𝚊𝚋 and 𝚏𝚊-𝙰𝚛𝚊𝚋, there exists a similar — albeit sociolinguistically conditioned — practice of leaving word-medial lax φ-vowels unwritten (§6.1). On the other hand, Stack (1849a) shuns both Arabic and Roman due to the alleged inventorial augmentation involved, while simultaneously defending similar enhancement in Devanagari and Landa. Also noteworthy is Stack’s use of Devanagari in his publications for Europeans, while simultaneously advocating a Landa-based writing system for educational and popular use by Sindhis. Burton, too, cites the familiarity of the Arabic script to educated Muslims and Amils in Sindh as an argument in its favour. However, it is unclear whether he advanced this argument out of genuine interest in Sindh’s people, or as a red herring to detract from his subjective preference of the Arabic script (Boivin, 2020, pp. 44–47). Overall, both Stack and Burton considered familiarity to Europeans a significant factor in selecting an official script for Sindhi.

A widespread sentiment articulated in the writings of colonial officials at the time, including Burton, Frere and Ellis, was the apparent reluctance of Muslim and Hindu Sindhis to read and write Sindhi in a script semiotically associated with the other religion (Aitken, 1907, p. 474). That is, Muslim Sindhis were expected to reject any Landa-based writing system for Sindhi, just as Hindu Sindhis, especially the trading community, would stay away from any proposed 𝚜𝚍-𝙰𝚛𝚊𝚋 system. On this issue, Frere wrote that:

[A]fter the most careful and dispassionate consideration, Mr. Ellis […] agreed that it was as hopeless to teach the Hindoos generally to read their own language in an Arabic character, as it would be to induce the Mahomedans to adopt generally a system of writing which they would consider as more than savouring of rank idolatry.

(Frere, in Government of Bombay, 1857, p. 105)

Frere’s and Ellis’ position sits at odds with the reality that Brahmi-derived Landa-based inventories had been in longstanding use among Muslim communities in Sindh — particularly the Ismailis and Memons — not just for record-keeping but also for liturgical writing. The officials’ view is also dissonant with the fact that the Hindu-Sikh community of Amils formed a significant section, if not the majority, of Persian-educated munshis in Sindh at the time. What comes across as particularly curious is that Ellis’ script committee included three Persian-literate Amil members, all of whom would go on to author several 𝚜𝚍-𝙰𝚛𝚊𝚋 textbooks, grammars and dictionaries in the following decades (Boivin, 2020). Yet, Frere and Ellis appeared reasonably convinced of their notions of the semiotic connotations of scripts and mutual religious prejudices prevailing in Sindh at the time. However, given that a majority of Sindh’s population was Muslim, Frere and Ellis were inclined towards officially adopting the Arabic script for Sindhi. Eventually, Ellis’ committee assented and recommended that 𝚜𝚍-𝙰𝚛𝚊𝚋 be officially adopted. Existing proto-standardised graphematic practices (§5.1.2) were supplemented with graphetic innovations to produce an augmented graph inventory. Finalised in 1853, the new inventory was officially adopted in 1857 as the basis for written Sindhi for administrative and, subsequently, educational purposes (Asani, 2003, p. 625; Khubchandani, 2007, p. 696).

Despite finalising the graph inventory of 𝚜𝚍-𝙰𝚛𝚊𝚋, Ellis’ committee did not focus on finer orthographic matters, such as diacritic placement, spellings of individual words and collation order (Lekhwani, 2011, p. 38). Several questions on these aspects of the 𝚜𝚍-𝙰𝚛𝚊𝚋 writing system remain debated even in present times (§6.3, §6.5).

5.2.2 Devanagari

In 1850, Stack published a translation of the Biblical Gospel of Matthew into 𝚜𝚍-𝙳𝚎𝚟𝚊 (Grierson, 1919, p. 13; Hooper, 1938, pp. 131–132), printed as a lithograph. Stack’s use of 𝚜𝚍-𝙳𝚎𝚟𝚊 was consistent with his position of advocating a Landa-based writing system for educational purposes but using Devanagari in his own works. Even after 𝚜𝚍-𝙰𝚛𝚊𝚋 was decreed official, debate on the issue persisted and supporters of alternative writing systems for Sindhi continued to make their case. Most prominent among supporters of Devanagari for Sindhi was the German Christian missionary Ernest Trumpp (§4.1). Trumpp (1857, p. 685) criticises 𝚜𝚍-𝙰𝚛𝚊𝚋 as being loaded “with a confusing heap of dots and other diacritical marks”. In his now-classic grammar of Sindhi, Trumpp (1872) devotes an entire chapter (pp. 1–30) to the development, characteristics and a critique of 𝚜𝚍-𝙰𝚛𝚊𝚋. This is likely the most detailed work on the subject from that period. In this work, Trumpp criticises the 1853 version of 𝚜𝚍-𝙰𝚛𝚊𝚋 (𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝟷𝟾𝟻𝟹), claiming that it was as unsystematic and indiscriminate in its application of dots as was the previous unstandardised practice (§5.1.2). Instead, he maintains that a writing system based on Devanagari, which he calls the “Sanskrit alphabet” (p. 1), would be the best suited for Sindhi.

Notably, a constant thread across several of Trumpp’s works is the supposed existence of religious prejudices between the Muslims and Hindus of Sindh, which echoes the sentiments expressed by Burton, Frere and Ellis. Trumpp associates scripts with religious groups to the extent that he labels the Arabic and Devanagari scripts ‘Hindu’ and ‘Muslim’, respectively (Trumpp, 1872). Like the British officials in the colonial administration, Trumpp believes that prevailing prejudices meant that one group would not learn the script of the other:

As the population of Sindh consists of Hindûs and Muhammadans, two distinct alphabets will be required for them. In respect to the Muhammadans, all are agreed that only the Arabic character will do for them […] The national alphabet for the Hindûs is the Sanscrit [Devanagari] character, […]

(Trumpp, 1858, pp. ii–iii; emphasis in original)

Yet, despite considering Devanagari to be linguistically better aligned with Sindhi phonology, Trumpp concedes that its adoption would be impractical in view of Sindh’s Muslim majority and their purported preference for a 𝚜𝚍-𝙰𝚛𝚊𝚋 system. However, since he was opposed to 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝟷𝟾𝟻𝟹, he advocates for the graph inventory and graph-phone correspondences of Urdu-Arabic (𝚞𝚛-𝙰𝚛𝚊𝚋) to be adopted for Sindhi, with minor modifications as needed (Trumpp, 1857; 1872). In Trumpp’s opinion, the graph inventory of 𝚞𝚛-𝙰𝚛𝚊𝚋 — which he refers to by the then-prevalent name ‘Hindustani’ — is comprehensive enough to allow for lexical items of Sanskritic as well as Perso-Arabic origin to be accurately transcribed. Consistent with his position, Trumpp’s (1872) grammar features Sindhi words in a modified 𝚞𝚛-𝙰𝚛𝚊𝚋 system as well as in a Devanagari-based system. For an analysis of Trumpp’s 𝚜𝚍-𝙰𝚛𝚊𝚋 and 𝚜𝚍-𝙳𝚎𝚟𝚊 inventories, see Sections 6.4 and 7.4, respectively.

While aligning somewhat with Ellis’ views, Trumpp’s views on needing separate writing systems based on religious affiliation comes across as dissonant with the fact that communities in Sindh read and wrote in a variety of writing systems, the isographs for which did not coincide with religious boundaries. More importantly, Trumpp’s views sit uneasily with the statistic that, in the late nineteenth century, less than ten per cent of Sindh’s population was literate (Aitken, 1907, p. 480; Boivin, 2020, p. 14). Against this background, it was probably unsurprising that Trumpp’s renditions of 𝚜𝚍-𝙰𝚛𝚊𝚋 or 𝚜𝚍-𝙳𝚎𝚟𝚊 faded away in the decades that followed. Thanks to government backing, 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝟷𝟾𝟻𝟹 went on to be adopted and used by educated Sindhis of all backgrounds. Use of 𝚜𝚍-𝙳𝚎𝚟𝚊 remained restricted to certain genres and contexts, chiefly liturgical and scholarly. These included Sindhi translations of Biblical texts, and the Sindhi volumes of George Grierson’s Linguistic Survey of India (1919).

As alluded to in Section 5.1.3, it was common to see British and European works describing the unstandardised Landa forms, and sometimes even Khojki and Khudawadi, as derivatives or corruptions of ‘Devanagari’ (Lepsius, 1863, p. 105). Such usage could be attributed to the British’s greater familiarity with the graphetically distinct variant of Devanagari and to the greater sociolinguistic prestige of this variant stemming from its use in writing Sanskrit. It is plausible that such indiscriminate usage of the term ‘Devanagari’ in British and European-authored works sowed the seeds of confusion in some quarters, particularly in the minds of colonial administrators who may not have been entirely familiar with Sindh’s grapholinguistic reality. At the other end of the analytical spectrum were those who opined that Devanagari in the modern sense was historically unused in Sindh (Covernton, 1906, p. 10). Regardless of whether Devanagari in the modern graphetic sense was used to write Sindhi in pre-British times, there remains no doubt on 𝚜𝚍-𝙳𝚎𝚟𝚊’s existence and use in the British era.

5.2.3 Khudawadi

Notwithstanding the adoption of 𝚜𝚍-𝙰𝚛𝚊𝚋 for official purposes in Sindh, the British administration retained concerns about its use in public education in the province. Consequently, Frere and Ellis sought to standardise a Landa-based graph inventory for Sindhi as the basis for a second writing system for the language, to be used in educating Hindu-Sikh children. This Landa-based writing system came to be known as Hindu Sindhi (Government of Bombay, 1857, p. 106). An initial version of this system was prepared in 1856, but was put on the back burner. In 1862, there was a brief attempt to introduce the so-called Hindu Sindhi writing system in a handful of schools, which failed due to the lack of printed material in it (Education Commission, Bombay, 1884, p. 25). Throughout the 1860s, the low uptake of government-sponsored education by the Vania trader community — the chief users of unstandardised Landa — prompted some officials to revisit the matter. Prominent among them was Narayan Jagannath Vaidya, one of the members of Ellis’ script committee who had since become Deputy Educational Inspector in Sindh. Vaidya was reportedly of the opinion that the low turnout of Vania children in government-run schools was due to instruction being provided in 𝚜𝚍-𝙰𝚛𝚊𝚋, which was of little use to them in their customary occupation of trade (Government of Bombay, 1864, pp. 85–86). According to Vaidya, providing instruction in a Sindhi-Landa system would act as an incentive for Vania children to such schools, given the widespread use and practical value of Landa in their community (Government of Bombay, 1865, p. 89). Still, other government officials remained sceptical. The British-appointed Commissioner in Sindh at the time, Samuel Mansfield, wrote:

Sind labours under great difficulties in the language introduced by Government for education. The only persons disposed to allow their children to learn it are the Government servants and those who wish to obtain Government employment [i.e., Amils]. The agricultural population [Sunni Muslims], if they educate their children at all, have them taught Persian only; and the trading community, Bunnia-Sindee, which is quite a distinct language [i.e., writing system] from that taught in the Government schools.

(Mansfield, in Government of Bombay, 1865, p. 8)

A similar sentiment was echoed by Vaidya’s immediate superior, acting Educational Inspector JG Moore. Despite having high regard for Vaidya’s abilities, Moore felt that introducing a Sindhi-Landa system in schools would not solve the problem at hand. According to him, Vania children had traditionally acquired a working knowledge of their community-specific Landa inventory from their elders. Consequently, there was no need for these children to attend government-run schools, rendering moot the question of writing system to be used in education (Government of Bombay, 1865, pp. 89–90). Yet, by 1869, support had re-emerged for introducing a Landa-based writing system for Sindhi for use in government schools. In February 1869, incumbent Commissioner in Sindh, William Merewether, issued a notification stating:

Government have decided upon introducing a uniform Bunya-Sindhi or Hindu-Sindhi alphabet, in the hope that: —

• (1.) By its use in our Schools, it may be the means of very largely increasing the attendance.
• (2.) By the publication of books in the new character, the education of the children of Native merchants and shopkeepers may be facilitated and knowledge largely diffused among that class.
• (3.) By the gradual introduction of the new character, the difficulty generally experienced by our Law Courts, in deciphering documents drawn up in the present Bunya-Sindhi character may be avoided.

(Merewether, in Government of Bombay, 1869, p. 213)

Also announced was the formation of a “Vernacular Literature Committee in Sind”, whose mandate, among other things, would include “fix[ing] a well-digested and uniform system of orthography both for the Arabic-Sindhi and Hindu-Sindhi characters” (Government of Bombay, 1869, p. 214). In finalising the ‘Hindu Sindhi’ or Sindhi-Landa system, the Committee adopted the existing proto-standardised Khudawadi graph inventory as the basis. The inventory was then supplemented with bound vowel graphs or vowel ‘diacritics’, whose graphetic shape was inspired by corresponding graphs in the Gurmukhi script (Boivin, 2020, p. 47). One graph, |𑋙| [ɾ(ə)], was taken from the Shikarpuri style of Landa (Grierson, 1919, p. 18). This new system came to be known as “improved Hindu Sindhi” (Grierson, 1919, p. 18). Compared to the previous proto-standardised 𝚜𝚍-𝚂𝚒𝚗𝚍, the new Sindhi-Khudawadi system featured additional graphs in its inventory, as well as a revamped and regularised set of graph-phone correspondences. Accordingly, the writing system hitherto referred to as proto-standardised Sindhi-Khudawadi will be assigned the language subtag 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍, while the so-called ‘improved Hindu-Sindhi’ will be denoted by the subtag 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝟷𝟾𝟼𝟿 (§2.9).

The first public school providing instruction in 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝟷𝟾𝟼𝟿 was set up in Karachi later in 1869 (Hughes, 1876, p. 373).⁶⁸ Textbooks in 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝟷𝟾𝟼𝟿 were authored by government officials, primarily Amil (Chapter 8). After initially being lithographed, books in 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝟷𝟾𝟼𝟿 were printed using movable type by 1874 (Government of Bombay, 1874, p. 262). Despite government support and funding, 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝟷𝟾𝟼𝟿 did not receive the public uptake that administrative officials expected. Although the initial couple of years after its introduction saw an increase in school attendance, there was “still a considerable want of confidence on the part of the Baniá class in the permanency of the New Character” (Government of Bombay, 1872, p. 442). By 1875, it was reported that school attendance by Vania children had begun to decline (Government of Bombay, 1875, p. 87), with many pupils — almost all boys — leaving midway to join their family occupation (Baillie, 1890, p. 161). In the meantime, with the demise of Narayan Jagannath Vaidya in 1873 (Baillie, 1890, p. 161), 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝟷𝟾𝟼𝟿 had lost its driving force. In 1881, Dayaram Gidumal Shahani, a civil servant of Amil background, recommended ending government support for 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝟷𝟾𝟼𝟿. Reiterating the observation of past civil servants, Shahani noted that Vania children had little use for the graphematically transparent 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝟷𝟾𝟼𝟿 in their community’s trading affairs, preferring instead the graphematically opaque but quick-to-write 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍. Shahani’s recommendation was approved by the Commissioner in Sindh, and government support for 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝟷𝟾𝟼𝟿 was eventually withdrawn (Dow, 1976, p. 55). By the early twentieth century, only a handful of schools continued to provide instruction in 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝟷𝟾𝟼𝟿, typically alongside 𝚜𝚍-𝙰𝚛𝚊𝚋 (Aitken, 1907, p. 479). Meanwhile, proto-standardised 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍 continued to remain popular among the Vania community for personal use, at least until Partition (Boivin, 2015; Falzon, 2004). Among printed works, the majority of twentieth-century works in 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝟷𝟾𝟼𝟿 comprise Sindhi translations of sections of the Bible (Figure 8.3).

In the scholarly literature, the British administration’s instituting of two distinct writing systems — and, consequently, educational systems — geared towards Muslim and Hindu-Sikh Sindhis is sometimes portrayed as deliberate, intending to drive a wedge between the two communities. For instance, Anand (1996, pp. 17–18) opines that the instituting of separate schools providing instruction in 𝚜𝚍-𝙰𝚛𝚊𝚋 and 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝟷𝟾𝟼𝟿 was consistent with the colonial-era British policy of divide et impera — Latin for ‘divide and conquer’. Similarly, Hugh Dow, a British civil servant who served as Governor of Sindh from 1941 to 1946, felt that the introduction of 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝟷𝟾𝟼𝟿 was “sectarian propaganda masquerading in the garb of educational enthusiasm” (Dow, 1976, p. 55). Such arguments may well have a considerable element of truth in them. However, there may also be a more mundane explanation for the grapholinguistic experimentation witnessed in early British-era Sindh. It emerges that the primary actors behind a standardised writing system for Sindhi were primarily non-Sindhis — Stack, Burton, Frere, Ellis, Trumpp and Vaidya, not to mention the various Educational Inspectors and Commissioners of British background. Notwithstanding the benevolence or malintent behind their actions, these non-Sindhi officials and intellectuals may have simply misunderstood the implicit-but-consistent patterns of writing system adoption prevalent among sections of the Sindhi community. For the Sindhis themselves, their preference for a particular writing system was dictated not by graphematic perfection or religious ideology, but by user-oriented and use-oriented pragmatism. This may explain why Frere’s and Ellis’ impression that Hindu-Sikh Sindhis would not learn 𝚜𝚍-𝙰𝚛𝚊𝚋, and Vaidya’s assumption that the same community would be drawn to an education in 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝟷𝟾𝟼𝟿, were both proven false.

5.2.4 Gurmukhi, Khojki and Roman

Despite official support for 𝚜𝚍-𝙰𝚛𝚊𝚋, and temporarily for 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝟷𝟾𝟼𝟿, the British era saw a slight increase in the use of Sindhi-Gurmukhi. In the 1850s, European missionaries translated select Christian scriptures into Sindhi, printed in 𝚜𝚍-𝙰𝚛𝚊𝚋, 𝚜𝚍-𝙳𝚎𝚟𝚊 and 𝚜𝚍-𝙶𝚞𝚛𝚞 (Grierson, 1919, p. 13; Hooper, 1938, pp. 131–132). New and revised Bible translations in 𝚜𝚍-𝙶𝚞𝚛𝚞 continued to appear until the early twentieth century (§8.3). The choice of 𝚜𝚍-𝙶𝚞𝚛𝚞 suggests a target readership of Hindu-Sikh women in Sindh (§5.1.4). In 1890, Hyderabad-based social reformer Navalrai Advani launched a monthly magazine in 𝚜𝚍-𝙶𝚞𝚛𝚞 dealing with women’s issues was launched (Jotwani, 1992, p. 368; Schimmel, 1964, p. 214). Titled Sudhar Patrika ‘Journal of Reform’, the choice of writing system for the magazine was determined by the literary competence of its intended audience (Gidumal, 1903, p. 354; Jetley, 1987a, p. 88).⁶⁹ Alongside 𝚜𝚍-𝙰𝚛𝚊𝚋 as the primary medium of instruction, a few girls’ schools in urban areas in Sindh also provided instruction in the Gurmukhi script. However, rates of female school attendance were low, and most women acquired proficiency in Gurmukhi at home from older female family members (Aitken, 1907, p. 481). In this regard, the transmission of Gurmukhi among Hindu-Sikh Sindhi women in the home domain represented a continuation of a pre-British tradition.

The British era also saw an uptick in Khojki-script publications, albeit not necessarily in the Sindhi language. Historically, the spoken language of the Nizari Ismaili community ranged from Siraiki to Kutchi on a continuum (Shackle & Moir, 1992, pp. 15, 43). By the early nineteenth century, the Nizari Ismaili community had spread from the Kutch region further into what is now Gujarat state in India. Consequently, multilingualism and multidialectalism became common, particularly in the Kathiawadi variety of Gujarati (Akhtar, 2016). Following the consolidation of British power in South Asia by the mid-1800s and the subsequent development of Bombay and Karachi as port cities, large sections of Kutchi and Kathiawadi-speaking Nizari Ismailis settled in these cities (Shackle & Moir, 1992, p. 11), particularly in Bombay (Asani, 2011, p. 105). The British-driven emergence of ‘standard’ Gujarati and Sindhi in spoken and written forms, and the increasing socioeconomic prestige of these languages, led to the Nizari Ismailis adopting these languages for formal and written purposes (Ivanow, 1931, p. 533 footnote), although they often retained Kutchi and Kathiawadi as their spoken community languages. Several Ismailis also emigrated to British colonies in the Middle East and East Africa (Akhtar, 2016), where they took their speech and writing practices.

The first books in Khojki were printed in Bombay in the late nineteenth century, initially using lithographic techniques and subsequently with movable type. Printing of these books was authorised by the spiritual leader of the Nizari Ismailis at the time, Aga Khan III. These books comprised recensions of the ginans as well as some novel content, albeit remaining within the theological domain (Shackle & Moir, 1992, p. 16). The first Khojki-script press, the privately-run Ghulam-i Husain Printing Press, was set up in the 1880s by one Aladin Ghulamhusen (Asani, 1987, p. 444). Subsequently, an ‘official’ press authorised by Aga Khan III was set up by Laljibhai Devraj, initially named the Khoja Sindhi Printing Press and, eventually, the Ismaili Printing Press (Asani, 1991, pp. 55–57). Besides pioneering Khojki printing using movable type, Devraj also attempted to develop and further standardise the Khojki graph inventory. While the Ghulam-i Husain Printing Press closed down not long after its inception, the Ismaili Printing Press continued publication until the 1930s (Asani, 1987, p. 444). Aside from the ginans, these presses produced some content in the Kutchi language in the Khojki script, which may be denoted by the language subtag 𝚔𝚏𝚛-𝙺𝚑𝚘𝚓. By the 1910s, however, most non-ginanic content issued from these presses began to appear in the Gujarati language, albeit in the Khojki script. In graphematic terms, the publications displayed a shift in manifestation from 𝚔𝚏𝚛-𝙺𝚑𝚘𝚓 to 𝚐𝚞-𝙺𝚑𝚘𝚓. The 1930s saw yet another metamorphosis in the written manifestation of Ismaili literature, with the Khojki script being increasingly replaced by the Gujarati script (Asani, 1991, p. 56; Moir, Shackle, & Mitha, n.d., p. 1). This meant that publications were now almost entirely in 𝚐𝚞-𝙶𝚞𝚓𝚛, with the presence of the Sindhi language or the Khojki script in modern non-ginanic Ismaili literature reduced to a minimum.

Various factors have been cited for the gradual shift from 𝚔𝚏𝚛-𝙺𝚑𝚘𝚓 to 𝚐𝚞-𝙺𝚑𝚘𝚓 and finally to 𝚐𝚞-𝙶𝚞𝚓𝚛. The first shift from 𝚔𝚏𝚛-𝙺𝚑𝚘𝚓 to 𝚐𝚞-𝙺𝚑𝚘𝚓 in the early twentieth century can be understood as reflecting the increasing socioeconomic prestige and use of the Gujarati language for the Ismaili community outside Sindh, including in the East African diaspora. Increased cost in procuring and maintaining Khojki-script types may also have played a part (Asani, 1991, p. 56). The second shift from 𝚐𝚞-𝙺𝚑𝚘𝚓 to 𝚐𝚞-𝙶𝚞𝚓𝚛 was likely an outcome of increasing community literacy in the Gujarati language and script, and a simultaneous decline in Khojki-script proficiency (Akhtar, 2016, p. 89). Moreover, despite Devraj’s efforts, there remained several gaps in his proposed systems for 𝚔𝚏𝚛-𝙺𝚑𝚘𝚓 and 𝚐𝚞-𝙺𝚑𝚘𝚓, leading to a gradual shift towards the established and graphematically more predictable 𝚐𝚞-𝙶𝚞𝚓𝚛. Additionally, a transliteration system for depicting Arabic-language Quranic text in the Gujarati script (𝚊𝚛-𝙶𝚞𝚓𝚛) had been developed by the early twentieth century (Akhtar, 2016, pp. 88–89; Pandey, 2009), which provided further impetus to adopt the Gujarati script for Ismaili liturgical publications. Partition dealt a decisive blow to printing in Khojki, since the Ismaili community, like the Sindhi community overall, was now split among Pakistan, independent India and the diaspora. For more on the timeline of Khojki-script publishing, see Asani (1991; 1992; 2011), Bruce (2015), Pandey (2009), Shackle and Moir (1992) and Virani (2022).

The most comprehensive attempt at using Roman to transcribe Sindhi is that of Grierson (1919, pp. ix–x), albeit only as an auxiliary system. In his monumental eight-volume Linguistic Survey of India (LSI; 1903–1928), Grierson depicts South Asian languages in a Roman-script writing system based on one initially proposed by the Welsh scholar William Jones in the late eighteenth century (Jones W. , 1799). Jones’ system loosely follows the principle of “vowels as in Italian, consonants as in English”, which was a concept popular among English missionaries at the time (Gleason, 1996, p. 778). Compared to Jones’ system, though, Grierson’s system is graphematically more sophisticated and transparent. It incorporates established conventions of distinguishing lax and tense γ-vowels with a macron (e.g., |a| and |ā|), and dental and retroflex γ-consonants with an underdot (e.g., |d| and |ḍ|). Grierson’s system also includes Sindhi-specific innovations, such as representing the language’s characteristic reduced vowels by superscript graphs (e.g., |a| and |ᵃ|). What makes the LSI invaluable in the present context is the presence of texts in various spoken varieties of Sindhi, transcribed in several writing systems. These include 𝚜𝚍-𝙰𝚛𝚊𝚋, 𝚜𝚍-𝙳𝚎𝚟𝚊, 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍, 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝟷𝟾𝟼𝟿, 𝚜𝚍-𝙻𝚊𝚝𝚗 and even Kutchi in the Gujarati script, or 𝚔𝚏𝚛-𝙶𝚞𝚓𝚛. The LSI does not feature any instances of 𝚜𝚍-𝙶𝚞𝚛𝚞 or 𝚔𝚏𝚛-𝙺𝚑𝚘𝚓, which may be considered indicative of the decline or diminished status of those systems by the early twentieth century.

5.2.5 Braille

Although the presence of Sindhi in Gurmukhi and Khojki began to diminish by the early twentieth century, the language also began to appear in a new script — Braille. In 1922, Parmanand Mewaram (PM) Advani, an educationist and principal of a school for Blind students in Karachi, designed a Braille-based system for encoding Sindhi (Ajwani, 1939, p. 94; Hall, 2013, pp. 90–91; Shaikh, 2017, p. 19).⁷⁰ Advani observed that Braille-based systems for other South Asian languages were also in use across schools for the Blind across South Asia, most of which were distinct and incompatible in terms of their graph-phone correspondences. He then began to advocate for a common Braille-based system that could accurately yet consistently encode all major South Asian languages for the benefit of Blind learners and readers. However, Advani himself was sighted, due to which his proposals received some pushback from members of the Blind community. In the 1940s, disability activist Lal Advani (1923–2005) designed an alternative Braille-based system capable of encoding all major South Asian languages, including Sindhi (Chander, 2005; Khandeker, 2018).⁷¹ Although unrelated to each other, the two Advanis shared a common background in being Amil Sindhis. Unlike PM Advani, however, Lal Advani was Blind.

Throughout the 1940s, the two Advanis’ proposals for a pan-Indian Braille-based writing system continued to be debated at the government level. The matter would be resolved only after Partition in 1947.

5.3 After 1947

From the 1920s, 𝚜𝚍-𝙰𝚛𝚊𝚋 emerged as the dominant writing system for Sindhi in the public sphere. While 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍 and 𝚜𝚍-𝙶𝚞𝚛𝚞 continued to be used for personal records and private correspondence, they gradually disappeared from the educational system and public eye. As a result, debates on Sindhi’s scripts seem to have lain low until after independence and the concurrent Partition of British India in 1947. Partition resulted in Sindh becoming a province of Pakistan and the emigration of most Hindu-Sikh Sindhis to India (§3.4). Today, in Sindh, 𝚜𝚍-𝙰𝚛𝚊𝚋 continues to be the unchallenged official writing system for the Sindhi language. While the Hyderabad-based Sindhi Language Authority (SLA) has brought out a few pedagogical publications in 𝚜𝚍-𝙳𝚎𝚟𝚊 (Hussain, 2011), the writing system is not used in Sindh in any official or educational capacity.

In independent India, the scars of Partition and its religious overtones prompted some Hindu-Sikh Sindhis to back Devanagari as the official script for Sindhi. The first calls to this effect were made shortly after Partition by a Bombay-based group called the Sindhi Sahitya Sabha ‘Sindhi Literary Assembly’ (Asani, 2003, p. 625; Daswani, 1979, p. 61; Cook M. A., 2016a), also referred to as the Sindhi Sahitya Sammelan ‘Sindhi Literary Conference’.⁷² In January 1949, the Secretary of the group, AG Mamtora, sent a resolution to the Indian Ministry of Education in New Delhi asking that government-funded primary schools start teaching Sindhi, but using Devanagari. The stated justification was “facilitating integration of the Sindhis who have migrated from Sindh” and “having a scientific script for the [Sindhi] language” (National Archives of India, 2018b, p. 65). Backing the Sabha’s position was the New Delhi-based Trust for Sindhi Women and Children, chaired by Sindhi politicians Jairamdas Daulatram and Choithram Gidwani (National Archives of India, 2018a, p. 92). On 25th and 26th August 1949, Gidwani presided over an All-India Sindhis (Displaced People) Convention in Bombay. The Convention endorsed the Sabha’s call for ensuring 𝚜𝚍-𝙳𝚎𝚟𝚊 instruction in government-funded primary schools. It also called on the state governments of Bombay (then comprising parts of present-day western Maharashtra and Gujarat) and Rajasthan to recognise Sindhi as a official language, given the large number of Sindhis who had settled in those states following Partition. On 10 October 1949, the Sindh Hindu Seva Samiti (also spelled Sewa and Samity) ‘Sindh Hindu Service Committee’, with Nanik Motwani as Secretary, wrote to the Ministry of Education requesting that the Convention’s resolutions be implemented forthwith (National Archives of India, 2018a, p. 57).⁷³ Opposing the calls for introducing 𝚜𝚍-𝙳𝚎𝚟𝚊 in primary schools was a group of Sindhi litterateurs and school officials, who wrote to the Minstry of Education asking for continued support for 𝚜𝚍-𝙰𝚛𝚊𝚋 (National Archives of India, 2018b, p. 66).

On 18 November 1949, the Indian Ministry of Education wrote to various state governments and universities around the country asking for objections to 𝚜𝚍-𝙳𝚎𝚟𝚊 being taught in government-funded schools and universities (National Archives of India, 2018a, p. 47). Since the question of teaching Sindhi primarily concerned the governments of Bombay, Rajasthan and the now-defunct Ajmer state, there was no noteworthy objection to the introduction of 𝚜𝚍-𝙳𝚎𝚟𝚊 (National Archives of India, 2018a, pp. 49–64). Accordingly, on 9 March 1950, the Ministry of Education notified all state governments that 𝚜𝚍-𝙳𝚎𝚟𝚊 now had official approval, and asked them to implement the decision in government-run institutions in the state offering Sindhi (National Archives of India, 2018b, p. 71). Shortly thereafter, the state government of Bombay confirmed that 𝚜𝚍-𝙳𝚎𝚟𝚊 would be used in primary schools starting from 1 March 1951. Besides asking all Sindhi-language teachers in the state to familiarise themselves with 𝚜𝚍-𝙳𝚎𝚟𝚊, the government also invited authors and publishers to prepare suitable 𝚜𝚍-𝙳𝚎𝚟𝚊 textbooks (National Archives of India, 2018b, p. 73).

In response, on 16 November 1950, the Bombay-based Committee for Sindhi Language and Script, whose Secretary was the noted Sindhi litterateur Kirat Babani, wrote to the Prime Minister of India strongly opposing the introduction of 𝚜𝚍-𝙳𝚎𝚟𝚊 and appealing for 𝚜𝚍-𝙰𝚛𝚊𝚋 to be reinstated in government-funded educational institutions (National Archives of India, 2018b, pp. 82–85). According to Babani, the government’s decision would effectively cut off Sindhi children from existing literature in 𝚜𝚍-𝙰𝚛𝚊𝚋. Noting that transliterating this literature into 𝚜𝚍-𝙳𝚎𝚟𝚊 would involve enormous labour, time and expenditure, he asked if the central or state government was prepared to take up the responsibility of the task. Highlighting the opposition to 𝚜𝚍-𝙳𝚎𝚟𝚊 by certain Sindhi educationists, Babani also underscored what he felt were irregularities in the pro-𝚜𝚍-𝙳𝚎𝚟𝚊 resolution passed by the All-India Sindhis (Displaced People) Convention, including the lack of unanimity behind the resolution as well as appeals to emotion that 𝚜𝚍-𝙰𝚛𝚊𝚋 was ‘Muslim’. That said, Babani’s justification of 𝚜𝚍-𝙰𝚛𝚊𝚋 contianed the following:

In Sindhi [i.e., 𝚜𝚍-𝙰𝚛𝚊𝚋], certain words with same pronunciation but with some what [sic] different spelling have different meanings. They cannot be written differently in Devnagri Script. There will be confusion in their meaning.

(Babani, 16 November 1950, in National Archives of India, 2018b, p. 84)

In terms of factual validity or lack thereof, the arguments in favour of 𝚜𝚍-𝙰𝚛𝚊𝚋 had much in common with the Sindhi Sahitya Sabha’s portrayal of Devanagari as a “scientific script”. Regardless, the appeals to emotion in both sides’ arguments had an impact on the government. On 2 January 1951, in an internal note in response to Babani’s letter, Tara Chand, Assistant Secretary to the Government of India, wrote:

I am not at all sure that our decision [to exclusively support 𝚜𝚍-𝙳𝚎𝚟𝚊] is technically correct. The Article 29(1) of the Constitution distinctly provides that “any section of the citizens in the territory of India… shall have the right to conserve” its distinct language and script. If a Section of Sindhis insists upon its right to regard the Arabic script in which Sindhi language has been written for centuries as its distinctive script, it is doubtful whether Govt. can force them to abandon it in favour of Deva Nagari.

(Chand, 2 January 1951, in National Archives of India, 2018b, p. 7)

Tara Chand’s observation was accepted by the Indian Minister of Education, Abul Kalam Azad, who added that the Sindhis’ preferred script for their traditional language would emerge with time (National Archives of India, 2018b, p. 8). Consequently, on 29 January 1951, the Indian Ministry of Education issued a press note stating that the government would allow both 𝚜𝚍-𝙰𝚛𝚊𝚋 and 𝚜𝚍-𝙳𝚎𝚟𝚊 to be used in state-supported educational institutions, with parents able to choose their desired medium. However, the government added the caveat that a particular writing system would be implemented only if at least forty students in the school, or ten students in a class, demanded it (National Archives of India, 2018b, p. 80).

Unsurprisingly, the central government’s co-recognition of 𝚜𝚍-𝙰𝚛𝚊𝚋 and 𝚜𝚍-𝙳𝚎𝚟𝚊 was swiftly opposed from all sides. The pro-𝚜𝚍-𝙳𝚎𝚟𝚊 lobby, spearheaded by Mamtora, Motwani, Gidwani and colleagues, argued that the government’s backtracking on its previous resolution created confusion. It also reiterated the subjective argument on how 𝚜𝚍-𝙳𝚎𝚟𝚊 would “further the cause of our language and literature, taking into consideration the genesis and philology of our language and its relations with other sister languages of the Indo-Aryan group” (National Archives of India, 2018b, p. 119). In attempting to demonstrate a tradition of writing in 𝚜𝚍-𝙳𝚎𝚟𝚊, the faction noted that the first printed dictionary of Sindhi (Stack, 1849b) was in 𝚜𝚍-𝙳𝚎𝚟𝚊. However, it also claimed that Sami’s 𝚜𝚍-𝙶𝚞𝚛𝚞 works (§5.1.4) and traders’ commercial records in 𝚜𝚍-𝚂𝚒𝚗𝚍, amounted to ‘varieties’ of 𝚜𝚍-𝙳𝚎𝚟𝚊. Finally, it characterised 𝚜𝚍-𝙰𝚛𝚊𝚋 supporters as “only a very small group of dissidents, some of whom have their own vested interests” (National Archives of India, 2018b, p. 120). On the other side, the pro-𝚜𝚍-𝙰𝚛𝚊𝚋 faction, spearheaded by Babani, claimed that approving both 𝚜𝚍-𝙰𝚛𝚊𝚋 and 𝚜𝚍-𝙳𝚎𝚟𝚊 would divide Indian Sindhis (National Archives of India, 2018b, pp. 100–102). Common to both parties was the argument that approving both writing systems would require additional resources and expenditure. On 23 February 1951, in response to delays by the state government of Bombay in recognising 𝚜𝚍-𝙰𝚛𝚊𝚋 as co-official, the Committee for Sindhi Language and Script submitted a strongly-worded petition to the President of India (National Archives of India, 2018b, pp. 130–226). Containing 1017 signatures, the petition demanded that the recognition of 𝚜𝚍-𝙳𝚎𝚟𝚊 be “withdrawn” and that of 𝚜𝚍-𝙰𝚛𝚊𝚋 be “restored”, while also reiterating the call for recognising Sindhi as an official language (National Archives of India, 2018b, p. 130). The vast majority of the signatories’ names, occupations and addresses were written in Roman — effectively in 𝚜𝚍-𝙻𝚊𝚝𝚗 — likely to ensure readability by non-Sindhi central government officials. That said, a few entries were in 𝚜𝚍-𝙰𝚛𝚊𝚋 and even 𝚜𝚍-𝚂𝚒𝚗𝚍.

In general, though, once 𝚜𝚍-𝙰𝚛𝚊𝚋 and 𝚜𝚍-𝙳𝚎𝚟𝚊 were jointly recognised by the central government, an implicit truce ensued and the script issue took a backseat. Both factions now shifted their focus to practical issues, such as ensuring that state governments complied with the central government’s order, and getting the Sindhi language recognised in the Indian Constitution. On the latter issue, efforts were already underway. On 21 January 1951, the residents of Pimpri camp, Poona, sent a petition to the President of India asking for Sindhi to be recognised as an official language in Indian states with significant Sindhi populations, as well as at the national level (National Archives of India, 2018b, pp. 109–115). Compared to the near-contemporaneous petition by the Committee for Sindhi Language and Script, the Pimpri residents’ petition only contained around two hundred and fifty signatures. However, the signatures on the latter were graphematically much more diverse than those on the former, with comparable numbers in 𝚜𝚍-𝙻𝚊𝚝𝚗, 𝚜𝚍-𝙰𝚛𝚊𝚋 and 𝚜𝚍-𝚂𝚒𝚗𝚍. Of the last category, most were in 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍, namely without γ-vowels. This offers some evidence that 𝚜𝚍-𝚂𝚒𝚗𝚍 was still in use in personal domains at the time of Partition. Notably, only four signatures were recognisably in 𝚜𝚍-𝙳𝚎𝚟𝚊. A selection of the signatures is shown in Figure 5.2.

In response to the petition, the Indian Ministry of Law replied that recognising Sindhi as a state language was up to individual state governments, while recognising Sindhi as an official language of India would require a constitutional amendment. Regarding the latter point, the Ministry noted that “this is hardly the time at which any proposal for an amendment of the Constitution can be made with any chance of success” (National Archives of India, 2018b, p. 22). Nevertheless, petitions for official recognition of Sindhi continued throughout the 1950s. These petitions were bolstered when state governments began gradually winding back special provisions made for Sindhi refugees, which included primary schools (National Archives of India, 2018b, pp. 306, 343). In the meantime, away from the heat of the public stage, bureaucratic discussions on matters of graphematic standardisation continued quietly. Over 1952 and 1953, suggestions were made on standardising the 𝚜𝚍-𝙳𝚎𝚟𝚊 graph inventory (National Archives of India, 2018b, pp. 362–363), which resulted in the now-ubiquitous 𝚜𝚍-𝙳𝚎𝚟𝚊 graphs |ॻ ॼ ॾ ॿ| for the φ-implosives [ɠ ʄ ɗ ɓ]. However, less attention was paid to the finer points of 𝚜𝚍-𝙳𝚎𝚟𝚊 spelling, the impacts of which are still felt to the present day (§7.5). The narrow focus on standardising 𝚜𝚍-𝙳𝚎𝚟𝚊’s graph inventory while sidelining matters of spelling was ironically reminiscent of the events surrounding the official adoption of 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝟷𝟾𝟻𝟹, almost exactly a hundred years prior (§5.2.1).

After more than fifteen years of having two official scripts but no official language status, Sindhi was finally recognised in the Indian Constitution in 1967. Paradoxically, Sindhi’s recognition as an official language of India had the effect of rekindling the script debate. With its new status making it eligible for greater government funding and support, both 𝚜𝚍-𝙰𝚛𝚊𝚋 and 𝚜𝚍-𝙳𝚎𝚟𝚊 supporters now wanted a greater share of the pie for teaching and publication in their favoured writing system (Daswani, 1979, p. 66). This resulted in the two factions attempting to elbow out the other where possible, with arguments laced with ever-increasing verbal sophistry and appeals to emotion. The pro-𝚜𝚍-𝙳𝚎𝚟𝚊 lobby has continued to claim that their favoured script is linguistically more suitable to the Sindhi language, and that there exists a historical and cultural link of the script with the language. In the process, script and religious identity are often conflated, and the Arabic script is identified with Islam. Along these lines, it is claimed that Devanagari was “buried underground by Muslim conquerors” (Asani, 2003, p. 625). It is also claimed that the Devanagari script is ubiquitous in India due to its use for writing Hindi. On this basis, it is argued that the adoption of 𝚜𝚍-𝙳𝚎𝚟𝚊 is in the interest of the Sindhi language’s long-term survival in India. On the other hand, the pro-𝚜𝚍-𝙰𝚛𝚊𝚋 lobby asserts that using their preferred script would continue to ensure the readability of past Sindhi literature, maintain a link between Indian and Pakistani Sindhis, and confer a distinct visual identity on the written form of their language. Using 𝚜𝚍-𝙳𝚎𝚟𝚊, they argue, would re-emphasise the hegemony of the Hindi language and Devanagari script over Indian Sindhis, especially over the younger generation. In the process, the pro-𝚜𝚍-𝙰𝚛𝚊𝚋 lobby paints the pro-𝚜𝚍-𝙳𝚎𝚟𝚊 lobby as sectarian, due to the latter’s use of anti-Islamic rhetoric in their propaganda.

The infighting and indecision on the Sindhi script issue in post-Partition India has been criticised by various authors, from different perspectives. Daswani (1979, p. 66) states that the script limbo has made the Sindhi community disillusioned with the issue, causing a large section of the younger generation to have no opinion in the matter. Anand (1996, p. 128) disapproves of the “compromise” of having two official scripts for Sindhi, adding that it exacerbates the precarious situation of a language whose speakers are geographically dispersed. On the other hand, scholars like Khubchandani have taken a broader view of the situation, and object to the very preoccupation with standardisation and codification of language and script. Khubchandani opines that the overall endeavour at “bringing order to chaotic diversity” (1984, p. 172) ultimately stems from a Western monolingual and homogenising mindset.

In the meantime, the Indian government has resolutely kept out of the Sindhi script debate, leaving the final decision to the community itself. Consequently, the language may be taught as 𝚜𝚍-𝙰𝚛𝚊𝚋, 𝚜𝚍-𝙳𝚎𝚟𝚊 or both in different Indian states. Government institutions, including the Central Institute of Indian Languages (CIIL), usually bring out their Sindhi-language publications in both 𝚜𝚍-𝙰𝚛𝚊𝚋 and 𝚜𝚍-𝙳𝚎𝚟𝚊. According to Daswani, this leads to “duplication and a waste of public funds” (1979, p. 67). Although the intensity of the script debate is no longer as intense as it was in the immediate aftermath of Partition, there remains a degree of one-upmanship between the two factions. For instance, the Sahitya Akademi, or the National Academy of Letters, only considers works in 𝚜𝚍-𝙰𝚛𝚊𝚋 to be eligible for its Sindhi literature awards. This has been attributed to 𝚜𝚍-𝙰𝚛𝚊𝚋 supporters holding influential positions in the Academy (Hardwani, personal communication, November 8, 2014). The indecision on the script issue also leads to occasional symbolic setbacks, such as Sindhi-language text being left out from Indian rupee currency notes (Young, 2009, pp. 165–166). Figure 5.3 shows the language panel from the reverse of an Indian ten-rupee note (Mahatma Gandhi Series, 1996–2016), which is conspicuously devoid of Sindhi-language text in any script.

Other writing systems such as 𝚜𝚍-𝚂𝚒𝚗𝚍, 𝚜𝚍-𝙶𝚞𝚛𝚞 and 𝚔𝚏𝚛-𝙺𝚑𝚘𝚓 have all but disappeared from everyday usage, both in Pakistan and India. Consequently, persons literate in these writing systems tend to be older. Grapholinguistic remnants in these systems dating from the pre-Partition era or early post-Partition years may still survive. In Sindh, there remain a few instances of 𝚜𝚍-𝚂𝚒𝚗𝚍 on pre-Partition public signage (Boivin, 2015). Similarly, religious publications in 𝚜𝚍-𝙶𝚞𝚛𝚞 targeted at Hindu-Sikh women may still exist, although most such literature would have likely been reprinted in 𝚜𝚍-𝙳𝚎𝚟𝚊 (Gidwani C. P., 2012). Notwithstanding the decline of 𝚜𝚍-𝙶𝚞𝚛𝚞, the use of Gurmukhi for the Punjabi language has flourished in the post-Partition era, with Punjabi-Gurmukhi (𝚙𝚊-𝙶𝚞𝚛𝚞) having been granted official status in India’s Punjab state. Gurmukhi’s cousin Khojki has not been as fortunate, and its adoption to write Gujarati in the form of 𝚐𝚞-𝙺𝚑𝚘𝚓 did not last long (§5.2.4). Unlike the rest of the Sindhi community, which was split between its Pakistani and Indian sections largely along religious lines, the Ismaili section of the community found itself in both countries (Shackle & Moir, 1992, p. 10). Ismailis in Sindh were now separated from their kin in Kutch, Gujarat and Bombay by the Indo-Pak international border. However, since Ismailis in Sindh were concentrated in Karachi, the Khojki script continued to be taught for a while in Karachi-based Ismaili religious schools in Pakistan. As a result, a small number of publications in 𝚐𝚞-𝙺𝚑𝚘𝚓 and 𝚔𝚏𝚛-𝙺𝚑𝚘𝚓 continued to appear (Asani, 1991, p. 57; Valliani, 2008, pp. 94–95). This lasted until 1975, when Ismaili schools in Pakistan stopped teaching Khojki citing student burden and the apparent disuse of Khojki in the global Ismaili diaspora (Tajddin Sadik Ali, 1989; Valliani, 2008, p. 95). Since then, Khojki-script publications have been largely restricted to ginanic works (Pandey, 2011b, p. 2) and primers prepared by lay and academic community members (Moir, Shackle, & Mitha, n.d.; Tajddin Sadik Ali, 1989). Despite the dwindling of Khojki in the twentieth century, the Ismaili community, especially those outside of Sindh, still refer to the script as ‘Sindhi’ (Asani, 1991, pp. 56–57).

Notwithstanding graphematically unstandardised practices and sociolinguistically subordinate status, 𝚜𝚍-𝙻𝚊𝚝𝚗 has emerged as the writing system most widely known by literate Sindhi speakers worldwide today. To a large extent, implicit community familiarity with 𝚜𝚍-𝙻𝚊𝚝𝚗 has been driven by the ubiquity of the Roman script and comparatively limited support for the Arabic and Devanagari scripts on computers and mobile devices worldwide. However, transcribing a language in an ad-hoc Roman-script-based writing system on electronic devices is commonly attested worldwide, regardless of language. As such, the use of ad-hoc 𝚜𝚍-𝙻𝚊𝚝𝚗 is sociolinguistically unremarkable, although it remains graphematically interesting. That said, there exists some community support exists for using 𝚜𝚍-𝙻𝚊𝚝𝚗 as a formal auxiliary system, to circumvent technological hindrances to writing Sindhi on electronic devices, as well as to create a writing system easily readable by Sindhis across the world. The most organised and prominent community group advocating for 𝚜𝚍-𝙻𝚊𝚝𝚗 as a sociolinguistically auxiliary writing system for Sindhi is the Romanized Sindhi group (RomanizedSindhi.org, 2010a). With a membership that spans several countries worldwide, the Romanized Sindhi group conducts regular workshops and outreach events worldwide to popularise its 𝚜𝚍-𝙻𝚊𝚝𝚗 orthography. At the time of writing this book, the Romanized Sindhi 𝚜𝚍-𝙻𝚊𝚝𝚗 orthography appeared on certain sections of the SLA’s website as a transliteration of 𝚜𝚍-𝙰𝚛𝚊𝚋 (Sindhi Language Authority, 2017a). The orthography does not appear to have been officially endorsed by the SLA, though. For a closer look at the Romanized Sindhi 𝚜𝚍-𝙻𝚊𝚝𝚗 orthography, see Section 12.3.

Regarding the issue of a Braille-based writing system for South Asian languages, the disagreements between the two Advanis’ systems persisted even after Partition. In the meantime, calls began to grow from Blind advocates from other countries to ensure harmony among Braille-based writing systems worldwide. Against this background, the government of newly independent India handed the issue over to UNESCO in 1949, requesting them to pursue the creation of a “world Braille” (Ministry of Education, India, 1952; Kabir, 1949; Mackenzie, 1954, pp. 9–10). UNESCO accepted the Indian government’s request, and initiated a series of conferences of international Braille experts, including the two Advanis. By 1954, a Braille-based system for South Asian languages that was largely harmonious with homoscriptal systems such as English-Braille and French-Braille was finalised, and officially adopted by the governments of independent India and Ceylon (today Sri Lanka) (Ministry of Education, India, 1952, p. 11; Mackenzie, 1954, p. 148). Pakistan devised a distinct Braille-based writing system, albeit only for Urdu (Mackenzie, 1954, pp. 37–38, 131–134). Since then, standardised Braille-based systems have been widely adopted and used in India and Pakistan, with the exception of Sindhi-Braille (UNESCO, 2013). The languishing of 𝚜𝚍-𝙱𝚛𝚊𝚒 appears particularly stark when one considers the pivotal role that Sindhis played in standardising Braille-based systems — not just in South Asia, but worldwide. For a detailed history of the events leading to the standardisation of World Braille, see Section 11.1.

5.4 Summary

The diachronic overview of Sindhi-language graphematics and graphosociolinguistics reveals rich and diverse traditions of writing the language, conditioned by the people who wrote it and the purposes for which they did. The writing systems of Sindhi, therefore, carry immense potential to test and advance graphematic and graphosociolinguistic theories as they currently exist. Each of the next seven chapters is exclusively devoted to describing and evaluating script-specific writing systems for Sindhi, including subvariants and orthographies. These are Arabic (𝚜𝚍-𝙰𝚛𝚊𝚋), Devanagari (𝚜𝚍-𝙳𝚎𝚟𝚊), Khudawadi (𝚜𝚍-𝚂𝚒𝚗𝚍), Gurmukhi (𝚜𝚍-𝙶𝚞𝚛𝚞), Khojki (𝚜𝚍-𝙺𝚑𝚘𝚓 and 𝚔𝚏𝚛-𝙺𝚑𝚘𝚓), Braille (𝚜𝚍-𝙱𝚛𝚊𝚒) and Roman (𝚜𝚍-𝙻𝚊𝚝𝚗) (ScriptSource, 2022b). Complementing the graphematic analysis, and concluding Part Two of the book, is a sociolinguistic investigation into contemporary community attitudes towards 𝚜𝚍-𝙰𝚛𝚊𝚋, 𝚜𝚍-𝙳𝚎𝚟𝚊 and 𝚜𝚍-𝙻𝚊𝚝𝚗 among Indian Sindhis.

☙ ❧

6 Arabic

If one accepts that Sindhi has been written in the Arabic script for almost a millennium now (§5.1), this would make Sindhi in the Arabic script (𝚜𝚍-𝙰𝚛𝚊𝚋) one of the oldest attested writing systems for the language. Over the centuries, 𝚜𝚍-𝙰𝚛𝚊𝚋 has been through a fascinating journey, undergoing several graphematic twists and turns and manifesting in various avatars. Today, 𝚜𝚍-𝙰𝚛𝚊𝚋 is recognised as the official writing system for the Sindhi language in Pakistan, and a co-official script, along with Devanagari, in India (Daswani, 1979).

6.1 Graphematic foundations

Broadly speaking, all variants of the 𝚜𝚍-𝙰𝚛𝚊𝚋 writing system are based on the Persian-Arabic writing system (𝚏𝚊-𝙰𝚛𝚊𝚋), itself derived from the Arabic-Arabic writing system (𝚊𝚛-𝙰𝚛𝚊𝚋). Consequently, Sindhi is often described in the grapholinguistic literature as being written in Perso-Arabic (Iyengar, 2018; Khubchandani, 2007, p. 696). Probably the most salient feature of 𝚜𝚍-𝙰𝚛𝚊𝚋 — indeed, of most Arabic-script-based writing systems worldwide — is dextrosinistrality, in that the general direction of writing is right-to-left.⁷⁴ A notable exception to this rule is that of numerals (or digits), which are written left to right. Accordingly, and assuming that numerals fall under the category of logograms (§2.8), 𝚜𝚍-𝙰𝚛𝚊𝚋 might be more precisely described as dextrosinistral when it comes to phonograms. Another salient graphematic feature of 𝚜𝚍-𝙰𝚛𝚊𝚋, again common to with most Arabic-based-systems worldwide, is inherent graphetic cursiveness. That is, most — but not all — graphs within a written word are joined to the next one. Such cursiveness causes graphs to manifest as distinct allographs or rasms (§2.7), depending on their position within a written word. In the context of Arabic-script writing, rasms that are considered allographs of each other are called POSITIONAL VARIANTS (Bauer, 1996). Table 6.1 shows the principal rasms in the Arabic-script graph inventory, grouped according to their complementary positional variants. Rasms in rows marked in grey are the calligraphic forms commonly used in 𝚜𝚍-𝙰𝚛𝚊𝚋, although considered allographs of the rasms in the row immediately above.

It emerges from Table 6.1 that a particular rasm may act as the so-called Initial or Medial positional variant of more than one Isolated variant. For example, the Isolated variants |ٮ ں ى| are graphetically distinct from each other, but their Initial and Medial variants are identical — |ٮـ| and |ـٮـ|, respectively. To distinguish graphetically identical rasms, dots called nuqtas are often added to rasms.⁷⁵ The number and position of nuqtas depends on the graph nventory in question. Nuqtas are considered intrinsic to the graph, making them graphosubsegmental in nature (Meletis, 2020). The graphosubsegmental nature of nuqtas is seen in Table 6.2, which contains a selection of phonograms and phonogram sequences from the 𝚊𝚛-𝙰𝚛𝚊𝚋 and 𝚏𝚊-𝙰𝚛𝚊𝚋 graph inventories. The table is designed as a matrix, with graphs representing consonants (henceforth γ-consonants) along the vertical axis and graphs representing vowels (henceforth γ-vowels) along the horizontal axis. The intersection of γ-consonants and γ-vowels contains the graph representing the corresponding phonological [CV] sequence.

Table 6.2 reveals that, in the 𝚊𝚛-𝙰𝚛𝚊𝚋 and 𝚏𝚊-𝙰𝚛𝚊𝚋 inventories, γ-consonants occupy their own graphosegmental space. Also occupying their own graphosegmental space are |ا ﺋ|, which are used as graphetic bases for γ-vowels in certain positions. While |ا| is a rasm, |ﺋ| is itself a complex graph composed of the rasm |ىـ| and a diminutive form of |ء|. In contrast, the γ-vowels |⁠◌َ ⁠◌ِ ⁠◌ُ|, as well as |⁠◌ْ|, are graphosubsegmental in that they only occur along with a base, be it a γ-consonant or one of |ا ﺋ|. Consequently, in the literature, γ-consonants in 𝚊𝚛-𝙰𝚛𝚊𝚋 and 𝚏𝚊-𝙰𝚛𝚊𝚋 are often described as “basic letters”, while γ-vowels, especially |⁠◌َ ⁠◌ِ ⁠◌ُ|, are termed “diacritics” (Bauer, 1996). In terms of phonosegmental representation, a φ-[CV] sequence is shown by a γ-consonant with the γ-vowel written above, below or to the right. A φ-[C] by itself, without a φ-vowel following, is shown by a γ-consonant with the graph |⁠◌ْ| or its allographs |⁠◌ۡ| or |⁠◌ٛ| above it. Outside of a φ-[CV] sequence, a φ-[V] by itself is shown by the corresponding γ-vowel written on the rasm |ا| when word-initial, and on the graph |ﺋ| when word-medial. For this reason, |ا ﺋ| are often termed “vowel holders” (Iyengar, 2018) or “vowel carriers” (Gnanadesikan, 2017a). Additional subsegmental graphs or ‘diacritics’ include the morphologically-salient |⁠◌ً ⁠◌ٍ ⁠◌ٌ| to indicate the Arabic adverbial suffixes [an ɪn ʊn], as well as |⁠◌ّ|, written over a γ-consonant to indicate gemination of the corresponding φ-consonant. The Sindhi and English names of the vowel carriers and subsegmental graphs described so far are listed in (6). Where multiple English names exist, the ones preferred in this book are shown in emphasised type.

In terms of orthographic and sociolinguistic convention, most subsegmental graphs are omitted from 𝚊𝚛-𝙰𝚛𝚊𝚋 and 𝚏𝚊-𝙰𝚛𝚊𝚋 texts, and included only when the author considers it important to accurately reflect the intended pronunciation in written form. Consequently, subsegmental graphs are typically included only in certain text genres, such as children’s books or religious publications (Bauer, 1996, p. 562). At the same time, graphs from the 𝚊𝚛-𝙰𝚛𝚊𝚋 inventory whose phonological values have become identical in 𝚏𝚊-𝙰𝚛𝚊𝚋, such as |ث س ص| and |ذ ز ض ظ| (see Table 6.2), are retained unchanged in the spellings of words. As a result, the 𝚏𝚊-𝙰𝚛𝚊𝚋 system features a many-to-one mapping between graphs and phones. Additionally, 𝚏𝚊-𝙰𝚛𝚊𝚋 contains the graphs |پ چ ژ گ| for the phonemes /p t͡ʃ ʒ ɡ/, which are present in the Persian language but not in Arabic. Other modifications in 𝚏𝚊-𝙰𝚛𝚊𝚋 include the substitution of |ك ي| with the allographs |ک ی| to denote /k j/, and conventionally eliminating the use of sukun in the vowel graphs |‏⁠◌ِی‎ ‏⁠◌ُو‎ ‏⁠◌َي‎ ‏⁠◌َو‎| /i u ej ow/ (Kaye, 1996). These 𝚏𝚊-𝙰𝚛𝚊𝚋-specific conventions are illustrated in Table 6.2 in the paradigms for |پ چ ژ ک گ ی|.

Numerals in the Arabic-script graph inventory and their allographs are shown in (7). Individual Arabic-based writing systems may prefer certain allographs (ScriptSource, 2022b; SIL International, 2022), although ‘mixing and matching’ of allographs remains common (Allana, 1993 [1964], p. 8).

Besides numerals, potential logographs in the Arabic-script graph inventory include the written forms of benedictions, the most prominent being |اللّٰه| ‘Allah’ and its subtly varying allograph |اللّٰہ|. Commonly characterised as a ‘ligature’ (§2.7), the graph |اللّٰه| tends to remain constant across Arabic-script-based writing systems, irrespective of the system’s specific graphematic properties.⁷⁶

The Arabic-script-based systems used for Sindhi adhere to the principles of dextrosinistrality and inherent cursiveness. However, they vary in terms of the actual graphs employed, their graphetic manifestations and the spellings of individual words. Based on common patterns of use, major subvariants of 𝚜𝚍-𝙰𝚛𝚊𝚋 may be identified. In the sections that follow, each such 𝚜𝚍-𝙰𝚛𝚊𝚋 variant will be assigned and referred to by a unique — albeit unofficial — IETF-style language subtag (§2.9). These derivative 𝚜𝚍-𝙰𝚛𝚊𝚋 variants are individually described in the sections that follow.

6.2 Early use and Abul Hasan’s norms

Although 𝚜𝚍-𝙰𝚛𝚊𝚋 is supposed to have first emerged a thousand years ago, idiosyncratic variation in graphematic conventions persisted for several centuries thereafter. As mentioned in Section 5.1, the earliest significant attempt at standardising 𝚜𝚍-𝙰𝚛𝚊𝚋 writing practices is that of Abul Hasan (d. 1711), a theologian hailing from the town of Thatta in southern Sindh. In the year 1700, Abul Hasan wrote a religious treatise in Sindhi rhyming verse, albeit with the Arabic-language title Muqaddimat as-Salat ‘Introduction to prayer’.⁷⁷ Written in a modified version of the Arabic script to account for Sindhi-specific φ-consonants (Schimmel, 1963, p. 229), Abul Hasan’s augmented graph inventory and graphematic practices exerted a significant influence on subsequent 𝚜𝚍-𝙰𝚛𝚊𝚋 Use. Hence, in addition to being considered “decisive” in the history of Sindhi literature (Schimmel, 1974, p. 18), Abul Hasan’s treatise also represents a foundational work in the history of 𝚜𝚍-𝙰𝚛𝚊𝚋 writing. His quasi-standardised graph inventory and graphematic practices, often referred to as “Abul Hasan Sindhi” (Allana, 1993 [1964]; Lekhwani, 2011, p. 28 footnote), remained in vogue until the British proclaimed a new Arabic-script-based system in 1853 (§6.3). Accordingly, Abul Hasan Sindhi will be assigned the language subtag 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚊𝚍.

In terms of content, most works composed in 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚊𝚍 were religious or didactic in nature (Rahman, 1999; Schimmel, 1963; 1974). Handwritten specimens of several such works have been compiled and presented in Baloch (1993), accompanied by a painstaking transliteration into present-day 𝚜𝚍-𝙰𝚛𝚊𝚋 spelling (§6.5). The lack of suitable movable types and the high cost of lithography restricted the number of books printed in 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚊𝚍. A section of the Bible, the Gospel of Matthew, was translated into Sindhi in 1825 at Serampore (in present-day West Bengal, India) in the “Arabic character” (Nida, 1972, p. 393). However, not much is known about its publication or of any extant copies (Grierson, 1919, p. 13; Hooper, 1938, p. 131). What the Sindhi Language Authority considers the “first book printed in the Sindhi language” (Hauze, 2016b) is a lithographed publication in 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚊𝚍 titled Hukayut Oosalaheen or stories of good men written in Sindee (Wulle Mahomed & Hafiz Sedeek, 1851).⁷⁸ An extract from the title page of this book is shown in Figure 6.1.

In the preface to the book, EP Arthur, Lieutenant of Police in Sindh, describes the work thus:

The Hukayut Oosalaheen consists of 20 Chapters in each of which some religious or moral duty is enforced & illustrates by a number of tales and allegories. It was translated from the Arabic language into Sindee by Wullee Mahomed at the Command of Hafiz Sedeek. The date of the translation is not known.

(Arthur, in Wulle Mahomed & Hafiz Sedeek, 1851, Preface, para. 1)

Arthur goes on to state that the colonial government of Sindh commissioned the work to be lithographed from a handwritten copy for the benefit of “those who are studying Sindee for more practical purposes”, namely for government officials needing to learn Sindhi for administrative purposes (§5.2.1). Regarding spelling and related graphematic practices, Arthur writes:

[I]n a language which has been so long neglected & of which so few even now pretend to know the correct Orthography and rules of Grammar many mistakes beyond those incidental to even the most careful transcription, are likely to meet the eye of the critical reader.

(Arthur, in Wulle Mahomed & Hafiz Sedeek, 1851, Preface, para. 3)

Arthur’s quote above reiterates the need to interpret 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚊𝚍 as a loose collection of quasi-standardised practices rather than as a full-fledged orthography in the modern sense. Table 6.3 lists the key graph variants used in 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚊𝚍, along with their usual linguistic values.

Based on Allana’s (1993 [1964]) description, 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚊𝚍 practices may be categorised into graphematic revitalisation, recycling and invention. graphematic revitalisation entailed more frequent use of existing graphematic devices such as the γ-vowel markers |⁠◌َ ⁠◌ِ ⁠◌ُ| to indicate the Sindhi φ-vowels [ə ɪ ʊ], respectively (see (6) and Table 6.2). A φ-consonant with no φ-vowel after it was overtly denoted in 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚊𝚍 with the jazm |⁠◌ۡ| or |⁠◌ٛ|, or occasionally by the allographic sukun |⁠◌ْ| (see (6)). As described in Section 6.1, these subsegmental graphs are not frequently used in 𝚊𝚛-𝙰𝚛𝚊𝚋 and 𝚏𝚊-𝙰𝚛𝚊𝚋. Hence, the conscious use of these graphs in 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚊𝚍 likely reflected writers’ implicit awareness of Sindhi’s vowel-heavy phonology and the resultant lexical and grammatical significance of φ-vowels in the language (§4.3.2, §4.4).

Graphematic recycling involved the deployment of graphs available in 𝚊𝚛-𝙰𝚛𝚊𝚋 but with different linguistic values assigned to them. For instance, the 𝚊𝚛-𝙰𝚛𝚊𝚋 subsegmental graphs |⁠◌ً ⁠◌ٍ ⁠◌ٌ| (see (6)) were repurposed in 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚊𝚍 to denote the Sindhi nasalised φ-vowels [ĩ ũ], respectively, in word-final position (Allana, 1993 [1964], pp. 92, 115). Less frequently, |⁠◌ً| was used to denote word-final [ə̃], and the compound graph |◌اً| for final [ã] (Grierson, 1919, p. 21). The subsegmental graph |⁠◌ٖ| occasionally used in Quranic texts was recycled in 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚊𝚍 to denote the Sindhi φ-vowel [e]. Most often, it manifested as the compound graph |⁠◌ٖيۡ|. Nasalised [ẽ], particularly in word-final position, was written |⁠◌﮼|. Curiously, no distinct graph was allocated for the φ-vowels [o] and [õ]. As seen in Table 6.3, the Sindhi φ-vowels [u o] were written identically, with the graph |⁠◌ـُوۡ|, while their nasalised counterparts [ũ õ] were also written homographically, using |⁠◌ٌ|. The Hukayut Oosalaheen (Wulle Mahomed & Hafiz Sedeek, 1851) features several instances of tanvin being used productively in this manner, of which frequently occurring ones are shown in (8):

Of the spellings shown in (8), |۽| and |۾| persist in present-day 𝚜𝚍-𝙰𝚛𝚊𝚋 as logograms. On occasion, |۽| is equated with the Roman-script logogram |&|, as seen in Unicode’s description of |۽| as U+06FD ARABIC SIGN SINDHI AMPERSAND (Unicode, 2024a). Notwithstanding their present-day use as logograms, |۽| and |۾| qualify as complex phonograms in 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚊𝚍, given the productive use of |‏◌ٍ‎ ‏◌ٌ‎| in that system. At the same time, the use of |‏◌ً‎ ‏◌ٍ‎ ‏◌ٌ‎| in 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚊𝚍 to denote φ-vowel nasalisation was not exclusive. Often, 𝚊𝚛-𝙰𝚛𝚊𝚋 loanwords with |‏◌ٍ‎ ‏◌ٌ‎| denoting [ɪn ʊn] were incorporated unmodified into 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚊𝚍 texts (Allana, 1993 [1964], pp. 92, 115). This put the onus on the reader to glean the phonological value of |‏◌ٍ‎ ‏◌ٌ‎| from the context.

Graphematic innovation in 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚊𝚍 entailed the development of new, distinct graphs to denote Sindhi-specific φ-consonants, especially aspirate and retroflex stops. As mentioned in Section 5.1, such graphs were typically formed by adding subsegmental nuqtas to existing 𝚊𝚛-𝙰𝚛𝚊𝚋 and 𝚏𝚊-𝙰𝚛𝚊𝚋 rasms and graphs, based on the 𝚏𝚊-𝙰𝚛𝚊𝚋 graphematic precedent of adding nuqtas to |ب ج ز| [b d͡ʒ z] to create |پ چ ژ| [p t͡ʃ ʒ]. New graphs in 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚊𝚍 were also created by rearranging nuqtas on existing graphs. For instance, 𝚊𝚛-𝙰𝚛𝚊𝚋 and 𝚏𝚊-𝙰𝚛𝚊𝚋 |ت| [t] was adopted in 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚊𝚍 to denote [t̪], after which the nuqtas were rearranged to create |ٺ| to denote Sindhi-specific [ʈ].⁸⁰ The creation of unique graphs for aspirate and retroflex φ-consonants in 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚊𝚍 by adding or rearranging nuqtas caught the attention of nineteenth-century European scholars, albeit not always in a positive manner. In his Essay on the Sindian Alphabets (1857), Trumpp dismisses such innovation as “load[ing] the Arabic alphabet with a confusing heap of dots and other diacritical marks” (p. 685). However, in his (1872) grammar, Trumpp describes the practice of creating new graphs by adding nuqtas to existing ones in a more measured and insightful manner:

The Sindhis had in this undertaking apparently the Sanskrit alphabet before their eyes, where the aspirates are written and treated as one sound. Accordingly they tried to express the aspiration of a letter by additional dots, which overloaded the few Arabic bases with diacritical signs. […] This attempt to adapt the Arabic characters to the sounds of a Prakrit language is very interesting, […]

(Trumpp, 1872, p. 3)

Despite these attempts at proto-standardisation, idiosyncratic practices persisted in the use of 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚊𝚍. Apart from decrying the addition of nuqtas as indiscriminate and unmethodical, Trumpp (1872, pp. 2–3) also notes that certain Sindhi φ-consonants were altogether ignored in the process, sometimes inexplicably so. For instance, velar [k kʰ ɡ ɠ ɡʱ] were all represented in 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚊𝚍 by |ک|, despite the availability of 𝚏𝚊-𝙰𝚛𝚊𝚋 |گ| [ɡ]. Similarly, separate graphs were not created for retroflex [ɽ] and [ɳ], which were represented simply by |ر| [ɾ] and |ن| [n]. Also employed were digraphs such as |نک| for [ŋ], which proved ambiguous given the multiple phonological values of |ک|. As a result, the linguistic values of the graphs in question remained variable.

Given the variation in the graph inventory and graph-phone correspondences of 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚊𝚍, it is unsurprising that the spellings of individual words also showed considerable variation. This was the case even with frequently occurring items. For instance, the graphematic composition of the Sindhi third-person singular copula form [aɦe] ‘is’ ranged from the now-normative |آهي| to the now-obsolete |اٰهِ| (Allana, 1993 [1964], p. 121). Notwithstanding such variation, these practices continued until the mid-nineteenth century, until British efforts at further standardisation resulted in the introduction of a modified graph inventory and linguistic values.

6.3 Barrow Ellis’ system

As outlined in Section 5.2, Commissioner Bartle Frere’s proclamation of Sindhi as the language of administration in British-ruled Sindh set into motion government efforts towards creating a standardised writing system for the language — at least more standardised than prevalent ones. Frere’s deputy, Assistant Commissioner Barrow Ellis, put together a committee to decide on the matter. After much debate, the committee decided to adopt a writing system based on the Arabic script with additional graphs for Sindhi-specific phones. The system was finalised in July 1853 (Government of Bombay, 1857, p. 104) and officially proclaimed on 29 August 1857 (Hauze, 2016g). The resultant 𝚜𝚍-𝙰𝚛𝚊𝚋 system has been described by Allana (1993 [1964]) as “Ellis’ orthography”.⁸¹ Colonial-era authors have dubbed the system the “Government Alphabet” (Trumpp, 1872, pp. 534–535) or the “Government Sindhī alphabet” (Grierson, 1919, p. 335), alluding to its bureaucratic origins. In terms of language subtags, the system will be referred to as 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝟷𝟾𝟻𝟹. As was the case with books printed in 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚊𝚍, the first books printed in the 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝟷𝟾𝟻𝟹 system were also lithographed, with publications using metal types appearing only in 1863 (Goldsmid, 1863, p. v).

6.3.1 Phonograms

An overview of γ-consonants in 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝟷𝟾𝟻𝟹 is shown in Table 6.4.

Illustrations of 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝟷𝟾𝟻𝟹 γ-vowels comprising subsegmental graphs combined with alif, hamza and |ب| [b] are shown in Table 6.5. Where variant forms exist, the choice of form may be determined by graphematic environment in certain instances, and freely by the writer in others.

Comparing Table 6.3 and Table 6.4, it is evident that the main distinction between the 𝚜𝚍-𝙰𝚛𝚊𝚋 systems of Abul Hasan and Barrow Ellis lay in the makeup of their respective graph inventories and graph-phone correspondences. In particular, 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝟷𝟾𝟻𝟹 featured modified graphetic forms of certain graphs, as well as more predictable correspondences between graphs and their linguistic values. The number of graphs listed in the collation order was also set at 52. Despite these measures, the new system was far from comprehensive. One of the earliest detailed critiques of 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝟷𝟾𝟻𝟹 is found in Trumpp (1872), who writes:

This new system [𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝟷𝟾𝟻𝟹], instead of striking at the root of the previous confusion [𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚊𝚍], merely endeavoured to make up some deficiencies of the old, while retaining all its errors, so that it cannot even boast of the compactness of the old system.

(Trumpp, 1872, p. 4)

Trumpp’s critique focuses on γ-consonants. Regarding the velars, he expresses surprise at the endorsement of |ڪ| [k] and |ک| [kʰ]. In 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚊𝚍, |ڪ| appeared only as a stylistic allograph of |ک| in Initial and Medial positions (Gacek, 2008, pp. 8, 43; 2009, p. 275).⁸² On this basis, Trumpp opposes |ڪ| being spun off into a distinctive phonogram. He also criticises the creation of |ڱ| [ŋ] and |ڳ| [ɠ] for various reasons, primarily for featuring a surfeit of nuqtas. Regarding the alveolo-palatals, retroflexes and labials, he notes the apparently unsystematic and, occasionally, unnecessary rearrangement of nuqtas on some graphs compared to the previous system, while also disapproving of the creation of |ڻ| /ɳ/. While he approves of the incorporation of 𝚞𝚛-𝙰𝚛𝚊𝚋 |‏گ‎ ‏گھ‎ ‏جھ‎ ‏ڙھ‎| /ɡ ɡʱ d͡ʑʱ ɽʱ/, he also questions this piecemeal inclusion as opposed to adopting the Hindustani graph inventory in its entirety for Sindhi:

[T]he old Sindhī system of writing [𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚊𝚍] did not answer its purposes, quite abstracted from its deficiency; but instead of emendating the old system by a different distribution of dots and inserting a few Hindūstānī letters, we consider it far more advisable, to adopt the whole Hindūstānī consonantal system, and to mark those sounds, which are peculiar to the Sindhī, by convenient dots.

(Trumpp, 1872, p. 5)

On this basis, Trumpp proposes a new variant of 𝚜𝚍-𝙰𝚛𝚊𝚋, with a graph inventory, linguistic values and spellings based on 𝚞𝚛-𝙰𝚛𝚊𝚋, described in Section 6.4.

An aspect of 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝟷𝟾𝟻𝟹 absent from Trumpp’s critique but raised by Allana (1993 [1964], p. 104 footnote) is that of aspirate φ-sonorants in Sindhi. Allana’s stance is that [mʱ nʱ ɳʱ lʱ ɽʱ ʋʱ] should be considered standalone phonemes in Sindhi phonology, similar to aspirate oral stops.⁸³ According to Allana, the interpretation of [mʱ nʱ ɳʱ lʱ ɽʱ ʋʱ] as clusters of a φ-sonorant and [ɦ] is mistaken, as is their digraphic representation as shown in (9):

Allana contends that representing each of [mʱ nʱ ɳʱ lʱ ɽʱ ʋʱ] with a digraph has given rise to gratuitous or redundant graphematic practices, such as affixing subsegmental γ-vowels multiple times on them (Allana, 1993 [1964], p. 149). As outlined in Section 4.3.1, Allana’s stance on the phonological status of [mʱ nʱ ɳʱ lʱ ɽʱ ʋʱ] in Sindhi is echoed by Nihalani (1999), but not by Cole (2001) or Khubchandani (2007). As Cole (2001) observes, there are few, if any, instances in Sindhi where [mʱ nʱ ɳʱ lʱ ɽʱ ʋʱ] contrast with the sequences [mɦ nɦ ɳɦ ɽɦ lɦ ʋɦ].⁸⁴ This may explain why visually distinct graphs may not have been considered for these phones. Indeed, it appears that no major South Asian writing system — including those of Sindhi — features dedicated graphs for any of these segments.⁸⁵ Nevertheless, Allana’s observation on the potential ambiguities caused by the digraphic representation of [mʱ nʱ ɳʱ lʱ ɽʱ ʋʱ] and the inconsistent addition of subsegmental graphs is valid, and continues to be observable in present-day 𝚜𝚍-𝙰𝚛𝚊𝚋 orthographic practices. As will be described in Section 6.5, such variable practices may have implications for 𝚜𝚍-𝙰𝚛𝚊𝚋 pedagogy and literacy.

With regard to the graphematic representation of φ-vowels, 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝟷𝟾𝟻𝟹 largely persisted with the 𝚡-𝚝𝚛𝚊𝚍 practice of explicitly denoting all γ-vowels, including any subsegmental elements. However, it also retained the shortcomings of the 𝚡-𝚝𝚛𝚊𝚍 variant, such as the graphematic underdifferentiation of [u] and [o], with both written |◌ـوُ|. A prominent example of this phenomenon is seen in Bartle Frere’s proclamation on 29 August 1857 (Hauze, 2016g) declaring Sindhi the official language of government work in Sindh. The title of the document features the spelling |پَڌَرۡناموُ| [pəd̪ʱəɾᵊnamo] ‘proclamation’, with [o] written |◌ـوُ|. Similar instances are also seen in the body text of the proclamation, such as |تَرجموُ| [t̪əɾᵊd͡ʑəmo] ‘translation’ and |مَهـيِـۡنوُ| [məɦino] ‘month’. Furthermore, while the new system persisted with using |⁠◌ٍ ⁠◌ٌ| for word-final [ĩ ũ (õ)] (Allana, 1993 [1964], p. 115), it also permitted |‏◌ـِين‎ ‏◌ـُون‎| as graphematic equivalents. This resulted in at least two possible ways to spell words containing these word-final φ-vowels.

The main reason why 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝟷𝟾𝟻𝟹 fell short of its intended aim of being a ‘standardised’ writing system for Sindhi was its failure to explicitly rule on the finer points of graphotactics (Lekhwani, 2011, p. 34). As Allana (1993 [1964], p. 129) notes, while Ellis’ system standardised the graph inventory to a large extent, it failed to usher in clarity on the spellings of individual words. As a result, words continued to be spelt in a variety of idiosyncratic ways, even by scholars. Such variable spellings extended to commonly occurring lexical items, such as [poᶦ] ‘afterwards’ appearing as |پوءِ| as well as |پوئي|, and [ʊn(ɦ)ənᶦ] ‘3PL.OBL’ being written as either |اُنَن| or |اُنَهنِ| (Allana, 1993 [1964], p. 130). Particularly vexing was the graphematic representation of nasalised final φ-vowels. Since word-final [ĩ ũ õ] could now be spelt |‏◌ـِين‎ ‏◌ـُون‎|, there was greater room for ambiguity on the phonological value of final |ن|. This issue was partly addressed by Shirt, Thavurdas and Mirza in their Sindhi-English dictionary (1879, p. iv), by adopting the following convention:

• |ن| with jazm |⁠◌ٛ| denotes [n]
• |ن| with a γ-vowel denotes [n] followed by a φ-vowel
• ‘bare’ |ن| denotes nasalisation of the preceding φ-vowel

In their dictionary, Shirt, Thavurdas and Mirza also make a rather telling statement on orthographic variation in 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝟷𝟾𝟻𝟹:

Some apology seems to be required for the many alternative forms that will be found in the Dictionary. Numerous as these may appear, it does not seem possible to eliminate them at present ; for they not only exist in the colloquial, but have all found a place in the written language. They cannot probably all survive the struggle for existence ; and the sooner the usage of some of them is discontinued, the better, but at present we can only speculate as to which forms will eventually prevail. This is a question which should form a subject of discussion by the Vernacular Literature Committee, and in future authors should confine themselves to those forms that meet with the approval of that body. Sindhi, though an old language, is still vigorous and growing, […]

(Shirt, Thavurdas, & Mirza, 1879, p. iii)

While the authors must be commended for attempting to describe rather than prescribe 𝚜𝚍-𝙰𝚛𝚊𝚋 graphematic practices, the “many alternative forms” they speak of is aptly illustrated in the dictionary entry for the Sindhi word [ɦoɗãɦᶷ̃]:

‏هوڏانهون‎ , ‏هوڏانهُون‎ , ‏هوڏانهِين‎ , ‏هوڏانئُن‎ , ‏هوڏانئِين‎ , ‏هوڏاهَنئُن‎ or ‏هوڏائون‎ adv. There, in that direction.

(Shirt, Thavurdas, & Mirza, 1879, p. 910)

All the same, Shirt et al.’s expectation that a graphematic natural selection would eventually prevail has indeed materialised to some extent. In the twenty-first century, 𝚜𝚍-𝙰𝚛𝚊𝚋 dictionaries commonly list only one spelling for the word [ɦoɗãɦᶷ̃], albeit one that is absent from Shirt et al.’s dictionary: |هوڏانهُن| (Baloch, 1998, p. 723; Lekhwani, 1996, p. 199; Rohra, Bijani, & Gurnani, 2011, p. 566).⁸⁶ Questions of spelling also extended to the nature of subsegmental graphs in a graphematic word. In the initial years of 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝟷𝟾𝟻𝟹, subsegmental graphs — most of which were γ-vowels — were generally written. Soon after, though, most subsegmental graphs, with the exception of tanvin, began to be omitted on a discretionary basis. For instance, a matriculation examination paper on Sindhi grammar from the year 1866 features 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝟷𝟾𝟻𝟹 text largely devoid of subsegmental graphs. However, the same paper also features questions like “Give the names of the short and long vowel marks [in 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝟷𝟾𝟻𝟹]” and “What are the names of the following marks, and what are the uses of each [in 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝟷𝟾𝟻𝟹]? ( ⁠◌ۡ ) ( ⁠◌ّ ) ( ⁠◌ٓ ) ( ⁠◌ٌ )” (University of Bombay, 1867, pp. xxxviii–xliii). The explicit recognition and inclusion of subsegmental graphs in formal education was accompanied by the implicit understanding that they would be left out of everyday 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝟷𝟾𝟻𝟹 writing. On the one hand, dropping subsegmental graphs from everyday use was consistent with 𝚊𝚛-𝙰𝚛𝚊𝚋 and 𝚏𝚊-𝙰𝚛𝚊𝚋 practice, as well as with abjadic unstandardised Landa writing (§5.1.1). On the other hand, the surface omission of subsegmental graphs from 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝟷𝟾𝟻𝟹 increased its graphematic opacity, eventually resulting in uncertainty on the underlying spellings of words. Swimming against the tide somewhat was the tanvin |⁠◌ٍ ⁠◌ٌ ⁠◌﮼|, which continued to be extensively used in nineteenth-century 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝟷𝟾𝟻𝟹 writing to denote word-final [ĩ ũ ẽ]. Figure 6.2 features an extract from a Sindhi translation of the Bible’s Gospel of John (American Bible Society, 1893, p. 26) in 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝟷𝟾𝟻𝟹. First translated in 1858 (Nida, 1972, p. 393), the text features several instances of tanvin being productively used, such as the word |ڎِناءٍ| [ɗɪnaĩ] ‘he gave’.⁸⁷

“For God so loved the world, that he gave his only begotten Son, that whosoever believeth in him should not perish, but have everlasting life.”

By the early twentieth century, though, the tanvin had given way to bare |ن| as the standard graphematic device to denote φ-vowel nasalisation.⁸⁸ This aligned with a large-scale, albeit incremental move, in the late nineteenth century towards resolving issues of graphematic irregularity in 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝟷𝟾𝟻𝟹. In 1888, the Education Inspector of Sindh, HP Jacobs, had a chart of the various spellings of certain words prepared. A chart of these words was distributed to schools throughout Sindh. However, this chart failed to address the spellings of words with final nasalised φ-vowels, as well as a few others. On 13 February 1913, educational experts along with members of the Sindhi Literature Committee passed a resolution to finalise standardised spellings for words omitted in the 1888 chart. This was followed up by a committee instituted in 1915, whose members included the renowned Sindhi author and translator Mirza Qalichbeg (Allana, 1993 [1964], pp. 131–135; Boivin, 2008c). Notably, the committee decided to retain the tanvin in the spellings of three commonly occurring words: |۽| [aᶦ̃ ~ aĩ] ‘and’, |۾| [mẽ] ‘in, within’ and |آءٌ| [aᶷ̃ ~ aũ] ‘1SG.NOM’ (see (8)). The decision to retain these spellings reflects their conventionalisation in the 𝚜𝚍-𝙰𝚛𝚊𝚋 graphosphere. Indeed, these three words appear to be the only entries featuring a tanvin in Shirt et al.’s (1879) and Mewaram’s (1910) Sindhi-English dictionaries.

Of these dictionaries, Mewaram’s (1910) work is now considered a lexicographic milestone in 𝚜𝚍-𝙰𝚛𝚊𝚋, and remains a key reference work more than a century after its publication (Addleton & Brown, 2010, p. xvi). It is unclear, though, whether and to what extent Mewaram’s dictionary might have contributed to the 1913–1915 reforms. By and large, though, the reforms succeeded in constraining the graphematic solution space for certain words in 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚊𝚍. At the same time, it is difficult to unequivocally assert that 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝟷𝟾𝟻𝟹 as updated in the 1915 reforms has prevailed as the undisputed variant of 𝚜𝚍-𝙰𝚛𝚊𝚋. The 1947 Partition of British India resulted in the split of the 𝚜𝚍-𝙰𝚛𝚊𝚋 graphosphere into Pakistani and Indian subgroups, and ongoing political tensions between Pakistan and independent India have resulted in these subgroups being relatively isolated from each other. Such mutual isolation has resulted in minor, albeit noticeable divergence in 𝚜𝚍-𝙰𝚛𝚊𝚋 graphematic practices between the two graphospheres, which will be described and evaluated in Section 6.5.

6.3.2 Logograms and collation order

The first book printed in 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝟷𝟾𝟻𝟹 using movable type (Goldsmid, 1863) contained the numeral forms |۱ ۲ ۳ ۴ ٥ ٦ ٧ ۸ ۹ ۰| ‘1 2 3 4 5 6 7 8 9 0’ (see (7)). These forms are identical to those commonly used in present-day 𝚊𝚛-𝙰𝚛𝚊𝚋 except for |۴| ‘4’, which usually appears as |٤| (Bauer, 1996, p. 562). Graphs used as punctuation in 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝟷𝟾𝟻𝟹 were largely adopted from the Roman-script inventory, although their use was highly variable. The Roman-script full stop | . | varied freely with the hyphen-like equivalent |۔| adopted from the 𝚞𝚛-𝙰𝚛𝚊𝚋 inventory (Grierson, 1919; University of Bombay, 1875, pp. xlvii–l). Asymmetrical graphs such as | , ; ? | initially appeared in their Roman-script orientations (University of Bombay, 1900, p. xlviii). They were eventually replaced by visually mirrored ones, presumably to harmonise with the dextrosinistral flow of 𝚜𝚍-𝙰𝚛𝚊𝚋.

Regarding the collation order of 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝟷𝟾𝟻𝟹, Allana (1993 [1964], p. 105) lists the following 52-graph sequence:

The above collation order is largely identical to the one listed in the very first 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝟷𝟾𝟻𝟹 printed using movable type (Goldsmid, 1863, pp. 28–29). However, the latter list omits hamza |ء|, resulting in a total of 51 graphs:

In their Sindhi-English dictionary, Shirt et al. (1879) use the following collation order:

This is also the order that Grierson (1919, p. 21) lists in the Linguistic Survey of India, terming it the earliest ‘standard’. This order differs from Goldsmid’s in the position of |ڱ| [ŋ]. It also comprises 52 graphs thanks to the addition of the 𝚏𝚊-𝙰𝚛𝚊𝚋 |ژ| [ʒ] (see Table 6.2), despite the corresponding phoneme being absent in Sindhi.

Mewaram (1910) uses the following collation order in his classic dictionary:

In this order, the status of |آ| [a] is ambiguous. Mewaram’s dictionary lists headwords commencing with |آ| separately, albeit within the section for |ا|. Overall, though, Mewaram’s collation order represents a return to Goldsmid’s order in terms of the position of |ڱ| and the omission of |ژ|. The omission of |ژ| has proven decisive, resulting in the disappearance of the graph from modern-day 𝚜𝚍-𝙰𝚛𝚊𝚋 inventories (Jhangiani, 1987).

6.4 Ernest Trumpp’s system

As described in the previous section, one of the most outspoken critics of Ellis’ system was the German Christian missionary Ernest Trumpp. One of the earliest — if the not the earliest — mentions of Trumpp’s dislike of 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝟷𝟾𝟻𝟹 is found in the brief ‘Essay on the Sindian Alphabets’, published in the Journal of the Royal Asiatic Society (1857). In this article, Trumpp criticises 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝟷𝟾𝟻𝟹 as being loaded “with a confusing heap of dots and other diacritical marks” (Trumpp, 1857, p. 685). As an alternative, Trumpp moots an alternative Arabic-script-based graph inventory and graph-phone correspondences based on 𝚞𝚛-𝙰𝚛𝚊𝚋. This system will be referred to henceforth by the language tag 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚞𝚖𝚙𝚙. Ideally, though, Trumpp was inclined towards using Devanagari for Sindhi. Besides, he subscribed to the existence of supposed religious prejudice between Muslim and Hindu Sindhis, and the consequent need for separate alphabets for the two communities (§5.2.2). On this basis, Trumpp proposes two distinct writing systems for Sindhi — one based on 𝚞𝚛-𝙰𝚛𝚊𝚋, and the other on Sanskrit-Devanagari (𝚜𝚊-𝙳𝚎𝚟𝚊). The latter is taken up in Section 7.4.

The first instance of 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚞𝚖𝚙𝚙 appearing in a printed work is likely in A Sindhi Reading-Book in the Sanscrit and Arabic Character (Trumpp, 1858). Comprising translations of Christian scriptures, the Sindhi translations in the book appear in 𝚜𝚍-𝙰𝚛𝚊𝚋 as well as in 𝚜𝚍-𝙳𝚎𝚟𝚊. This reflected Trumpp’s view that two separate writing systems were apparently needed for Muslim and Hindu Sindhis. In a German-language article on Sindhi grammar published in the Zeitschrift der Deutschen Morgenländischen Gesellschaft (ZDMG), Trumpp provides an overview of his proposed 𝚜𝚍-𝙰𝚛𝚊𝚋 and 𝚜𝚍-𝙳𝚎𝚟𝚊 systems, while making no bones about his feelings on 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝟷𝟾𝟻𝟹:

Die vorstehenden Alphabete [𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚞𝚖𝚙𝚙 & 𝚜𝚍-𝙳𝚎𝚟𝚊-𝚡-𝚝𝚛𝚞𝚖𝚙𝚙] habe ich, sowie sie gegeben sind, erst selbst verfertigt; es sind also keineswegs althergebrachte Systeme. […] Das arabisirte Alphabet, für die Muhammedaner bestimmt, habe ich zum ersten male schneiden und damit drucken lassen; ich habe dabei das Hindūstānī-Alphabet zu Grunde gelegt, und die für das Sindhī nothigen Laute [sic] eingeschoben. Die Muhammedaner in Sindh gebrauchen verschiedene arabische Alphabete, theils mit, theils ohne diacritische Zeichen, aber so unwissentschaftlich zusammengestellt, dass sie einer Revision dringend bedürfen.

Um der herrschenden Confusion ein Ende zu Machen, hat kurz vor meiner Ankunft in Sindh, ein englischer Regierungsbeamter [Frere? Ellis?] ein neues arabisirtes Alphabet componirt, und es dem Lande aufgedrungen, das aber das albernste von allen genannt werden muss.

The alphabets presented here [𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚞𝚖𝚙𝚙 & 𝚜𝚍-𝙳𝚎𝚟𝚊-𝚡-𝚝𝚛𝚞𝚖𝚙𝚙] are in no way established systems. Rather, they have been designed by me […] The Arabic alphabet, intended for Mohammedans, has been cast and used in print for the first time by myself.⁸⁹ I have based it on the Hindūstānī alphabet, adding those sounds [sic] necessary for Sindhī. The Mohammedans in Sindh use several Arabic alphabets, sometimes with, sometimes without diacritical marks, but in a manner so unscientific that it had to be urgently revised.

In order to bring the reigning confusion to an end, an English government official [Frere? Ellis?] had an Arabic alphabet prepared and imposed upon the country. This alphabet, however, must be regarded as the most ridiculous of them all.

(Trumpp, 1861, pp. 697–698 footnote)

Trumpp also employs his 𝚜𝚍-𝙰𝚛𝚊𝚋 system in subsequent ZDMG articles. The first (1862) is a continuation of his Sindhi grammar from the previous year, while the second is an overview of Shah Abdul Latif’s poetry (1863). The latter work was eventually published as a separate compilation (Trumpp, 1866), marking the first appearance of Shah Latif’s poetry in print (Boivin, 2020, p. 95). Trumpp’s ZDMG articles on Sindhi grammar were also compiled and published as a book (1872), which is now considered a classic in the field of Sindhi linguistics.

On the question of script to be used for Sindhi, Trumpp writes in his grammar that “[t]he number of the Indian alphabets should not be augmented, but rather, wherever possible, be restricted, as they only serve as barriers to mutual intercourse” (Trumpp, 1872, p. 6). On this basis, he opines that, since the Sindhi language is “restricted to the comparatively small province of Sindh”, it makes sense to adopt the more widely used “Hindustani alphabet” (𝚞𝚛-𝙰𝚛𝚊𝚋), with suitable modifications as necessary to suit Sindhi phonology.

Compared to its first appearance in print (Trumpp, 1858), 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚞𝚖𝚙𝚙 varied ever so slightly in subsequent works until finalised (Trumpp, 1872). The description that follows is based on the final version, and emphasises the points of divergence from 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝟷𝟾𝟻𝟹.

6.4.1 Phonograms

Figure 6.3 is an extract from Trumpp’s grammar (1872), showing the inventory of γ-consonants in his proposed 𝚜𝚍-𝙰𝚛𝚊𝚋 system.

As seen in Figure 6.3, 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚞𝚖𝚙𝚙 is founded upon the 𝚞𝚛-𝙰𝚛𝚊𝚋 graph inventory of the time. The most prominent evidence of this fact is the consistent use of |ھ| to denote an aspirated φ-stop. Similarly conspicuous is the choice of the 𝚞𝚛-𝙰𝚛𝚊𝚋 |ٿ ڐ ڙ| for [ʈ ɖ ɽ], respectively, which, for much of the nineteenth century, were in free variation with the now-standard |ٹ ڈ ڑ| (Lepsius, 1863, p. 101; Shakespear, 1813, p. viii). Phonologically, 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚞𝚖𝚙𝚙 spellings reflect Siroli pronunciations in overtly denoting the clusters [ʈɾ ɖɾ], thus distinguishing them from simple [ʈ ɖ] (§4.3.1). Notable is the depiction of |ڙھ| as a compound graph for [ɽʱ], illustrated by its equivalence with 𝚜𝚍-𝙳𝚎𝚟𝚊 |ढ़|. The inclusion of |ڙھ| as a standalone entity in the inventory suggests that Trumpp considered [ɽʱ] phonemic (see Table 4.1). Also noteworthy is the presence of |ن࣪| for [ɳ]. Since [ɳ] is not well established in classical Urdu, |ن࣪| may be considered a marginal graph in 𝚞𝚛-𝙰𝚛𝚊𝚋 (Shakespear, 1813, p. 22). However, the phonological salience of [ɳ] in Sindhi meant that |ن࣪| was prominent in the 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚞𝚖𝚙𝚙 inventory. In contrast, the 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚞𝚖𝚙𝚙 inventory does not feature an equivalent for 𝚜𝚊-𝙳𝚎𝚟𝚊 |ष|, which denotes [ʂ(a)] in Sanskrit. The omission was deliberate, since Trumpp felt that the articulation of [ʂ] as such in Sindhi “is completely ignored by the common people” (Trumpp, 1872, p. 18). The graphs |ٻ ڄ ڎ| for the implosives [ɓ ʄ ɗ] are retained from 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚊𝚍. However, [ɠ] is depicted by the Pashto-inspired |ڰ| (cf. Pashto |ګ| [ɡ]).⁹⁰ Completing the inventory are the graph sequences |‏نگ‎ ‏نڄ‎| for [ŋ ɲ]. Trumpp admits to the ambiguity inherent in these graph sequences, but puts the onus on the reader to navigate them:

In Hindūstānī an independent guttural ṅ [IPA ŋ] is not to be met with ; we have therefore been compelled to circumscribe it [in 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚞𝚖𝚙𝚙] by the compound نگ ng […] But one difficulty still will remain, that the guttural simple ṅ can thus not be distinguished from the guttural ṅ preceding a letter of its own varga [i.e., phonological category], as: اَنگُ (अङु) aṅu, body, and اَنگُ (अंगु, or more properly: अङ्गु) angu (aṅgu) a limb. […] Practically the difficulty will be easily surmounted by any careful student, […]

(Trumpp, 1872, p. 11)

Allana (1993 [1964], p. 124) reiterates and critiques the above ambiguity. However, in doing so, Allana depicts the graph sequence for [ɲ] as |نج|, rather than Trumpp’s |نڄ|. This typographical mix-up may well be an inadvertent illustration of what Trumpp has termed “the reigning confusion” across 𝚜𝚍-𝙰𝚛𝚊𝚋 graph inventories (Trumpp, 1861, pp. 697–698 footnote).

Regarding consonant gemination, Trumpp recommends using tashdid |⁠◌ّ| only for written disambiguation, offering the examples of |اُنَ| [ʊnᵊ] ‘3SG.OBL’ and |اُنَّ| [ʊnnᵊ] ‘wool’. However, he also states that, “in the old Sindhī writings the tašdīd is hardly ever to be met with […]” (1872, p. 29). Overall, Trumpp makes no explicit statements on whether phonological gemination in Sindhi should be considered contrastive (§4.3.1). Aside from these modifications, all γ-consonants originating in 𝚊𝚛-𝙰𝚛𝚊𝚋 and 𝚏𝚊-𝙰𝚛𝚊𝚋 are retained unchanged. This includes 𝚏𝚊-𝙰𝚛𝚊𝚋 |ژ| [ʑ ~ ʒ] (see Figure 6.3), which was absent from the initial version of 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚞𝚖𝚙𝚙 (Trumpp, 1858).

Figure 6.4 shows the inventory of γ-vowels in 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚞𝚖𝚙𝚙.

In explicitly endorsing |◌ٖي| for [e], 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚞𝚖𝚙𝚙 adopts graphematic practice from 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚊𝚍 and 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝟷𝟾𝟻𝟹. However, there existed no comparable precedent for denoting [o]. Against this background, 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚞𝚖𝚙𝚙’s representation of [o] as |◌ۉ| counts as an innovation. Graphetically, |◌ۉ| is a compound of |و| and jazm |◌ٛ|, which precludes the latter from being unambiguously used in 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚞𝚖𝚙𝚙 as an allograph of sukun |◌ْ| to show the absence of a postconsonantal φ-vowel [Ø]. Hence, 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚞𝚖𝚙𝚙 only uses sukun |◌ْ| to denote [Ø].⁹¹ Regardless of nomenclature, |◌ْ| is also used in the 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚞𝚖𝚙𝚙 graphs |اَيْ| and |اَوْ|, denoting the diphthongs [ɛ ~ əɪ̯] and [ɔ ~ əʊ], respectively. Strictly speaking, |اَيْ| and |اَوْ| could also be read as [əj] and [əw], respectively, although doing so does not result in a phonemic contrast in Sindhi. Finally, 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚞𝚖𝚙𝚙 explicitly deprecates the practice of using zer |◌ِ| [ɪ] to show the absence of a vowel [Ø] (Trumpp, 1872, p. 29). Yet, the use of |◌ِ| as a quasi-allograph of sukun |◌ْ| has persisted into the twenty-first century.

In terms of word-medial sequences of γ-vowels, 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚞𝚖𝚙𝚙 codifies and explicitly endorses the use of hamza, including the distribution of its allographs. His recommendations are consistent with present-day 𝚜𝚍-𝙰𝚛𝚊𝚋 usage:

The sign ــٔـ (هَمْزَه hamzah, i.e. punction) is used in the midst and at the end of words as a vicarious base for ا; when two vowels, short or long, meet in a word, the second vowel must be supported by the base ى, furnished with hamzah, e.g.: […]هِنئَرَ hĩara, now; […] A final short vowel, preceded by ā, ū, ō is usually supported by the sign hamzah alone, as بهاءُ bhā-u, brother;

(Trumpp, 1872, p. 29)

Trumpp includes a laborious — and occasionally inconsistent — explanation on how to use |ن| to denote nasalisation of a φ-vowel on the one hand, and the alveolar nasal φ-consonant [n] on the other (pp. 25–28). Regardless of the explanation, examples from his grammar reveal the following graphematic patterns and phonological values:

• bare |ن| denotes φ-vowel nasalisation, and is written immediately after the corresponding γ-vowel;
• |ن| suprafixed with sukun |◌ْ| represents the nasal φ-consonant [n];
• |ن| affixed with a subsegmental γ-vowel represents [n] followed by the corresponding φ-vowel.

In this regard, Trumpp’s representation of φ-vowel nasalisation is near-identical to that advocated by Shirt et al. (1879), although it remains unclear whether the latter was inspired in any way by the former. At any rate, 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚞𝚖𝚙𝚙’s conventions on the use of |ن| were more consistent than in 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚊𝚍 and 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝟷𝟾𝟻𝟹 (Trumpp, 1872, p. 28). Aside from its inconsistent use, Trumpp speaks out against using tanvin to mark φ-vowel nasalisation due to the potential for a clash with its 𝚊𝚛-𝙰𝚛𝚊𝚋 value:

Die Sindhī’s [sic] selbst haben versucht, einen auslautenden nasalisirten Vocal durch das arabische Tanvīn auszudrücken, z. B. sie schreiben کِيَاࣨ = किआंऊं; abgesehen davon, dass das arabische Tanvīn keine Nasalisation im Sinne der Prākritsprachen ausdrückt, tritt dabei auch noch der Misstand ein, dass die kurzen oder langen nasalisirten Vocale gar nicht mehr in der Schreibweise auseinander gehalten werden können. […] Allein die Form ۽ = ē̃ „und“ haben wir beibehalten, obschon sie an und für sich ein Monstrum ist, als eine bequeme Abkürzung, wie etwa unser & statt et; eigentlich sollte es اين oder اَيْن geschrieben werden.

The Sindhis themselves tried to mark a final nasalised vowel using the Arabic Tanvīn. Thus, they write کِيَاࣨ = किआंऊं. Besides the fact that the Tanvīn does not represent nasalisation of the Prākrit kind, the use of Tanvīn also means that short and long nasalised vowels can no longer be clearly distinguished in writing. […] We have only retained the form ۽ = ē̃ “and”, which is a monstrosity that, nevertheless, serves as a convenient abbreviation, much like our & for et. Ideally, it should have been written اين or اَيْن.

(Trumpp, 1863, p. 257)

Thus, with the exception of logographic |۽| [ɛ̃ ~ əĩ] ‘and’, 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚞𝚖𝚙𝚙 deprecates the use of tanvin to denote φ-vowel nasalisation. Consequently, 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚊𝚍 |۾| [mẽ] ‘in’ becomes 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚞𝚖𝚙𝚙 |مٖين| (Trumpp, 1866, p. 731). In general, 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚞𝚖𝚙𝚙 discourages graphematic recycling from 𝚊𝚛-𝙰𝚛𝚊𝚋. By extension, it recommends retaining source writing conventions in borrowings, as in the use of medial |آ| in 𝚜𝚍-𝙰𝚛𝚊𝚋 |قُرآنُ| [qʊɾanᶷ ~ kʊɾanᶷ] ‘Quran’ (Trumpp, 1872, p. 28).

Overall, 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚞𝚖𝚙𝚙 consistently and rigorously denotes all graphetically subsegmental elements. Although Trumpp emphasises the utility of retaining subsegmental graphs in works intended for a European readership, he concedes that they may well prove superfluous in works “destined for the use of natives” (p. 23).

6.4.2 Logograms and collation order

While 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚞𝚖𝚙𝚙 contains no explicit specifications on numeral forms per se, Trumpp’s works feature the graphs |۱ ۲ ۳ ۴ ٥ ٦ ٧ ۸ ۹ ۰| ‘1 2 3 4 5 6 7 8 9 0’ (Trumpp, 1866; 1872, pp. 157–168). These are identical to those employed by Goldsmid (1863), making the numeral inventories of 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝟷𝟾𝟻𝟹 and 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚞𝚖𝚙𝚙 identical. Also absent from 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚞𝚖𝚙𝚙 are rulings on punctuation marks, although a four-dot separator |⁛| is consistently used as the equivalent of the 𝚎𝚗-𝙻𝚊𝚝𝚗 full stop. In fact, Trumpp employs this graph not just in his 𝚜𝚍-𝙰𝚛𝚊𝚋 works, but also his 𝚜𝚍-𝙳𝚎𝚟𝚊 texts (Trumpp, 1858; 1866; §7.4, p. 201). Other prominent logograms in 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚞𝚖𝚙𝚙 include the aforementioned |۽| [ɛ̃ ~ əĩ] ‘and’, and the benediction |اللّٰه| ‘Allah’. Trumpp does, however, augment the latter with subsegmental phonograms to denote Sindhi-specific noun declensions (§4.4), as in |اللّٰهُ| [əl(l)aɦᶷ] ‘Allah (nominative case)’ and |اللّٰهَ| [əl(l)aɦᵊ] ‘Allah (oblique case)’ (Trumpp, 1866). The juxtaposition of the graphosegmental logogram |اللّٰه| with the graphosubsegmental phonograms |◌َ ◌ُ| [ᵊ ᶷ], which also occurs in 𝚊𝚛-𝙰𝚛𝚊𝚋, presents an intriguing case of hybridity that tests the limits of prevailing grapholinguistic terminology and theory.

The comprehensiveness of 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚞𝚖𝚙𝚙 is well-illustrated by the inclusion of a collation order, with graphs sorted based on an 𝚊𝚛-𝙰𝚛𝚊𝚋-inspired collation order (Trumpp, 1872, pp. 8–9):

Compared with the graphs shown in Figure 6.3 and Figure 6.4, the 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚞𝚖𝚙𝚙 collation order above does not list aspirate γ-consonants per se. This is likely due to Trumpp’s interpretation of aspirate γ-consonants as multigraphs comprising a γ-consonant and |ھ| (1872, p. 3). The sequence also omits hamza |ء|.

Apart from the odd incongruity, 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚞𝚖𝚙𝚙 emerges as clear and consistent in terms of graph inventory and graph-phone correspondences. Its regularity stands in stark contrast to the variation observed in the use of 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝟷𝟾𝟻𝟹, especially when one considers that the latter was instituted by a committee of scholars and administrators. Despite its graphematic robustness, though, 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚞𝚖𝚙𝚙 was extensively employed only by Trumpp himself. At most, the system was acknowledged by other authors, and the graph inventory reproduced for purposes of demonstration (Lepsius, 1863, pp. 103–105). It is possible that other Arabic-script-based writing systems may have borrowed elements of the 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚞𝚖𝚙𝚙 graph inventory (Jukes, 1900, p. v), although this remains conjectural. Apart from references by third parties, 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚞𝚖𝚙𝚙 does not seem to have been used in a book-length publication by any author other than Trumpp himself. Nor does there seem to be evidence of 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚞𝚖𝚙𝚙 being used in Sindhi-language education or everyday writing. Thus, 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚞𝚖𝚙𝚙 represents a case of a writing system fading into oblivion despite being well-designed and internally consistent. The fate of 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚞𝚖𝚙𝚙 reinforces Meletis’ (2020, pp. 16–17) observation that the linguistic fit of a writing system does not guarantee its adoption by the community.

6.5 Post-Partition practices ⁹²

As described in Section 6.3, the 1915 reforms of 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝟷𝟾𝟻𝟹 (helped constrain variation in spelling and, in turn, establish the foundations of a quasi-conventional orthography (Allana, 1993 [1964], p. 136). Although areas of ambiguity remained, increasing levels of literacy in 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝟷𝟾𝟻𝟹 led to standardised spellings becoming established over time. However, by the early twentieth century, the 𝚊𝚛-𝙰𝚛𝚊𝚋-inspired practice of omitting subsegmental graphs had caught on in 𝚜𝚍-𝙰𝚛𝚊𝚋. Due to the ‘out of sight, out of mind’ effect of omitting subsegmental γ-vowels, attempts at spelling standardisation were restricted almost entirely to matters of γ-consonants. Questions on which γ-vowels to use in certain words were largely overlooked (Allana, 1993 [1964], p. 136; Khubchandani, 1969, p. 204). A major event impacting the graphematic evolution of 𝚜𝚍-𝙰𝚛𝚊𝚋 was Partition in 1947, which resulted in the bifurcation of the Sindhi graphosphere into its Pakistani and Indian circles. Although 𝚜𝚍-𝙰𝚛𝚊𝚋 literary, pedagogical and lexicographic activity resumed shortly thereafter on both sides of the new international border, ensuing political tensions between the Pakistani and Indian governments impeded the cross-border exchange of 𝚜𝚍-𝙰𝚛𝚊𝚋 material. As a result, minor distinctions have emerged between Pakistani and Indian 𝚜𝚍-𝙰𝚛𝚊𝚋 graphematic practices in the post-Partition era.

Preeminent among post-Partition 𝚜𝚍-𝙰𝚛𝚊𝚋 lexicographic works in Pakistan is Baloch’s five-volume Jame Sindhi Lughat ‘Comprehensive Sindhi Dictionary’.⁹⁴ Described as a “monumental” work (Khubchandani, 2007, p. 685), the five volumes of this dictionary were released over two decades, from 1961–1981. This was followed by a second edition starting in the 1990s (Addleton & Brown, 2010, p. xvi). A one-volume version of the dictionary has also been published (Baloch, 1998). Other specialised lexicons and thesauri in 𝚜𝚍-𝙰𝚛𝚊𝚋 have also been published over the years, many of which the Sindhi Language Authority has digitised and compiled into a searchable database on its website (Sindhi Language Authority, 2021a). The SLA has also produced an autodidactic online tutorial for spoken and written Sindhi, with the written form presented in 𝚜𝚍-𝙰𝚛𝚊𝚋 as well as 𝚜𝚍-𝙻𝚊𝚝𝚗 (§12.3). Albeit few in number, grapholinguistic works on 𝚜𝚍-𝙰𝚛𝚊𝚋 have also been published, the most prominent being Allana’s (1993 [1964]) monograph on the history and development of 𝚜𝚍-𝙰𝚛𝚊𝚋. Among 𝚜𝚍-𝙰𝚛𝚊𝚋 lexicographic works published in post-Partition India, Lekhwani’s trilingual Basic Sindhi-English-Hindi dictionary (1996) and bilingual An Intensive Course in Sindhi textbook (1997) are worthy of mention. Besides their lexicographic value, they also exhibit a high level of graphematic rigour. Lekhwani’s writing conventions have been adopted with slight modifications in the Indian Institute of Sindhology’s Trilingual Sindhi-Hindi-English dictionary (Rohra, Bijani, & Gurnani, 2011). The publication of these dictionaries and related pedagogical works has had a stabilising effect on the graph inventory of 𝚜𝚍-𝙰𝚛𝚊𝚋 on both sides of the Indo-Pak border, particularly regarding γ-consonants. Yet, open questions still remain on matters of subsegmental graphs, allography and spelling.

6.5.1 Graphematic vowel nasalisation

In their 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝟷𝟾𝟻𝟹 Sindhi-English dictionary, Shirt et al. (1879) adopted the practice of graphematically denoting a nasalised φ-vowel with bare |ن|, devoid of any subsegmental graphs. Such practice was identical to that observed in 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚞𝚖𝚙𝚙 (Trumpp, 1872). In the post-Partition era, Allana has endorsed this practice, and categorically proscribes adding jazm |⁠◌ۡ| over |ن| to denote φ-vowel nasalisation:

‏سنڌي ڪتابن ۾ ’گهڻي نون‘ کي ’ساڪن‘ سمجهي، اُن جي مٿان جزم ڏني ويندي آهي. جزم جو اهڙي حالت ۾ استعمال سراسر غلط آهي.‎

Based on the interpretation of ‘nasal nun’ as ‘quiescent’ [vowelless], Sindhi books often affix jazm over nun. Such use of jazm is utterly wrong.

(Allana, 1993 [1964], p. 119 footnote)

Despite advocacy by various authors, the written representation of φ-vowel nasalisation in 𝚜𝚍-𝙰𝚛𝚊𝚋 has been variable. For instance, the SLA’s online Sindhi tutorial (Sindhi Language Authority, 2015b) uses |◌نۡ| to denote φ-vowel nasalisation, and bare |ن| to indicate consonantal [n] with no subsequent φ-vowel. This convention is the converse of that endorsed by Shirt et al., Trumpp’s and Allana. That said, despite swapping the graphematic practices in question, the SLA tutorial is internally consistent in their application. Overall, though, the use of jazm in 𝚜𝚍-𝙰𝚛𝚊𝚋 of late seems to be reducing on both sides of the border, even in pedagogical works (see for example Varyani & Thakwani, 2003). As a result, bare |ن| may indicate either φ-vowel nasalisation, or consonantal [n] with no φ-vowel after it.

Overall, the omission of subsegmental graphs in 𝚜𝚍-𝙰𝚛𝚊𝚋 results in homographs and, consequently, an increase in graphematic opacity. Thus, |تَوھان| [t̪əʋʱã] ‘2PL.OBL’ and |تَوھِينَ| [t̪ɔɦinᵊ] ‘insult’ may both appear as underspecified |توھين|, while |مِينھُن| [mĩɦᶷ̃] ‘rain’ and |مينھِن| [mẽɦᶦ̃] ‘she-buffalo’ become graphematically indistinguishable as |مينھن|. It is evident that omitting subsegmental graphs in this manner puts the onus on the reader to correctly decode the spoken values of the underspecified written strings from context. This necessitates a high level of proficiency in spoken Sindhi as well as in written 𝚜𝚍-𝙰𝚛𝚊𝚋.

6.5.2 Graphematic consonant gemination

The disagreement on the presence of geminate φ-consonants in spoken Sindhi is reflected in its writing systems. In 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚞𝚖𝚙𝚙, Trumpp (1872, p. xxxiii footnote) uses tashdid only to avoid potential homographs. Trumpp cites the examples of |اُنَ| ‘his/her/its’ and |اُنَّ| ‘wool’, wherein the graphovocalised forms indicate the pronunciations [ʊnᵊ] and [ʊnnᵊ], respectively. However, in Mewaram’s 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝟷𝟾𝟻𝟹 dictionary (1910, p. 25), the entry for ‘wool’ is spelt |اُنَ|. In the post-Partition era, scholars differ in their opinion on using tashdid. Allana (1993 [1964]) recommends that tashdid “should not be used in 𝚜𝚍-𝙰𝚛𝚊𝚋 orthography, except in Arabic-origin words” (p. 114).⁹⁵ In contrast, and as outlined in Section 4.3.1, Lekhwani recommends avoiding tashdid altogether in 𝚜𝚍-𝙰𝚛𝚊𝚋. This seemingly insignificant difference in guidelines results in words having noticeably distinct written manifestations in practice. Revisiting the example of the Arabic-origin loanword meaning ‘respect’ in Sindhi (§4.3.1), the common Sindhi pronunciation of this word varies along a spectrum, ranging from the quasi-Arabic [ɪzzət̪] (Khubchandani, 2007, p. 697) to the Sindhi-ised [ɪzət̪ᵊ] (Lekhwani, 1996, p. 12). Theoretically, when fully graphovocalised, the pronunciation [ɪzzət̪] would be represented as |عِزَّتۡ|, with tashdid included. In contrast, [ɪzət̪ᵊ] would be written |عِزَتَ|, with tashdid absent. When subsegmental graphs are omitted, however, the result in both instances is identical, namely the underspecified |عزت|. Dictionaries may differ in the fully graphovocalised forms they cite, or may simply cite the underspecified forms, without subsegmental graphs. As a result, standard fully-graphovocalised 𝚜𝚍-𝙰𝚛𝚊𝚋 spellings for several Sindhi words are yet to emerge.

6.5.3 Graphematic allography

Variation also prevails in the graphematic manifestation of φ-vowels and φ-consonants. In the context of vowels, variation is seen primarily in the nature and position of subsegmental elements. As shown in Table 6.2, the φ-vowels [i u] are shown in 𝚜𝚍-𝙰𝚛𝚊𝚋 by the complex graphs |‏◌ِي‎ ‏◌ُو‎|, respectively. However, [i u] may also appear as |‏◌يِ‎ ‏◌وُ‎|, differing in the placement of the subsegmental elements zer |◌ِ| and pesh |◌ُ|, respectively. A similar pair of allographs is also observed in the written representation of the φ-vowels [ɛ ɔ]. In 𝚜𝚍-𝙰𝚛𝚊𝚋, the commonly used complex graphs for these vowels are |‏◌َي‎ ‏◌َو‎|, respectively. However, | ‏◌َﺋِ‎ ‏◌ؤ‎| are also seen, as evinced by Figure 6.5. The image features a signboard at an intersection in the Khar suburb of Mumbai, named after an Indian Sindhi educationist.⁹⁶ The signboard displays the name of the intersection in 𝚜𝚍-𝙳𝚎𝚟𝚊, 𝚜𝚍-𝙰𝚛𝚊𝚋 and 𝚎𝚗-𝙻𝚊𝚝𝚗-𝙸𝙽.⁹⁷ In the 𝚜𝚍-𝙰𝚛𝚊𝚋 text, the Sindhi word [t͡ɕɔ̃kᶷ] ‘square, intersection’ appears in its graphematically underspecified form as |چؤنڪ|, with [ɔ] written as |◌ؤ|. In contrast, Mewaram (1910, p. 211), Lekhwani’s (1996, p. 57) and the SLA all list the word as |چَونڪُ|, in which [ɔ] is written |◌َو|. Allana (1993 [1964], pp. 117–118) and Lekhwani (1996, p. vi) consider | ‏◌َﺋِ‎ ‏◌ؤ‎| and to be allographs in free variation.

Similar variability is observed in the realisation of Sindhi [e], which may appear in 𝚜𝚍-𝙰𝚛𝚊𝚋 as |◌ٖي| or |◌يٖ|. In the post-Partition era, Pakistani 𝚜𝚍-𝙰𝚛𝚊𝚋 works have predominantly featured the former (Sindhi Language Authority, 2015b), while Indian 𝚜𝚍-𝙰𝚛𝚊𝚋 works have tended towards the latter (Lekhwani, 1996; 1997; Shahaney, 1967 [1906]; Varyani & Thakwani, 2003). Occasionally, though, a 𝚜𝚍-𝙰𝚛𝚊𝚋 work may depart from the trend typical of its location of publication, and feature cross-border practices instead. At the same time, the subsegmental element |◌ٖ| appears to be falling out of use on both sides of the border, even in fully graphovocalised texts. This has resulted in unintended graphematic standardisation, with Sindhi [e] being written simply as |◌ي| in most recent 𝚜𝚍-𝙰𝚛𝚊𝚋 publications (Lekhwani, 1996; 1997; Sindhi Language Authority, 2015b; Varyani & Thakwani, 2003). Graphetically, what distinguishes |◌ي| from visually similar graphs is the absence of a subsegmental graphetic component. Such distinctiveness-by-absence is reminiscent of the use of bare |ن| to indicate φ-vowel nasalisation. At the same time, the gradual exclusion of the element |◌ٖ| from |◌ي| is akin to the fading away of jazm from |ن|. In both instances, the elimination of the subsegmental graph may cause ambiguity and increase the system’s opacity.

Allography in representing φ-consonants primarily concerns [ɦ] and its written forms. With reference to Table 6.1, 𝚊𝚛-𝙰𝚛𝚊𝚋 and 𝚏𝚊-𝙰𝚛𝚊𝚋 denote the glottal fricative [ɦ] with the rasms |ه ﻫ ﻬ ﻪ|. The occurrence of these allographs or positional variants is conditioned by graphematic environment, putting the variants in mutually complementary distribution. In 𝚜𝚍-𝙰𝚛𝚊𝚋, however, consonantal [ɦ] is denoted by |ﮬ| in all graphematic positions (Dalwani, 2020; Kew, 2005; Motivational Skills By Sufi Shar, 2020; ScriptSource, 2022c). Moreover, 𝚜𝚍-𝙰𝚛𝚊𝚋 uses the rasm |ﻬ| as an invariant component in the digraphs |‏ﮔﻬ‎ ‏ﺟﻬ‎ ‏ﻣﻬ‎ ‏ﻧﻬ‎ ‏ﮢﻬ‎ ‏ﻟﻬ‎| [ɡʱ d͡ʑʱ mʱ nʱ ɳʱ lʱ] (see (9)) to denote φ-aspiration [ʱ]. Thus, in 𝚜𝚍-𝙰𝚛𝚊𝚋, the rasms |ﻬ ﮭ| cannot be considered allographic positional variants of the same underlying written unit. Certain 𝚜𝚍-𝙰𝚛𝚊𝚋 works adhere strictly to this allographic distinction (Lekhwani, 1996; NCPSL, 2005; 2014; 2015; Varyani & Thakwani, 2003). For instance, Lekhwani (1996) spells the word [d͡ʑəɦazᶷ] ‘ship’ as |جَھازُ|, using the rasm |ﮭ| to denote consonantal [ɦ]. In contrast, the headword [d͡ʑʱaɾᶦ] ‘flock of birds’ is transcribed |جهارِ|, where the rasm |ﻬ| marks aspiration. Extracts of these words as printed in Lekhwani (1996) are shown in Figure 6.6.

However, adherence to these allographic distributions is by no means universal, and 𝚜𝚍-𝙰𝚛𝚊𝚋 texts may mix and match these rasms somewhat unpredictably. That is, a 𝚜𝚍-𝙰𝚛𝚊𝚋 text may use |ﻬ| to mark segmental [ɦ] (Shahaney, 1967 [1906], p. 9), or employ |ﻫ| to denote aspiration [ʱ] (Baloch, 1962; Lekhwani, 1997). This results in |‏ﮔﻬ‎ ‏ﺟﻬ‎| [ɡʱ d͡ʑʱ] appearing instead as |‏گھ‎ ‏جھ‎|, respectively. Figure 6.5 contains an example of this phenomenon in the word |جھمٽمل| [d͡ʑʱəməʈᵊməlᶷ], in which [d͡ʑʱ] is written as |جھ| instead of |ﺟﻬ|. Compounding matters is the fact that the rasms |‏و‎ ‏ر‎| do not join the following rasm (see Table 6.1). Consequently, the digraphs for [ɽʰ ʋʱ] almost always appear as |‏ڙھ‎ ‏وھ‎|, and hardly ever as |‏ڙﻬ‎ ‏وﻬ‎|. The appearance of the rasm |ھ| in the digraphs |‏ڙھ‎ ‏وھ‎| further muddies the waters when discussing the potential phonemic status of [ɽʰ ʋʱ] in Sindhi.

In addition, in 𝚜𝚍-𝙰𝚛𝚊𝚋, consonantal [ɦ] in word-final position may occasionally be written with the 𝚊𝚛-𝙰𝚛𝚊𝚋-style rasm |◌ﻪ|. Thus, the Sindhi word [ɦəmeɕəɦᵊ] ‘always’ is spelt by Lekhwani (1996, p. 194) as |هَميشَھَ|, but by Shahaney (1967 [1906], p. 21) as |هَميشَهَ|. The former spelling denotes final [ɦ] with |◌ﮫ|, while the latter uses |◌ﻪ|. No discernible patterns are observed behind such variation, and must be attributed to font limitations and idiosyncratic preference. Furthermore, in 𝚜𝚍-𝙰𝚛𝚊𝚋, the rasm |◌ﮧ| (see Table 6.1) is used word-finally in a handful of mostly monosyllabic words as a phonologically empty graph. Examples include |نَہ| [nə] ‘no’, |تَہ| [t̪ə] ‘that (complementiser)’, |بِہ| [bᶦ] ‘also’ and |جَئِہ| [d͡ʑɛ] ‘victory’. In such words, |◌ﮧ| is effectively ‘silent’, due to which it is sometimes termed imperceptible ‘h’ (Shahaney, 1967 [1906], pp. 21–22). However, the rasm |◌ﮧ| may sometimes be substituted with the 𝚊𝚛-𝙰𝚛𝚊𝚋-style rasm |◌ﻪ|. As a result, the 𝚜𝚍-𝙰𝚛𝚊𝚋 words listed above may appear as |‏نَه‎ ‏تَه‎ ‏بِه‎ ‏جَئِه‎|. This phenomenon is almost entirely restricted to electronic 𝚜𝚍-𝙰𝚛𝚊𝚋 texts, and is traceable to the absence of the appropriate Unicode character for |◌ﮧ| on 𝚜𝚍-𝙰𝚛𝚊𝚋 keyboards and input method editors (IMEs).

Thus, despite attempts by linguists, lexicographers and litterateurs to spin off 𝚊𝚛-𝙰𝚛𝚊𝚋 allographs into contrastive graphs in 𝚜𝚍-𝙰𝚛𝚊𝚋, universal acceptance has proven elusive. Yet, the allographic variation in the context of |ن| and |ﻫ| in 𝚜𝚍-𝙰𝚛𝚊𝚋 is not unique. A similar phenomenon is observed by Shackle (2007, p. 655) in the context of Arabic-script Punjabi (𝚙𝚊-𝙰𝚛𝚊𝚋). Despite proposals to standardise the use of |ن|, |ﻫ| and related graphs in 𝚙𝚊-𝙰𝚛𝚊𝚋, the forms in question continue to be used somewhat indiscriminately to denote disparate phonological values. Notwithstanding the variation described, its real-world impacts remain unclear. Sociolinguistically, the varying shapes of multigraphs like |‏ﮔﻬ‎ ‏ﺟﻬ‎| might be inconsequential to fluent readers of 𝚜𝚍-𝙰𝚛𝚊𝚋. Indeed, such readers may even fail to perceive the graphetic differences described above. However, the pedagogical impacts of such variation on learners of 𝚜𝚍-𝙰𝚛𝚊𝚋 are yet to be studied rigorously. The variation in question may also affect data sorting and collation. Moreover, it raises questions on whether the rasm |ﻬ|is potentially a grapheme in 𝚜𝚍-𝙰𝚛𝚊𝚋, or is simply a nongraphemic component of a multigraph.

Related to the subject of allography is that of calligraphic practice in and the visual appearance of 𝚜𝚍-𝙰𝚛𝚊𝚋. Typographically, 𝚜𝚍-𝙰𝚛𝚊𝚋 is characterised by the conventional use of the naskh calligraphic style (Shackle, 2014a). These features visually distinguish 𝚜𝚍-𝙰𝚛𝚊𝚋 from neighbouring Arabic-script-based writing systems such as 𝚞𝚛-𝙰𝚛𝚊𝚋, 𝚙𝚊-𝙰𝚛𝚊𝚋 and Kashmiri-Arabic (𝚔𝚜-𝙰𝚛𝚊𝚋), all of which are typically written and printed in the Nastaliq calligraphic hand. Although there is no graphematic restriction on 𝚜𝚍-𝙰𝚛𝚊𝚋 being written and printed in Nastaliq, doing so may be considered sociolinguistically anomalous (§2.6). Even within the Naskh style, 𝚜𝚍-𝙰𝚛𝚊𝚋 calligraphic tradition prefers certain rasmic forms, such as | ر م| over | ر م| (Gacek, 2009, pp. 318–319; ScriptSource, 2022d; SIL International, 2022; see also Table 6.1).

6.5.4 Spelling and orthography

As described earlier, the practice of omitting subsegmental graphs (see (6)) from quotidian 𝚜𝚍-𝙰𝚛𝚊𝚋 writing has perpetuated variability in when and how they manifest, even in scholarly works. For instance, the Sindhi word [d͡ʑɛɦᵋɽo] ‘which (correlative)’ is written in 𝚜𝚍-𝙰𝚛𝚊𝚋 without subsegmental graphs as |جھڙو|. However, if subsegmental graphs are to be included, their nature and position is unclear. Mewaram’s dictionary (1910, p. 181) graphovocalises |جھڙو| [d͡ʑɛɦᵋɽo] in two ways, |جِھَڙو| and |جَھِڙو|, and lists them as equivalent alternatives. In comparison, post-Partition publications suggest the emergence of a slight divergence in graphovocalisation based on geography. In Pakistan, the SLA’s online 𝚜𝚍-𝙰𝚛𝚊𝚋 dictionary (Sindhi Language Authority, 2021c) only lists the first form cited in Mewaram, namely |جِھَڙو|. In contrast, the Indian 𝚜𝚍-𝙰𝚛𝚊𝚋 dictionaries of Lekhwani (1996, p. 62) and Rohra et al. (2011, p. 201) only list the second one, |جَھِڙو|. The disagreement on the nature and position of subsegmental graphs in the underlying graphematic form of [d͡ʑɛɦᵋɽo] may have phonological implications. If |جِھَڙو| is considered the underlying form of the word, the implied phonemic makeup would be /d͡ʑɪɦəɽo/. In contrast, |جَھِڙو| maps on to /d͡ʑəɦɪɽo/.

Approaching the issue from a phonoprimary perspective proves similarly fraught. Although some scholars assert that all Sindhi words are phonemically vowel-final (§4.3.2), there is equivocation on whether this principle applies to recent loanwords that have entered the language. Consequently, it remains undecided whether an English-origin loanword like ‘mobile’ should be transcribed in 𝚜𝚍-𝙰𝚛𝚊𝚋 as |موبائِلۡ| or |موبائِلُ|. In terms of the underlying phonemic forms indicated, the former spelling denotes consonant-final /mobaɪ̯l/, while the latter spelling denotes vowel-final /mobaɪ̯lʊ/.

The question of the nature and position of subsegmental graphs also extends to the absence of a word-medial φ-vowel and its written representation. As Trumpp observes, in colloquial Sindhi speech, the lax φ-vowel [ɪ] freely alternates with post-consonantal [Ø] (§6.4). In the absence of established spellings, writers may graphovocalise 𝚜𝚍-𝙰𝚛𝚊𝚋 words based on their own speech. Consequently, a particular lexical item spelt by one author with zer |◌ِ| may be spelt by another with jazm |◌ۡ|. Allana (1993 [1964], pp. 137–145) recommends denoting post-consonantal [Ø] only with jazm |◌ۡ| and not with zer |◌ِ|. He recommends that the latter should be restricted to denote [ɪ]. However, this practice presupposes clarity on the underlying or phonemic makeup of words, which proves elusive.

In contrast to the equivocation on lax γ-vowels and subsegmental graphs in 𝚜𝚍-𝙰𝚛𝚊𝚋 spellings, the status of γ-consonants is far clearer thanks to their graphetic segmentality and graphematic obligatoriness. As seen in previous Sections, the graph inventories of 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚊𝚍, 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝟷𝟾𝟻𝟹 and 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚞𝚖𝚙𝚙 have retained the majority of γ-consonant graphs from the 𝚊𝚛-𝙰𝚛𝚊𝚋 inventory, even when the phonological values of these graphs have changed or no longer contrast in Sindhi. In addition to graph retention, 𝚜𝚍-𝙰𝚛𝚊𝚋 also exhibits graphematic retention in retaining the source spellings of loanwords from the homoscriptal 𝚊𝚛-𝙰𝚛𝚊𝚋. This aligns with similar practice in 𝚏𝚊-𝙰𝚛𝚊𝚋 (§6.1). Share and Daniels (2016) refer to the retention of source spellings, even when phonologically unwarranted, as orthographic inertia. Pending agreement on the exact definition and scope of the term orthography, the alternative term graphematic inertia could be used. Illustrating the phenomenon of graphematic inertia is the 𝚊𝚛-𝙰𝚛𝚊𝚋 word |تعليم| ‘education’. In standard Arabic, this word is pronounced [tɑʕliːm]. This word has been borrowed into Sindhi, with its pronunciation indigenised to [t̪alimᵊ] or [t̪ɛlimᵊ]. However, its spelling in 𝚜𝚍-𝙰𝚛𝚊𝚋 has remained unchanged from the source 𝚊𝚛-𝙰𝚛𝚊𝚋 spelling. In general, Arabic and Persian loanwords that have entered Sindhi are assimilated phonologically but not graphematically. At most, there may be minor graphematic adaptations to align with Sindhi morphophonology. Such graphematic inertia sometimes serves a disambiguating purpose, as seen in Table 6.6.

Graphematic inertia also means that the spellings of such loanwords in 𝚜𝚍-𝙰𝚛𝚊𝚋 take on a logographic dimension and may need to be memorised. This includes spellings that retain elements such as 𝚊𝚛-𝙰𝚛𝚊𝚋 |ال| [aɫ] ‘the’ and 𝚏𝚊-𝙰𝚛𝚊𝚋 |آباد| [ɒbɒd] ‘city, town’. Examples include 𝚜𝚍-𝙰𝚛𝚊𝚋 |بِالِڪُلِ| [bɪlᶦkʊlᶦ] ‘totally’ and |حَيدَرآبادُ| [ɦɛd̪əɾaba̪d̪ᶷ] ‘Hyderabad’. In these spellings, the occurrence of the strings |ال| and |آباد| are not entirely predictable based on their corresponding pronunciations, and must be explicitly learnt.

6.5.5 Logograms and collation order

Aside from logography arising from graphematic inertia, the most prominent logograms in present-day 𝚜𝚍-𝙰𝚛𝚊𝚋 are |۽| [ɛ̃ ~ əĩ] ‘and’, |۾| [mẽ] ‘in’ and |اللّٰه| ‘Allah’. The last logogram is also retained when part of a personal name, as in 𝚜𝚍-𝙰𝚛𝚊𝚋 |عبداللّٰه| [əbd̪ʊl(l)aɦᶷ] ‘Abdullah’. Other benedictory logograms related to the Islamic faith and borrowed from 𝚊𝚛-𝙰𝚛𝚊𝚋 may also appear in 𝚜𝚍-𝙰𝚛𝚊𝚋 texts, such as the basmala |﷽| ‘In the name of Allah, the compassionate, the merciful’. Notably, when fully graphovocalised within a 𝚜𝚍-𝙰𝚛𝚊𝚋 text, both |عَبۡدُاللّٰهُ| [əbd̪ʊl(l)aɦᶷ] and the invocation |بِاسۡمِ‌اللّٰهِ| [bɪsmɪl(l)aɦᶦ] ‘In the name of Allah’ would feature the graphosegmental logogram |اللّٰه| together with the graphosubsegmental phonograms |◌ُ ◌ِ| [ʊ ɪ], respectively. As noted in Section 6.4.2, such co-occurrence — despite not being unique to 𝚜𝚍-𝙰𝚛𝚊𝚋 — is yet to be rigorously described and theorised.

On numerals, certain sources claim that present-day 𝚜𝚍-𝙰𝚛𝚊𝚋 comprises a subset of numeral graphs distinct from those in 𝚊𝚛-𝙰𝚛𝚊𝚋 (ScriptSource, 2022c; SIL International, 2022). However, this is far from clear-cut, and modern 𝚜𝚍-𝙰𝚛𝚊𝚋 publications may feature any combination of the Arabic-script numeral superset listed in (7). The variation in numeral graphs is comparable to that seen with the rasms |ﻬ ﮭ| in that a discernible pattern proves elusive. Salient punctuation graphs in the present-day 𝚜𝚍-𝙰𝚛𝚊𝚋 inventory include |. ⹁ ⁏ ⸮| as full stop, comma, semicolon and question mark, respectively (Priest Evans, 2020).

After a century and a half of variation in the collation order of 𝚜𝚍-𝙰𝚛𝚊𝚋, strides have been made towards standardisation in the recent past, especially in Pakistan. Baloch (2007, p. 10) uses the following sequence in the Jame Sindhi Lughat, which has also been adopted by the SLA in its 𝚜𝚍-𝙰𝚛𝚊𝚋 tutorials (Sindhi Language Authority, 2015b):

Compared to the collation orders followed by Mewaram and Grierson (§6.3.2), Baloch’s order differs mainly in the position of |ڦ| [pʰ], which now occurs after the graphetically similar |ف| [f] rather than after the phonologically similar |پ| [p]. Notably, it lists hamza |ء| as a distinct graph while subsuming |آ| under |ا|. In India, on the other hand, scholarly works continue to demonstrate slight variation in the 𝚜𝚍-𝙰𝚛𝚊𝚋 collation order. Below are a few collation orders used in works by prominent Indian Sindhi authors:

Hardwani (1991, p. x):

Lekhwani (1997, p. xii):

Varyani & Thakwani (2003, p. x):

Khubchandani (2007, p. 697):

Although the sort order by itself is typically not considered graphematically significant, idiosyncratic variation in this regard can have consequences for lexicographic works. In an increasingly digital world, it also has implications for computer-aided sorting.

6.6 Analysis

6.6.1 Graphematic typology

This chapter has shown that most 𝚜𝚍-𝙰𝚛𝚊𝚋 variants, but especially 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝟷𝟾𝟻𝟹 and 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚞𝚖𝚙𝚙, provide for almost all φ-vowels in the Sindhi language — including the absence of a φ-vowel — to be explicitly indicated in writing. In this regard, the various 𝚜𝚍-𝙰𝚛𝚊𝚋 systems appear to have much in common with the various 𝚜𝚍-𝙳𝚎𝚟𝚊 systems that will be encountered in Chapter 7. Against this background, what should 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝟷𝟾𝟻𝟹 and 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚞𝚖𝚙𝚙 be typologically categorised as? Is the label of ‘abjad’, traditionally applied to 𝚊𝚛-𝙰𝚛𝚊𝚋, appropriate for 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝟷𝟾𝟻𝟹 and 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚞𝚖𝚙𝚙? If not, is there is a case for revisiting the label of abjad for 𝚊𝚛-𝙰𝚛𝚊𝚋, too?

The label ‘abjad’ insinuates that the writing system in question comprises phonograms that solely or primarily denote φ-consonants, and leaves φ-vowels entirely or largely unmarked. However, as Table 6.2 reveals, the inventories of 𝚊𝚛-𝙰𝚛𝚊𝚋 and 𝚏𝚊-𝙰𝚛𝚊𝚋 comprise a comprehensive set of phonograms to explicitly indicate all φ-vowels in their target languages. While graphs for φ-vowels in 𝚊𝚛-𝙰𝚛𝚊𝚋 and 𝚏𝚊-𝙰𝚛𝚊𝚋 are usually omitted by choice, such omission is based on sociolinguistic discretion and not on graphematic limitations. The problematic nature of classifying 𝚊𝚛-𝙰𝚛𝚊𝚋, 𝚏𝚊-𝙰𝚛𝚊𝚋 and their daughter systems as abjads is neatly highlighted by Sproat’s (2010a, p. 53) deft example of |f u cn rd ths u cn b trnd as a scrtry nd gt a gd jb| (§5.1.1). Sproat’s example, transcribed in unstandardised 𝚎𝚗-𝙻𝚊𝚝𝚗, omits most non-initial γ-vowels when compared to its ‘fully-vowelled’ standardised version. Yet, most readers would hesitate to classify 𝚎𝚗-𝙻𝚊𝚝𝚗 as an abjad based on this example. At the same time, the very same readers might well accept the classification of 𝚊𝚛-𝙰𝚛𝚊𝚋, 𝚏𝚊-𝙰𝚛𝚊𝚋 and 𝚜𝚍-𝙰𝚛𝚊𝚋 variants as abjads based on similar reasoning. The classification of several Arabic-script-based writing systems as abjads affirms the observation in Section 2.10 that writing systems are often assigned a particular graphematic label based on their sociolinguistic characteristics.

If 𝚊𝚛-𝙰𝚛𝚊𝚋, 𝚏𝚊-𝙰𝚛𝚊𝚋 and the major 𝚜𝚍-𝙰𝚛𝚊𝚋 variants described in this chapter are not abjads, then what graphematic type should they be categorised under? Since they are capable of denoting most or all φ-vowels in their target language, they should be considered vowelled segmentaries (Figure 2.2). The specific subtype of vowelled segmentary can be determined by addressing the questions underlying the classification in Figure 2.3, namely:

• Which postconsonantal φ-vowels can be graphematically represented?
• How are postconsonantal φ-vowels graphematically represented?

In 𝚊𝚛-𝙰𝚛𝚊𝚋, 𝚏𝚊-𝙰𝚛𝚊𝚋, 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝟷𝟾𝟻𝟹 and 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚞𝚖𝚙𝚙, all φ-vowels in the target language can be reasonably represented in an overt manner, making these systems plenaries. Furthermore, in each of these systems, postconsonantal φ-vowels are represented in writing using bound allographs, making the systems alphasyllabaries. The latter finding aligns with Bright’s (1999, p. 50 footnote 2) classification of “fully pointed” or graphovocalised 𝚊𝚛-𝙰𝚛𝚊𝚋 and he-Hebr typologically alphasyllabic. Thus, in terms of graphematic type, 𝚊𝚛-𝙰𝚛𝚊𝚋, 𝚏𝚊-𝙰𝚛𝚊𝚋 and their derivatives 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝟷𝟾𝟻𝟹 and 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚞𝚖𝚙𝚙, are all plenar alphasyllabaries.

At this juncture, it is worth noting that the distinction between an alphabet and an alphasyllabary in Figure 2.3 is based on γ-vowels in postconsonantal position showing up as bound allographs as opposed to their free forms. Here, the complementary distribution of the two allograph sets is graphematically determined, rather than graphetically. As shown for 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝟷𝟾𝟻𝟹 in Table 6.5, the ‘free’ γ-vowel allographs are those formed by combining the ‘bound’ γ-vowel allographs |‏◌َ‎ ‏◌ا‎ ‏◌ِ‎ ‏◌ِي‎ ‏◌ُ‎ ‏◌ُو‎ ‏◌ي‎ ‏◌َي‎ ‏◌و‎ ‏◌َو‎| [ə a ɪ iː ʊ u e ɛː o ɔː] with a vowel holder element as the graphetic base. Put differently, the classification of a γ-vowel as free or bound depends on whether the graphetic base is a phonologically ‘full’ one such as |ب| [b] or the phonologically empty alif or hamza. Hence, despite 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝟷𝟾𝟻𝟹’s free and bound γ-vowel allographs overlapping in their graphetic composition, their occurrence in distinct graphematically-conditioned patterns justifies the label ‘alphasyllabary’. Moreover, a similar pattern — albeit now obsolete — has been attested in 𝚜𝚍-𝙳𝚎𝚟𝚊 (§7.7.1, footnote 107), wherein a set of invariant graphetic elements that denote φ-vowels are combined with varying graphetic bases to form free and bound vowel allographs in complementary distribution. The fact that 𝚜𝚍-𝙳𝚎𝚟𝚊, too, is an alphasyllabary (§7.7.1) further vindicates classifying 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝟷𝟾𝟻𝟹 and 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚞𝚖𝚙𝚙 as alphasyllabaries.

6.6.2 Graph inventory

In Section 2.2, a graph was defined as the basic unit of the written modality of language. According to this definition, every glottographic element counts as a graph, be it rasmic or graphetically complex in nature. Hence, ‘graph’ proves to be a useful catch-all term for a particular glottographic element, especially when its exact graphematic status is still uncertain.

Table 6.1 lists most of the rasms underlying the various 𝚜𝚍-𝙰𝚛𝚊𝚋 inventories. Most of them, such as |ٮ|, are underspecified and require further graphetic augmentation before potentially qualifying as phonograms in one or more of the 𝚜𝚍-𝙰𝚛𝚊𝚋 inventories. Thus, the rasm |ٮ| needs to be augmented with dots or nuqtas — as in |ب پ ت ث ٽ ٺ| — in order to become identifiable with a particular linguistic value and approach phonogram status.

Yet, determining which rasms and augmented graphs in the various 𝚜𝚍-𝙰𝚛𝚊𝚋 inventories qualify as graphemes in those inventories proves tricky. Per Meletis’ (2020, pp. 94–96) criteria outlined in Section 2.5, a graph needs to be minimal, contrastive and have linguistic value in order to qualify as a grapheme in a particular writing system. The rasm |ٮ| is not a grapheme as it is devoid of linguistic value. However, do the various nuqta patterns added to this rasm qualify as graphemic? That is, are |◌﮳ ◌﮶ ◌﮹ ◌﮽| graphemic in 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝟷𝟾𝟻𝟹 or 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚞𝚖𝚙𝚙? Arguing in the affirmative proves to be an uphill task, since these nuqta patterns are not associated with identifiable linguistic values of any kind. In fact, Meletis (2020, p. 294) himself suggests that nuqtas in Arabic-script inventories are not graphemes per se, but simply “distinguishing features” that serve to “heterogenize” rasms. Thus, despite nuqtas in the 𝚜𝚍-𝙰𝚛𝚊𝚋 inventories — and Arabic-script inventories in general — creating contrast, evidence is thin for them qualifying as graphemes in themselves. In acting as nongraphemic distinguishing features, nuqta patterns share certain characteristics with phonological features such as aspiration.

Subsegmental graphs in 𝚜𝚍-𝙰𝚛𝚊𝚋 such as |‏◌َ‎ ‏◌ِ‎ ‏◌ُ‎| [ə ɪ ʊ] can safely be considered graphemes, as they neatly satisfy the three graphemic criteria. Their graphosubsegmentality does not prejudice their grapheme status (Meletis, 2020, p. 96 footnote 102). The sukun |◌ْ| and jazm |◌ۡ| also make the cut as potential graphemes, thanks to their linguistic value of [ᶦ ~ Ø]. The jury is still out on tashdid |⁠◌ّ|, as its eligibility for grapheme status will depend on whether φ-consonant gemination is considered a ‘linguistic value’.

Curiously, it is segmental graphs whose grapheme status proves the most vexing. Prima facie, a graph like alif |ا| may be considered rasmic and linguistically underspecified, becoming fully specified only when combined with a suitable γ-vowel. At the same time, |◌ا| is used as the bound γ-vowel for [a], as in |با| [ba]. Moreover, in modern 𝚜𝚍-𝙰𝚛𝚊𝚋, certain authors may denote the Sindhi phonological sequence [ba] with the 𝚊𝚛-𝙰𝚛𝚊𝚋-style graphematic sequence |بَا| (Table 6.2). Allana (1993 [1964], p. 113 footnote) explicitly recommends against this practice in modern 𝚜𝚍-𝙰𝚛𝚊𝚋, but it remains unclear what the specific rule — if any — is on representing the φ-vowel [a] in postconsonantal position. As evident, the verdict on |◌َا| as a legitimate graphematic sequence in 𝚜𝚍-𝙰𝚛𝚊𝚋 to denote [a] will affect the grapheme status of alif |ا| per se. If one considers the fully specified bound γ-vowel form of [a] to be |◌َا|, bare alif |ا| becomes linguistically underspecified and, hence, nongraphemic. But if one follows Allana’s recommendation and writes the bound γ-vowel for [a] as bare alif |◌ا| without |◌َ|, it would make bare alif a grapheme. This conundrum highlights the crucial definitional role played by orthography in determining graphemes within a particular system.

A similar dilemma is encountered when using rasms to form γ-consonants. At first glance, one may assume that a graph like |د|, with linguistic value [d̪], satisfies the three criteria of minimality, contrast and linguistic value and, accordingly, attains grapheme status. However, a closer investigation throws a spanner into the works. Given the presence of sukun or jazm in the inventories of 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝟷𝟾𝟻𝟹 and 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚞𝚖𝚙𝚙, one may argue that, strictly speaking, the linguistic value [d̪] is actually denoted by a combination of |د| and jazm — namely |دۡ|. Adopting this stance would make |د| a linguistically underspecified form that becomes fully specified only when combined with a suitable bound γ-vowel or jazm. However, if an orthographic rule were to be introduced that eliminates jazm from the 𝚜𝚍-𝙰𝚛𝚊𝚋 inventory with no replacement, bare |د| would ostensibly acquire the linguistic vale of [d̪] and, hence, graduate to grapheme status. This situation reiterates how the status of a graph as a grapheme is conditioned by the specific orthographic module in play within the relevant writing system. Overall, these findings reaffirm the existence of graphemes at the level of writing system, and not at the level of script.

A comparably intricate issue is the status of the graphetic element |ﻬ| and its alternative form |ﻫ|, which in theory denotes aspiration [ʱ] in [ɡʱ d͡ʑʱ]. Building on Figure 6.6, If one argues that the authentic written forms of [ɡʱ d͡ʑʱ] are |‏ﮔﻬ‎ ‏ﺟﻬ‎|, and that |‏گھ‎ ‏جھ‎| are incorrect, then |ﻬ| emerges as a grapheme distinct from |ﻫ|. However, and as described in Section 6.5, vacillation between these graph sets is widespread. On this basis, if the two sets are simply considered free variants of each other, the elements |ﻬ| and |ﻫ| would emerge as allographs of each other in more-or-less free variation. Constraining and regulating the alternation would require appropriate orthographic specification, thereby reiterating the salience of a system’s orthographic module in determining its graphemes.

What about the status of |ﻬ| and |ﻫ| in 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚞𝚖𝚙𝚙? Since 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚞𝚖𝚙𝚙 is based on the 𝚞𝚛-𝙰𝚛𝚊𝚋 graph inventory, it deviates from 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝟷𝟾𝟻𝟹 in consistently indicating phonological aspiration [ʰ ʱ] by adding a graphetic element to the unaspirated form of γ-consonants. However, the graphetic element may appear either as |ﻬ| or as |ﻫ| (Figure 6.3), albeit in complementary distribution conditioned by the graphetic properties of the previous and following graphs. Hence, |ﻬ| and |ﻫ| do not seem to be graphemes in 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚞𝚖𝚙𝚙 either, although one may argue that they are allographs of the same grapheme.

The analysis in this subsection shows that the orthography plays a crucial role in determining not only the specific subset of graphs that constitutes the inventory of a particular writing system, but also which graphs within the inventory are deemed to be graphemes. At any rate, the graph inventory and orthography are inextricably linked, with changes in one impacting the other. By extension, a graph potentially acquires grapheme status only in the context of a particular writing system and, often, in the presence of an identifiable orthography. Graphemes do not exist in a vacuum, or even at the script level. Consequently, in order to reliably bestow grapheme status to elements in the various 𝚜𝚍-𝙰𝚛𝚊𝚋 variants described in this Chapter, one needs to first address related questions on the structure of a writing system and its constituent modules, particularly orthography. As it turns out, the answers to these questions will require further scholarly debate and consensus. Thus, reliably identifying graphemes in the various 𝚜𝚍-𝙰𝚛𝚊𝚋 systems described here will have to remain a topic for future research.

6.6.3 Graphematics and orthography

So far, 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚊𝚍, 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝟷𝟾𝟻𝟹 and 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚞𝚖𝚙𝚙 have simply been referred to by the generic label ‘(sub)variant’. In the context of the Modular Theory (MT), are each of the above variants distinct graphematic systems? Or are they better characterised as distinct orthographies? Furthermore, and to the extent the above variants qualify as orthographies, which of them are systematic and which conventional?

Unlike orthographies imposed in the context of European writing systems, graphematic or orthographic specifications in the context of 𝚜𝚍-𝙰𝚛𝚊𝚋 — especially 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝟷𝟾𝟻𝟹 — was gradual and organic, taking place over decades. Moreover, in the twenty-first century, the contemporary avatar of 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝟷𝟾𝟻𝟹 exhibits elements of a top-down orthography — as in the spellings of certain words — that coexist with variable elements such as the graphetic composition of γ-vowels. Although such idiosyncratic variation within the graphematic solution space has been described in Section 6.5, it is yet to be rigorously incorporated into and accounted for by suitable descriptive-analytical frameworks.

A latent theme emerging from the description in Section 6.5 is that of subtly differing graphematic practices between Pakistani and Indian 𝚜𝚍-𝙰𝚛𝚊𝚋 use. This observation prompts one to ask whether they should be considered distinct 𝚜𝚍-𝙰𝚛𝚊𝚋 subvariants or orthographies. In other words, is there an argument to be made for the existence of 𝚜𝚍-𝙰𝚛𝚊𝚋-𝙿𝙺 and 𝚜𝚍-𝙰𝚛𝚊𝚋-𝙸𝙽 based on mutual graphematic differences? Although initial evidence may point in that direction, additional evidence is desirable. For instance, there exist geographic tendencies in whether the bound γ-vowels for [i u] appear as |‏◌ِي‎ ‏◌ُو‎| or |‏◌يِ‎ ‏◌وُ‎|, but they remain tendencies and are far from being firmly established. Indeed, writing practices characteristic of one region are often found in the other, sometimes in works by the same author or even in the same text. In any event, identifying and grouping distinct patterns in writing practice requires guidelines on how different the patterns need to be from each other. Even if sufficient evidence emerges for the existence of distinct 𝚜𝚍-𝙰𝚛𝚊𝚋-𝙿𝙺 and 𝚜𝚍-𝙰𝚛𝚊𝚋-𝙸𝙽 subvariants, the question remains on what they should be termed as — graphematic systems, orthographies, or something else. These questions reinforce the inextricable intertwining of the graphematic and orthographic modules with the graph inventory, including in matters of graphetic composition as illustrated by |‏◌ِي‎ ‏◌ُو‎| and |‏◌يِ‎ ‏◌وُ‎|.

6.6.4 Logography

In contrasting phonography with logography, Section 2.8 cited the examples of 𝚎𝚗-𝙻𝚊𝚝𝚗 ⟨deer⟩ and ⟨dear⟩ (Gnanadesikan, 2017a, p. 15), wherein two homophonous lexical items are distinguished by distinct spellings. Both spellings are licensed by 𝚎𝚗-𝙻𝚊𝚝𝚗’s graphematic solution space, but remain somewhat arbitrary selections. With regard to the Japanese writing system, particularly the Kanji inventory, Honda (2021, p. 628) notes that the correspondence between individual Kanji characters and phonological elements in the Japanese language “relies entirely on conventional orthography”. In other words, there is a strong logographic element in the selection of Kanji to represent certain Japanese-language morpholexical items. Honda’s observation on the somewhat arbitrary graph-phone correspondences seen in logographic writing is consistent with that of Sproat and Gutkin (2021, pp. 5–7), who note that logography is characterised by the notion of different words being spelt differently, even if phonologically similar.

An impressionistic appraisal of the logographic element in 𝚜𝚍-𝙰𝚛𝚊𝚋, particularly 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝟷𝟾𝟻𝟹 in its post-Partition stage of evolution, places it in a league similar to 𝚎𝚗-𝙻𝚊𝚝𝚗. As seen in Table 6.6, the use of distinct elements and sequences in the written representations of words is helpful in lexico-semantic disambiguation, particularly when γ-vowels may be omitted. However, the instances of 𝚜𝚍-𝙰𝚛𝚊𝚋 logography seen in Table 6.6 are distinct from those in, say, Japanese-Kanji, in that they result from organic historical phonological mergers and accompanying graphematic inertia than from planned top-down orthographic decisions. As elaborated in the paragraphs preceding Table 6.6, 𝚜𝚍-𝙰𝚛𝚊𝚋 features numerous instances of graphematic retention from 𝚊𝚛-𝙰𝚛𝚊𝚋, of which several pairs of graphematically distinct lexical items happen to be realised homophonously in spoken Sindhi. Comparing and contrasting different types of logography across writing systems is a fruitful area of research, for which Sproat and Gutkin (2021) propose an innovative method of measurement.

6.6.5 Sociolinguistics

Chapter Six has shed light on how, over the centuries, individual 𝚜𝚍-𝙰𝚛𝚊𝚋 variants simultaneously possessed areas of graphematic clarity and ambiguity. The findings in this chapter also corroborate the MT’s view on the orthographic module being a theoretically optional component of a writing system, and reiterate the cosmetic nature of standardised collation orders and related finishing touches. Yet, despite orthography and standardisation being understood as optional in scholarly circles, they assume importance for lay users. This is particularly true under contemporary sociolinguistic mores, under which literacy is prized and written languages are accorded greater cultural value than oral ones (Coulmas, 2014, p. 1; Fishman, 1997, p. 154). Moreover, as Bunčić (2016a, p. 16) states, “the invention of the printing press made people think that every language had (and had to have) a uniform orthography”. Jaffe echoes this sentiment in observing that:

it is not only important [from the layperson’s perspective] to have an orthography, but it is also critical for that orthography to have prescriptive power – to be standardised and authoritative […]

(Jaffe, 2000, p. 506)

Consequently, in a literate society, standardisation or lack thereof in matters of orthography, graph inventory and collation order add to or detract from the popular image of a language. Subtle variation in writing practices may not impact proficient readers of 𝚜𝚍-𝙰𝚛𝚊𝚋. However, such variation does carry ramifications for 𝚜𝚍-𝙰𝚛𝚊𝚋 pedagogical material, including dictionaries and primers. In this regard, the lack of orthographic prescriptivism may negatively impact learners of 𝚜𝚍-𝙰𝚛𝚊𝚋. Ambiguity over spelling-related minutiae may seem insignificant in isolation, but may cumulatively result in learner frustration and consequent loss of motivation. Moreover, in a modern world where orthographic consistency is highly regarded and variability disparaged, a writing system devoid of a prescriptive orthographic module may be looked down upon and deemed unrefined. Subjective impressions of defectiveness and crudeness may also be weaponised, as evinced by the arguments against 𝚜𝚍-𝙰𝚛𝚊𝚋 in post-Partition India (§5.3). The real-world sociolinguistic significance of an orthographic module, therefore, exists in a tension with its supposed graphematic dispensability.

☙ ❧

7 Devanagari

The twentieth century saw the rise of 𝚜𝚍-𝙰𝚛𝚊𝚋 as the preeminent writing system for Sindhi, whose position was further cemented following Partition (§5.3). At the same time, Partition also resulted in the rise of a writing system that remains the only one to have noticeably challenged 𝚜𝚍-𝙰𝚛𝚊𝚋’s sociolinguistic status in recent times. Recognised by the Indian government as co-official with 𝚜𝚍-𝙰𝚛𝚊𝚋, Sindhi-Devanagari (𝚜𝚍-𝙳𝚎𝚟𝚊) has, since Partition, seen a remarkable rise in interest among scholars and laypersons alike. The writing system has also been at the core of a vigorous and often acrimonious script debate in post-Partition India, with some of its supporters seeking to establish it as the sole official writing system of Sindhi in India. As a result, 𝚜𝚍-𝙳𝚎𝚟𝚊 has been the topic of much academic and popular debate, resulting in the spread of fact as well as fiction on the writing system’s graphematics and historical sociolinguistics.

7.1 Graphematic foundations

Despite being graphetically unstandardised and graphematically variable, early 𝚜𝚍-𝙳𝚎𝚟𝚊 writing was broadly based on a Sanskrit-Devanagari (𝚜𝚊-𝙳𝚎𝚟𝚊) canonical template. Consequently, all 𝚜𝚍-𝙳𝚎𝚟𝚊 variants, whether historical or contemporary, align with 𝚜𝚊-𝙳𝚎𝚟𝚊’s graphematic typology in being abugidic and alphasyllabic in nature. In the post-Partition era, writing practices from Hindi-Devanagari (𝚑𝚒-𝙳𝚎𝚟𝚊) have also significantly influenced those in 𝚜𝚍-𝙰𝚛𝚊𝚋. Table 7.1 contains a selection of 𝚜𝚍-𝙳𝚎𝚟𝚊 and 𝚑𝚒-𝙳𝚎𝚟𝚊 phonograms that form the graphematic foundation of 𝚜𝚍-𝙳𝚎𝚟𝚊. Like Table 6.2, Table 7.1 is also designed as a matrix, with γ-consonants shown along the vertical axis and γ-vowels along the horizontal axis. The intersection of a row and column contains the graph representing the corresponding phonological [CV] sequence.

Table 7.1 reveals several graphe(ma)tic similarities between 𝚜𝚊-𝙳𝚎𝚟𝚊 and 𝚑𝚒-𝙳𝚎𝚟𝚊 on the one hand, and 𝚊𝚛-𝙰𝚛𝚊𝚋 and 𝚏𝚊-𝙰𝚛𝚊𝚋 on the other (see Table 6.2). In all these systems, γ-consonants occupy their own segmental space and can occur on their own. In contrast, the γ-vowels | ा ⁠ि ⁠ी ⁠ु ⁠ू ⁠े ⁠ै ⁠ो ⁠ौ| and the graph |⁠्| [Ø] can only occur together with a γ-consonant.⁹⁸ Hence, the dichotomous description of 𝚊𝚛-𝙰𝚛𝚊𝚋 and 𝚏𝚊-𝙰𝚛𝚊𝚋 γ-consonants and γ-vowels as “basic letters” and “diacritics”, respectively (Bauer, 1996) has also been applied to 𝚜𝚊-𝙳𝚎𝚟𝚊 and 𝚑𝚒-𝙳𝚎𝚟𝚊 γ-consonants and γ-vowels (Bright, 1996, p. 376).

However, the two sets of writing systems differ slightly when it comes to denoting φ-[CV] sequences. In most Devanagari-based writing systems, there exists one φ-vowel which, when occurring after post-consonantally, is represented implicitly in writing, without any overt graphetic element. This φ-[V] is commonly known as the ‘default’ or ‘inherent’ vowel, denoted here by φ-[V₀]. The phonological quality of [V₀] varies with language. In 𝚜𝚊-𝙳𝚎𝚟𝚊, φ-[V₀] is [a], whereas in 𝚑𝚒-𝙳𝚎𝚟𝚊 and 𝚜𝚍-𝙳𝚎𝚟𝚊, it is [ə] (§4.3.2). Thus, 𝚜𝚊-𝙳𝚎𝚟𝚊 [ba] and 𝚑𝚒-𝙳𝚎𝚟𝚊 and 𝚜𝚍-𝙳𝚎𝚟𝚊 [bə] are denoted simply by |⁠ब|, where the presence of postconsonantal φ-[V₀] is implied by the absence of a corresponding graph. As a result, graphs denoting φ-[CV₀] are considered graphematically primary or ‘basic’ (Bright, 1996, p. 376). A φ-[CV] sequence is denoted by combining the graph for φ-[CV₀] with the appropriate bound γ-vowel. Thus, |बी| [bi(ː)] is obtained by concatenating |ब| [ba ~ bə] and |⁠ी| [i(ː)]. A pure φ-[C] with no subsequent φ-vowel is written with the corresponding graph for φ-[CV₀] combined with |⁠्|. While the graph |⁠्| may be known by different names depending on the language, it is increasingly referred to by the generic term virama in the context of Indic scripts.⁹⁹ In acting as a ‘vowel killer’, the virama is the Indic counterpart of the Arabic sukun. Alternatively, in some instances, a pure φ-[C] may be written with a graphetically ‘truncated’ allograph. Thus, [b] may appear as |ब्| or |ब्‍◌|, conditioned by graphetic factors, the graphematic environment and orthographic guidelines. Additionally, the writing system — or, more precisely, its orthography — may, in certain environments, permit the use of a graph denoting φ-[CV₀] to also denote φ-[C]. For instance, 𝚑𝚒-𝙳𝚎𝚟𝚊 allows the use of |ब| [bə] to denote [b] in specific word positions, especially finally. The motivation behind this graphematic practice is the phonological rule of schwa deletion (see Table 4.5). However, the practice also adds a layer of opacity to 𝚑𝚒-𝙳𝚎𝚟𝚊’s graph-phone correspondences compared to the biunique mappings of 𝚜𝚊-𝙳𝚎𝚟𝚊.

Outside of a φ-[CV] sequence, a φ-[V] by itself, which includes φ-[V₀], is written with a distinct set of γ-vowel allographs, |अ आ इ ई उ ऊ ए ऐ ओ औ|. These allographs can occur on their own and occupy their own graphosegmental space. For nomenclatural convenience, graphs denoting φ-[CV₀], such as |ब| [ba ~ bə], will be referred to as free forms of γ-consonants. Similarly, |अ आ इ ई उ ऊ ए ऐ ओ औ| will be termed the free allographs of γ-vowels. Graphetically truncated or ‘half’ forms such as |ब्‍◌| [b], as well as the ‘diacritics’ |⁠ा ⁠ि ⁠ी ⁠ु ⁠ू ⁠े ⁠ै ⁠ो ⁠ौ| will be called bound γ-consonants and γ-vowels, respectively. As a bound graph, the virama |⁠्| [Ø] bears similarities to the bound γ-vowels. To sum up, in 𝚜𝚊-𝙳𝚎𝚟𝚊 and 𝚑𝚒-𝙳𝚎𝚟𝚊, free graphs such as |ब| or |ई| can occur on their own, and are comparable to the Isolated forms of 𝚊𝚛-𝙰𝚛𝚊𝚋 and 𝚏𝚊-𝙰𝚛𝚊𝚋. Bound graphs such as |ब्‍◌| or |⁠ी| can only occur together with a free graph as the graphe(ma)tic base, and may be subject to additional orthographic restrictions. Generally, multiple bound graphs cannot co-occur on the same base. Exceptions include the chandrabindu or anunasika |⁠ँ|, the anusvara |⁠ं| and the visarga |⁠ः|.¹⁰⁰

The chandrabindu is rare in 𝚜𝚊-𝙳𝚎𝚟𝚊 but widely used in 𝚑𝚒-𝙳𝚎𝚟𝚊. In the latter system, the graph is used to denote the nasalisation suprasegmental on a φ-vowel. Graphematically, the chandrabindu is written above a free or bound graph. However, if competing for segmental space with another supralinear graphetic element, the chandrabindu may be visually truncated and appear identically to anusvara |⁠ं|, as shown in (10):

For consistency and ease, some 𝚑𝚒-𝙳𝚎𝚟𝚊 writers may denote φ-vowel nasalisation in all instances by the anusvara-style |⁠ं|. Such use may clash with the use of anusvara |⁠ं| to denote other phonological values. Used only occasionally in 𝚜𝚊-𝙳𝚎𝚟𝚊 but extensively in 𝚑𝚒-𝙳𝚎𝚟𝚊, the anusvara indicates the nasal segment in a homorganic nasal-oral φ-consonant cluster. Coupled with its use as an allograph of chandrabindu, the phonological values of anusvara in 𝚑𝚒-𝙳𝚎𝚟𝚊 encompass [ŋ (ɲ) ɳ n n̪ m ◌ ̃], with its precise value determined from written context. At the same time, the nasal φ-consonants [ŋ (ɲ) ɳ n n̪ m] may also be written with the allographs |ङ् ञ्‍◌ ण्‍◌ न्‍◌ म्‍◌| (see Table 7.1). The sequences listed in (11), all of which adhere to 𝚑𝚒-𝙳𝚎𝚟𝚊 graphematic principles, illustrate how the anusvara may be homophonous with |ङ् ञ्‍◌ ण्‍◌ न्‍◌ म्‍◌| depending on written environment:

The examples in (11) show that the presence of anusvara allows for some homophonous heterography when representing nasal-oral φ-consonant clusters in 𝚑𝚒-𝙳𝚎𝚟𝚊. This property has led scholars to describe the anusvara as a graphematic abbreviation of sorts for a nasal γ-consonant. Such descriptions have ranged from the terse “shorthand notation” (Salomon, 2007, p. 84) to the poetic “compendium scripturæ” (Grierson, 1881, p. 2). However, individual writing systems may prefer one spelling over the other. For instance, spellings with overt nasal γ-consonants are more frequent in 𝚜𝚊-𝙳𝚎𝚟𝚊. In 𝚑𝚒-𝙳𝚎𝚟𝚊, [ɳ n m] may appear either as |ण्‍◌ न्‍◌ म्‍◌| or as |⁠ं|. However, [ŋ ɲ] only rarely appear as |ङ् ञ्‍◌|, and are almost always written using with |⁠ं|. This graphematic distribution is likely due to [ɳ n m] being phonemic in Hindi, with [ŋ ɲ] only surfacing as allophones of [n]. Ultimately, though, the final choice of spelling is determined by various factors, including word etymology and personal preference. Also playing a part is the sociolinguistic prestige of Sanskrit in South Asia, which may induce some 𝚑𝚒-𝙳𝚎𝚟𝚊 writers to prefer the 𝚜𝚊-𝙳𝚎𝚟𝚊-style spellings (Salomon, 2007, p. 84).

The final bound graph in this class, the visarga |⁠ः|, is frequently found in 𝚜𝚊-𝙳𝚎𝚟𝚊, where it denotes syllable-final [ɦ]. It is rare in 𝚑𝚒-𝙳𝚎𝚟𝚊, and found only in Sanskrit borrowings in formal registers.

The inventory of γ-consonants also includes several graphs that are commonly termed ‘ligatures’ despite being rasmically distinct. As outlined in Section 2.7, a graph based on a distinct rasm but representing a φ-consonant cluster is often termed a ligature. Thus, the motivation behind classifying such graphs as ligatures is not their written manifestation but their spoken values. In 𝚑𝚒-𝙳𝚎𝚟𝚊, commonly occurring graphs in this category include |क्ष त्र ज्ञ श्र क्त| [kɕ(ə) t̪ɾ(ə) ɡj(ə) ɕɾ(ə) kt̪(ə)]. Of these, |क्ष त्र ज्ञ| are commonly listed in 𝚜𝚊-𝙳𝚎𝚟𝚊 and 𝚑𝚒-𝙳𝚎𝚟𝚊 symbol charts, but not |श्र क्त| or others.

In terms of typographical preference, certain rasms or graphs may be preferred over others as allographs. Although phonologically inconsequential, the choice of allographs may be semiotically loaded (Salomon, 2007, p. 84). The allographs in question, grouped by their indexical associations, are shown in (12).

Notwithstanding the indexical associations mentioned in (12), it is common to see allographs from both styles featured in a particular Devanagari-script text, whether due to idiosyncratic preference or typographical limitations. There may also be writing traditions that draw in a predictable manner from both groups of allographs. For instance, contemporary Marathi-Devanagari (𝚖𝚛-𝙳𝚎𝚟𝚊) texts tend to feature modern-style |अ ख झ ण क्ष| but traditional-style |ल श त्र|.

In the twentieth century, the 𝚑𝚒-𝙳𝚎𝚟𝚊 graph inventory has been augmented to unambiguously denote Hindi phones absent in Sanskrit, as well as to mark certain Perso-Arabic-origin phones that have been absorbed into Hindi phonology. Graphetically, the new graphs have been created by augmenting existing ones with a subfixed nuqta, as shown in (13):

While graphs for Hindi-specific phones absent from Sanskrit have been added to 𝚑𝚒-𝙳𝚎𝚟𝚊’s inventory, the converse has not occurred. Graphs for Sanskrit-specific phones absent from Hindi, such as |ऋ ⁠ृ| [ɾ̩] and |ष| [ʂ] have not been deleted from 𝚑𝚒-𝙳𝚎𝚟𝚊’s inventory and are retained as they are. Such graphematic inertia is reminiscent of similar practice in 𝚏𝚊-𝙰𝚛𝚊𝚋 and 𝚜𝚍-𝙰𝚛𝚊𝚋, wherein 𝚊𝚛-𝙰𝚛𝚊𝚋 graphs and spellings are retained near-unchanged (§6.5.4). Masica opines that the graphematic inertia observed in South Asian writing systems based on the Arabic and Devanagari scripts is motivated by the liturgical nature of the principal languages associated with these scripts — Arabic and Sanskrit, respectively:

[R]edundant Arabic graphemes (with no attempt at distinctive articulation) are preserved in the spelling of Arabic loanwords in most languages using the Perso-Arabic script, just as redundant Neo-Brahmi letters are preserved in the spelling of Sanskrit loanwords in most languages using Neo-Brahmi scripts – although there are more such letters in Perso-Arabic. Attempts to simplify the orthographies by spelling words as they sound […] have occasionally been made, but have not met with general acceptance, since the Arabic language is as sacrosanct for Muslims as Sanskrit is for Hindus.

(Masica, 1991, p. 151)

The seemingly sacrosanct nature of 𝚜𝚊-𝙳𝚎𝚟𝚊 has also encouraged the retention of certain benedictory logograms, particularly |ॐ| [om] (§2.8). Punctuation graphs in 𝚜𝚊-𝙳𝚎𝚟𝚊 are minimal, with the salient ones being the sentence separator |।| and paragraph separator |॥|. In 𝚑𝚒-𝙳𝚎𝚟𝚊, which follows punctuation principles similar to those in 𝚎𝚗-𝙻𝚊𝚝𝚗, |।| has been repurposed as a full stop. However, other modern Devanagari-script-based writing systems such as 𝚖𝚛-𝙳𝚎𝚟𝚊 have replaced |।| with Roman-script |.| Overall, most punctuation graphs commonly used in modern 𝚑𝚒-𝙳𝚎𝚟𝚊 and 𝚖𝚛-𝙳𝚎𝚟𝚊 have been borrowed from the Roman-script inventory, such as |, ; ? !|. Numeral graphs in the Devanagari inventory, along with prevalent allographic forms, are shown in (14). As with the phonographic allographs shown in (12), the choice of numeral allograph may be conditioned by various graphematic and sociolinguistic factors, and defies neat prediction.

Modern Sindhi-Devanagari (𝚜𝚍-𝙳𝚎𝚟𝚊) makes use of many of the graphs and writing practices described above, while also deviating from them in places. Moreover, 𝚜𝚍-𝙳𝚎𝚟𝚊 also features certain unique practices. Details of 𝚜𝚍-𝙳𝚎𝚟𝚊’s graphematic makeup, along with its historical evolution, follow in the sections below.

7.2 Early use

As outlined in Section 5.1, script inventories graphetically resembling present-day Devanagari forms have been used to write Sindhi and related speech varieties for at least two centuries, if not more. Prominent instances of early 𝚜𝚍-𝙳𝚎𝚟𝚊 use are seen in the works of Qazi Qadan and Pran Nath. That said, the use of 𝚜𝚍-𝙳𝚎𝚟𝚊 experienced noticeable spikes in use at two distinct moments in time, the first immediately following the British annexation of Sindh in the mid-1800s, and the second following Partition in 1947. The first surge in 𝚜𝚍-𝙳𝚎𝚟𝚊 writing was driven largely by British and European interest in Sindhi, typically with administrative or Christian-missionary aims in mind. One of the earliest such 𝚜𝚍-𝙳𝚎𝚟𝚊 publications dates from 1834, when parts of the Bible’s Gospel of Matthew were translated into Kutchi (ISO-639 𝚔𝚏𝚛) and transcribed using Devanagari (Nida, 1972, p. 212). Graphematically, the writing system in question is not 𝚜𝚍-𝙳𝚎𝚟𝚊 but 𝚔𝚏𝚛-𝙳𝚎𝚟𝚊. The 𝚔𝚏𝚛-𝙳𝚎𝚟𝚊 extract is reproduced and retranscribed in Figure 7.1.

• 1 When he [Jesus] saw the crowds he went up the mountain. He sat down and his disciples joined him.
• 2 He started to teach them, saying:
• 3 Fortunate are the spiritually poor, for theirs is heaven’s kingdom.
• 4 Fortunate are the grieving, for they shall be consoled
• 5 Fortunate are the kindly, for they shall be given the world.

Apart from the now-archaic spellings, Figure 7.1 features at least two noteworthy graphematic practices. The first is the use of |ष ख| as freely varying allographs of each other. In 𝚜𝚊-𝙳𝚎𝚟𝚊, |ष| has the linguistic value [ʂ(a)], and |ख| has the value [kʰ(a)]. However, Figure 7.1 features the 𝚔𝚏𝚛-𝙳𝚎𝚟𝚊 spellings |षुदा| [kʰʊd̪a ~ xʊd̪a] ‘god’ (5:3) and |षुश| [kʰʊɕ ~ xʊɕ] ‘happy’ (5:4) on the one hand, and |सीखा| [sɪkʰa ~ sikʰa] ‘teaching’ (5:2) on the other. The practice of using |ष ख| in a mutually substitutable manner was common among Devanagari-based writing systems used in north-western South Asia, until at least the early twentieth century (Mandal, 2021). In 𝚜𝚍-𝙳𝚎𝚟𝚊, this convention was also followed by George Stack (§7.3).

The second noteworthy graphematic practice in Figure 7.1 is the contrastive use of |ॸ| [ɖ(ə)] and |ड| [ɽ(ə), ɽʱ(ə)]. Viewed against a modern-day Kutchi pronunciation, verse 1 reads as follows:

|एं ऊ माडू एं जी मंढली ॸिसी हिकडे ॸूंगरतें चडी वियो एं जॸें ऊ वेठो उनज चेला उन वट आया|

[ẽ u maɽu ẽ d͡ʑi məɳɖ(ʱ)ᵊli ɖɪsi ɦɪkᵊɽe ɖũɡəɾᵊt̪ẽ t͡ɕəɽ(ʱ)i ʋɪjo ẽ d͡ʑəɖẽ u ʋeʈʰo ʊnᵊd͡ʑᵊ t͡ɕela ʊnᵊ ʋəʈᵊ aja]

Here, the Kutchi words |ॸिसी| [ɖɪsi] ‘having seen’, |ॸूंगरतें| [ɖũŋɡəɾᵊt̪e] ‘on a mountain’ and |जॸें| [d͡ʑəɖɛ̃] ‘when’ feature the graph |ॸ| for [ɖ(ə)].¹⁰³ In contrast, |माडू| [maɽu] ‘men, people’, |हिकडे| [ɦɪkᵊɽe] ‘a, one’ and |चडी| [t͡ɕəɽ(ʱ)i] ‘having climbed’ feature |ड| for [ɽ(ə) ~ ɽʱ(ə)]. Graphematically, such contrastive use of |ॸ| [ɖ(ə)] and |ड| [ɽ(ə)] is most closely associated with Devanagari-script Marwari (𝚖𝚠𝚛-𝙳𝚎𝚟𝚊), although |ॸ| fell out of use in 𝚖𝚠𝚛-𝙳𝚎𝚟𝚊 by the early twentieth century (Pandey, 2010b). Despite unclear graphetic origins, Pandey speculates that |ॸ| may be derived from a homoglyphic and homophonic graph in the Sharada script traditionally used for Kashmiri (Pandey, 2005; 2010b, p. 3). Graphematically, the use of distinct graphs for [ɖ] and [ɽ] in Kutchi has parallels in other writing systems for Kutchi, particularly Khojki (see Chapter 10).

Aside from the graphematic significance of individual graphs and their linguistic values, the 𝚔𝚏𝚛-𝙳𝚎𝚟𝚊 extract also proves sociolinguistically significant. As the compiler, Samuel Bagster, notes in his work:

A translation of the Gospel of St. Matthew into Cutchee [i.e., Kutchi] was executed by the late Rev. James Gray, one of the chaplains at Bombay; and in 1835 a small edition, consisting of 500 copies, was printed. This edition was, however, found to be of little or no service, from the circumstance of its being issued in the [Devanagari] character, with which the people of Cutch are unacquainted.

(Bagster, 1851, p. 99)

Thus, the 𝚔𝚏𝚛-𝙳𝚎𝚟𝚊 specimen in Figure 7.1 may not be reflective of actual in-group or emic usage. This is consistent with the observation that, in the nineteenth century, it was Europeans who were the primary drivers and users of Sindhi-Kutchi in Devanagari. Exemplifying such etic use of 𝚜𝚍-𝙳𝚎𝚟𝚊 are the works of grammarian, lexicographer and translator, George Stack.

7.3 George Stack’s system

The first book-length works to feature Sindhi written in Devanagari were the pioneering grammar and dictionaries of Captain George Stack, a military officer in the East India Company. Initially posted to Bombay in 1843–44 shortly after the Company’s annexation of Sindh, Stack began acquainting himself with the Hindustani and Sanskrit languages during his tenure there. In 1845, he was transferred to Hyderabad, Sindh, as Deputy Collector, following which he acquired a remarkable level of proficiency in Sindhi within a relatively short time (Encyclopedia Sindhiana, n.d., p. 7130). By March 1847, he had prepared a grammar of Sindhi, which, however, went to press only in 1849 (Stack, 1849a, p. vii). Besides containing several pages of “Stories for Exercise” in typeset 𝚜𝚍-𝙳𝚎𝚟𝚊 (pp. 134–153), Stack’s grammar also contains extensive samples of handwritten 𝚜𝚍-𝙳𝚎𝚟𝚊. Stack’s grammar was accompanied by an English-Sindhi dictionary (1849b), followed by a posthumously-published Sindhi-English dictionary (1855).

Stack’s rationale for choosing Devanagari to transcribe Sindhi is outlined in the introduction to his grammar (§5.2.1). To reiterate, Stack opts for Devanagari since it was the script most familiar to British officers of the East India Company. To Stack, the fact that Devanagari was not widely used for the language by Sindhis themselves was apparently of less importance. Barrow Ellis corroborates this sentiment in his foreword to Stack’s Sindhi-English dictionary (1855):

The reasons which induced Captain Stack to print in the Devanagri [sic] character he has slightly touched on in his Preface to the English and Sindhi dictionary. Lest his views be misunderstood I may mention that he frequently assured me, that he adopted the character merely as one generally known to those officers who might be supposed to be the persons most likely to use these books — and not with any view to its permanent adoption by Sindhis. Had I myself been required to decide the point, I do not think I should have adopted the same character, but it was obviously better to act and do something which might be useful to a large class of students, than to talk and wait for the termination of an apparently interminable controversy to settle a Sindhí character for universal adoption.

(Ellis, in Stack, 1855, p. vi; emphasis added)

To be borne in mind is the fact that the above foreword was published after the official adoption of 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝟷𝟾𝟻𝟹, in which Ellis himself played a key role (§6.3). Notwithstanding the etic use of Devanagari, Stack’s 𝚜𝚍-𝙳𝚎𝚟𝚊 system was largely consistent across his works, and subject only to minor variations. Such consistency was all the more noteworthy given that Stack’s 𝚜𝚍-𝙳𝚎𝚟𝚊 system was largely an ex-nihilo creation, without precedent to refer to:

I have had considerable difficulty in fixing the Orthography. Many of the words I have given will be found to be pronounced and written very differently by different persons. I have tried to choose the most common forms of spelling, or to follow the general manner of pronunciation, without at the same time going into actual vulgarisms; but my choice was very arbitrary, and it will doubtless be thought that I am often incorrect.

(Stack, 1849b, p. iv)

Following the IETF format for language tags, Stack’s Sindhi-Devanagari system will be referred to as 𝚜𝚍-𝙳𝚎𝚟𝚊-𝚡-𝚜𝚝𝚊𝚌𝚔. Table 7.2 provides an overview of the graph inventory and graph-phone correspondences of 𝚜𝚍-𝙳𝚎𝚟𝚊-𝚡-𝚜𝚝𝚊𝚌𝚔 as listed in Stack’s (1849a) grammar.

Table 7.2 reveals that the γ-vowels in 𝚜𝚍-𝙳𝚎𝚟𝚊-𝚡-𝚜𝚝𝚊𝚌𝚔 are derived from 𝚜𝚊-𝙳𝚎𝚟𝚊 in terms of linguistic values and graphotactic behaviour. The latter aspect is especially evident in the complementary distribution of free and bound γ-vowel allographs (Stack, 1849a, p. 127). In terms of γ-consonants, Table 7.2 shows that 𝚜𝚍-𝙳𝚎𝚟𝚊-𝚡-𝚜𝚝𝚊𝚌𝚔 does not contain separate graphs for the φ-consonants [q x ɣ z f]. This is consistent with Stack’s intention to spell Sindhi words according to “the general manner of pronunciation, without […] going into actual vulgarisms” (1849b, p. iv; see also quote above). Thus, Stack spells Sindhi [komᵊ] ‘tribe, race’ as |कोम| (1855, p. 68). Compare this spelling with the modern 𝚜𝚍-𝙳𝚎𝚟𝚊 spelling |क़ौम| (Rohra, Bijani, & Gurnani, 2011, p. 115), which reflects the pronunciation [qɔmᵊ]. Nevertheless, given that the everyday pronunciation of the word is [komᵊ], the newer spelling |क़ौम| may actually be considered an etymologically-inspired spelling (cf. Arabic [qawm] ‘ibid.’). On the other hand, 𝚜𝚍-𝙳𝚎𝚟𝚊-𝚡-𝚜𝚝𝚊𝚌𝚔 does provide for separate graphs for implosive [ɠ(ə) ʄ(ə) ɗ(ə) ɓ(ə)]: |ग़ ज़ ड़ ब़|, respectively. However, |ड़| is also used to denote plosive [ɖ(ə)]. Stack deems it unnecessary to graphematically differentiate implosive [ɗ(ə)] and plosive [ɖ(ə)], claiming that “the distinction though plain to Natives, is not at first apparent to European ears, and is of little importance” (1849b, p. 9 footnote). Stack’s graphematic underdifferentiation of [ɖ(ə)] and [ɗ(ə)] has been criticised by Lepsius (1863, pp. 105–106) and Trumpp, with the latter terming the underdifferentiation “very embarrassing” (1858, p. iii) and “a great mistake” (1872, p. 16). According to Trumpp, “[a] Sindhi will never confound [ɖ(ə)] and [ɗ(ə)]; they are in his mouth thoroughly distinct from each other” (1872, p. 16).¹⁰⁴

Despite conflating [ɖ(ə)] and [ɗ(ə)], 𝚜𝚍-𝙳𝚎𝚟𝚊-𝚡-𝚜𝚝𝚊𝚌𝚔 represents [ɽ(ə)] distinctly, with undotted |ड|. Thus, the pattern of graphematically distinguishing (im)plosive |ड़| [ɗ(ə) ɖ(ə)] from the approximant |ड| [ɽ(ə)] bears similarities to that seen in Figure 7.1, in which |ॸ| [ɗ(ə) ɖ(ə)] is graphematically distinguished from |ड| [ɽ(ə)]. Furthermore, Stack’s system represents aspirate [ɖʱ(ə)] with |ढ|, but writes [ɽʱ(ə)] with the graph sequence |ड्ह|. Noteworthy is the use of distinct graphs |ट़ ॾ| for the φ-clusters [ʈɾ(ə) ɖɾ(ə)], respectively, which Stack justifies by citing similar practice in Sindhi-Khudawadi (𝚜𝚍-𝚂𝚒𝚗𝚍).¹⁰⁵ By extension, [ɖʱɾ(ə)] is written |ढ़|. It is conceivable that the distinct representation of [ʈɾ(ə) ɖɾ(ə)] in 𝚜𝚍-𝙳𝚎𝚟𝚊-𝚡-𝚜𝚝𝚊𝚌𝚔 inspired similar practice in Trumpp’s system (§7.4).

Although listed as a distinct graph in the inventory, Sanskritic |ष| is hardly used in Stack’s grammar (1849a, pp. 8, 127) and English-Sindhi dictionary (1849b). However, |ष| appears occasionally in his Sindhi-English dictionary (1855), as in the entries |संतोखु, संतोषु| [sən̪t̪okʰᶷ, sən̪t̪oɕᶷ] ‘contentment’ (p. 385) and |कष्टु| [kəɕʈᶷ] ‘wretchedness, misery’ (p. 50). In contrast, Stack’s English-Sindhi Dictionary (1849b) only lists |संतोखु| (p. 26), and lacks an entry corresponding to |कष्टु|. Phonologically, 𝚜𝚍-𝙳𝚎𝚟𝚊-𝚡-𝚜𝚝𝚊𝚌𝚔’s equating of the graphs |ष ख| reflects the phonological merger of Sanskritic retroflex [ʂ] with New Indo-Aryan [kʰ] (Grierson, 1881, p. 3; Masica, 1991, p. 105). Also, while |ष| is marginal in 𝚜𝚍-𝙳𝚎𝚟𝚊-𝚡-𝚜𝚝𝚊𝚌𝚔, it is altogether absent from 𝚜𝚍-𝙳𝚎𝚟𝚊-𝚡-𝚝𝚛𝚞𝚖𝚙𝚙 (see Figure 6.3; see also Section 6.4, p. 163).

To represent φ-nasalisation, 𝚜𝚍-𝙳𝚎𝚟𝚊-𝚡-𝚜𝚝𝚊𝚌𝚔 provides for the use of the anunasika graph |⁠ँ|, also known as chandrabindu. In his English-Sindhi dictionary (1849b, p. 10), Stack claims to use the anunasika distinctly from the anusvara |⁠ं|, restricting the latter to denoting a φ-nasal homorganic with the following φ-obstruent. However, his works reveal somewhat vacillating use of these graphematic devices. The Sanskritic graphs |ऋ ⁠ृ| [ɾ̩] and |⁠ः| [ɦ] (see Table 7.1) are absent from 𝚜𝚍-𝙳𝚎𝚟𝚊-𝚡-𝚜𝚝𝚊𝚌𝚔.

In terms of graphetic form and typography, Stack’s (1849a) grammar features different allographs in print and handwriting. The φ-vowel [ə] and the φ-consonants [d͡ʑʱ(ə) ɳ(ə)] are printed as |अ झ ण|, but handwritten as |अ झ ण|, respectively. Also, while the print samples indicate [ɽʱ(ə)] with |ड्ह|, the handwritten samples feature the form |ड्ह|. Figure 7.2 shows the handwritten forms of |चाड्ह्याईं| [t͡ɕaɽʱᶦjãĩ] ‘he/she raised’ and |झूनाग़ड्हु| [d͡ʑʱunaɠəɽʱᶷ] ‘Junagadh (a region of present-day Gujarat)’ as they appear in Stack’s work.

Overall, graphetic combinations of γ-consonants or so-called ligatures are widely attested in 𝚜𝚍-𝙳𝚎𝚟𝚊-𝚡-𝚜𝚝𝚊𝚌𝚔, although Stack (1849b, p. 128) specifies that such graphetic combination or ligation is optional. This is affirmed by the freely varying use of atomic |ड्ह| and ligated |ड्ह| for [ɽʱ(ə)].

Stack also translated the Biblical Gospel of Matthew into Sindhi using 𝚜𝚍-𝙳𝚎𝚟𝚊-𝚡-𝚜𝚝𝚊𝚌𝚔, which was published in 1850 (Grierson, 1919, p. 13; Nida, 1972, p. 393).

7.4 Ernest Trumpp’s system

Following closely in the heels of 𝚜𝚍-𝙳𝚎𝚟𝚊-𝚡-𝚜𝚝𝚊𝚌𝚔 was an alternative graph inventory and graphematic system for 𝚜𝚍-𝙳𝚎𝚟𝚊, designed by Ernst Trumpp. Aside from the graph inventory of 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚞𝚖𝚙𝚙, Figure 6.3 and Figure 6.4 also feature that of 𝚜𝚍-𝙳𝚎𝚟𝚊-𝚡-𝚝𝚛𝚞𝚖𝚙𝚙, with linguistic values shown in Lepsius’ Standard Alphabet. Trumpp’s system differs from Stack’s in very minor ways, almost entirely in the domains of γ-consonants and their phonological values. The distinctions in graph-phone correspondences between 𝚜𝚍-𝙳𝚎𝚟𝚊-𝚡-𝚜𝚝𝚊𝚌𝚔 and 𝚜𝚍-𝙳𝚎𝚟𝚊-𝚡-𝚝𝚛𝚞𝚖𝚙𝚙 are shown in (15):

The examples in (15) reveal that, at least for the graphs in question, 𝚜𝚍-𝙳𝚎𝚟𝚊-𝚡-𝚝𝚛𝚞𝚖𝚙𝚙 was somewhat more regular than 𝚜𝚍-𝙳𝚎𝚟𝚊-𝚡-𝚜𝚝𝚊𝚌𝚔. Graphetically, though, the use of |ॾ| for [ɗ(ə)] disrupts the visual harmony in the set of γ-implosives |ग़ ज़ ॾ ब़|. The Sanskritic graphs |ऋ ष ⁠ः| along with any bound forms (see Table 7.1) are excluded from 𝚜𝚍-𝙳𝚎𝚟𝚊-𝚡-𝚝𝚛𝚞𝚖𝚙𝚙, although Trumpp (1872, p. 18) acknowledges Stack’s sporadic use of |ष|. Similar to 𝚜𝚍-𝙳𝚎𝚟𝚊-𝚡-𝚜𝚝𝚊𝚌𝚔, 𝚜𝚍-𝙳𝚎𝚟𝚊-𝚡-𝚝𝚛𝚞𝚖𝚙𝚙 uses both |⁠ँ| and |⁠ं| to denote φ-vowel nasalisation in a somewhat arbitrary manner. In terms of graphetic features, Figure 6.4 shows the 𝚜𝚍-𝙳𝚎𝚟𝚊-𝚡-𝚝𝚛𝚞𝚖𝚙𝚙 γ-vowel inventory with [ə a o ɔ] represented by |अ आ ओ औ|, respectively. These forms may be considered allographs in free variation with |अ आ ओ औ|, respectively. Similar to 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚞𝚖𝚙𝚙, 𝚜𝚍-𝙳𝚎𝚟𝚊-𝚡-𝚝𝚛𝚞𝚖𝚙𝚙 features the distinctive four-dot graph |⁛| as a full stop or sentence separator (§6.4).

As with 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚞𝚖𝚙𝚙, 𝚜𝚍-𝙳𝚎𝚟𝚊-𝚡-𝚝𝚛𝚞𝚖𝚙𝚙 was apparently not used by any author other than its creator. At most, the system was employed as a transliteration or transcription in the odd scholarly work, most notably in Grierson (1919). Building on Trumpp (1858, p. vi) and Lepsius (1863, p. 103), Grierson (1919, p. 22) augments the 𝚜𝚍-𝙳𝚎𝚟𝚊 repertoire as shown in (16):

As evident from (16), Grierson reverts to the 𝚜𝚍-𝙳𝚎𝚟𝚊-𝚡-𝚜𝚝𝚊𝚌𝚔 paradigm of using |ग़ ज़ ड़ ब़| for the φ-implosives [ɠ(ə) ʄ(ə) ɗ(ə) ɓ(ə)], and redeploys |ॾ| for [ɽ(ə)]. Notably, Grierson adds the graphs |स़ ह़ ख़ (स़) अ़ फ़ क़| as canonical equivalents of 𝚊𝚛-𝙰𝚛𝚊𝚋 (left-to-right) |ث ح خ (ص) ع ف ق|, respectively. He also uses these graphs in his 𝚜𝚍-𝙳𝚎𝚟𝚊 text samples (1919, pp. 99, 107), making it a rare instance of 𝚊𝚛-𝙰𝚛𝚊𝚋 |ث ح ص ع| being distinctly and reversibly transliterated in 𝚜𝚍-𝙳𝚎𝚟𝚊 running text. Despite his eye for detail, deviations from the above norms are found in Grierson’s work.¹⁰⁶ Yet, they remain few in number, and might well be unintended typographical errors. Grierson also replaces the four-dot sentence separator |⁛| with the 𝚜𝚊-𝙳𝚎𝚟𝚊-style sentence separator |।| (§7.1).

7.5 Post-Partition practices​¹⁰⁷

In early twentieth-century Sindh, as 𝚜𝚍-𝙰𝚛𝚊𝚋 became increasingly adopted as the primary writing system for Sindhi, publications in 𝚜𝚍-𝙳𝚎𝚟𝚊 remained limited in number. However, Partition gave 𝚜𝚍-𝙳𝚎𝚟𝚊 a new lease of life. As outlined in Section 5.3, a considerable support base for 𝚜𝚍-𝙳𝚎𝚟𝚊 began to emerge in post-Partition India, leading to official recognition as well as increasing standardisation of the writing system. However, several areas of graphematic variation continue to persist in 𝚜𝚍-𝙳𝚎𝚟𝚊.

In the post-Partition era, almost all 𝚜𝚍-𝙳𝚎𝚟𝚊 publications have originated from India. Pakistani 𝚜𝚍-𝙳𝚎𝚟𝚊 publications are most likely to be of a pedagogical nature, and published by linguistic or literary institutions such as the Sindhi Language Authority (Hussain, 2011). Such publications exhibit graphematic regularity and variation in line with Indian 𝚜𝚍-𝙳𝚎𝚟𝚊 publications.

7.5.1 Graph inventory

Steps towards standardising 𝚜𝚍-𝙳𝚎𝚟𝚊’s graph inventory began in the early 1950s. Since Hindi and Sindhi share their phonologies to a large extent, the 𝚑𝚒-𝙳𝚎𝚟𝚊 graph inventory was implicitly adopted as the basis for 𝚜𝚍-𝙳𝚎𝚟𝚊. On the suggestion of linguist Suniti Kumar Chatterji, the Sindhi implosives [ɠ(ə) ʄ(ə) ɗ(ə) ɓ(ə)] began to be written |ॻ ॼ द॒ ॿ| (National Archives of India, 2018b, pp. 362–363). Of these, |द॒| has since been replaced by |ॾ|. However, Chatterji’s suggestion of representing the free γ-vowels as |अ आ अि अी अु अू अे अै ओ औ|, with |अ| used as a rasmic base akin to 𝚜𝚍-𝙰𝚛𝚊𝚋’s alif |ا|, did not meet with similar success.¹⁰⁸ Efforts towards graphematic standardisation in 𝚜𝚍-𝙳𝚎𝚟𝚊 were buttressed by vigorous Sindhi-language literary activity in the two decades immediately after Partition. Described as defying expectations (Lekhwani, personal communication, December 8, 2014), such literary activity further stimulated the publication of several 𝚜𝚍-𝙳𝚎𝚟𝚊 lexicographic (Hardwani, 1991; Lekhwani, 1996; Rohra, Bijani, & Gurnani, 2011) and pedagogical works (Lekhwani, 1997; Varyani & Thakwani, 2003). These events and publications have cumulatively precipitated a gradual standardisation of the 𝚜𝚍-𝙳𝚎𝚟𝚊 graph inventory and, to a lesser extent, its graphematic practices.

Yet, 𝚜𝚍-𝙳𝚎𝚟𝚊’s journey thus far towards standardisation has not been smooth, nor has it been exhaustive. Just as the 𝚊𝚛-𝙰𝚛𝚊𝚋 and 𝚏𝚊-𝙰𝚛𝚊𝚋 graph inventories and graphematic practices exert pressure on 𝚜𝚍-𝙰𝚛𝚊𝚋, so too do 𝚜𝚊-𝙳𝚎𝚟𝚊 and 𝚑𝚒-𝙳𝚎𝚟𝚊 influence 𝚜𝚍-𝙳𝚎𝚟𝚊. This results in the retention of graphs from the former inventories in 𝚜𝚍-𝙳𝚎𝚟𝚊, even when phonologically superfluous in the context of modern spoken Sindhi (Khubchandani, 2007, p. 699). Thus, the 𝚜𝚊-𝙳𝚎𝚟𝚊 free graphs |ऋ ष| (Sanskrit: [ɾ̩ ʂa]) and their bound allographs |⁠ृ ष्‍◌| are pronounced [ɾɪ ɕ(ə)], respectively, making them homophonous with 𝚜𝚍-𝙳𝚎𝚟𝚊 |रि श|. Similarly, 𝚜𝚊-𝙳𝚎𝚟𝚊 |ज्ञ| [ɟɲə] is realised in 𝚜𝚍-𝙳𝚎𝚟𝚊 as [ɡjə], identically to |ग्य|. Certain authors may choose to retain the Sanskritic |ऋ ष ज्ञ| and their allographs in their 𝚜𝚍-𝙳𝚎𝚟𝚊 work, while others may opt for the phonologically more transparent |रि श ग्य|. Also inherited in the 𝚜𝚍-𝙳𝚎𝚟𝚊 inventory are modified graphs for the Perso-Arabic phones [q x ɣ z f] (see (13)). Based on the precedent of 𝚑𝚒-𝙳𝚎𝚟𝚊, the 𝚜𝚍-𝙳𝚎𝚟𝚊 inventory features |क़ ख़ ग़ ज़ फ़| for [q(ə) x(ə) ɣ(ə) z(ə) f(ə)], created by subfixing a nuqta to |क ख ग ज फ| [k(ə) kʰ(ə) ɡ(ə) d͡ʑ(ə) pʰ(ə)]. However, [q x ɣ z f] are unstable in modern Indian Sindhi and vary unpredictably with [k kʰ ɡ d͡ʑ pʰ] (§4.3.1). As a result, popular or casual 𝚜𝚍-𝙳𝚎𝚟𝚊 writing may substitute underdotted |क़ ख़ ग़ ज़ फ़| with the graphetically simpler |क ख ग ज फ|. Scholarly works in 𝚜𝚍-𝙳𝚎𝚟𝚊 usually retain the underdotted forms |क़ ख़ ग़ ज़ फ़|, although they may allude to their near-homophony with |क ख ग ज फ| to various extents. For instance, a 𝚜𝚍-𝙳𝚎𝚟𝚊 primer published by the NCPSL (Varyani & Thakwani, 2003, p. x) explicitly considers |क़ क| to be homophonous, with both pronounced [k(ə)]. The Indian Institute of Sindhology (Rohra, Bijani, & Gurnani, 2011) has attempted to lay down guidelines on the status and use of homophonous graphs in 𝚜𝚍-𝙳𝚎𝚟𝚊 (§7.6).

Comparing homophonous graphs in 𝚜𝚍-𝙰𝚛𝚊𝚋 with 𝚜𝚍-𝙳𝚎𝚟𝚊, it emerges that such instances are fewer in 𝚜𝚍-𝙳𝚎𝚟𝚊. In contrast, the distribution and use of homophonous graphs in 𝚜𝚍-𝙳𝚎𝚟𝚊 writing appears more unpredictable. Such variation has implications for 𝚜𝚍-𝙳𝚎𝚟𝚊 lexicography. For instance, the Sanskrit-derived Sindhi lexical root [ʋɪɡjanᶷ] ‘science’ is spelt |विज्ञानु| in one dictionary (Lekhwani, 1996, p. 168) but |विग्यानु| in another (Hardwani, 1991, p. 442). Moreover, differences in graph inventory may also be seen in works by the same author. In most of his 𝚜𝚍-𝙳𝚎𝚟𝚊 works, Hardwani (1991; 2013) lists and uses the underdotted graphs |क़ ख़ ग़ ज़ फ़ ड़ ढ़| [q(ə) x(ə) ɣ(ə) z(ə) f(ə) ɽ(ə) ɽʱ(ə)] (see (13)). However, in a more recently published primer (Hardwani, 2017), he only includes |ड़ ढ़| as part of the 𝚜𝚍-𝙳𝚎𝚟𝚊 inventory, and excludes augmented graphs to denote phones of Perso-Arabic origin. In the foreword to his (2017) publication, Hardwani writes:

मां देवनागरी सिंधी खे अरबी सिधीअ खां अलॻ भाषा समुझंदो आहियां, छो त ॿिन्ही जी अखरमाला अलॻ-अलॻ आहे ऐं शब्द-भंडारु पंहिंजे पंहिंजे नमूने जो आहे. […] इन करे हिन किताब में अरबी सिंधी अखरमाला वारा ख़, ग़, ज़, फ़ वर्ण या अखर कोन खंया विया आहिनि. ऐं इएं करण सां सिंधी भाषा पढ़ण-लिखण में कोई अड़चन पैदा कान थींदी. देवनागरी सिंधीअ जी हीअ पंहिंजी वर्णमाला / अखरमाला आहे.

I consider Devanagari-Sindhi to be a separate language [sic] from Arabic-Sindhi, since their letter inventories and vocabularies are distinct. […] Accordingly, the Arabic-Sindhi-inspired letters ख़, ग़, ज़, फ़ are omitted from this book. Doing so does not create any difficulties in reading and writing the Sindhi language. After all, this letter inventory is intrinsic to Devanagari-Sindhi.

(Hardwani, 2017, p. 2)

Hardwani’s motivation for omitting |क़ ख़ ग़ ज़ फ़| based on semiotic connotations rather than on phonological grounds exemplifies the kinds of graphosociolinguistic battles permeating the Sindhi script debates in post-Partition India (§5.3). That said, while Hardwani is upfront about his ideological stance, not all 𝚜𝚍-𝙳𝚎𝚟𝚊 writers are. As a result, it is often unclear whether grapholinguistic practices seen in a particular 𝚜𝚍-𝙳𝚎𝚟𝚊 publication are attributable to sociolinguistic ideology, technological limitations or plain human oversight.

7.5.2 Graphematic allography

In the early years after Partition, as efforts towards standardising the 𝚜𝚍-𝙳𝚎𝚟𝚊 inventory were in still in its infancy, it was common to see a large degree of allography in 𝚜𝚍-𝙳𝚎𝚟𝚊 publications involving the sets shown in (12)(14) and (14). The most frequent instances of allography involved the phonograms for [ə d͡ʑʱ(ə) ɳ(ə) l(ə) ɕ(ə)], whose forms alternated between the ‘traditional’ forms |अ झ ण ल श| and their ‘modern’ counterparts |अ झ ण ल श|, respectively. Less frequently, variation was also seen between the free γ-vowel forms |ए| and |अे|, both denoting [e]. Other allographic sets in free variation involved graphs denoting φ-consonant clusters, such as |क्त ष्ट्र| and |क्‍त ष्‍ट्र| for [kt̪(ə) ɕʈɾ(ə)]. Although not allography per se, homophonous heterography was also seen in the depiction of [ɳ n̪ m] in a φ-consonant cluster either with an anusvara |ं| or with the bound allographs |ण्‍◌ न्‍◌ म्‍◌|. Notably, variation was often ‘free’ to the extent that the allographs in question often appeared in the same publication, on the same page, or even within the same sentence, with no evident conditioning factors. Table 7.3 features such instances of free variation from three 𝚜𝚍-𝙳𝚎𝚟𝚊 publications from the 1960s: a work of fiction (Bhambhani, 1964), a work of non-fiction (Jagtiani, n.d.) and a government-sanctioned school textbook (Maharashtra State Board of Secondary Education, 1967).

Notwithstanding the allographic variability demonstrated in Table 7.3, the publications in question were consistent in featuring |द॒| [ɖ(ə)]. This graph was subsequently supplanted by |ॾ|. One of the texts (Maharashtra State Board of Secondary Education, 1967) also writes [ɛ̃] consistently as |अैं|, an allograph that has since given way to |ऐं|.

The use of anusvara |⁠ं| has since become the preferred graphematic option in modern 𝚜𝚍-𝙳𝚎𝚟𝚊 to indicate [ɳ n̪ m] in a homorganic nasal-oral φ-consonant cluster. At the same time, the use of anusvara to indicate nasalisation of a φ-vowel has also become conventionalised in 𝚜𝚍-𝙳𝚎𝚟𝚊, with chandrabindu |⁠ँ| becoming implicitly deprecated. As with 𝚑𝚒-𝙳𝚎𝚟𝚊 (see (10)), the polyvalent use of anusvara introduces opacity into an otherwise relatively transparent writing system. For instance, a nonfluent reader of Sindhi may decode 𝚜𝚍-𝙳𝚎𝚟𝚊 |नांगु| either as [nãɡᶷ] or as [naŋɡᶷ], both of which are plausible. Similarly, |हंसु| may be justifiably interpreted either as [ɦə̃sᶷ] or as [ɦənsᶷ]. However, the ‘correct’ or desired pronunciations are [nãɡᶷ] and [ɦənsᶷ], meaning ‘cobra’ and ‘swan’, respectively. Given the lack of graphematic cues, retrieving the intended phonological forms requires the reader to be proficient in spoken Sindhi.

7.5.3 Spelling and orthography

Despite 𝚜𝚍-𝙳𝚎𝚟𝚊’s overall similarity to 𝚑𝚒-𝙳𝚎𝚟𝚊, the two differ in certain aspects. One of the key areas of difference concerns the representation of lax φ-vowels [ə ɪ ʊ] in certain word positions, in which they often manifest as their reduced allophones [ᵊ ᶦ ᶷ]. As a phonological feature, reduced φ-vowel allophones are largely absent from most other languages written using Devanagari, including Hindi.¹⁰⁹ However, due to the greater sociolinguistic prominence of 𝚑𝚒-𝙳𝚎𝚟𝚊, readers may decode 𝚜𝚍-𝙳𝚎𝚟𝚊 while applying the graphematic and orthographic rules of 𝚑𝚒-𝙳𝚎𝚟𝚊, leading to ill-formed results.

As outlined in Section 7.1, the inherent φ-[V₀] in 𝚑𝚒-𝙳𝚎𝚟𝚊 and 𝚜𝚍-𝙳𝚎𝚟𝚊 is [ə]. However, aside from some Sanskrit neologisms, Hindi phonotactics do not permit [ə] in word-final position (Gumperz, 1958, p. 216) or in a [VC_CV] phonological environment (see the Indo-Aryan schwa deletion rule in Table 4.5). Consequently, in 𝚑𝚒-𝙳𝚎𝚟𝚊, writers abstain from explicitly denoting the absence of [ə] in these word positions, say, by affixing a virama onto a φ-[CV₀] graph, or by using the appropriate bound form of the φ-[CV₀] graph. Thus, in 𝚑𝚒-𝙳𝚎𝚟𝚊, [ə] is left graphematically unsuppressed in word-final and φ-[VC_CV] environments, as the phonotactics of Hindi preclude [ə] from occurring in these positions. A fluent reader of 𝚑𝚒-𝙳𝚎𝚟𝚊 is implicitly aware of this orthographic practice, and mentally suppresses [ə] where required when decoding 𝚑𝚒-𝙳𝚎𝚟𝚊 text. In contrast, Sindhi phonotactics permit [ə] in word-final and φ-[VC_CV] environments, usually in its reduced form [ᵊ]. Hence, if a reader unwittingly applies 𝚑𝚒-𝙳𝚎𝚟𝚊 orthographic conventions when decoding 𝚜𝚍-𝙳𝚎𝚟𝚊, they may end up erroneously suppressing [ə]. Table 7.4 lists certain homographic words common to 𝚜𝚍-𝙳𝚎𝚟𝚊 and 𝚑𝚒-𝙳𝚎𝚟𝚊. Despite their identical spelling and meaning, the words differ in pronunciation in terms of final [ə].

In addition to the restrictions on [ə], Hindi phonotactics also do not permit lax [ɪ ʊ] in word-final position (Gumperz, 1958, p. 216; Schmidt, 2007, p. 341). Consequently, in 𝚑𝚒-𝙳𝚎𝚟𝚊, the bound γ-vowels |⁠ि ⁠ु| in word-final position are phonologically decoded as tense [i u]. On the other hand, Sindhi phonotactics do permit lax [ɪ ʊ] in word-final position, typically as the reduced allophones [ᶦ ᶷ]. Thus, in 𝚜𝚍-𝙳𝚎𝚟𝚊, |⁠ि ⁠ु| or |इ उ| at the end of a word should be decoded as reduced [ᶦ ᶷ]. These graphs should also be decoded as reduced [ᶦ ᶷ] when corresponding to φ-[VC_CV] environments (see Table 4.5 and Table 4.6). The 𝚜𝚍-𝙳𝚎𝚟𝚊 examples listed in Table 7.5 illustrate this phenomenon.

Also implicated in the subtle graphematic differences between 𝚑𝚒-𝙳𝚎𝚟𝚊 and 𝚜𝚍-𝙳𝚎𝚟𝚊 are the written forms of loanwords from Sanskrit. Modern literary Hindi is rich in in Sanskrit-derived lexicon, including neologisms (King R. D., 2001). Due to the revered status of Sanskrit in India (Deshpande, 2016), writers are often loath to tinker with the spellings of loanwords. As a result, the source 𝚜𝚊-𝙳𝚎𝚟𝚊 spellings of loanwords are typically retained in 𝚑𝚒-𝙳𝚎𝚟𝚊, although their pronunciation is usually adapted to suit the phonotactics of modern Hindi. In other words, graphematic authenticity takes precedence over phonological authenticity. Consequently, 𝚜𝚊-𝙳𝚎𝚟𝚊 |शक्ति| ‘strength’ and |वस्तु| ‘thing’ are used in 𝚑𝚒-𝙳𝚎𝚟𝚊 with their graphematic composition unchanged, but are phonologically rendered [ɕəkt̪i] and [ʋəst̪u], respectively (Shapiro, 2007, p. 284). This reflects the 𝚑𝚒-𝙳𝚎𝚟𝚊 rule of interpreting final |⁠ि ⁠ु| as [i u]. In modern Sindhi, Sanskrit loanwords tend to be pronounced in line with their Hindi pronunciations. Thus, the Sindhi pronunciation of the Sanskritic words for ‘strength’ and ‘thing’ would also be [ɕəkt̪i] and [ʋəst̪u]. However, their written forms, |शक्ति| and |वस्तु|, pose a graphematic problem. According to the rules of 𝚜𝚍-𝙳𝚎𝚟𝚊, final |⁠ि ⁠ु| should be decoded as reduced [ᶦ ᶷ], which result in the ill-formed [ɕəkt̪ᶦ] and [ʋəst̪ᶷ], respectively. Notwithstanding the misalignment between graphematic and phonological forms, graphosociolinguistic pressures may oblige some 𝚜𝚍-𝙳𝚎𝚟𝚊 authors to persist with the source 𝚜𝚊-𝙳𝚎𝚟𝚊 spellings, potentially giving rise to unpredictable spelling variation and ambiguity for readers, especially learners. The resultant unpredictable variation is aptly illustrated by Table 7.3, which reveals the spellings |शक्‍ति| and |शक्ती| appearing in the very same 𝚜𝚍-𝙳𝚎𝚟𝚊 school textbook.

An argument in favour of having etymologically-determined uniform spellings for words across 𝚜𝚊-𝙳𝚎𝚟𝚊, 𝚑𝚒-𝙳𝚎𝚟𝚊 and 𝚜𝚍-𝙳𝚎𝚟𝚊 is learner convenience and the conceivable facilitation of literacy transfer. The disadvantage of such spellings is the likelihood for conflict with 𝚜𝚍-𝙳𝚎𝚟𝚊 guidelines, due to which certain written forms would need to be learnt as exceptions. This results in an increased learning curve.

Complicating matters is the emergence of the New Variety of Sindhi and its divergent pronunciations from the Old Variety. As outlined in Section 4.3.2, the reduced allophones [ᵊ ᶦ ᶷ] are absent or noncontrastive in the New Variety. As a result, in the Sindhi New Variety, the words listed in Table 7.4 tend to be pronounced almost like their Hindi equivalents, while those in Table 7.5 are generally pronounced [dɪl], [pʊt͡ɕʰ(ᵊ)] and [ɡʊz(ᵊ)ɾəɳ(ᵊ)], respectively. Hence, a Sindhi New Variety speaker who is dominant in 𝚑𝚒-𝙳𝚎𝚟𝚊 may be doubly disadvantaged when attempting to decode certain 𝚜𝚍-𝙳𝚎𝚟𝚊 spellings, which tend to be based on Old Variety pronunciations. That said, a 𝚑𝚒-𝙳𝚎𝚟𝚊-dominant reader might find it more intuitive to decode 𝚜𝚊-𝙳𝚎𝚟𝚊-based spellings such as |शक्ति| and |वस्तु|, thereby mitigating the disadvantage somewhat.

Given that 𝚜𝚍-𝙳𝚎𝚟𝚊 is not biunique in its graph-phone correspondences, it is evident that its graphematic solution space needs to be constrained by a widely accepted orthography. The absence of such a normative orthography has resulted in the proliferation of idiosyncratic spellings in 𝚜𝚍-𝙳𝚎𝚟𝚊 texts. Variable spellings are particularly noticeable in certain phonological contexts. For instance, [ɽʱ(ə)] may be spelt |ड़ह ड़्ह ढ़ ढ़्ह ढ़ह| (see (16)), reflecting the uncertainty in its phonemic status (§4.3.1). Similar uncertainty on the existence of φ-consonant gemination in Sindhi results in multiple possible spellings for a given word, reminiscent of the analogous phenomenon in 𝚜𝚍-𝙰𝚛𝚊𝚋 (§6.5.2). Thus, in Hardwani’s 𝚜𝚍-𝙳𝚎𝚟𝚊 dictionary (1991, p. 403), the Sindhi word for ‘wool’, pronounced [ʊnᵊ ~ ʊnnᵊ], has two equivalent spellings listed for it: |उन| and |उन्न|. In addition, speakers of Sindhi varieties that feature the clusters [ʈɾ ɖɾ] may reflect these pronunciations in their 𝚜𝚍-𝙳𝚎𝚟𝚊 writing. Thus, the Sindhi festival of Cheti Chand, pronounced [t͡ɕeʈi t͡ɕəɳɖᶷ] in the Vicholi variety and [t͡ɕeʈɾi t͡ɕəɳɖɾᶷ] in Siroli, might be transcribed in by speakers of the latter as |चेट्री चंड्रु| (The Sindhu World, 2021). The Indian Institute of Sindhology has attempted to resolve some of these open questions through its recent lexicographic works (Rohra, Bijani, & Gurnani, 2011), the details of which are described in Section 7.6.

7.5.4 Logograms and collation order

Logograms in 𝚜𝚍-𝙳𝚎𝚟𝚊, including numerals and punctuation marks, largely overlap with those in modern 𝚑𝚒-𝙳𝚎𝚟𝚊. Exceptions include the graph for a full stop. While some 𝚜𝚍-𝙳𝚎𝚟𝚊 works in the immediate post-Partition era used the 𝚑𝚒-𝙳𝚎𝚟𝚊-inspired |।| as a sentence separator, most present-day 𝚜𝚍-𝙳𝚎𝚟𝚊 publications feature the Roman-script full stop |.|.

In terms of collation order, there exist minor differences in the works of different authors. For instance, Lekhwani’s dictionary (1996, p. vi) does not explicitly list |ऋ| and |⁠ं| in the enclosed 𝚜𝚍-𝙳𝚎𝚟𝚊 character chart, although it uses |⁠ं| extensively in spelling 𝚜𝚍-𝙳𝚎𝚟𝚊 entries. On the other hand, Khubchandani (2007, p. 698) includes not only |ऋ| and |⁠ं| in his sequence of graphs, but also the Sanskritic visarga |⁠ः|, which is rarely, if ever, used in modern 𝚜𝚍-𝙳𝚎𝚟𝚊. The Sanskritic |ष| tends to be explicitly listed in most 𝚜𝚍-𝙳𝚎𝚟𝚊 sort orders.

In sum, there exist aspects of graphematic regularity as well as fuzziness in 𝚜𝚍-𝙳𝚎𝚟𝚊. These aspects may reflect corresponding regularity and fuzziness in 𝚜𝚍-𝙰𝚛𝚊𝚋, or may be traceable to the sociolinguistic impact of 𝚜𝚊-𝙳𝚎𝚟𝚊 and 𝚑𝚒-𝙳𝚎𝚟𝚊. As in 𝚜𝚍-𝙰𝚛𝚊𝚋, 𝚜𝚍-𝙳𝚎𝚟𝚊 authors may adopt writing practices that reflect their ideological affiliation. The resultant variation is tolerated, and may even go unnoticed, by fluent readers. However, such variation may snowball into confusion and disenchantment for learners of 𝚜𝚍-𝙳𝚎𝚟𝚊. In the context of variation in 𝚖𝚛-𝙳𝚎𝚟𝚊, Deshpande (2016, p. 72) states that “seemingly trivial issues of signs and dots gradually emerge as part of larger ones about literacy, historicity, [and] community”. This statement appears to hold true for 𝚜𝚍-𝙰𝚛𝚊𝚋 and 𝚜𝚍-𝙳𝚎𝚟𝚊 as well.

7.6 Indian Institute of Sindhology’s system

In the twenty-first century, the Indian Institute of Sindhology’s (IIS) has attempted to constrain and regulate certain aspects of graphematic variation in 𝚜𝚍-𝙳𝚎𝚟𝚊 by means of its Standard Trilingual Dictionary (Sindhi – Hindi – English) (Rohra, Bijani, & Gurnani, 2011). The trilingual-triscriptal structure of this work is similar to that of Lekhwani (1996), with headwords listed in 𝚜𝚍-𝙳𝚎𝚟𝚊 and 𝚜𝚍-𝙰𝚛𝚊𝚋, and meanings provided in Hindi (𝚑𝚒-𝙳𝚎𝚟𝚊) and English (𝚎𝚗-𝙻𝚊𝚝𝚗). The graphematic rules listed in the dictionary (pp. xix–xx) are comprehensive enough to warrant them assigned a distinct language subtag: 𝚜𝚍-𝙳𝚎𝚟𝚊-𝚡-𝚒𝚒𝚜.

7.6.1 Graph inventory and allography

Although attempting to take a stance on certain areas of graphematic ambiguity, the graph inventory of 𝚜𝚍-𝙳𝚎𝚟𝚊-𝚡-𝚒𝚒𝚜 remains largely consistent with post-Partition 𝚜𝚍-𝙳𝚎𝚟𝚊 practice. For instance, in Sanskrit-origin words, 𝚜𝚍-𝙳𝚎𝚟𝚊-𝚡-𝚒𝚒𝚜 explicitly treats |ऋ ज्ञ क्ष ष| as graphematically equivalent to |रि ग्य क्श श| (Rohra, Bijani, & Gurnani, 2011, p. xix), making the two sets of allographs — or homophonous heterographs — in free variation. Accordingly, |ऋ ज्ञ क्ष ष| do not receive their own dictionary sections, although spellings involving these graphs are listed separately in headword entries as equivalent alternatives. A selection of such listings is shown in (17):

Furthermore, |ऋ ज्ञ क्ष ष| are not used to spell non-Sanskritic words. Thus, Sindhi [ɾikᶦɕa] ‘rickshaw’ is only spelt |रिक्शा| (p. 455).

Spellings with γ-consonant clusters are permitted in 𝚜𝚍-𝙳𝚎𝚟𝚊-𝚡-𝚒𝚒𝚜. However, universal rules are absent, and spellings seem to have been decided on an individual basis. Thus, 𝚜𝚍-𝙳𝚎𝚟𝚊-𝚡-𝚒𝚒𝚜 writes [pᶦjaɾᶷ] as |प्यार‌ु| instead of |पियारु| (pp. xix–xx), but [kabɪlᶦjətə] ‘capability’ as |क़ाबिलियत|, not |क़ाबिल्यत| (p. 98).

Regarding graphs for the Perso-Arabic phones [q x ɣ z f], 𝚜𝚍-𝙳𝚎𝚟𝚊-𝚡-𝚒𝚒𝚜 considers the nuqta-bearing graphs |क़ ख़ ग़ ज़ फ़| to be distinct from dotless |क ख ग ज फ|. Notably, words beginning with |ख़ ग़ ज़ फ़| receive their own sections, while words beginning with |क़| are listed under the heading for |क| and ordered after it. Thus, the entry |काफ़िर‌ु| [kafɪɾᶷ] ‘infidel’ appears before |क़ाफ़िलो| [qafɪlo] ‘convoy’, which in turn precedes |काफ़ी| [kafi] ‘enough’ (p. 98). The 𝚜𝚍-𝙰𝚛𝚊𝚋 graphs |ث ح ذ ص ض ط ظ ع غ| do not have canonical equivalents in 𝚜𝚍-𝙳𝚎𝚟𝚊, and 𝚜𝚍-𝙰𝚛𝚊𝚋 words featuring these graphs are sorted based on their 𝚜𝚍-𝙳𝚎𝚟𝚊 spelling.

A salient graphetic feature of 𝚜𝚍-𝙳𝚎𝚟𝚊-𝚡-𝚒𝚒𝚜 is the representation of [ɾʊ ɾu] as the compositionally transparent |र‌ु र‌ू| rather than the customary |रु रू|. While the latter forms are plausible within the graphematic solution space of 𝚜𝚍-𝙳𝚎𝚟𝚊, they are rarely seen in any prevalent Devanagari-based writing system.

7.6.2 Spelling and orthography

Gemination of γ-consonants is not marked in 𝚜𝚍-𝙳𝚎𝚟𝚊-𝚡-𝚒𝚒𝚜, and loanwords are (re)spelt to reflect this rule. Thus, 𝚑𝚒-𝙳𝚎𝚟𝚊 |सत्ता| [sət̪t̪a] becomes 𝚜𝚍-𝙳𝚎𝚟𝚊-𝚡-𝚒𝚒𝚜 |सता| [sət̪a] ‘political power’.

Indeclinable adjectives that are pronounced in the Sindhi New Variety chronolect with a final φ-consonant or epenthetic [ᵊ] are (re)spelt accordingly. Thus, the Sindhi words for ‘happy’, ‘clever’ and ‘interesting’, listed in Lekhwani (1996) as |ख़ुशि|, |चालाकु| and |दिलचस्पु|, are respelt in 𝚜𝚍-𝙳𝚎𝚟𝚊-𝚡-𝚒𝚒𝚜 as |ख़ुश|, |चालाक| and |दिलचस्प|, respectively, to better reflect their New Variety pronunciations as [xʊɕ(ᵊ)], [t͡ɕalak(ᵊ)] and [d̪ɪlᵊt͡ɕəsp(ᵊ)].

Certain nouns may also be respelt to better reflect their New Variety φ-vowel makeup. Thus, 𝚜𝚊-𝙳𝚎𝚟𝚊 |नीति| ‘policy, ethics’, realised in Sindhi as [nit̪i], is spelt in 𝚜𝚍-𝙳𝚎𝚟𝚊-𝚡-𝚒𝚒𝚜 as |नीती|. Conversely, the Sindhi word for ‘fragrance’, whose New Variety pronunciation is [xʊɕᵊbu], has been respelt from the hitherto prevalent |ख़ुशबूइ| as 𝚜𝚍-𝙳𝚎𝚟𝚊-𝚡-𝚒𝚒𝚜 |ख़ुशबू|. That said, 𝚜𝚍-𝙳𝚎𝚟𝚊-𝚡-𝚒𝚒𝚜 is discretionary in graphematically indicating noun gender. Native or assimilated nouns whose Old Variety pronunciation and spelling comprised a final lax vowel indexing grammatical gender (§4.4) are not respelt to match their New Variety pronunciations. Thus, Old Variety [d̪ɪlᶦ] ‘heart’ and [kɪt̪abᶷ] ‘book’, traditionally spelt |दिलि| and |किताबु|, are not respelt in 𝚜𝚍-𝙳𝚎𝚟𝚊-𝚡-𝚒𝚒𝚜 to reflect their New Variety pronunciations [d̪ɪl(ᵊ)] and [kɪt̪ab(ᵊ)], respectively. Their spellings retain the final bound γ-vowels |⁠ि ⁠ु| that index feminine and masculine gender, respectively. On the other hand, unassimilated English-origin nouns for inanimate referents, which are pronounced in the New Variety with a final φ-consonant or epenthetic [ᵊ], are spelt in 𝚜𝚍-𝙳𝚎𝚟𝚊-𝚡-𝚒𝚒𝚜 with a final γ-consonant without a virama. Thus, English ‘fashion’, pronounced [fɛɕən(ᵊ)] in the New Variety, is spelt |फ़ैशन|, with final |न| [n(ə)]. A comparison of 𝚜𝚍-𝙳𝚎𝚟𝚊-𝚡-𝚒𝚒𝚜 spellings with their equivalents in Lekhwani (1996) is shown in (18):

Words featuring [ɛ ɔ] are consistently spelt in 𝚜𝚍-𝙳𝚎𝚟𝚊-𝚡-𝚒𝚒𝚜 with |⁠ै ⁠ौ|. Thus, [pɛso] ‘money’ is |पैसो| and [t͡ɕɔk(ᵊ)] ‘square, intersection’ is |चौंकु|, with their 𝚜𝚍-𝙰𝚛𝚊𝚋 equivalents listed as |پَسو| and |چَونڪُ|. By doing so, 𝚜𝚍-𝙳𝚎𝚟𝚊-𝚡-𝚒𝚒𝚜 deprecates the previously-used alternative spellings |पइसो پَئِسو| and |चउंकु چَؤنڪُ|, respectively (see Figure 6.5). Words whose New Variety pronunciation no longer features [ɛ ɔ] are respelt accordingly. Thus, the word for ‘education’, pronounced [t̪ɛlimᵊ] in the Old Variety and spelt |तइलीम| (Lekhwani, 1996, p. 80), tends to be pronounced [t̪alim(ᵊ)] in the New Variety (see also §6.5.4). Accordingly, 𝚜𝚍-𝙳𝚎𝚟𝚊-𝚡-𝚒𝚒𝚜 spells the word as |तालीम| to reflect the latter pronunciation.

7.6.3 Collation order

Although headwords in the IIS’ dictionary are listed in 𝚜𝚍-𝙳𝚎𝚟𝚊-𝚡-𝚒𝚒𝚜 as well as in 𝚜𝚍-𝙰𝚛𝚊𝚋, they are ordered according to the following Devanagari-based collation order (Rohra, Bijani, & Gurnani, 2011, p. xxiv):

This collation order is similar to the one used in Lekhwani (1996, p. vi). Graphs in parentheses do not have their own dictionary sections. While the situation of |क़| has been described earlier, the other parenthesised graphs do not occur word-initially in 𝚜𝚍-𝙳𝚎𝚟𝚊-𝚡-𝚒𝚒𝚜. Words containing anusvara are sorted thus: |अंगु| [əŋɡᶷ] ‘number’ → |अंसु| [ənsᶷ] ‘component’ → |अगरि| [əɡəɾᶦ] ‘if’ → |असुली| [əsᶷli] ‘real’ (pp. 1–26). Graphs based on the same rasm are sorted with the graphetically basic form first, followed by the underlined or underdotted form. If a rasm has an underlined as well as an underdotted counterpart, there appears to be no pattern on which one occurs first, as seen in the sequences |क क़|, |ख ख़|, |ग ग़ ॻ|, |ज ॼ ज़|, |ड ड़ ॾ|, |ढ ढ़|, |फ फ़| and |ब ॿ|.

7.7 Analysis

7.7.1 Graphematic typology

The description of 𝚜𝚍-𝙳𝚎𝚟𝚊’s gradual evolution since the early nineteenth century reveals that, despite various developments and refinements in its graphematic makeup, the writing system has continued to be abugidic and alphasyllabic in nature. In this regard, 𝚜𝚍-𝙳𝚎𝚟𝚊’s typology has remained identical to that of its founding systems, 𝚑𝚒-𝙳𝚎𝚟𝚊 and 𝚜𝚊-𝙳𝚎𝚟𝚊 (§7.1; see also Figure 2.3). Moreover, 𝚜𝚍-𝙳𝚎𝚟𝚊’s conceptual Swarakhadi-style free γ-vowel allographs|अ आ अि अी अु अू अे अै ओ औ|, with |अ| as a common base, are compositionally near-identical to 𝚜𝚍-𝙰𝚛𝚊𝚋’s free γ-vowels, which use alif |ا| as a common base. This graphetic and graphematic commonality between 𝚜𝚍-𝙳𝚎𝚟𝚊 and 𝚜𝚍-𝙰𝚛𝚊𝚋 further reinforces the alphasyllabic nature of both systems (§6.6.1).

7.7.2 Graph inventory

Following the description of 𝚜𝚍-𝙳𝚎𝚟𝚊’s evolution over the years, including additions and modifications to its graph inventory, the focus turns to the graphematic categorisation of the inventory’s elements. When it comes to identifying rasms, the most evident ones are the bound γ-vowels |⁠ा ⁠ि ⁠ी ⁠ु ⁠ू ⁠े ⁠ै|, which canonically denote the Sindhi vowels [a ɪ i ʊ u e ɛ], respectively. Since they are minimal, contrastive and have an identifiable linguistic value, these bound γ-vowels also appear to qualify as graphemes in 𝚜𝚍-𝙳𝚎𝚟𝚊. In contrast, the bound γ-vowels |⁠ो ⁠ौ| [o ɔ] are not graphetically minimal, being composed of the elements |⁠ा| and |⁠े| or |⁠ै|. As a result, |⁠ो ⁠ौ| fail to qualify as graphemes in 𝚜𝚍-𝙳𝚎𝚟𝚊. Indeed, Meletis (2020, p. 102 footnote 112) uses similar reasoning in the context of graphetically complex γ-vowels in the Tamil (𝚝𝚊-𝚃𝚊𝚖𝚕) writing system, arguing that such γ-vowels do not constitute graphemes based on the tripartite definitional criteria on hand.

Deciding the status of other graphs in the inventory is less straightforward. A look at Table 7.1 suggest that certain graphs, such as the free γ-vowels |अ इ उ ए| and the free γ-consonants |ङ ट ठ ड ढ र| [ŋ(ə) ʈ(ə) ʈʰ(ə) ɖ(ə) ɖʱ(ə) ɾ(ə)] seem likely candidates for grapheme status. A handful of rasms, such as the free γ-consonants such as |क फ| [k(ə) pʰ(ə)] and their associated bound or ‘half’ forms |क्‍◌ फ्‍◌| [k pʰ], also fulfil the grapheme criteria individually. In fact, all bound γ-consonant forms — whether graphetically predictable forms like |ब्‍◌| [b] or unexpected ones like |र्◌| [ɾ] — appear to qualify as graphemes in themselves, since they are minimal, contrastive and have linguistic value. The majority of 𝚜𝚍-𝙳𝚎𝚟𝚊’s ‘full’ γ-consonants, though, appear to be graphetically composed of their associated half form and the element |⁠ा|.¹¹⁰ Thus, |ब| [b(ə)] may be interpreted as a graphetic combination of |ब्‍◌| and |⁠ा|. However, bound γ-consonants and |⁠ा| have already been identified as potential graphemes themselves. As a result, those free γ-consonants in the 𝚜𝚍-𝙳𝚎𝚟𝚊 inventory that graphetically comprise a bound γ-consonant and |⁠ा| would fail the minimality test and emerge as nongraphemes. The matter is further complicated by the fuzziness around the presence of φ-consonant clusters in Sindhi phonology (§4.3.1) and, consequently, the tenability of using bound or half forms of γ-consonants in 𝚜𝚍-𝙳𝚎𝚟𝚊.

Other areas of uncertainty include the graphs |इ ई| [ɪ i]. By itself, |इ| is likely a grapheme in 𝚜𝚍-𝙳𝚎𝚟𝚊. However, |ई| may be considered a graphetic fusion of the graphemes |इ| and |र्◌⁠| [ɾ], in the process failing the minimality criterion. At the same time, both |उ ऊ| [ʊ u] would satisfy all three grapheme criteria since the tail-like element on |ऊ| is not a grapheme by itself. Although the tail-like elements on |ई| and |ऊ| are not identical, they are, nevertheless, very similar in visual appearance and in their function as distinguishing elements compared to |इ| and |उ|, respectively. Faced with this line of reasoning, can one justifiably reconcile granting grapheme status to |ऊ| while denying it to |ई|, particularly when the visual similarity of the tail-like element on |ई| with |र्◌⁠| [ɾ] is likely a coincidental outcome of typographical modularisation?

The analysis presented here suggests that the tripartite defining criteria for a grapheme must continue to be applied to and tested across several writing systems. No doubt, Meletis’ proposed criteria constitute a significant step forward in ‘reclaiming’ the concept of a grapheme. Nevertheless, further testing and refinement are essential for the criteria to remain theoretically sound and capable of delivering consistent and plausible verdicts on identifying graphemes across a diverse range of writing systems. As was the case with 𝚜𝚍-𝙰𝚛𝚊𝚋, the current state of grapholinguistic theory does not allow us to conclusively determine the graphemes in the 𝚜𝚍-𝙳𝚎𝚟𝚊 inventory. Although seemingly anti-climactic, such an outcome is just as well, since the status of complex graphs in 𝚜𝚍-𝙳𝚎𝚟𝚊, including so-called ligatures, presents much food for thought for ongoing holistic evaluations in this regard.

7.7.3 Graphematics and orthography

In Table 7.3 , the 𝚜𝚍-𝙳𝚎𝚟𝚊 complex graphs |क्त ष्ट्र| shine the spotlight on the concept and scope of ligatures Devanagari-based writing systems, including their mandatory or optional nature (§2.7). In 𝚜𝚊-𝙳𝚎𝚟𝚊 and 𝚑𝚒-𝙳𝚎𝚟𝚊, the phonological sequences [kt̪(ə) ɕʈɾ(ə)] may be written either as |क्त ष्ट्र| or as |क्‍त ष्‍ट्र|, with the only implications being stylistic or semiotic in nature (see Example (12)). On the other hand, 𝚜𝚊-𝙳𝚎𝚟𝚊 and 𝚑𝚒-𝙳𝚎𝚟𝚊 require that the phonological sequences [ɕɾ(ə) kɕ(ə)] always be written |श्र क्ष|, and not as |श्‍र क्‍ष|. Viewed from the lens of the Modular Theory, the graphematic solution spaces of 𝚜𝚊-𝙳𝚎𝚟𝚊 and 𝚑𝚒-𝙳𝚎𝚟𝚊 license |श्र क्ष| as well as |श्‍र क्‍ष| as potential representations of [ɕɾ(ə) kɕ(ə)]. However, the implicit orthographies of these writing systems constrain the available solutions and mandate |श्र क्ष| as the only ‘correct’ choice. Some of these writing practices and constraints, such as those governing |श्र क्ष|, have been adopted more or less unchanged into post-Partition 𝚜𝚍-𝙳𝚎𝚟𝚊. At the same time, certain 𝚜𝚊-𝙳𝚎𝚟𝚊 and 𝚑𝚒-𝙳𝚎𝚟𝚊 traditions have fallen away or been tacitly weakened in 𝚜𝚍-𝙳𝚎𝚟𝚊. For instance, 𝚜𝚊-𝙳𝚎𝚟𝚊 and 𝚑𝚒-𝙳𝚎𝚟𝚊 consider |ज्ञ ग्य| to be graphematically distinct and not interchangeable, whereas 𝚜𝚍-𝙳𝚎𝚟𝚊 treats them as freely varying allographs for [ɡj(ə)] (§7.5.1).

The advent of 𝚜𝚍-𝙳𝚎𝚟𝚊-𝚡-𝚒𝚒𝚜 and its guidelines has resulted in the legitimisation of certain graphs and graph sequences previously considered inauthentic. These include the compositionally more transparent sequences |क्‍ष| and |र‌ु र‌ू| as alternatives for |क्ष| [kɕ(ə)] (see Example (17)) and |रु रू| [ɾʊ ɾu]. In addition, |क्‍ष| is optional — albeit preferred — while |र‌ु र‌ू| is across the board. These developments reiterate the observation in Section 2.7, that the concept of ‘ligature’ is primarily graphematic in nature and based on the homophony or phonological equivalence between two graphs or graph sequences. Graphetically speaking, a ‘ligature’ may turn out to be simpler than its compositionally transparent equivalents; a deconstruction of |क्ष| and |क्‍ष| reveals that the former comprises the rasms |क्ष्‍◌ ⁠ा| and the latter |क्‍◌ ष्‍◌ ⁠ा|. This reality sits somewhat uneasily with the popular interpretation of a ligature as a supposedly complex combination of graphs. In addition, the fact that such rulings on ligatures were conceived of and promulgated in a top-down manner by a language body — the IIS — suggests that ligatures owe their existence in part to the orthographic module. Moreover, the fact that 𝚜𝚍-𝙳𝚎𝚟𝚊-𝚡-𝚒𝚒𝚜 considers some ligatures optional and others mandatory further strengthens the idea that assigning ‘ligature’ status to a graph is heavily conditioned by the graphematic and orthographic modules of a writing system.

The discussion on orthographic module brings us to the topic of 𝚜𝚍-𝙳𝚎𝚟𝚊-𝚡-𝚒𝚒𝚜’s credentials. As the product of a language body featuring a codified set of writing rules, 𝚜𝚍-𝙳𝚎𝚟𝚊-𝚡-𝚒𝚒𝚜 may be considered a conventional orthography as defined in Section 2.9. Although promulgated in a top-down manner, 𝚜𝚍-𝙳𝚎𝚟𝚊-𝚡-𝚒𝚒𝚜 does not introduce any graphematic forms or spellings not already licensed by 𝚜𝚍-𝙳𝚎𝚟𝚊’s graphematic solution space. Although 𝚜𝚍-𝙳𝚎𝚟𝚊-𝚡-𝚒𝚒𝚜 may still leave specific questions answered, the availability of an accompanying dictionary as a reference greatly mitigates the number and extent of spelling-related grey areas.

If 𝚜𝚍-𝙳𝚎𝚟𝚊-𝚡-𝚒𝚒𝚜 is deemed an orthography, why is the same label not assigned to 𝚜𝚍-𝙰𝚛𝚊𝚋? Considering that the IIS trilingual-triscriptal dictionary (Rohra, Bijani, & Gurnani, 2011) features Sindhi-language headwords in 𝚜𝚍-𝙳𝚎𝚟𝚊 as well as 𝚜𝚍-𝙰𝚛𝚊𝚋 (§7.6), is there an argument to be made for the existence of a 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚒𝚒𝚜? If yes, what about a potential 𝚜𝚍-𝙳𝚎𝚟𝚊-𝚡-𝚜𝚕𝚊, considering that the SLA in Pakistan performs a role very similar to that of the IIS in India? The answers to these questions will depend on the precise definition of orthography, as well as the related concept of orthography reform. According to Neef (2021), the existence of an orthography implies the availability of a “codified and binding norm” of graphematic rules and spellings, while orthographic reform presupposes “[a]n authority that has the power to change the norm”. As it turns out, the guidelines listed by the IIS in its trilingual-triscriptal dictionary pertain only to 𝚜𝚍-𝙳𝚎𝚟𝚊, and not to 𝚜𝚍-𝙰𝚛𝚊𝚋. Although the guidelines make numerous references to 𝚜𝚍-𝙰𝚛𝚊𝚋 as a point of comparison, no recommendations are made on 𝚜𝚍-𝙰𝚛𝚊𝚋 spelling rules. Per Neef’s conceptualisation of orthography, the absence of a codified set of rules targeting 𝚜𝚍-𝙰𝚛𝚊𝚋 leaves insubstantial evidence for the existence of a 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚒𝚒𝚜 orthography.

Along similar lines, the absence of an identifiable set of codified norms governing 𝚜𝚍-𝙰𝚛𝚊𝚋 undermines the argument in favour of a 𝚜𝚍-𝙳𝚎𝚟𝚊-𝚡-𝚜𝚕𝚊 orthography, while also precluding the existence of reforms. Whereas the SLA has implicitly regulated variation in 𝚜𝚍-𝙰𝚛𝚊𝚋 over the decades through its dictionaries and publications, the institution does not appear to have disseminated a consolidated set of rules governing 𝚜𝚍-𝙰𝚛𝚊𝚋 graphematics and spelling. Evidently, stating that the SLA has yet to set out orthographic norms for 𝚜𝚍-𝙰𝚛𝚊𝚋 strikes one as disingenuous considering the copious amounts of 𝚜𝚍-𝙰𝚛𝚊𝚋 material published under the SLA’s auspices. This palpable incongruity reiterates the earlier observation on the existence of an orthography depending on one’s individual understanding of the concept, and the consequent need for a comprehensive definition of the term.

7.7.4 Sociolinguistics

Besides the existence of an identifiable norm, Neef’s conceptualisation of orthography and its reform also presupposes the existence of a language body that has authority and power over the norm and its modifications. Whether the language body is a governmental or semi-governmental agency or a private lexicographer, the authority it commands is crucial in determining the success of its orthographies. This brings us to examining the authority that language bodies such as the IIS have, and how such authority may be reasonably measured or assessed. As it turns out, reliable tools and frameworks for evaluating the authority of a language body and its orthographies are scant. Then again, it seems uncontroversial to state that 𝚜𝚍-𝙳𝚎𝚟𝚊-𝚡-𝚒𝚒𝚜 is not binding in any manner. In fact, knowledge of this ‘norm’ appears low among lay Sindhis in India, evinced by the persistence of unstandardised 𝚜𝚍-𝙳𝚎𝚟𝚊 practices on social media and messaging apps even after a decade of 𝚜𝚍-𝙳𝚎𝚟𝚊-𝚡-𝚒𝚒𝚜’s release. The situation is unsurprising, since the emic relegation of 𝚜𝚍-𝙳𝚎𝚟𝚊 usage primarily to unofficial domains of use takes away any educational, legal or social incentive to follow specific orthographic norms. Thus, the perceived authority of 𝚜𝚍-𝙳𝚎𝚟𝚊-𝚡-𝚒𝚒𝚜 is undermined by the sociolinguistic inconspicuousness of 𝚜𝚍-𝙳𝚎𝚟𝚊 writing in the Indian Sindhi graphosphere.

In sum, 𝚜𝚍-𝙳𝚎𝚟𝚊-𝚡-𝚒𝚒𝚜 emerges as a reasonably lucid and comprehensive set of graphematic guidelines that, nevertheless, lacks sociolinguistic authority. From an orthographic perspective, the situation with the IIS and 𝚜𝚍-𝙳𝚎𝚟𝚊 is the converse of that involving the SLA and 𝚜𝚍-𝙰𝚛𝚊𝚋. While the former falls short in terms of authority, the latter does so with regard to codification. This finding prompts one to ask: what kind of relative weightage should the components of codification and authority have when deciding whether a collection of graphematic norms deserves the label ‘orthography’?

☙ ❧

8 Khudawadi

In the twenty-first century, written Sindhi is most often associated with the Arabic and Devanagari scripts. However, in the mid-nineteenth century, what locals and Europeans considered the “Sindhi” script were the Landa inventories prevalent in the region (Stack, 1849a, p. 2; Trumpp, 1858, p. ii). By extension, the moniker ‘Sindhi’ was also applied to the proto-standardised and best-known offshoots of the Landa inventories of Sindh — Khojki and Khudawadi (§5.1.4). Inadvertently reinforcing this association is the four-letter ISO 15924 script code for Khudawadi — 𝚂𝚒𝚗𝚍 (Pandey, 2010f). Consequently, the language subtag for the Sindhi-Khudawadi writing system is 𝚜𝚍-𝚂𝚒𝚗𝚍.

As suggested by Figure 5.1, the proto-standardised Landa inventories of Punjab and Sindh formed a geographically-conditioned graphetic continuum, ranging from Gurmukhi in the north to Khojki in the south. In the geographical and graphetic middle lay the Multani script of southern Punjab, and Khudawadi. Multani was the proto-standardised Landa inventory used to write the Siraiki language (𝚜𝚔𝚛-𝙼𝚞𝚕𝚝). The linguistic closeness of the Siraiki and Sindhi languages, the graphetic likeness of the Multani and Khudawadi inventories, and the typological similarity of the 𝚜𝚔𝚛-𝙼𝚞𝚕𝚝 and 𝚜𝚍-𝚂𝚒𝚗𝚍 writing systems as used in the mid-nineteenth century led to the two often being depicted alongside each other in nineteenth century European-authored works (Bagster, 1851, p. 12; Faulmann, 1880a; Taylor, 1883, pp. 338–339).

By 1870, however, there emerged what was claimed to be a new-and-improved avatar of 𝚜𝚍-𝚂𝚒𝚗𝚍, which was typologically very distinct from its predecessor. Even so, the new variant of 𝚜𝚍-𝚂𝚒𝚗𝚍 was rejected by its target population in favour of the previous variant. Hence, 𝚜𝚍-𝚂𝚒𝚗𝚍 represents a noteworthy example of a transparent writing system being rejected in favour of an opaque one. This makes 𝚜𝚍-𝚂𝚒𝚗𝚍 and its variants worthy subjects of further graphematic and sociolinguistic investigation.

8.1 Graphematic foundations and early use

Compared to 𝚜𝚍-𝙰𝚛𝚊𝚋 or 𝚜𝚍-𝙳𝚎𝚟𝚊, which are ultimately based on the relatively transparent systems of 𝚊𝚛-𝙰𝚛𝚊𝚋 and 𝚜𝚊-𝙳𝚎𝚟𝚊, respectively, early unstandardised variants of 𝚜𝚍-𝚂𝚒𝚗𝚍 were comparatively opaque. That said, the opacity of unstandardised 𝚜𝚍-𝚂𝚒𝚗𝚍 was distinct from the kind caused by discretionary omission of subsegmental graphs in 𝚜𝚍-𝙰𝚛𝚊𝚋. As a general rule, all variants of 𝚜𝚍-𝙰𝚛𝚊𝚋 have graphematic provisions to ensure a high degree of phonological transparency. However, writers often choose not to harness the full graphematic repertoire available. In contrast, early versions of 𝚜𝚍-𝚂𝚒𝚗𝚍 simply lacked the requisite provisions for reflecting spoken Sindhi in a transparent manner. Put differently, the opacity of 𝚜𝚍-𝙰𝚛𝚊𝚋 was and is sociolinguistically conditioned, while that of early 𝚜𝚍-𝚂𝚒𝚗𝚍 was graphematically conditioned. The opaque variant of 𝚜𝚍-𝚂𝚒𝚗𝚍 was the only one in popular use until 1869 (§8.2), based on which it has been assigned the language subtag 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍. Table 8.1 shows the most common 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍 phonograms reproduced from Stack’s (1849a) handwritten chart and Bagster’s (1851, p. xliv) plate. Graph-phone correspondences in the table follow Stack (1849a), with Bagster’s graphs included primarily for graphetic comparison.

Stack (1849a)	Bagster (1851)	Modern typeface	Phonological value(s) in IPA
		𑊰	ə a ɪ i ʊ u e ɛ o ɔ
		𑊲	ɪ i ʊ u e ɛ o ɔ
not listed		𑊴	ɪ i ʊ u e ɛ o ɔ
		𑊺	k (q)
		𑊻	kʰ (x)
		𑊼	ɡ ɡʱ (ɣ)
		𑊽	ɠ
	not listed	𑊿	ŋ ɲ
		𑋀	t͡ɕ
		𑋁	t͡ɕʰ
		𑋂   𑋘	d͡ʑ d͡ʑʱ (j) (z)
		𑋃	ʄ
		𑋅	ŋ ɲ
		𑋆	ʈ
same as above		–	ʈʰ
		𑋈	ʈɾ ɖɾ
	not listed	𑋊   ॥	ɽ (ɽʱ) ɾ (ɾʱ)
	not listed	𑋋	ɖ ɖʱ
		𑋶	ɗ
		𑋌	ɳ (ɽ̃)
		𑋍	t̪
		𑋎	t̪ʰ
		𑋏	d̪ d̪ʱ
same as above		–	(d̪ʱ)
		𑋑	n
		𑋒	p
		𑋓	pʰ (f)
		𑋔	b bʱ
		𑋕	ɓ
		𑋗	m
		𑋘	d͡ʑ d͡ʑʱ (j)
	not listed	॥	ɾ
under ‘Shikarpuri’		𑋙	ɾ
		𑋚	l
		𑋛	ʋ
		𑋝	s (ɕ)
		𑋞	ɦ

Graphetically, certain graphs in the 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍 inventory have cognates not just in other Landa-based inventories, but also in Devanagari-based inventories. Some of these cognate graphs in different inventories reflect historical phonological mergers. Thus, 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍 |𑊻| [kʰ] is graphetically cognate with 𝚜𝚊-𝙳𝚎𝚟𝚊 |ष| [ʂə], reflecting the realisation of Sanskritic [ʂ] as New Indo-Aryan [kʰ] (see §7.3, p. 199).

In terms of inventory, and like its Landa-derived sister writing systems, 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍 features only three γ-vowels: |𑊰 𑊲 𑊴|. Of these, only |𑊰| was used word-initially, irrespective of the φ-vowel in question (Coulmas, 1996a, p. 282; Stack, 1849a, pp. 2, 3 footnote). Crucially, there were no bound γ-vowel allographs. While the second member of a φ-[VV] sequence was occasionally — and idiosyncratically — written with any one of |𑊰 𑊲 𑊴|, medial and final φ-vowels occurring in a φ-[CV] sequence were almost never written. In this regard, Pandey’s (2012, p. 2) summary of the 𝚜𝚔𝚛-𝙼𝚞𝚕𝚝 writing system is entirely applicable to 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍:

[Siraiki-Multani] is closer in structure to an abjad than an alphasyllabary. There is no VIRAMA. There are […] no dependent vowel signs. [Graphematic] Vowels are generally not written unless they occur in isolation, in word-initial position, or in the final position of monosyllabic words. [Graphematic] Consonants theoretically possess the inherent [φ-]vowel /a/, but as [γ-]vowels are not explicitly marked, the actual syllabic value of a [γ-]consonant in running text is ambiguous and must be inferred from context. […] there are no [γ-]conjuncts.

(Pandey, 2012, p. 2)

Thus, Pandey characterises 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍 as structurally abjadic, aside from the occasional use of free γ-vowels in non-initial position. He describes the latter practice as “alphabetic” (Pandey, 2011c, pp. 2–3). He further opines that the opacity of 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍 writing requires the reader to decode it at the morphological level (Pandey, 2011c, p. 3). In several instances, graphematic and sociolinguistic context may also be required. The need for additional context to accurately decipher abjadic 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍 writing is succinctly illustrated by Stack (1849a, p. 2) using the graph sequence |𑊼𑋚|:

Thus 𑊼𑋚 might stand for गोलो [ɡolo], गोली [ɡoli], गुलु [ɡʊlᶷ], गलो [ɡəlo], गेलि [ɡelᶦ], गिलो [ɡɪlo], गिला [ɡɪla], गोलु [ɡolᶷ], गुलू [ɡʊlu], or for any other word that might happen to be in use, formed by pronouncing these two [γ-]Consonants with any of the vowel sounds.¹¹¹

(Stack, 1849a, p. 2)

Faulmann (1880a) compares the absence of bound γ-vowels in 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍 and 𝚜𝚔𝚛-𝙼𝚞𝚕𝚝 with the conventional omission of subsegmental graphs in writing systems based on the Arabic and Hebrew scripts:

Die vorstehenden Schriften sind […] beachtenswert, […] weil sie in gleicher Weise wie die semitischen Schriften die Vokale in der Mitte der Wörter nicht schreiben, z. B. Sindisch 𑋍𑋞𑋘 tuhidžo, […]

These scripts [𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍 and 𝚜𝚔𝚛-𝙼𝚞𝚕𝚝] are […] noteworthy for their omission of word-medial [bound γ-]vowels exactly as the Semitic scripts do, e.g., Sindhi 𑋍𑋞𑋘 tuhidžo [t̪ʊ̃ɦᶦ̃d͡ʑo] […]

(Faulmann, 1880a, p. 121)

In a subsequent work, Faulmann (1880b, pp. 454–455) also compares the three free vowel graphs of the Khudawadi and Multani inventories with the Hebrew-Hebrew (𝚑𝚎-𝙷𝚎𝚋𝚛) inventory of |ו י א| [a i u]. On this basis, Faulmann portrays writing systems based on the Khudawadi and Multani inventories as a bridge between Semitic and Indic systems. However, and as mentioned at the start of this section, it is necessary to distinguish the absence of subsegmental graphs in the Semitic-origin 𝚜𝚍-𝙰𝚛𝚊𝚋 from the superficially similar phenomenon in the Indic-origin 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍. First, in 𝚜𝚍-𝙰𝚛𝚊𝚋, only a subset of subsegmental γ-vowels — those denoting [ə ɪ ʊ] — are sociolinguistically omissible. All other γ-vowels need to be written, else the text would be viewed as incorrect or unacceptable. In contrast, 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍 did not possess bound γ-vowels at all. Even the three free γ-vowels in its inventory, |𑊰 𑊲 𑊴|, were sociolinguistically omissible in medial and final position. More importantly, 𝚜𝚍-𝙰𝚛𝚊𝚋 featured near-biunique correspondences between vowel graphs and phones. This meant that a 𝚜𝚍-𝙰𝚛𝚊𝚋 text could be quite transparent as long as the writer included all γ-vowels. In 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍, however, the number of γ-vowels in the script fell far short of the number of φ-vowels in spoken Sindhi. This made it graphematically impossible to write 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍 in a transparent manner, even if the writer so desired. Hence, the underlying factors conditioning the absence of γ-vowels were distinct in the Semitic 𝚜𝚍-𝙰𝚛𝚊𝚋 and the Indic 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍, despite the superficial similarities of their end results.

The number of graphs in 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍’s inventory was also insufficient for unambiguously representing Sindhi’s φ-consonants (see Table 8.1). Some of these cases of apparent graphematic underdifferentiation were phonologically justifiable. For instance, the lack of contrast in spoken Sindhi between [ɽ ɾ] and their aspirated counterparts [ɽʱ ɾʱ] (§4.3.1) would have likely prompted the interchangeable use of |𑋊 ॥|. A similar absence of phonemic distinction between [s ɕ] in native Sindhi vocabulary at the time (Trumpp, 1872, p. xvii) would have rendered distinct graphs for these phones redundant, with |𑋝| sufficing for the purpose. The phonological merger of Sanskritic [j] into Middle Indo-Aryan [d͡ʑ] and the consequent loss of a distinctive [j] in the latter group of languages (Grierson, 1881, p. 3; Trumpp, 1872, p. xxviii) would have prompted the use of |𑋘| for [d͡ʑ] and [j] alike. Also attributable to historical-etymological reasons is the emergence of |𑋈| for [ʈɾ ɖɾ]. Trumpp (1858, p. iii) traces the origin of Sindhi [ʈɾ] to Sanskrit [t̪ɾ], and, in doing so, demonstrates the relation of Khudawadi |𑋈| to Devanagari |त्र| (Sanskrit [t̪ɾə]). Bagster (1851, p. 12), too, lists |𑋈| [ʈɾ ~ t̪ɾ] separately from |𑋆| [ʈ] and |𑋍| [t̪]. What remains unclear, though, is why |𑋈| was also used for [ɖɾ].

Notably, 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍 typically features only one graph for each pair of voiced φ-plosives — [ɡ ɡʱ], [d͡ʑ d͡ʑʱ], [ɖ ɖʱ], [d̪ d̪ʱ] and [b bʱ] — but assigns dedicated graphs to each of the φ-implosives [ɠ ʄ ɗ ɓ] (Stack, 1849a, p. 1). Such a distribution was also seen in other unstandardised Landa-based scripts, including Multani and Khojki (see Chapter 10). Masica (1991) traces this phenomenon to a supposed a recalibration of canonical phone-graph correspondences:

[…] it would appear that in the various mercantile scripts of Laṇḍā type used in Sind, the distinctive Sindhi [φ-]implosives were written with the symbol originally standing for the voiced plosive, while the ordinary voiced and aspirate voiced plosives (g/gh, etc.) were both written with another set of symbols and not clearly distinguished from one another.

(Masica, 1991, p. 470)

Furthermore, in a homorganic nasal-stop φ-consonant cluster such as [mb], the nasal element was usually not written. Likewise, the nasalisation of φ-vowels was also graphematically unmarked. Again, such practice was in line with those observed in Landa-script-based writing systems, as attested by Pandey in the context of Marwari-Mahajani (𝚖𝚠𝚛-𝙼𝚊𝚑𝚓) (Pandey, 2011c, p. 3):

[Phonological] Nasalization is not represented using special signs, such as ANUSVARA. The letter 𑅧 NA is used in cases where nasalization is explicitly recorded. In several cases, words are written simply with [γ-]nasalization deleted, eg. Devanagari हुंडी huṃḍī and Mahajani 𑅱𑅠 hḍ.

(Pandey, 2011c, p. 3)

An early specimen of 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍 is found in Prinsep’s (1837) review of Wathen’s Sindhi grammar (§5.1), on a facsimile of a Sindhi hundi — an indigenous bill of exchange. The 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍 text on the hundi, along with Prinsep’s auxiliary transcriptions into 𝚜𝚍-𝙳𝚎𝚟𝚊 and 𝚜𝚍-𝙻𝚊𝚝𝚗 are reproduced in Figure 8.1.

The 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍 text in Figure 8.1, accompanied by an IPA transcription and gloss based on Prinsep’s (1837) interpretation, is shown in (19):¹¹²

This example of a 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍 text succinctly illustrates the need to rely on graphematic and sociolinguistic context to unpack its meaning, without which the text assumes a cryptic nature. As shown in (19), Prinsep (1837, p. 352) is able to decode and decipher the first line of the salutation and the countersign. However, he is unable to offer a cogent decipherment of the second line, offering instead this explanation:

The second line has probably a meaning also, but not a single word of it can be found in [Wathen’s] vocabulary; nor can the natives be persuaded to divulge it, whether from superstitious prejudice or from ignorance.

(Prinsep, 1837, p. 353)

The 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍 text on the hundi also exemplifies several of the characteristics outlined in the previous section. As Figure 8.1 reveals, a postconsonantal φ-vowel, i.e., one in a φ-[CV] sequence, is not marked in writing. In a φ-[VV] sequence, the second φ-[V] is unpredictably denoted with |𑊰 𑊲 𑊴|. A voiced φ-plosive and its aspirated version are indicated by the same graph, while φ-implosives have distinct graphs. Nasalisation of φ-vowels and the nasal component in a homorganic nasal-oral φ-consonant cluster are not indicated. Also absent are any regular patterns of word spacing or punctuation.

One of the largest extant corpora of early 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍 writing is found in an unpublished handwritten wordlist compiled by British orientalist and administrator Edward Backhouse Eastwick. The wordlist comprises a list of English headwords and their Sindhi equivalents transcribed in 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍, accompanied by an 𝚎𝚗-𝙻𝚊𝚝𝚗-influenced 𝚜𝚍-𝙻𝚊𝚝𝚗 transcription. In a later publication (Eastwick, 1883, p. 63 footnote), the author claims his wordlist contains approximately 2200 entries. He also states its completion date as 1840, although the handwritten manuscript of the worldlist is dated 1843. A few years after its completion, Eastwick (1849, p. 119) sheds further light on the preparation of his wordlist or “Vocabulary”:

In preparing the Vocabulary, I, of course, had to learn the [Khudawadi] character, which has this peculiarity, that only initial [γ-]vowels (with a very few exceptions,) are written: consequently, there is the greatest difficulty in decyphering [sic] writings, for only the [γ-]consonants appear, and you must insert the [φ-]vowels as you think will best suit the sense. Thus, the word Pirín [pᶦɾĩ], “beloved,” is written exactly like pare [pəɾe], “beyond,” for only the p and the r are represented [𑋒𑋙]. The consequence of such an elliptical mode of writing is, that even the natives make egregious blunders in extracting the pith of the queer little epistles with which their correspondents favour them.

(Eastwick, 1849, p. 119)

Eastwick’s wordlist also groups the Sindhi terms into the speech varieties of “Sár” and “Lár”, corresponding to upper and lower Sindh, respectively. However, the latter category of words is only sparsely populated. A selection of entries from Eastwick’s list, along with their present-day pronunciations in standard Sindhi, is shown in Table 8.2.

Table 8.2 shows that the 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍 entries in Eastwick (1843a) largely agree with the graph inventory and phonological values outlined in Table 8.1. Minor areas of divergence in Eastwick’s work include the transcription of [ɾ] with the Shikarpuri allograph |𑋙| (Grierson, 1919, p. 18; Stack, 1849a, p. 8), instead of the traditional Khudawadi graph |॥|. Minor graphetic variation is also seen in the shapes of |𑋅 𑊿| [ŋ ɲ], with Eastwick’s work depicting the dot element in these graphs with a circle (Grierson, 1919, p. 101). Overall, though, the 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍 spellings in Eastwick (1843a) largely correspond with their modern-day pronunciations, with unexpected spellings usually reflecting phonologically plausible variations. For instance, in present-day Sindhi, the word for ‘to know’ is pronounced [ʄaɳəɳᶷ]. Theoretically, the corresponding 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍 transcription of this pronunciation would be |𑋃𑋌𑋌|. In contrast, Eastwick’s wordlist (1843a, p. 33) provides the spelling |𑋃𑋙𑋌|. In Eastwick’s spelling, the presence of |𑋙| in place of medial |𑋌| can be easily explained as reflecting an idiolectal surface realisation [ʄaɾ̃əɳᶷ], which is consistent with the [ɽ ~ ɾ] alternation common in spoken Sindhi. An analogous alternation between [ɳ] and [ɽ̃] has also been attested (§4.3.1, Consonant allophony).

Graphematically, Table 8.2 affirms the abjadic practice prevalent in 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍, with most φ-vowels not marked in writing. Exceptions include the entries |𑋞𑊰| [ha] ‘yes’ and |𑋔𑊲| [bʱi] ‘also’, where the sole φ-vowel in these monosyllabic words is explicitly marked using a free γ-vowel. Typologically, such practice is effectively alphabetic (Pandey, 2011c, pp. 2–3).

Eastwick’s Sindhi wordlist was subsequently published in the Journal of the Asiatic Society of Bengal (Eastwick, 1843b; credited as ‘J. B. Eastwick’) but with the Sindhi text printed only in Roman. The omission of the 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍 is attributed to unavailability of metal types for the script (Eastwick, 1843b, p. 1). Half-a-century later, Eastwick’s Sindhi wordlist was republished, this time accompanied by a Punjabi one (Eastwick, 1883). However, this trilingual wordlist, too, was published entirely in the Roman script, despite Khudawadi and Gurmukhi types being available by then.

Looking back from a present-day perspective, one may be tempted to regard Eastwick’s wordlist as amateurish. Indeed, the work has received its share of criticism, with Hoernle (1885, p. 162) dismissing it as “a promiscuous collection of words”. Vindicating Hoernle’s description are entries like ‘penguin’, cited as “Píán 𑋒𑊰𑋑” in the handwritten manuscript (Eastwick, 1843a, p. 46) and as “peean” in the print version (Eastwick, 1843b, p. 14). Evidently, Eastwick did not consider it linguistically or ecologically anomalous to find a word for ‘penguin’ in the language of a subtropical, semi-arid region in the northern hemisphere. The Sindhi gloss for ‘penguin’ might be a misinterpretation of the Sindhi word for ‘pelican’, [peɳᶦ] (Mewaram, 1910, p. 160; Stack, 1855, p. 248), although this remains conjectural.

Despite its shortcomings, Eastwick’s vocabulary remains invaluable as a historical-grapholinguistic artefact. Moreover, the biscriptal listing of Sindhi entries in 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍 and 𝚜𝚍-𝙻𝚊𝚝𝚗 acts as a parallel text. Despite its anglicised nature and opaqueness, the 𝚜𝚍-𝙻𝚊𝚝𝚗 text often helps fill in graphematic and phonological gaps in the abjadic 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍 entries, especially when it comes to vowels.

At this stage, it is worth revisiting the apocryphal jokes on misinterpretations arising from abjadic 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍 writing described in Section 5.1.1. The examples cited are reiterated in (20), transcribed in the IPA as well as in 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍. To reiterate, the humour in these jokes hinges on the multiple ways in which the 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍 writing may be decoded. Evidently, such a supposition implies that all of the possible phonological interpretations are necessarily written identically in 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍. That is, the phonological strings mentioned in the jokes are homographic in 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍. However, a closer look at the 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍 transcriptions in (20) suggests otherwise, raising doubts on the graphematic plausibility of these jokes.

The transcriptions in (20) reveal that, in the first joke, the φ-word [puʈʰe] ‘cover (oblique)’ would be written |𑋒𑋆| or |𑋒𑋇| in 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍 (see Eastwick (1843a, p. 5) for the homographic cognate [pʊʈʰᶦ]). Further, [puʈ(ɾ)ə] ‘son (oblique)’ would likely appear as |𑋒𑋈| (Eastwick, 1843a, p. 53). This emerges from the fact that, in 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍, [ʈʰ] and [ʈ(ɾ)] were written with distinct graphs (see Table 8.1). Moreover, writers usually adhered to the graphematic distinction between the two. In the second joke, [əd͡ʑᶦmeɾᶦ] ‘to Ajmer’ would be spelt |𑊰𑋂𑋗𑋙|, and [əʄᶷ məɾi] ‘today (he/she) died’ as |𑊰𑋃𑋗𑋙|. Here, too, the φ-consonants [d͡ʑ] and [ʄ] would usually be graphematically distinguished in 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍 writing. Casting further doubt on the authenticity of these wordplays is the fact that similar witticisms are found in the context of other abjadic writing systems such as Hindi-Mahajani (Beames, 1872, p. 56).¹¹³ Hence, while amusing tales on the multiple interpretations of supposed homography in 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍 have become culturally iconic, their graphematic veracity remains dubious.

8.2 ‘Improved Hindu Sindhi’

Section 5.2 revealed the belief prevalent among several colonial-era European administrators and intellectuals that Hindu-Sikh Sindhis would not learn 𝚜𝚍-𝙰𝚛𝚊𝚋 due to religious prejudice and the Arabic script’s semiotic associations with Islam. Having witnessed the prevalence of 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍 among the mercantile Vania caste of Hindu-Sikh Sindhis, Barrow Ellis decided to standardise 𝚜𝚍-𝚂𝚒𝚗𝚍 and mandate it in government schools as an incentive for Vania children to attend. By 1869, a revised and standardised 𝚜𝚍-𝚂𝚒𝚗𝚍 graph inventory and graph-phone correspondences had been finalised under the supervision of then-Deputy Educational Inspector of Sindh, Narayan Jagannath Vaidya. This particular subvariant of Sindhi-Khudawadi will be referred to by the language subtag 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝟷𝟾𝟼𝟿 (§5.2.3).

Colonial sources often describe 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝟷𝟾𝟼𝟿 as “(improved) Hindu Sindhi” (Hughes, 1876, p. 373) or “(improved) Hindi Sindhi” (Grierson, 1919, p. 19). Pandey (2011a, p. 2) states that the adjective ‘Hindu’ or ‘Hindi’ refers to the Indic origin of the underlying Khudawadi script, while also distinguishing the writing system from 𝚜𝚍-𝙰𝚛𝚊𝚋 or ‘Arabic Sindhi’.

8.2.1 Phonograms

Table 8.3 shows the main phonograms in 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝟷𝟾𝟼𝟿 and their phonological values.

Compared to 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍 (Table 8.1), 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝟷𝟾𝟼𝟿’s graph inventory and graph-phone correspondences comprise several salient modifications. The most significant one was to assign the rasms used for γ-consonants the phonological value φ-[CV₀], where [V₀] was [ə]. This established 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝟷𝟾𝟼𝟿’s status as an abugida (see Figure 2.3) and distinguished it typologically from the abjadic 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍. Graph-phone correspondences were also regularised. The graph|𑋂| [d͡ʑə] was listed as distinct from |𑋘| [jə], which implicitly reinforced the contrastiveness of the φ-consonants they denoted. Similarly, |𑋈| [ɖə] was codified separately from |𑋋| [ɖʱə]. In addition, [ʈɾə ɖɾə ɖʱɾə] were considered allophones of [ʈə ɖə ɖʱə], respectively, based on which they were simply written |𑋆 𑋈 𑋋|. This meant the 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍 contrast between |𑋈| [ʈɾ ~ t̪ɾ] and |𑋆| [ʈ] was effectively neutralised in 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝟷𝟾𝟼𝟿. |𑋊| was mapped to [ɽə], although no dedicated graph was assigned to [ɽʱə]. The graph |𑋙| was taken from the Shikarpuri inventory of Landa (§5.2.3) and assigned to [ɾə]. The Khudawadi-style graph |॥| [ɾə] was rejected, presumably to avoid graphetic confusion with |𑋌| [ɳə ~ ɽ̃ə] or with the logograms for fractions (§8.2.3). Finally, new graphs were created from existing ones where needed. The rasms |𑊼 𑋂 𑋏 𑋔| denoting the voiced unaspirated φ-plosives [ɡə d͡ʑə d̪ə bə] were graphetically modified as |𑊾 𑋄 𑋐 𑋖| to create unique graphs for the voiced aspirate φ-plosives |𑊾 𑋄 𑋐 𑋖| [ɡʱə d͡ʑʱə d̪ʱə bʱə]. Likewise, a curl was added to the rasm |𑋝| [sə] to create |𑋜| [ɕə]. Based on precedent from 𝚙𝚊-𝙶𝚞𝚛𝚞 and perhaps 𝚑𝚒-𝙳𝚎𝚟𝚊 (see (11)), the rasms |𑊻 𑊼 𑋂 𑋓| [kʰə ɡə d͡ʑə pʰə] were graphetically augmented with a subfixed nuqta to create |𑊻𑋩 𑊼𑋩 𑋂𑋩 𑋓𑋩| [xə ɣə zə fə]. Although graphematically licensed, |𑊺𑋩| [qə] was not commonly attested. The graphs for the implosives were largely unchanged from 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍.

The new system also addressed one of the principal criticisms of 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍 by introducing distinct γ-vowels into the inventory to match the number of φ-vowels in spoken Sindhi. The rasms |𑊰 𑊲 𑊴| were laid down as the free γ-vowels for [ə ɪ ʊ], including their reduced allophones [ᵊ ᶦ ᶷ]. This was accompanied by a full set of bound γ-vowels |⁠𑋠 ⁠◌𑋡 ⁠◌𑋢 ⁠◌𑋣 ⁠◌𑋤 ⁠◌𑋥 ⁠◌𑋦 ⁠◌𑋧 ⁠◌𑋨| corresponding to [a ɪ i ʊ u e ɛ o ɔ]. The graphetic shapes of the bound γ-vowels were inspired by those in the Gurmukhi inventory (Pandey, 2010d, p. 24). The inventory of γ-vowels in 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝟷𝟾𝟼𝟿 was completed by appropriately concatenating the rasms |𑊰 𑊲 𑊴| with the aforementioned bound allographs to create the free forms |𑊱 𑊳 𑊵 𑊶 𑊷 𑊸 𑊹| [a i u e ɛ o ɔ]. The complementary distribution of free and bound γ-vowels made 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝟷𝟾𝟼𝟿 an alphasyllabary (see Figure 2.3). Rounding off the inventory of bound γ-vowels was a virama |⁠𑋪|.

Other bound graphs in 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝟷𝟾𝟼𝟿 included an anusvara |⁠𑋟|. Like its present-day 𝚑𝚒-𝙳𝚎𝚟𝚊 counterpart (see (13)), the 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝟷𝟾𝟼𝟿 anusvara could denote either the nasalisation suprasegmental of a φ-vowel, or the nasal segment in a homorganic nasal-oral φ-consonant cluster. In the context of the latter, it became system-internal practice to write phonological sequences like [ns nɕ] not with anusvara but with ‘full’ |𑋑𑋪| [n]. Thus, the Sindhi word [mʊn(ᶦ)ɕi] ‘secretary, scribe’ was written |𑋗𑋣𑋑𑋪𑋜𑋢| (Thanvardas, 1873) rather than |𑋗𑋣𑋟𑋜𑋢|. Analogous to its counterpart in 𝚑𝚒-𝙳𝚎𝚟𝚊 and 𝚜𝚍-𝙳𝚎𝚟𝚊, the anusvara in 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝟷𝟾𝟼𝟿 can co-occur with free as well as bound γ-vowels. Certain Sindhi primers in 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝟷𝟾𝟼𝟿 (Thanvardas, 1873) also listed a bound visarga |⁠◌𑈺| as a canonical equivalent of the 𝚜𝚊-𝙳𝚎𝚟𝚊 visarga. However, outside of didactic charts of the 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝟷𝟾𝟼𝟿 alphasyllabary, the visarga does not seem to have been used in everyday writing. Grierson (1919, p. 19) does not list a visarga as part of the 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝟷𝟾𝟼𝟿 graph inventory.

8.2.2 Spelling and orthography

Early publications in 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝟷𝟾𝟼𝟿 include the primers |𑋞𑋡𑋟𑋏𑋤 𑋝𑋡𑋟𑋐𑋢 𑋒𑋞𑋪𑋙𑋡𑋘𑋧𑋟 𑊺𑋡𑋍𑋠𑋔𑋣| ‘Hindu-Sindhi First Book’ (1873) and |𑋝𑋡𑋟𑋐𑋢 𑋕𑋡𑊸 𑊺𑋡𑋍𑋠𑋔𑋣| ‘Sindhi Second Book’ (1871) by Udharam Thanvardas Mirchandani and |𑋝𑋡𑋟𑋐𑋣𑋂𑋢 𑋗𑋣𑊻𑋩𑋪𑋍𑋝𑋙𑋪 𑋍𑋠𑋙𑋢𑊻𑋩| ‘A Brief History of Sindh’ (1871) by Alumal Trikamdas Bhojwani.¹¹⁴ All these works were intended as school textbooks for Vania children. Since types for 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝟷𝟾𝟼𝟿 were cast only after 1872 (Government of Bombay, 1874, p. 262), the aforementioned books were all lithographed. Figure 8.2 shows an extract from Thanvardas (1873, p. 10), followed by an IPA transcription of the ‘spelling pronunciation’ suggested by the text, as well as of the modern standard Sindhi pronunciation where required.

𑋑𑋡𑋍𑋣	𑋗𑊺𑋪𑋍𑋔𑋗𑋥𑋟	𑋛𑋅𑋪𑋂𑋥 –	𑊰𑋛𑋚𑋪	𑊱𑊻𑋩𑋤𑋟𑋏𑊻𑋥	𑋝𑋚𑋠𑋗𑋣
nɪt̪ᶷ	mək(ᵊ)t̪əbᵊmẽ	ʋəɲ(ᶦ)d͡ʑe	əʋəl(ᶦ)	axun̪d̪ᵊkʰe	səlamᶷ
𑊺𑋡𑋂𑋥	𑋒𑋧𑊲	𑋒𑋙𑋪𑋞𑋌	𑋛𑋥𑋞𑋪𑋂𑋥 –	𑋒𑋙𑋪𑋞𑋌𑋂𑋥	𑋛𑋥𑋚𑋥
kɪd͡ʑe ~ kəd͡ʑe	poᶦ	pəɽʱəɳᵊ	ʋeɦ(ᶦ)d͡ʑe	pəɽʱəɳᵊd͡ʑe	ʋele
𑋕𑋢𑊰	𑊺𑋟𑋞𑋡𑋟	𑊽𑋠𑋚𑋪𑋞𑋡𑋂𑋥	𑋐𑋡𑋘𑋠𑋑𑋣	𑋑𑊺𑋡𑋂𑋥	𑊰𑋟𑊲𑋟
ɓiᵊ	kɛ̃ɦᵋ̃	ɠalʱᶦd͡ʑo	d̪ʱᶦjanᶷ	nəkɪd͡ʑe ~ nəkəd͡ʑe	ə̃ɪ̃ ~ ɛ̃
𑊺𑋟𑋞𑋡𑋟	𑋁𑋧𑊺𑋙𑋝𑋠𑋟	𑋔𑋡	𑋑𑊽𑋠𑋚𑋪𑋞𑋠𑊲𑋂𑋥–
kɛ̃ɦᵋ̃	t͡ɕʰokəɾᵊsã	bᶦ	nəɠalʱaᶦd͡ʑe

“Go to school every day. First, greet your teacher, then sit down to study. When studying, do not think about anything else, nor talk with another boy.”

Compared to the abjadic 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍 (see Figure 8.1), the most striking visual aspect of 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝟷𝟾𝟼𝟿 as demonstrated by Figure 8.2 is the presence of bound γ-vowels. Also present in Figure 8.2 are so-called ligatures, or complex graphs derived from two distinct rasms. An example is || [pəɽʱəɳᵊ] ‘to study (oblique)’, in which [ɽʱə] is denoted by a so-called ligature of |𑋙| [ɽə] and |𑋞| [ɦə]. Indeed, the three 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝟷𝟾𝟼𝟿 textbooks in question (Trikamdas, 1871; Thanvardas, 1871; 1873) feature several so-called ligatures involving the rasms |⁠𑋞| [ɦə] and |𑋘| [jə] as the secondary graphetic element. This practice may have been influenced by the 𝚙𝚊-𝙶𝚞𝚛𝚞 practice of using the bound allographs |੍ਹ| [ɦə] and |⁠੍ਯ| [jə] to denote [ɦə] and [jə] as the second element in a φ-consonant sequence (see §8.3 & Table 9.1). At first glance, the attestation of ligatures based on |𑋞| [ɦə] and |𑋘| [jə] appears to contradict Pandey’s (2011a, p. 3) statement that 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝟷𝟾𝟼𝟿 does not feature “half-forms and ligatures”. The discrepancy may be explained if one looks at subsequent publications in 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝟷𝟾𝟼𝟿 printed using movable type. The typeset publications do not feature complex graphs or ligatures of the kind attested in the earlier lithographed textbooks. The eschewing of complex graphs may have been a deliberate typographical decision aimed at simplifying the printing process. The legacy of dispensing with complex graphs in the standardised Khudawadi graph inventory and favouring atomic graphs continues to influence present-day Khudawadi typography. At the time of writing this book, no publicly available Unicode-compliant Khudawadi fonts allowed for the automatic rendering of complex graphs involving |𑋞| [ɦə] and |𑋘| [jə].

Aside from the aforementioned bound allographs for |𑋞| [ɦə] and |𑋘| [jə], therefore, 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝟷𝟾𝟼𝟿 was almost completely devoid of noteworthy complex graphs of any kind. Consequently, another visually conspicuous aspect of early 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝟷𝟾𝟼𝟿 text is the liberal use of the virama. That said, if one assumes that the use of a virama necessarily indicates the absence of a φ-vowel, the presence of a virama in word-final position contradicts the phonological maxim that all Sindhi words end in a φ-consonant (§4.3.2). For instance, in Figure 8.2, the spelling |𑊰𑋛𑋚𑋪| ‘first’ graphematically corresponds to the pronunciation [əʋəl], which is φ-consonant-final. In contrast, present-day 𝚜𝚍-𝙰𝚛𝚊𝚋 and 𝚜𝚍-𝙳𝚎𝚟𝚊 feature the spellings |اَوَلِ| and |अवलि|, respectively, both of which correspond to the φ-vowel-final pronunciation [əʋəlᶦ] (Rohra, Bijani, & Gurnani, 2011, p. 24; Sindhi Language Authority, 2021d). Needless to say, colloquial pronunciation may vary on a spectrum between the two.

Other noteworthy spelling practices in Figure 8.2 include the absence of a graphomorphemic boundary in |𑋗𑊺𑋪𑋍𑋔𑋗𑋥𑋟| [məkᵊt̪əbᵊmẽ] ‘in(to) school’, |𑊱𑊻𑋩𑋤𑋟𑋏𑊻𑋥| [axun̪d̪ᵊkʰe] ‘to the teacher’ and |𑋑𑊽𑋠𑋚𑋪𑋞𑋠𑊲𑋂𑋥| [nəɠalʱaᶦd͡ʑe] ‘should not talk’. In these spellings, the postpositions |𑋗𑋥𑋟| [mẽ] ‘in(to)’ and |𑊻𑋥| [kʰe] ‘to’, and the negation particle |𑋑| [nə] ‘no, not’ are graphematically fused with their nominal or verbal stems. This contrasts with practice in later 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝟷𝟾𝟼𝟿 texts (see Figure 8.3) as well as in present-day 𝚜𝚍-𝙰𝚛𝚊𝚋 and 𝚜𝚍-𝙳𝚎𝚟𝚊 writing, in which postpositions and stems are almost always separated by a graphematic space. Along similar lines, Figure 8.2 features the Sindhi word for ‘and’, [ɛ̃ ~ əĩ], spelt as |𑊰𑋟𑊲𑋟|. However, a few pages later in the same publication, the word is spelt |𑊰𑋟𑊳𑋟| (Thanvardas, 1873, p. 25). The spelling |𑊷𑋟| (see Figure 8.3) was eventually settled on, which is graphematically analogous to the now-obsolete 𝚜𝚍-𝙳𝚎𝚟𝚊 |अैं| (§7.5.2).

Notwithstanding government efforts to standardise and popularise 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝟷𝟾𝟼𝟿, the Vania community persisted with 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍 in their personal lives (§5.2.3). As a result, 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝟷𝟾𝟼𝟿 never got off the ground, and government publications in 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝟷𝟾𝟼𝟿 had nearly ceased by 1910 (Hotchand & Bhojwani, 1982 [1915], p. 165 footnote). Occasional publications in the system, typically by community outsiders, continued to appear into the early twentieth century. A Sindhi translation of the Bible’s Gospel of Matthew in 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝟷𝟾𝟼𝟿 was published in 1913 from Lahore, transliterated from existing 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝟷𝟾𝟻𝟹 and 𝚜𝚍-𝙶𝚞𝚛𝚞 editions (Nida, 1972, p. 393). An extract of this translation is shown in Figure 8.3.

𑋍	𑋍𑋤𑋟	𑋒𑋡𑋙𑋪𑋖𑋤	𑋒𑋟𑋞𑋡𑋟𑋂𑋥	𑋒𑋙𑋡𑋗𑋥𑋝𑋙	𑊻𑋥	𑋒𑋤𑋃𑋡;
t̪ə	tũ	pɪɾᶦbʱu	pɛ̃ɦᵋ̃d͡ʑe	pəɾᶦmesəɾᵊ	kʰe	puʄᶦ
𑊷𑋟	𑋙𑋣𑊽𑋧	𑊴𑋑𑋪𑋞𑋢𑊰	𑋂𑋢	𑊳	𑋜𑋥𑋛𑋠	𑊺𑋙𑋡 –
ɛ̃	ɾuɠo	ʊnʱiᵊ	d͡ʑi	i	ɕeʋa	kəɾᶦ

“Worship the Lord your God and serve Him only”

Serving as a particularly useful source of Sindhi texts is the Linguistic Survey of India (Grierson, 1919), thanks to its collection of parallel texts in 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍, 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝟷𝟾𝟼𝟿, 𝚜𝚍-𝙳𝚎𝚟𝚊-𝚡-𝚜𝚝𝚊𝚌𝚔 and 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝟷𝟾𝟻𝟹, accompanied by a Roman transliteration. Extracts of the Biblical Parable of the Prodigal Son in 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍 and 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝟷𝟾𝟼𝟿 as published in Grierson (1919, p. 101) are shown in Figure 8.4.¹¹⁵

Figure 8.4 reveals that Grierson’s 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝟷𝟾𝟼𝟿 text features graph-phone correspondences and spelling principles similar to those in Thanvardas (1871; 1873) and Trikamdas (1871). Where it differs from the latter is in being devoid of complex graphs or ligatures of any kind. That said, the text samples in Grierson (1919) were intentionally commissioned. Hence, one may consider these texts ‘manicured’ and not reflective of everyday writing practice. At the same time, the point may be somewhat moot, as 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝟷𝟾𝟼𝟿 was never really used in everyday writing. Consequently, asserting the graphematic authenticity of any 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝟷𝟾𝟼𝟿 text proves problematic.

Logograms and collation order

Compared to 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝟷𝟾𝟻𝟹, 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝟷𝟾𝟼𝟿 was largely free of logographic elements, aside from numeral and punctuation graphs. Numeral graphs in 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝟷𝟾𝟼𝟿 were identical to those used in 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍. In his review of Wathen’s Sindhi grammar, Prinsep (1837, p. 352) appears to state that certain 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍 graphs do double duty as phonograms and numerals. On this basis, he draws a parallel with the Divehi-Thaana writing system, in which ten phonograms are graphetically derived from Arabic-script numerals:¹¹⁶

[W]ith one or two exceptions the letters are merely represented by ciphers, combinations of numbers, and fractional parts : for example 𑋌 (¾ths) for n ; 𑋀 (4) for ch ; &c. &c. !

Having on a former occasion noticed the singular application of the Arabic numerals to the alphabet of the Maldive islands, we were struck by the apparent similarity of the process here pointed out at the opposite extremity of India!

Prinsep (1837, p. 352)

Indeed, in handwritten specimens (Stack, 1849b, p. 8; Thanvardas, 1871; 1873; Trikamdas, 1871), 𝚜𝚍-𝚂𝚒𝚗𝚍 numeral graphs may look very similar to certain γ-consonants. However, the graphetic resemblance is coincidental, and the Khudawadi numeral graphs are clearly related to their equivalents in the Devanagari and Gujarati inventories. Also coincidental is the graphetic similarity of the 𝚜𝚍-𝚂𝚒𝚗𝚍 rasms |॥| and |𑋌| (see Table 8.1) with then-prevalent numerical logograms |꠱| ‘half’ and |꠲| ‘three-fourths’, respectively (Pandey, 2007). The graphetic similarities between 𝚜𝚍-𝚂𝚒𝚗𝚍 numeral graphs and certain phonograms are shown in (21):

Early specimens of 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝟷𝟾𝟼𝟿 (Thanvardas, 1871; 1873; Trikamdas, 1871) are particularly light on punctuation, as seen in the sentence |𑋗𑋤𑋟𑊻𑋥 𑊴𑋅 𑋚𑊽𑋢 𑊱𑋞𑋥 𑋒𑋠𑋌𑋢 𑋉𑋥| [mũ kʰe ʊɲᵊ ləɠi aɦe paɳi ɖe] ‘I am thirsty give me water [sic]’ (Thanvardas, 1873, p. 7). That said, all three sources consistently use a hyphen-like graph |–| as a full stop or sentence separator (see Figure 8.2). While some later publications persisted with this graph (see Figure 8.3), others adopted an 𝚎𝚗-𝙻𝚊𝚝𝚗-style full stop (see Figure 8.4).

Being unstandardised, 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍 did not feature a specific collation order per se. The 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝟷𝟾𝟼𝟿 sequence featured in Thanvardas (1873, p. 1) is as follows:

Noteworthy in Thanvardas’ collation order is his listing of the free γ-vowels after the γ-consonants, and not before as is commonly done based on the Sanskritic tradition. Also conspicuous is his inclusion of visarga within the compound graph |𑊰:|. However, and as mentioned in Section 8.2.1, Thanvardas likely included the visarga due to sociolinguistic pressure from the 𝚜𝚊-𝙳𝚎𝚟𝚊 inventory, only to give the 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝟷𝟾𝟼𝟿 symbol chart a semblance of paradigmatic completeness. In no 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝟷𝟾𝟼𝟿 publication does the visarga seem to appear in the body text.

8.3 Analysis

8.3.1 Graphematic typology

In the previous two chapters, we saw that, regardless of disparity in their graph inventories and graphematic-orthographic nuances, all subvariants of 𝚜𝚍-𝙰𝚛𝚊𝚋 and 𝚜𝚍-𝙳𝚎𝚟𝚊 were typologically alike. Nuances aside, all derivatives of 𝚜𝚍-𝙰𝚛𝚊𝚋 emerge as plenar alphasyllabaries, while those of 𝚜𝚍-𝙳𝚎𝚟𝚊 are abugidic alphasyllabaries. Can the same be said for 𝚜𝚍-𝚂𝚒𝚗𝚍’s subvariants, namely 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍 and 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝟷𝟾𝟼𝟿? Or does the evidence require slotting them into separate typological categories?

The descriptions of unstandardised Sindhi-Landa in general (§5.1.4) and unstandardised Sindhi-Khudawadi or 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍 in particular (§8.1) leaves little doubt on 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍’s abjadic nature. Apart from the γ-vowel rasms |𑊰 𑊲 𑊴|, of which |𑊰| was used most often, there was simply no graphematic provision in 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍 to denote Sindhi’s φ-vowels with any degree of predictability. Typologically, therefore, 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍 meets the definition of a true consonantary or abjad to a far greater extent than contemporary 𝚊𝚛-𝙰𝚛𝚊𝚋 or 𝚑𝚎-𝙷𝚎𝚋𝚛 do. On the other hand, “improved Hindu Sindhi” or 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝟷𝟾𝟼𝟿, which was graphematically based on the Indic model and whose graphetics drew heavily on Gurmukhi, clearly falls into a distinct typological category. The incorporation of graphematic principles such as an ‘inherent’ or implicit [ə] in γ-consonants and the complementary use of free and bound vowel allographs conditioned by environment establish 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝟷𝟾𝟼𝟿’s credentials as an abugidic alphasyllabary.

Thus, 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍 and 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝟷𝟾𝟼𝟿 represent a case of intrasystemic typological difference in that they fall into distinct typological categories despite involving the same language and script components. This finding reiterates the fact that the same language and script pair can be mapped onto each other in different ways, resulting in distinct graph-phone correspondences and, consequently, distinct typological categorisations. Such a conclusion also justifies preferring the term ‘graphematic typology’ to ‘script typology’ (§2.4), since it is the mapping between language and script that determines the system’s categorisation. Although relatively uncommon, intrasystemic differences in graphematic typology are by no means unusual. In contemporary times, a prominent instance of intrasystemic typological difference is that of Berber languages written in the Tifinagh script (𝚋𝚎𝚛-𝚃𝚏𝚗𝚐). Traditional 𝚋𝚎𝚛-𝚃𝚏𝚗𝚐 is largely devoid of graphematic devices to indicate φ-vowels and, therefore, lies on the abjadic end of the typological spectrum. In contrast, a recently devised 𝚋𝚎𝚛-𝚃𝚏𝚗𝚐 system involving an expanded graph inventory and reconfigured graph-phone mappings — known as neo-Tifinagh — allows for φ-vowels to be distinctly represented. This makes neo-Tifinagh an alphabet (Brenzinger, 2007; Buckley, 2010).

At this stage, it is worth revisiting the commonly attested characterisation of writing in unstandardised Landa — including 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍 — as ‘shorthand’ writing.¹¹⁷ While this may be acceptable as a sociolinguistic or contextual label, how accurate is it as a graphematic label? As outlined in Section 2.10.2, a shorthand presupposes the existence of a longhand counterpart. However, this chapter has revealed that 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍 existed without a corresponding longhand. Even after the introduction of a relatively standardised 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝟷𝟾𝟼𝟿, unstandardised 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍 continued to prevail independently, including after the standardised subvariant died out. Thus, the two cannot be considered shorthand and longhand counterparts of each other.

Another characteristic of a shorthand system listed by Daniels (1996b, p. 815) is its intended capturing of speech as well as writing. In other words, shorthand systems often serve a dual purpose, serving as an abbreviated notation for phonological and graphematic units. Evidently, the graphematic units are those of the corresponding longhand, which users are also assumed to be literate in. However, in the context of 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍, there was no longhand to speak of. Essentially, Sindhi Vanias who wrote in 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍 were using it only as a reference to speech units — namely segments, syllables and lexemes of spoken Sindhi. Moreover, the common depiction of a shorthand as a special or marked variant of a longhand fails to capture the use of 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍 as the customary, default or unmarked form among Sindhi Vanias. Indeed, for most Sindhi Vanias, it was the only form of writing they were literate in. The phenomenon of exclusive literacy in 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍 is comparable to the hypothetical scenario of users being literate in abjadic or near-consonantal 𝚊𝚛-𝙰𝚛𝚊𝚋 or 𝚑𝚎-𝙷𝚎𝚋𝚛 but not in their graphovocalised alphasyllabic variants. Thus, while unstandardised 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍 may be justifiably described as abjadic or near-abjadic, it does not fulfil the graphematic criteria to be called a shorthand.

8.3.2 Graph inventory

As outlined in Section 5.1.4, the emergence of the term ‘Khudawadi’ reflects the fact that, notwithstanding the variability in its graph-phone correspondences, 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍 did feature a recognisable and, hence, a visually coherent graph inventory. In contrast, Khudawadi’s predecessors — the unstandardised Landa inventories — do not lend themselves as easily to being described as a visually consistent collection of graphs. This raises the terminological question of how the unstandardised Landa inventories should be depicted. Can the Landa inventories be considered a single ‘script’ manifesting as multiple typographical and stylistic variants? Or are they better understood as a family of scripts, wherein each typographical or stylistic variant — such as Khudawadi — is considered a distinct ‘script’?

In terms of nomenclature, the ISO 639 standard for language names and codes has a provision for the category of ‘macrolanguage’. For instance, Norwegian (ISO 639 code 𝚗𝚘) is considered a macrolanguage that encompasses the ‘languages’ Bokmål (𝚗𝚋) and Nynorsk (𝚗𝚗). In contrast, the ISO 15924 standard for script names and codes does not provide for a category of ‘macroscript’, which the Landa inventories would presumably fit under. Although ISO 15924 does list the option of 𝚉𝚣𝚣𝚣 as a ‘code for uncoded script’, using this code for the Landa inventories might imply that Landa is distinct from Khudawadi, in the process blurring the intrinsic relation between them. From the perspective of Unicode, Pandey (2010a) proposes that fonts for regional varieties of Landa not already encoded in Unicode utilise the Khudawadi block as the foundation for their characters, with glyphs designed to reflect the desired Landa variant. In terms of encoding, such an approach is akin to designing Fraktur or Gaelic fonts based on the Unicode codepoints for Roman-script characters, or Nastaliq fonts that utilise the same Arabic-script Unicode codepoints as Naskh fonts do. Yet, from a graphematic-typological perspective, there is also an argument for considering the Fraktur, Gaelic or Nastaliq calligraphic hands as distinct graphetic inventories in their own right and, therefore, as independent scripts (§2.6).

In terms of practical impact, designing fonts for visually distinct Landa variants based on the Khudawadi Unicode block may make it tricky to, say, simultaneously display the Khudawadi and Shikarpuri variants of a certain graph using the same font. For instance, someone wishing to depict Khudawadi-based 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍 |॥| [ɾ] and Shikarpuri-derived 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝟷𝟾𝟼𝟿 |𑋙| [ɾə] with the same font would have to resort to typographical workarounds if both glyphs are encoded on the same Unicode codepoint. Whereas Unicode-compliant fonts can allow for the same underlying character to be rendered with different surface glyphs or ‘stylistic sets’ by tweaking certain parameters, harnessing this capability to reliably display, say, the Shikarpuri or Memon variants of a particular Landa graph in the same context might require some thought.

8.3.3 Graphematics and orthography

The distinct typological categories and graph inventories of 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍 and 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝟷𝟾𝟼𝟿 imply that there is a noticeable difference in their graph-phone correspondences. Hence, the question posed in Section 6.6.1 is worth asking here, too: how should 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍 and 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝟷𝟾𝟼𝟿 be characterised within the context of the Modular Theory? Are they simply distinct sets of graph-phone mappings, or is there a superior terminological alternative? Also, given the deliberate design and top-down promulgation of 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝟷𝟾𝟼𝟿 — reminiscent of 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝟷𝟾𝟻𝟹 as well as 𝚜𝚍-𝙳𝚎𝚟𝚊-𝚡-𝚒𝚒𝚜 — is there a case for it to be classified as an orthography? If yes, and as alluded to in Section 6.6.3, the definition of an orthography would have to account for its potential impact on the graph inventory.

With regard to the graph inventory, the fact that the visarga was nominally included in 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝟷𝟾𝟼𝟿 primers as part of the symbol chart reinforces the maxim that a collation order is based on tradition and convention, and not reflective of graphematic structure or sociolinguistic use. Neither does the collation order of a writing system necessarily comprise all graphs employed in that system, nor does it conclusively indicate the graphematic status of those graphs. At most, a collation order may be considered graphematically relevant in that it is often influenced by the orthographic module.

8.3.4 Sociolinguistics

The inclusion of visarga in 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝟷𝟾𝟼𝟿’s collation order, despite likely never being used, demonstrates how sociolinguistic pressure can influence the elements included in a graph inventory. Moreover, the presence of visarga in 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝟷𝟾𝟼𝟿’s inventory bears some resemblance to graphematic retention in 𝚜𝚍-𝙰𝚛𝚊𝚋 and 𝚜𝚍-𝙳𝚎𝚟𝚊 in being conditioned by a dominant writing system’s inventory and practices.

Finally, the fact that a graphematically comprehensive writing system such as 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝟷𝟾𝟼𝟿 turned out to be a sociolinguistic dud and faded into oblivion is starkly reminiscent of the fate of 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚞𝚖𝚙𝚙. Both these cases affirm Meletis’ (2020, pp. 16–17) observation that a robust linguistic fit is no guarantee of a writing system’s sociolinguistic fit and, consequently, of its success.

☙ ❧

9 Gurmukhi

In a sense, Gurmukhi is to the Punjab what Khudawadi is to Sindh — a Landa-origin script native to the region and traditionally used to transcribe local speech varieties (see Figure 5.1). At the same time, Gurmukhi differs from Khudawadi in that the former is still widely used in the twenty-first century, while the latter has become obsolete, at least in everyday use (Pandey, 2011a, p. 2). The graphetic roots of Gurmukhi in its present form lie in a Punjabi-origin Landa inventory used to transcribe the devotional hymns of the Sikh religion (Mann, 2001, p. 5). Since Sikhism originated in the Punjab, it was logical that Sikh liturgical compositions were written down in the local Landa inventory. The initial collection of Sikh hymns transcribed in Gurmukhi appeared in the early seventeenth century under the name Adi Granth ‘First Book’ (Shackle, 2007, p. 638). The final, enlarged collection of hymns is now called the Guru Granth Sahib, and is considered the sacred scripture of the Sikhs (Kohli, 1961, p. 12). Legend has it that the second Sikh spiritual master or guru, Guru Angad, improved the unstandardised Punjabi-Landa inventory in vogue at the time by furnishing it with additional graphs (Grierson, 1916, p. 624). As a result, the refined graph inventory came to be known as Gurmukhi ‘from the mouth of the Guru’ (Salomon, 2007, p. 92). In contrast, academic consensus is on the script being a descendant of Brahmi and on individual graphs in the inventory existing in a recognisable form prior to Guru Angad’s times (Bahri, 2011, p. 181; Kohli, 1961, pp. 13–14 footnote). Regardless of its origin, Gurmukhi has come to be revered by Sikhs, according it a status comparable to that of the Arabic script among Muslims and Devanagari among north Indian Hindus (Jain, 2007, pp. 59–60; Mann, 2001, p. 5). In post-Partition India, Gurmukhi has also been recognised in the Indian state of Punjab as the official script for the Punjabi language (Jain, 2007, p. 59; Shackle, 2007, p. 640), and is widely used in administration, education and mass media in that state.

The close association of the Gurmukhi script with the Punjabi language on the one hand, and with Sikh scriptures on the other, sometimes leads to people describing the Guru Granth Sahib as containing Punjabi-language hymns. In reality, the hymns of the Guru Granth Sahib are composed in a mixture of north-western Indo-Aryan speech varieties, often with discernible Persian influence (Grierson, 1916, pp. 618, 624; Kohli, 1961; Mann, 2001, p. 5). In graphematic terms, the writing system used in these hymns reflects not just Punjabi in Gurmukhi (𝚙𝚊-𝙶𝚞𝚛𝚞) but multiple languages written in the script, among them Sindhi (𝚜𝚍-𝙶𝚞𝚛𝚞; see Example (24)). Regardless, the pervasiveness of popular indexical associations often results in the terms ‘Gurmukhi’ and ‘Punjabi’ being interchangeably used to describe the language and the script, respectively. Despite the seemingly intertwined and inextricable links between language and script, they must be treated as distinct. Grierson (1916, p. 624) provides a pithy summary of the situation in noting that “[t]here is no more a ‘Gurmukhī’ language than there is a ‘Dēva-nāgarī’ one”. Besides, he reiterates that several languages have been written in Gurmukhi, which includes Sindhi.

9.1 Graphematic foundations

It is likely that early unstandardised Gurmukhi-based transcriptions of regional lects would have followed quasi-abjadic principles, much like 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍. Notwithstanding its legendary origin, Gurmukhi’s use in liturgical contexts and the desire to record and reproduce Sikh hymns in a phonologically accurate manner encouraged a gradual, organic process of graphetic and graphematic standardisation. Thus, compared to 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍, Gurmukhi-based writing systems departed from their abjadic origins and moved towards the abugidic end of the typological spectrum at a much earlier stage of evolution.

The Sindhi-Gurmukhi (𝚜𝚍-𝙶𝚞𝚛𝚞) graph inventory is based on the Punjabi-Gurmukhi (𝚙𝚊-𝙶𝚞𝚛𝚞) inventory, the main elements of which are shown in Table 9.1.

The graph inventory of 𝚙𝚊-𝙶𝚞𝚛𝚞 began to undergo standardisation by the nineteenth century, not least due to its increasing appearance in print. Figure 9.1 features 𝚙𝚊-𝙶𝚞𝚛𝚞’s free γ-vowels as seen in an early nineteenth-century publication.

Typographically, the free γ-vowels in Figure 9.1 bear a strong resemblance to their present-day counterparts in Table 9.1. In terms of graphetic composition, however, the free γ-vowel forms |ੲੈ| [æ] and || [ɔ] were variable in their rasmic base and also appeared as |ਐ ਔ|, respectively (Stack, 1849a, p. 4). While the latter forms have become standard in the modern Gurmukhi inventory, the graph || persists as part of the benedictory logogram |ੴ| ‘God is one’.

Reviews of handwritten (Stack, 1849a, pp. 3–8) and printed samples (Faulmann, 1880a, p. 123) of the Gurmukhi inventory over the course of the nineteenth century reveals a steady but palpable process of graphetic standardisation. Additionally, in the context of 𝚙𝚊-𝙶𝚞𝚛𝚞, several new γ-consonants were created by subfixing a nuqta to existing rasms. These included |ਖ਼ ਗ਼ ਜ਼ ਫ਼ ਸ਼| for Perso-Arabic-origin [x(ə) ɣ(ə) z(ə) f(ə) ɕ(ə)], and |ਲ਼| for indigenous [ɭ(ə)] (Gill, 1996, p. 396; Shackle, 2007, p. 651). As was the case in 𝚜𝚍-𝙳𝚎𝚟𝚊 and 𝚜𝚍-𝚂𝚒𝚗𝚍, a graphematically plausible |ਕ਼| [q(ə)] did not become well-established in 𝚙𝚊-𝙶𝚞𝚛𝚞, likely due to the marginal nature of the corresponding phone [q] in Punjabi. Along similar lines, the absence of Sanskritic [ʂ] in Punjabi likely reduced the motivation to create a standalone 𝚙𝚊-𝙶𝚞𝚛𝚞 equivalent of 𝚜𝚍-𝙳𝚎𝚟𝚊 |ष|, with the graphetically related graph |ਖ| being allocated instead to [kʰ(ə)] (§7.3, p. 199; §8.1, p. 221).

Present-day 𝚙𝚊-𝙶𝚞𝚛𝚞 has the characteristics of an abugida, in that so-called consonant graphs in the inventory actually denote φ-[CV₀]. In modern Punjabi, the so-called inherent φ-vowel [V₀] has the value [ə]. The inventory also contains a bound virama |੍| [Ø] to suppress φ-[V₀], in line with other Indic abugidas. In practice, though, the virama is rarely used in modern 𝚙𝚊-𝙶𝚞𝚛𝚞, and writers and readers typically rely on graphematic environment to encode and decode the presence of φ-[V₀] (Masica, 1991, p. 149; Shackle, 2007, p. 652). Like 𝚜𝚍-𝚂𝚒𝚗𝚍, 𝚙𝚊-𝙶𝚞𝚛𝚞 is poor in bound allographs or ‘half forms’ of γ-consonants (Shackle, 2007, p. 652). At most, the graph inventory provides for |⁠੍ਯ ⁠੍ਰ ⁠੍ਵ ⁠੍ਹ| as bound allographs of |ਯ ਰ ਵ ਹ| [j(ə) ɾ(ə) ʋ(ə) ɦ(ə)]. The bound forms are used when the corresponding phones occur as the second member of a consonant cluster. These bound allographs may also be used for other purposes, such as marking phonological tone in Punjabi. In the past, |⁠੍ਰ| was occasionally used to form the compound graph |ਲ੍ਰ| for [ɭ(ə)] (Grierson, 1916, p. 627), although this has now given way to |ਲ਼|.

Modern 𝚙𝚊-𝙶𝚞𝚛𝚞’s abugidic characteristics also derive from its free and bound γ-vowels occurring in complementary distribution. Notably, free γ-vowels in 𝚙𝚊-𝙶𝚞𝚛𝚞 are formed using just three rasms as graphetic bases (see Table 9.1; see also (23)). The distribution of graphetic rasms forms a pattern with the qualities of the φ-vowels represented. Thus, |ਅ| is used as the base for the low φ-vowels [ə a æ ɔ], |ੲ| for the high and mid-high front φ-vowels [ɪ i e], and |ੳ| for the high and mid-high back φ-vowels [ʊ u o] (Gill, 1996, p. 396; Salomon, 2007, p. 93). The rasm |ਅ| also serves as the fully-specified free γ-vowel for [ə], while graphetically truncated |ੳ| in the form of |ਓ| stands for [o]. The other eight free γ-vowels are formed by combining the bound allographs |⁠ਾ ⁠ਿ ⁠ੀ ⁠ੁ ⁠ੂ ⁠ੇ ⁠ੈ ⁠ੋ ⁠ੌ| with one of the three rasmic bases, resulting in |ਆ ਇ ਈ ਉ ਊ ਏ ਐ ਔ| [a ɪ i ʊ u e æ ɔ].

In terms of graphetic compositionality, the 𝚙𝚊-𝙶𝚞𝚛𝚞 rasms |ਅ ੲ ੳ| most closely resemble 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝟷𝟾𝟼𝟿 |𑊰 𑊲 𑊴|. Echoing conjecture in the context of 𝚜𝚍-𝚂𝚒𝚗𝚍 (§8.1), scholars have postulated that the presence in Gurmukhi of three rasms as graphetic bases for free γ-vowels reflects Semitic influence (Masica, 1991, pp. 150–151; Salomon, 2007, p. 93; Shackle, 2007, p. 651). On this basis, authors have described the rasms |ਅ ੲ ੳ| as “vowel carriers” or “vowel bearers” (Gill, 1996, pp. 395–396; Masica, 1991, pp. 150–151), which are terms used more commonly in the context of Semitic writing systems (§6.1).

Like 𝚑𝚒-𝙳𝚎𝚟𝚊 |ं| and |⁠ँ|, 𝚙𝚊-𝙶𝚞𝚛𝚞 contains two bound graphs to denote nasality, known as tippi |⁠ੰ| and bindi |⁠ਂ|. Both graphs may denote either φ-vowel nasalisation or the nasal component of a nasal-oral φ-consonant cluster. However, the distribution of |⁠ੰ| and |ਂ| is distinct from 𝚑𝚒-𝙳𝚎𝚟𝚊 |⁠ं| and |⁠ँ|. Unlike the latter, the occurrence of 𝚙𝚊-𝙶𝚞𝚛𝚞 |⁠ੰ| and |⁠ਂ| is determined purely by the graphematic environment (Grierson, 1916, p. 627; Shackle, 2007, p. 653), resulting in complementary distribution as shown in (22):

A noteworthy bound graph in 𝚙𝚊-𝙶𝚞𝚛𝚞 is |⁠ੱ|, known as addak.¹¹⁸ The addak indicates gemination of the φ-consonant represented by the graph immediately following, as in |ਪੱਥਰ| [pət̪t̪ʰəɾ] ‘stone’ (Grierson, 1916, p. 627). Evidently, the graphotactic distribution of the addak precludes indicating word-initial φ-consonant gemination. However, this does not pose a problem for 𝚙𝚊-𝙶𝚞𝚛𝚞, since Punjabi phonology prohibits word-initial φ-consonant gemination. Scholars have attributed the introduction of the addak in the nineteenth century to Christian missionaries (Shackle, 2007, p. 654). Authors have also highlighted its graphetic similarity with the functionally near-identical 𝚊𝚛-𝙰𝚛𝚊𝚋 tashdid (Masica, 1991, p. 149; Salomon, 2007, p. 93). Despite its availability, the use of addak in 𝚙𝚊-𝙶𝚞𝚛𝚞 remains inconsistent (Shackle & Skjærvø, 2006, p. 548). Even rarer is the use of 𝚙𝚊-𝙶𝚞𝚛𝚞 visarga |⁠ਃ|. It is occasionally used when one intends to authentically transcribe Sanskrit words into Gurmukhi (𝚜𝚊-𝙶𝚞𝚛𝚞). In older liturgical texts, the visarga may be used as an abbreviation marker.

The numeral graphs in 𝚙𝚊-𝙶𝚞𝚛𝚞 are |੦ ੧ ੨ ੩ ੪ ੫ ੬ ੭ ੮ ੯| ‘0 1 2 3 4 5 6 7 8 9’. Emic graphs for punctuation include |। ॥|, which are graphetically and graphematically identical to their 𝚜𝚊-𝙳𝚎𝚟𝚊 counterparts (§7.1). Modern 𝚙𝚊-𝙶𝚞𝚛𝚞 writing makes extensive use of 𝚎𝚗-𝙻𝚊𝚝𝚗 numeral and punctuation graphs.

The traditional 𝚙𝚊-𝙶𝚞𝚛𝚞 symbol chart is shown in (23) (Gill, 1996, p. 396; Shackle, 2007, p. 652). It comprises thirty-five graphetic rasms, due to which 𝚙𝚊-𝙶𝚞𝚛𝚞 was also historically referred to as paintih ‘the thirty-five’ (Punjabi: [pæ̃t̪í]) (Gill, 1996, p. 395; Shackle, 2007, p. 651). The chart departs from the traditional Sanskritic collation order in commencing not with fully-specified free γ-vowels but with the underspecified γ-vowel rasms |ੳ ਅ ੲ|. Occasionally, these three rasms may appear in the Sanskrit-influenced sequence |ਅ ੲ ੳ| (Grierson, 1916, p. 625). The three γ-vowel rasms are succeeded by |ਸ ਹ| [s(ə) ɦ(ə)], followed by other γ-consonants in the conventional Sanskritic order. The list concludes with |ੜ| [ɽ(ə)].

In denoting the retroflex flap [ɽ(ə)] with the distinct rasm |ੜ|, 𝚙𝚊-𝙶𝚞𝚛𝚞 contrasts with 𝚑𝚒-𝙳𝚎𝚟𝚊’s graphetically augmented |ड़| while harmonising with 𝚜𝚍-𝚂𝚒𝚗𝚍’s |𑋊|. The presence of a distinct rasm for [ɽ(ə)] in 𝚙𝚊-𝙶𝚞𝚛𝚞 and 𝚜𝚍-𝚂𝚒𝚗𝚍 also serves as a subtle graphematic reminder of the Landa origins of their constituent inventories.

9.2 Early use

The rise of Gurmukhi as a distinct inventory more or less coincides with its use in Sikh liturgical writing. Hence, the hymns in the seventeenth-century Adi Granth may be considered the earliest specimens of Gurmukhi-based writing. The graph inventory used in the Adi Granth was devoid of underdotted γ-consonants and the addak, since they were introduced only in the nineteenth century. Also absent or rarely marked was nasalisation on γ-vowels (Shackle, 2007, p. 653), which was consistent with practice in pre-standardised Landa (§5.1.1). On the other hand, certain graphematic practices used in the Adi Granth have persisted and been codified into modern 𝚙𝚊-𝙶𝚞𝚛𝚞 orthography. Prominent among them is the representation of word-medial [j] with the bound γ-vowel |ਿ| [ɪ] (Shackle, 2007, p. 653).

At the same time, the Adi Granth comprised hymns composed in various north-western Indo-Aryan speech varieties, including Sindhi. Thus, the collection also contains some of the earliest instances of Sindhi in Gurmukhi (𝚜𝚍-𝙶𝚞𝚛𝚞). Kohli (1961, p. 62) has identified verses in the Adi Granth whose language lies on a continuum between the Siraiki of southern Punjab and Siroli of northern Sindh (§4.2). He notes that these verses are collectively called dakhne [ɖəkʰ(ə)ɳe] (cf. Sindhi [ɗəkʰəɳᶷ] ‘south’), which alludes to their ‘southern’ idiom. An example of such a verse in the Adi Granth (later Guru Granth Sahib) is shown in (24), with recognisably Sindhi elements bolded and accompanied by an IPA transcription and English gloss.

To a speaker of modern Sindhi, several items of the lexicon used in (24) may come across as archaic, or at least highly poetic and literary in nature. Of the lexical items still in use, some may appear in so-called non-standard or ‘dialectal’ forms. For instance, item #7, |ਮਹਿੰਜਾ| [məɦᶦ̃d͡ʑa] ‘my (sing.)’ is related to present-day Siroli [mə̃ɦᶦ̃d͡ʑo] ‘ibid.’ but distinct from Vicholi or standard Sindhi [mʊ̃ɦᶦ̃d͡ʑo] ‘ibid.’. Linguistically and geographically, the appearance of these lexical items is consistent with the Siraiki-Siroli idiom of the verses. Also consistent are the graphematic features in the verses with what is expected of the time. Thus, items #1 |ਤੂ| and #10 |ਪਿਰੀ|, if transparently interpreted, suggest the pronunciations [t̪u] and [pᶦɾi]. In contrast, the equivalent present-day Sindhi pronunciations of these words are [t̪ũ] and [pᶦɾĩ], respectively. The discrepancy is neatly explained if one considers that φ-vowel nasalisation was unpredictably marked in writing at the time.

9.3 British-era and post-Partition practices

Prior to the British conquest of Sindh, works in 𝚜𝚍-𝙶𝚞𝚛𝚞 were limited in number, which was consistent with the fact that written Sindhi was not widespread at the time. Probably the best-known 𝚜𝚍-𝙶𝚞𝚛𝚞 composition of this time is the poetry of celebrated Sindhi poet Chainrai Dataramani, popularly known by his pen-name ‘Sami’ (1743?–1850) (Jotwani, Shackle, & Das, 2015). Hardwani (2015, p. 2) speculates that Sami’s choice of Gurmukhi was natural, since he wrote most of his poetry in the Punjabi cultural centre of Amritsar. That said, Sami’s handwritten manuscripts were compiled and published only by the late 1800s (Jetley, 1989). It remains to be researched whether the graphematic features seen in published versions of Sami’s 𝚜𝚍-𝙶𝚞𝚛𝚞 writing are unchanged from the original, or have undergone subtle modifications along the way. For instance, a selection of his verses published in 1938 (Hauze, 2016f) features the composite graph |ੲੈਂ| [ɛ̃ ~ əĩ] ‘and’ instead of the 𝚙𝚊-𝙶𝚞𝚛𝚞-inspired |ਐਂ| (§9.1). Although |ੲੈਂ| is licensed by the graphematic solution spaces of both 𝚜𝚍-𝙶𝚞𝚛𝚞 and 𝚙𝚊-𝙶𝚞𝚛𝚞, the latter’s orthography implicitly licenses the composite graph |ਐਂ| (see Table 9.1). Hence, knowing whether the choice of |ੲੈਂ| in Sami’s published works was his own, or is a subsequent modification, would be valuable from a graphematic perspective.

Publication in 𝚜𝚍-𝙶𝚞𝚛𝚞 gained momentum after the British takeover of Sindh, with administrators, scholars and Christian missionaries professing interest in the writing system. In his Sindhi grammar, Stack (1849a, pp. 3–8) included handwritten samples of Gurmukhi graphs in his comparative chart. Despite eventually opting for Devanagari in his work, Stack (1849a, p. vi) describes Gurmukhi as “assimilating with the Sindhi somewhat more than Devenagari [sic] did, and […] being also more known to Hindoos in Sindh”. Stack’s contemporary and Christian missionary Andrew Burn translated the Gospel of John into 𝚜𝚍-𝙶𝚞𝚛𝚞, which was lithographed and published in Karachi in 1859 (Blumhardt, 1893, pp. 4–5; Nida, 1972, p. 393). An extract from Burn’s translation (John 3:16) is shown in Figure 9.2.

“For God so loved the world, that he gave his only begotten Son, that whosoever believeth in him should not perish, but have everlasting life.”

New and revised translations of Biblical scriptures in 𝚜𝚍-𝙶𝚞𝚛𝚞 appeared in 1877, 1899, 1908 and 1917 (Nida, 1972, pp. 393–394). The 1877 edition of the Gospel of John, published by Oxford University Press, was a typographical landmark in the history of 𝚜𝚍-𝙶𝚞𝚛𝚞 publishing, as it was the first 𝚜𝚍-𝙶𝚞𝚛𝚞 publication to be printed using movable type (Afshar, 2020, p. 161). Other 𝚜𝚍-𝙶𝚞𝚛𝚞 publications in the late nineteenth century ranged from government-issued textbooks for girls’ schools to magazines aimed at women readers (§5.2.4). Publication in 𝚜𝚍-𝙶𝚞𝚛𝚞 began to wane in the twentieth century as 𝚜𝚍-𝙰𝚛𝚊𝚋 came to be increasingly accepted and used in Sindh. Indeed, among the numerous heterographic Sindhi-language text samples provided in the Linguistic Survey of India, 𝚜𝚍-𝙶𝚞𝚛𝚞 text samples are conspicuous by their absence.

An overview of accessible 𝚜𝚍-𝙶𝚞𝚛𝚞 works from the pre-Partition era reveals large degrees of inventorial overlap with 𝚙𝚊-𝙶𝚞𝚛𝚞. Stack’s comparative chart of scripts (1849a, pp. 3–8) does not contain distinct 𝚜𝚍-𝙶𝚞𝚛𝚞 graphs for the Sindhi implosives [ɠ ʄ ɗ ɓ]. Authors of Christian literature in 𝚜𝚍-𝙶𝚞𝚛𝚞 were likely the first to attempt creating distinct graphs for the φ-implosives. Thus, the text in Figure 9.2, dating from 1858, features the spellings |ਜਗ਼ਤ| [d͡ʑəɠət̪ᵊ] ‘world’ and |ਜ਼ਣ੍ਯਲੁ| [ʄəɳ(ᶦ)jəlᶷ] ‘born’. Here, the rasms for [ɡ(ə) d͡ʑ(ə)] have been graphetically subfixed with a nuqta to create |ਗ਼ ਜ਼| [ɠ(ə) ʄ(ə)]. By the late nineteenth century, however, the underdotted |ਖ਼ ਗ਼ ਜ਼ ਫ਼| began to be increasingly used to denote [x(ə) ɣ(ə) z(ə) f(ə)]. Consequently, new graphs needed to be created for [ɠ(ə) ʄ(ə)], if not for the entire φ-implosive set. A Christian publication from the early twentieth century featuring the Lord’s Prayer in 𝚜𝚍-𝙶𝚞𝚛𝚞 (Gilbert & Rivington, 1905, p. 136) denotes [ʄ(ə)] with two nuqtas subfixed on |ਜ|. The prayer does not contain any instance of [ɠ(ə)], although one may hypothesise that a similar graphetic pattern might have been used. On occasion, [ɓ(ə)] was transcribed as |ਭ਼| (Nida, 1972, p. 393), by underdotting the rasm |ਭ| [bʱ(ə)] rather than |ਬ| [b(ə)]. Overall in pre-Partition 𝚜𝚍-𝙶𝚞𝚛𝚞 texts, the relatively consistent adoption of |ਖ਼ ਗ਼ ਜ਼ ਫ਼| for [x(ə) ɣ(ə) z(ə) f(ə)] contrasted with the vacillation on graphs for the φ-implosives, despite the latter being native to the Sindhi language. In this regard, the 𝚜𝚍-𝙶𝚞𝚛𝚞 graph inventory contrasted with 𝚜𝚍-𝚂𝚒𝚗𝚍 in not having clear-cut graphs for the φ-implosives, in the process signalling Gurmukhi’s origins outside of Sindh.

The then-repertoire of 𝚜𝚍-𝙶𝚞𝚛𝚞 γ-vowels, too, was largely isographic with that of 𝚙𝚊-𝙶𝚞𝚛𝚞. Areas of inventorial and graphematic divergence included the greater prevalence in 𝚜𝚍-𝙶𝚞𝚛𝚞 of |ੲੈਂ| [ɛ̃ ~ əĩ] ‘and’, be it in Sami’s verses or in Christian publications. Graphematically, 𝚜𝚍-𝙶𝚞𝚛𝚞 was distinguished by its use of bound γ-vowels. Conditioned by the notion that most — if not all — Sindhi φ-syllables are vowel-final, as well as by the unpopularity of the virama in 𝚙𝚊-𝙶𝚞𝚛𝚞 writing, 𝚜𝚍-𝙶𝚞𝚛𝚞 writers often used |⁠ਿ| to indicate [ᶦ ~ Ø]. This practice is homologous with the 𝚜𝚍-𝙰𝚛𝚊𝚋 practice of using zer |⁠◌ِ| as a near-allograph of jazm |⁠◌ٛ| to mark [ᶦ ~ Ø], itself dating from pre-British times (Stack, 1849a, p. 128). As with all the writing system variants seen thus far, 𝚜𝚍-𝙶𝚞𝚛𝚞, too, did not graphematically distinguish [ə ɪ ʊ] from their reduced allophones [ᵊ ᶦ ᶷ], and used the same set of γ-vowels for both. Nasalisation of φ-vowels generally followed the 𝚙𝚊-𝙶𝚞𝚛𝚞 distribution (see (22)), although idiosyncratic deviations were common. Evidence for an attested or recommended collation order in 𝚜𝚍-𝙶𝚞𝚛𝚞 proves elusive. It remains to be seen whether the preferred template is the 𝚙𝚊-𝙶𝚞𝚛𝚞 sort order (see (23)) or a 𝚜𝚍-𝙳𝚎𝚟𝚊-inspired one (§7.5.4, §7.6.3).

Publication in 𝚜𝚍-𝙶𝚞𝚛𝚞 largely ceased by the mid-twentieth century, with only the odd publication appearing in post-Partition India. Unsurprisingly, they remain rare and hard to come by. One such publication is an undated 𝚜𝚍-𝙶𝚞𝚛𝚞 booklet on Hindu spiritual themes, published in Ulhasnagar, Maharashtra (Kalyani, n.d.). The booklet largely follows the pre-Partition graphematic practices described above. An extract of a passage from the booklet is shown in (25), followed by an IPA transcription of the 𝚜𝚍-𝙶𝚞𝚛𝚞 spelling pronunciations. Where spelling pronunciations are divergent from standard Sindhi pronunciations, an IPA transcription of the latter is also provided.

Aside from adhering to the pre-Partition 𝚜𝚍-𝙶𝚞𝚛𝚞 practices described earlier, the extract in (25) also largely follows 𝚙𝚊-𝙶𝚞𝚛𝚞 practices. In line with (22), tippi |⁠ੰ| and bindi |⁠ਂ| are in complementary distribution conditioned entirely by graphematic environment. In other words, both |⁠ੰ| and |⁠ਂ| may denote the nasal component of a homorganic nasal-oral φ-consonant cluster, or the nasalisation of a φ-vowel. Thus, in (25), tippi denotes homorganic [ŋ] in |ਸ਼ੰਕਰ| [ɕəŋkəɾᵊ] ‘Shiva’, but φ-vowel nasalisation [◌̃] in |ਮੁੰਹਿੰਜੇ| [mʊ̃ɦᶦ̃d͡ʑe] ‘my (oblique)’. Elsewhere in the booklet, the graphematic conditioning at play is further illustrated by the examples |ਕੰਦੇ| [kən̪d̪e] ‘while doing’ but |ਵੇਂਦਾ| [ʋen̪d̪a] ‘(they) will go’, in which both |ੰ| and |ਂ| denote homorganic [n̪] (Kalyani, n.d., p. 2). That said, the booklet also reveals deviations from the 𝚙𝚊-𝙶𝚞𝚛𝚞 distribution of these graphs. Most of them appear to be oversights that lie within the realm of graphematic plausibility. Thus, in |ਮੁਂਹਿੰਜੋ| (p. 2) and |ਤੁਂਹਿੰਜੋ| (p. 6), an expected |⁠ੰ| is substituted by |⁠ਂ|, while, in |ਤੁਹਿੰਜੇ| (p. 18), neither is present. However, there is also the odd deviation from the norm occurring consistently throughout the booklet, thereby appearing intentional. The most prominent of these is the spelling of the Sindhi pronoun [t̪ũ] ‘thou’, which is consistently written |ਤੂਂ| rather than the expected |ਤੂੰ| predicted by (22). Also occurring consistently is |ੲੈਂ| [ɛ̃ ~ əĩ] ‘and’, which, although distinct in its rasmic base from 𝚙𝚊-𝙶𝚞𝚛𝚞-style |ਐਂ|, follows the norm in featuring a bindi.

The booklet does not use distinct graphs for the φ-implosives, representing them homographically with their φ-plosive counterparts using |ਗ ਜ ਡ ਬ|. Also underdifferentiated are the Perso-Arabic-origin [x ɣ z f] from [kʰ ɡ d͡ʑ pʰ], with rasms |ਖ ਗ ਜ ਫ| used for both sets. On occasion, the booklet also uses |ਭ| [bʱ(ə)] where one would expect |ਬ| [b(ə)], albeit unpredictably. Thus, 𝚜𝚍-𝙶𝚞𝚛𝚞 |ਜਵਾਬੁ| [d͡ʑəʋabᶷ] ‘reply’ appears in (25) as |ਜਵਾਭੁ|, suggesting the non­standard pronunciation [d͡ʑəʋabʱᶷ].

The booklet extensively uses the handful of bound γ-consonants available in the Gurmukhi inventory: |੍ਯ ੍ਰ ੍ਵ ੍ਹ|. Accordingly, Sindhi [əʋʱĩ] ‘you (pl.)’ is transcribed as |ਅਵ੍ਹੀਂ| (p. 6). Where possible, the graphematic makeup of Sanskrit loanwords is retained. Thus, Sanskrit [kɾod̪ʱᵊ] ‘anger’ is transcribed |ਕ੍ਰੋਧ| (p. 6), whose spelling suggests the pronunciation [kɾod̪ʱ] rather than the assimilated Sindhi pronunciation [k(ᶦ)ɾod̪ʱᶷ]. The booklet does not appear to use the virama at all, preferring instead to use |ਿ| to indicate [ᶦ ~ Ø]. Curiously, the latter practice also extends to English words transcribed into 𝚜𝚍-𝙶𝚞𝚛𝚞. For instance, the name of the publisher, “L. Kishinchand and Sons, Booksellers”, is transcribed on the cover as |ਏਲਿ. ਕਿਸ਼ਿਨਿਚੰਦੁ ੲੈਂ ਸੰਸਿ ਭੁਕੁਸੇਲਰਿਸਿ|, suggesting the spelling pronunciation [elᶦ kɪɕɪnᶦt͡ɕən̪d̪ᶷ ɛ̃ sənsᶦ b(ʱ)ʊkᶷseləɾᶦsᶦ]. With regard to logograms, the booklet uses the standard 𝚙𝚊-𝙶𝚞𝚛𝚞 set of numeral graphs, and, aside from |।| as full stop, uses 𝚎𝚗-𝙻𝚊𝚝𝚗 punctuation throughout.

9.4 Analysis

Compared to Sindhi’s Arabic, Devanagari and Khudawadi-based writing systems, 𝚜𝚍-𝙶𝚞𝚛𝚞 emerges as comparatively less normativised. Yet, the large degree of overlap between Sindhi and Punjabi phonology (§4.3) and a reasonably high level of pre-existing regularity in 𝚙𝚊-𝙶𝚞𝚛𝚞’s appears to have mitigated the need for large-scale customisations to 𝚜𝚍-𝙶𝚞𝚛𝚞. Consequently, some of the observations made in this section are based on, and also apply to, 𝚙𝚊-𝙶𝚞𝚛𝚞.

The lack of significant graphematic modifications specific to 𝚜𝚍-𝙶𝚞𝚛𝚞 render it typologically identical to 𝚙𝚊-𝙶𝚞𝚛𝚞 in being an abugidic alphasyllabary. This puts 𝚜𝚍-𝙶𝚞𝚛𝚞 in the same graphematic category as 𝚜𝚍-𝙳𝚎𝚟𝚊 and 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝟷𝟾𝟼𝟿, while also aligning it with 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝟷𝟾𝟻𝟹 and 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚞𝚖𝚙𝚙 in terms of their alphasyllabic nature. Graphematic commonalities among these systems are also evident in the principles governing the composition of their free γ-vowels. The three rasms |ਅ ੲ ੳ| used in 𝚙𝚊-𝙶𝚞𝚛𝚞 and 𝚜𝚍-𝙶𝚞𝚛𝚞 as graphetic bases to create free γ-vowels is analogous to 𝚜𝚍-𝚂𝚒𝚗𝚍’s use of |𑊰 𑊲 𑊴| and 𝚜𝚍-𝙰𝚛𝚊𝚋’s |ء ﺋ ا|. This principle is also followed in spirit by 𝚜𝚍-𝙳𝚎𝚟𝚊 in its use of |अ इ उ ए| as graphetic bases for free γ-vowels.

Against this background, can the use of three graphetic bases to create free vowel graphs be legitimately attributed to Semitic influence, as has also been speculated in the context of 𝚜𝚍-𝚂𝚒𝚗𝚍 (§8.1)? Or is the occurrence of three graphetic bases somewhat unremarkable, since 𝚜𝚍-𝙳𝚎𝚟𝚊 — and several Devanagari-based systems for Indo-Aryan languages — use only four graphetic bases to form distinct free graphs for their ten-or-so vowel phonemes (Table 4.3)? A comprehensive conclusion on this question will require targeted scrutiny, but the initial evidence emerging from the cross-systemic investigation of Sindhi’s writing systems points raises doubts on the extent of Semitic influence.

One area where 𝚜𝚍-𝙶𝚞𝚛𝚞 falls short in terms of graphematic precision is in possessing distinct graphs for Sindhi’s φ-implosives. This gap in 𝚜𝚍-𝙶𝚞𝚛𝚞’s graph inventory reflects the phonological fact that contrastive φ-implosives are absent among most north-western Indo-Aryan languages (§4.3.1). While certain authors, especially translators of Christian scriptures, attempted to create unique graphs for Sindhi’s φ-implosives by means of nuqtas, none of these innovations caught on. In theory, the noncontrastive nature of phonological gemination in Sindhi frees up the addak |⁠ੱ| to potentially be used to denote φ-implosives. As will be seen in Chapter 12, a similar rationale has been used in the context of 𝚜𝚍-𝙻𝚊𝚝𝚗 to distinctly represent Sindhi’s φ-implosives (§12.2). However, as mentioned in Section 9.1, the graphotactic requirement for an addak to occur before the impacted graph poses a problem when the impacted graph is word-initial. Consequently, using an addak would not permit representing word-initial φ-implosives in Sindhi. Illustrating this limitation is the Sindhi word [ɖɪsi] ‘having seen’, which, in Example (25) is transcribed |ਡਿਸੀ| using plosive |ਡ| [ɖ(ə)].

In theory, the graphotactic requirements around 𝚜𝚍-𝙶𝚞𝚛𝚞’s addak may be modified to make it occur above the impacted graph, thereby aligning it with its homologues — 𝚜𝚍-𝙰𝚛𝚊𝚋’s tashdid and 𝚔𝚏𝚛-𝙺𝚑𝚘𝚓’s tashdid (Chapter 10). However, the lack of a 𝚜𝚍-𝙶𝚞𝚛𝚞 graphosphere with a critical mass of reader-writers, not to mention a 𝚜𝚍-𝙶𝚞𝚛𝚞 language body with authority, meant that the ingredients for graphematic refinement and orthographic consolidation were simply absent.

☙ ❧

10 Khojki

In the tapestry of Sindhi’s writing systems, the script now known as Khojki occupies a unique position. As Figure 5.1 shows, Khojki traces its origin to the unstandardised Landa inventories used in Sindh. Indeed, of the fourteen or so unstandardised Landa inventories prevalent in Sindh, Khojki and Khudawadi were the only ones that went on to be augmented, standardised and recognised as full-fledged scripts. In graphematic terms, Khojki’s ontogenesis and visual characteristics make it a sister script to Khudawadi.

In sociolinguistic terms, though, Khojki has more in common with Gurmukhi — its Landa ‘cousin’ from Punjab. While Khudawadi — as part of 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝟷𝟾𝟼𝟿 — underwent standardisation thanks to government intervention, the driving force behind standardising Gurmukhi and Khojki was their status as liturgical scripts. The status enjoyed by Gurmukhi within the Sikh community is mirrored by that of Khojki among the Nizari Ismailis of South Asia (Asani, 1987). The Nizari Ismailis are Shia Muslims who trace their roots primarily to the present-day regions of Sindh, Kutch, Gujarat and southern Punjab (Shackle & Moir, 1992). Indeed, the position of Khojki vis-à-vis the Nizari Ismaili community and the Sindhi language bears stark similarities with that of Gurmukhi’s status within the Sikh community and the Punjabi language.

The graph inventory today labelled Khojki was used in medieval times to transcribe Ismaili devotional hymns, known as ginans (§5.1.4).¹¹⁹ Like much medieval folklore of the region, the ginans originated as oral poetic compositions and were transmitted as such (Nanji, 1972, p. 24). The language of the ginans is a mélange of early New Indo-Aryan varieties of Sindhi, Kutchi, Punjabi, Gujarati and Hindi-Urdu with considerable influence from Sanskrit, Persian and Arabic (Asani, 1991, p. xi; Shackle & Moir, 1992). The “composite poetic idiom” of the ginans has been compared with that used in the Guru Granth Sahib (Shackle & Moir, 1992, p. 42). Additions, modifications and recensions make it difficult to date the ginans with certainty (Schimmel in Asani, 2002, p. xvi). While Shackle and Moir (1992, p. 15) consider the earliest ginans known today to have been orally composed in the sixteenth century, they admit that individual ginans may well comprise older elements. The oldest extant ginan in written form, which is also the oldest extant text in the Khojki script, dates from 1737 (Khakee, 1972, p. 5).

As with Gurmukhi, academic consensus on the Khojki script’s evolution from unstandardised Landa and, ultimately, from Brahmi stands in contrast to popular legends on its origin. Just as Sikh tradition talks of Gurmukhi having been created or polished by Guru Angad, Ismaili tradition considers Khojki to have been invented or refined by the fifteenth-century Ismaili missionary or pir, Pir Sadruddin (Allana, 1993 [1964], p. 37; Asani, 1987, p. 439). Moreover, just as the Guru Granth Sahib’s mixed idiom has not prevented the Gurmukhi script from becoming near-synonymous with the label ‘Punjabi’, so too has the variegated linguistic nature of the ginans not impeded the Khojki script from becoming intimately associated with the label ‘Sindhi’ (Virani, 2022). Thus, while a dispassionate academic argument may posit that Khojki is no more ‘Sindhi’ than Gurmukhi is ‘Punjabi’, popular opinion indicates otherwise.

10.1 Nomenclature

The term Khojki is the adjectival form of Khoja [kʰod͡ʑa ~ xod͡ʑa] — an alternative and commonly used appellation for South Asian Nizari Ismailis thought to be a phonological indigenisation of the Persian [xʷɒd͡ʒe] ‘master’ (Asani, 1991, p. 64; Shackle & Moir, 1992, p. 7; but see Akhtar, 2016, pp. 33–34).¹²⁰ Whereas the term Khoja has been in use for centuries, the name Khojki is arguably of relatively recent origin. Historically, the script did not have a specific name, existing as it did on a graphetic continuum of unstandardised Landa inventories (§5.1.1). Nineteenth-century European authors tend to describe the script as a community-specific calligraphic hand of Sindhi-style Landa, on par with Khudawadi. In his comparative chart of graph inventories for the Sindhi language, Stack (1849a, pp. 3–9) includes the inventory of graphetic forms today identified as Khojki under the title ‘Khwajas’. Early twentieth-century publications in the script printed using metal types simply refer to the script as ‘Sindhi’ (Devraj, 1910).

According to Virani (2022), one of the earliest mentions in writing of a term resembling [kʰod͡ʑᶦki ~ xod͡ʑᶦki] is in a Gujarati language-and-script (𝚐𝚞-𝙶𝚞𝚓𝚛) report on the annual meeting of the Khoja Ismaili Library of Karachi. The report lists the library’s Khojki-script book holdings under the label |સીંધી ખોજકી| [sin̪d̪ʱi kʰod͡ʑ(ə)ki]. A decade later, another 𝚐𝚞-𝙶𝚞𝚓𝚛 community publication featured the term |ખોજાકી| [kʰod͡ʑaki], which varied slightly in form from the previous instance. Within the academic sphere, Virani attributes the introduction of the 𝚎𝚗-𝙻𝚊𝚝𝚗 form |Khojki| to the Russian scholar Wladimir Ivanow, who wrote in 1936:

It is generally accepted that in the earliest times, they [the ginans] were not written, but simply committed to memory by the faithful; it was only much later that they were written down in Sindhi (Khojki) characters.

(Ivanow, 1936, p. 29)

In another research article a couple of years later, Ivanow (1938, pp. 57–59) once again uses the term |Khojki| to describe the liturgical script of the Ismailis. Over the following two decades, use of the name increased steadily within scholarship. Baloch’s (1962) 𝚜𝚍-𝙰𝚛𝚊𝚋 monograph contains mentions of |خوجڪي| [xod͡ʑᶦki] (pp. 107, 115), while Schimmel’s (1964) German-language article contains the Latin-script spelling |Khojki| (p. 241). Overall, though, Bhalloo and Akhtar (2018, p. 322) credit the spread of the term in academia to the first edition of Allana’s (1993 [1964]) 𝚜𝚍-𝙰𝚛𝚊𝚋 monograph on Sindhi orthography. Appearing to validate this claim is the steady rise in the scholarly use of |Khojki| and its equivalents in the years after Allana’s work, as attested in Noorally (1971), Khakee (1972), Nanji (1972) and Asani (1984). That said, the 𝚜𝚍-𝙰𝚛𝚊𝚋 form used by Allana (1993 [1964]) to describe the script is actually |خواجڪي| [xᶷwad͡ʑᶦki] and its variant |خواجڪو| [xᶷwad͡ʑᶦko] (pp. 37–39), both meaning ‘of the Khwajas’.¹²¹.

In this regard, Virani (2022) underscores the term’s origins as an externally introduced neologism. In particular, he observes that the uptake of |Khojki| has been the most vigorous in Western academia. Other sources corroborate the spread of |Khojki| within English-language scholarship (Tajddin Sadik Ali, 1989). In popular Ismaili literature, though, the term often appears juxtaposed with ‘Sindhi’, the latter being the historically prevalent emic name for the script. For Virani, the piecemeal and tentative adoption of the term |Khojki| in lay contexts points to implicit community awareness that the term is an external coinage. Virani himself refers to the script as |Khwājah Sindhi|, occasionally appending ‘Khojki’ for clarity.

In the context of the present book, adopting the term ‘Khwajah Sindhi’ for the Ismaili community’s traditional script may risk inadvertent confusion with the Sindhi language, and in the process impact reader comprehension. It would also warrant an analogous approach to the name ‘Khudawadi’ for reasons of consistency, since the latter script, too, was emically known simply as ‘Sindhi’. The interdependent nature of these onomastic considerations necessitates a thoroughgoing holistic treatment, which is beyond the scope of this book. Consequently, I retain the term ‘Khojki’ in this book purely for reasons of practicality and expedience.

Likewise, unambiguously identifying the writing system — or systems — that Khojki forms a part of proves contentious. As mentioned, much early Khojki-script writing, including the earliest ginans, are in a mixed idiom comprising various New Indo-Aryan speech varieties. If the idiom of a Khojki-script ginan is identifiably Sindhi, the resultant writing system may be labelled 𝚜𝚍-𝙺𝚑𝚘𝚓. Similarly, Khojki-script ginans whose language component is predominantly Kutchi or Gujarati may be designated 𝚔𝚏𝚛-𝙺𝚑𝚘𝚓 and 𝚐𝚞-𝙺𝚑𝚘𝚓, respectively. Despite the multilingual and fluid nature of the Khojki-script ginans, there are observable patterns and conventions to be found in their graph inventory and graph-phone correspondences. Collectively, therefore, the writing system featured in written ginans may be referred to as 𝚖𝚞𝚕-𝙺𝚑𝚘𝚓.¹²² By the nineteenth century, however, the increasing socioeconomic status of the Gujarati language within the Ismaili community led to a shift towards publishing primarily in Gujarati, albeit still in the Khojki script (§5.2.4). In fact, apart from the ginans, most Khojki-script content issued from the printing presses set up at the turn of the twentieth century was in 𝚐𝚞-𝙺𝚑𝚘𝚓. Notably, many of these 𝚐𝚞-𝙺𝚑𝚘𝚓 publications were labelled as being in ‘Sindhi’ (Devraj, 1910), reflecting the then-prevalent name for the script and affirming Virani’s (2022) arguments in this regard.

Since the language of most twentieth-century Khojki-script publications is not Sindhi but Gujarati, it may be argued that an analysis of such works lies outside the scope of this book. Furthermore, the liturgical significance of Khojki for the Ismaili community has led to the script and its content being extensively researched from graphetic, graphematic and sociolinguistic perspectives (Allana, 1993 [1964]; Asani, 1987, 1991, 1992; Bruce, 2015; Khakee, 1972; Moir, Shackle & Mitha, n.d.; Shackle & Moir, 1992, Tajddin Sadik Ali, 1989). Yet, the close indexical association of the label ‘Sindhi’ with the Khojki script is a strong argument in favour of the latter’s inclusion in a book on Sindhi’s writing systems. Given the considerable volume of grapholinguistic research that already exists on Khojki (albeit not under the rubric of grapholinguistics), this chapter will be selective in its scope, and prioritise those aspects of the script and its derivative writing systems that reinforce the findings of the other chapters in this book.

10.2 Graphematic foundations and early use

As with 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍, Khojki-based writing exhibits a high degree of variation, be it in graph-phone correspondences or in the visual appearance of individual graphs. However, the neat graphematic distinction between 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍 and 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝟷𝟾𝟼𝟿 does not have a parallel in Khojki-based — that is, 𝚖𝚞𝚕-𝙺𝚑𝚘𝚓 — subvariants. Instead, graph-phone correspondences and graphetic features are seen to evolve gradually along a chronological continuum. At the same time, the advent of Khojki-script printing in using metal types at the turn of the twentieth century ushered in several graphetic innovations and graphematic standardisations in 𝚖𝚞𝚕-𝙺𝚑𝚘𝚓. The rise of Khojki-script printing, thus, represents a watershed moment in the evolution of 𝚖𝚞𝚕-𝙺𝚑𝚘𝚓, and will be used as a reference point to delineate the system’s subvariants (Shackle & Moir, 1992, p. 36) into 𝚖𝚞𝚕-𝙺𝚑𝚘𝚓-𝚡-𝚝𝚛𝚊𝚍 and 𝚖𝚞𝚕-𝙺𝚑𝚘𝚓-𝚡-𝚖𝚘𝚍𝚎𝚛𝚗. The former subvariant encompasses the highly variable graph inventory, graph-phone correspondences and spellings observed in manuscripts and early lithographs, while the latter denotes the quasi-standardised graph inventory, graph-phone correspondences and spellings attested after Khojki began to be printed using movable type. Consequently, Khojki-script texts composed in the twentieth and twenty-first centuries tend to reflect 𝚖𝚞𝚕-𝙺𝚑𝚘𝚓-𝚡-𝚖𝚘𝚍𝚎𝚛𝚗’s inventory and graph-phone correspondences, unless explicitly intended to be a facsimile of a 𝚖𝚞𝚕-𝙺𝚑𝚘𝚓-𝚡-𝚝𝚛𝚊𝚍 text.

Despite the variability inherent in 𝚖𝚞𝚕-𝙺𝚑𝚘𝚓-𝚡-𝚝𝚛𝚊𝚍 inventories, an overall visual coherence was attestable. Moreover, manuscripts in the system were often appended with a tabular chart of forty graphs — whence the name chaliha akhari ‘forty-letter script’ (§5.1.4). An idealised chaliha akhari table as listed in Shackle and Moir (1992, p. 35) is reproduced in (26).

Manuscripts would traditionally commence with the 𝚜𝚍-𝙺𝚑𝚘𝚓 benediction |𑈀𑈧𑈬𑈪 𑈙𑈲 𑈁𑈪𑈬𑈦| [əlaɦ(ᵊ) t̪o aɦaɾ(ᵊ)] ‘with the help of Allah’ (Asani, 1992, p. 44). Alternatively, the benediction might be appended to the table, in which case it was often recast as the five-unit graphematic string |𑈀 𑈧𑈬 𑈙𑈲 𑈪𑈬 𑈦| to align with the five-by-eight matrix (Shackle & Moir, 1992, p. 35). Table 10.1 shows the handwritten forms of Khojki graphs and their phonological values as featured in Stack’s Sindhi grammar (1849a, pp. 3–8). Also shown are the phonological values of the graphs attested in other sources (Asani, 1991, pp. 51–72; Khakee, 1972, pp. 598–607; Pandey, 2011b; Shackle & Moir, 1992, pp. 35–42). Depending on graphematic environment, every graph that canonically denotes a φ-consonant may also correspond to φ-[CV₀], where φ-[V₀] is the inherent or default φ-vowel for the language represented. For most languages transcribed in Khojki, including Sindhi-Kutchi and Gujarati, the default vowel is [ə].

Stack (1849a)	Modern typeface	Common value(s) in IPA	Alternative values in IPA
	𑈀	ə
	𑈁	a
	𑈄 (𑈀𑈭 )	ɪ	e
	𑈂	i	ɪ e j
		ʊ	u o
	𑈃	u	ʊ o
	𑈄 (𑈀𑈭 )	e	ɪ (ɛ)
	𑈀𑈄	əi̯ (ɛ)
		o	(ɔ)
	𑈀	əʊ̯ (ɔ)
	𑈈	k	(q)
	𑈉	kʰ	(x)
	𑈊	ɡ	ɡʱ ɠ (ɣ)
	𑈋	ɠ	ɡ ɡɾ
	𑈌	ɡʱ	ɡ ɠ
	𑈓 (𑈍)	ŋ ɲ
	𑈎	t͡ɕ
	𑈏	t͡ɕʰ
	𑈐	d͡ʑ	d͡ʑʱ ʄ (z)
	𑈑	ʄ	d͡ʑ d͡ʑʱ ɲ (z)
	(𑈐?)	d͡ʑʱ	d͡ʑ ʄ (z)
	𑈔 (𑈕)	ʈ ʈɾ ʈʰ
	𑈙𑈵𑈦	ʈ ʈɾ	t̪ɾ
	𑈘 (𑈖)	ɽ	ɖ ɳ (ɽ̃)
	𑈜	ɗ	ɖ d̪ d̪ʱ
	𑈛𑈵𑈦	ɖɾ	ɗ d̪ d̪ɾ d̪ʱɾ
	𑈗	ɖʱ	ɖ
	𑈙	t̪
	𑈚	t̪ʰ
	𑈛	d̪	d̪ʱ
	𑈝	d̪ʱ	d͡ʑʱ d̪ (z)
	𑈞	n (n̪)
	𑈟	p
	𑈠	pʰ	(f)
	𑈡	b	ɓ bʱ
	𑈢	ɓ	b bʱ
	𑈣	bʱ	b ɓ
	𑈤	m
	𑈦	ɾ
	𑈧	l
	𑈨	ʋ
	𑈩	s	(ɕ)
	𑈪	ɦ

Absent from Stack’s chart are 𝚖𝚞𝚕-𝙺𝚑𝚘𝚓-𝚡-𝚝𝚛𝚊𝚍’s bound γ-vowels, as well as variant forms of free γ-vowels. These are shown in (27), along with their respective phonological values (Asani, 1991, pp. 51–72; Khakee, 1972, pp. 598–607; Pandey, 2011b; Shackle & Moir, 1992, pp. 35–42).

Taken together, Table 10.1 and (27) reveal how the resemblance between Khojki and its Landa relatives goes deeper than mere visual or graphetic harmony. Similar to writing systems based on Khudawadi and Gurmukhi, 𝚖𝚞𝚕-𝙺𝚑𝚘𝚓-𝚡-𝚝𝚛𝚊𝚍, too, employs three distinct graphetic bases to form graphs for low, high front and high back φ-vowels (§5.1.4). In this regard, 𝚖𝚞𝚕-𝙺𝚑𝚘𝚓 |𑈀 𑉀 𑈆| are equivalent to 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍 |𑊰 𑊲 𑊴| and 𝚙𝚊-𝙶𝚞𝚛𝚞 |ਅ ੲ ੳ|, respectively. Where 𝚖𝚞𝚕-𝙺𝚑𝚘𝚓-𝚡-𝚝𝚛𝚊𝚍 diverges from 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍 while further resembling early 𝚙𝚊-𝙶𝚞𝚛𝚞 is in its use of bound γ-vowels in postconsonantal position (Shackle & Moir, 1992, p. 37). The use of bound γ-vowels can be attributed to the sustained use of the writing system in liturgical contexts, which incentivised organic development of transparency and consistency in graph-phone correspondences. Indeed, the use of bound γ-vowels distinguished 𝚖𝚞𝚕-𝙺𝚑𝚘𝚓-𝚡-𝚝𝚛𝚊𝚍 from the various unstandardised Landa-based writing systems used in pre-British Sindh. In his Sindhi grammar, Stack (1849a, p. 2) notes that, of all the communities in Sindh that wrote in Landa-based inventories at the time, “[t]he Khwájás alone occasionally use the lákaná, or medial vowel marks”.¹²³

Yet, the distribution between free and bound γ-vowels was not mutually exclusive or exhaustive. Early manuscripts typically exhibited an inventory of six bound γ-vowels, shown in (27) (Shackle & Moir, 1992, p. 37). However, the phonological values of these bound γ-vowels were variable, as was their supposed allography with free γ-vowels. For instance, a particular 𝚖𝚞𝚕-𝙺𝚑𝚘𝚓-𝚡-𝚝𝚛𝚊𝚍 text might feature bound |⁠𑈭 ⁠𑈮| [ɪ ~ i] but only free |𑈂|, resulting in a theoretical many-to-one correspondence between bound and free γ-vowels. Moreover, |⁠𑈭| might also denote values other than [ɪ ~ i], resulting in further disparity with the idealised Indic alphasyllabary paradigm. With time, though, the crystallisation of graph values and distribution resulted in 𝚖𝚞𝚕-𝙺𝚑𝚘𝚓-𝚡-𝚝𝚛𝚊𝚍 gradually featuring mutually exclusive free and bound γ-vowel pairs. Where necessary, new graphs were developed by combining rasmic bases. Thus, |𑈄| [ɪ ~ e] is likely the result of a graphetic merger between |𑈀| and |𑈭 | (Khakee, 1972, p. 604). Similarly, |𑈃| and the near-identical || appear to be fusions of |𑈆| and |⁠𑈯|. Sociolinguistic pressure on the inventory’s scope also resulted in the introduction of Devanagari-inspired |⁠𑈱 ⁠𑈳| for [əɪ̯ ~ ɛ] and [əʊ̯ ~ ɔ], respectively. These bound γ-vowels were then used to backform the corresponding free γ-vowels |𑈅 𑈇| as alternatives to the digraphic |𑈀𑈄 𑈀𑈆|, respectively. That said, |⁠𑈱 ⁠𑈳| and their free counterparts are attested only rarely, with most texts employing |⁠𑈰 ⁠𑈲| [e o] and their free allographs in their stead (Asani, 1991, pp. 51–72; Khakee, 1972, pp. 598–607; Pandey, 2011b; Shackle & Moir, 1992, pp. 35–42).

Notwithstanding the gradual normativisation of 𝚖𝚞𝚕-𝙺𝚑𝚘𝚓-𝚡-𝚝𝚛𝚊𝚍 free γ-vowels, allographs and graphetic variations were particularly noticeable in the composition of graphs denoting high back φ-vowels. For instance, Allana (1993 [1964], p. 39) lists | | as the free γ-vowels for [ʊ ~ u]. While || may be considered a graphetic allograph of |𑈃| or || from Stack’s chart (Table 10.1), || has no parallel. Compositionally, it comprises the Khojki rasm |𑈆| juxtaposed with Gujarati-script |⁠ૂ| within the same graphosegmental space. Notably, Khojki || appears compositionally equivalent to the Gujarati-script free γ-vowel |ઉૂ |, which is listed by Bagster (1851, p. xlv) as the 𝚐𝚞-𝙶𝚞𝚓𝚛 graph for [u]. Much like Khojki ||, Gujarati-script |ઉૂ | features the rasm |ઉ| [ʊ ~ u] combined with the bound γ-vowel |ૂ| [u]. While the Khojki-script graph || listed by Allana is graphetically and graphematically plausible, Asani (1991, pp. 58–59) notes that it does not appear to have been attested in any 𝚖𝚞𝚕-𝙺𝚑𝚘𝚓-𝚡-𝚝𝚛𝚊𝚍 manuscript. Asani makes a similar observation regarding || [o] listed by Stack (Table 10.1).

The phonological polyvalency of 𝚖𝚞𝚕-𝙺𝚑𝚘𝚓-𝚡-𝚝𝚛𝚊𝚍’s γ-consonants has also been the subject of much scholarly attention. Based on Asani (1991, pp. 60–62), Pandey (2011b, p. 3) attributes the use of a single graph with multiple phonological values to the gradual spread of the proto-standardised Khojki inventory across regions with different dominant languages:

A reason for such [graphematic] ambiguity is the use of Khojki outside of Sindh by scribes unfamiliar with the Sindhi language. When non-Sindhi speakers wrote Khojki, certain characters [i.e., graphs] that represented sounds specific to Sindhi, such as the implosive consonant letters [denoting ɠ(ə) ʄ(ə) ɗ(ə) ɓ(ə)] lost their original phonetic values. As a result of this neutralization, the original glyphs for these letters were reassigned to other consonants [i.e., consonant phones] within the same articulation class.

(Pandey, 2011b, p. 3)

Pandey, thus, observes that Khojki texts with content in Sindhi or Kutchi — that is, texts where the writing system is 𝚜𝚍-𝙺𝚑𝚘𝚓 or 𝚔𝚏𝚛-𝙺𝚑𝚘𝚓 — tend to feature the graphs |𑈋 𑈑 𑈜 𑈢| with the phonological values [ɠ(ə) ʄ(ə) ɗ(ə) ɓ(ə)]. However, the phonological inventories of Gujarati, Hindi-Urdu and related South Asian languages do not feature φ-implosives. As a result, when Gujarati or Hindi-Urdu content was transcribed in Khojki, the graphs |𑈋 𑈑 𑈜 𑈢| were often ‘recycled’ and used to denote other φ-consonants in the relevant languages (Asani, 1991, p. 61; Shackle & Moir, 1992, p. 36). Thus, |𑈜|, which typically denoted [ɗ] in a 𝚜𝚍-𝙺𝚑𝚘𝚓 or 𝚔𝚏𝚛-𝙺𝚑𝚘𝚓 text, often took on the value [d̪] in a 𝚐𝚞-𝙺𝚑𝚘𝚓 text. Against this background, Asani (1991, pp. 61–62) notes that a correct phonological interpretation of a Khojki-script text often necessitates knowledge of the geographical region in which the text originated, in order to arrive at the likely speech variety represented in the text. This allows for the text to be decoded by applying graph-phone correspondences typical of that language. The phenomenon of distinct graph-phone correspondences based on the language denoted holds true not just for manuscripts, but also for print publications in Khojki. Thus, in the 𝚔𝚏𝚛-𝙺𝚑𝚘𝚓 text titled |𑈩𑈙𑈨𑈰𑈘𑈮 𑈨𑈜𑈮 𑈙𑈚𑈬 𑈞𑈭𑈴𑈗𑈮| [sət̪(ᵊ)ʋeɳi ʋəɗi t̪ət̪ʰa nɪɳɖʱi], |𑈜𑈮| denotes [ɗi], whereas in the 𝚐𝚞-𝙺𝚑𝚘𝚓 text titled |𑈞𑈬𑈞𑈲 𑈜𑈩 𑈀𑈨𑈙𑈬𑈦 𑈟𑈮𑈦 𑈩𑈜𑈦𑈜𑈮𑈞 𑈞𑈲 𑈦𑈎𑈰𑈧𑈲| [nano d̪əs əʋ(ə)t̪aɾ piɾ səd̪əɾd̪in no ɾət͡ɕelo], |𑈜 𑈜𑈮| indicate [d̪ə d̪i], respectively (Bruce, 2015, p. 45).

Yet, while graph-phone correspondences in Khojki-script texts display observable patterns of conditioning by the language being encoded, such patterns are not predictable or consistent. This means that a Khojki-script text featuring content in a language other than Sindhi-Kutchi — and which, therefore, does not feature φ-implosives — may feature |𑈋 𑈑 𑈜 𑈢| assigned to a host of different φ-consonants, depending on the scribe and their dominant language. For instance, the graph |𑈑| may denote [d͡ʑ(ə)] or [d͡ʑʱ(ə)] in a 𝚐𝚞-𝙺𝚑𝚘𝚓 text. In addition, Khakee (1972, p. 602) notes that a similar-looking graph is used to indicate [ɲ] in the oldest extant Khojki-script text, dating from 1737. Graphematically, the practice of using the same graph for [ʄ] as well as [ɲ] bears similarities to 𝚜𝚍-𝙰𝚛𝚊𝚋-𝚡-𝚝𝚛𝚊𝚍’s use of |ڃۡ| for [ʄ] and the graphetically similar |نڃۡ| for [ɲ] (Table 6.3). One may also conjecture that the graphetic similarity between 𝚐𝚞-𝙶𝚞𝚓𝚛 |ઝ| [d͡ʑʱ(ə)] and |ઞ| [ɲ(ə)] may have played a part here.

Related to |𑈑| is the graphetically similar |𑈝|, depicted in Stack’s chart as denoting [d̪ʱ(ə)]. However, a similar-looking graph has also been attested in manuscripts with the value [d͡ʑʱ(ə)] (Allana, 1993 [1964], p. 39; Asani, 1992, p. 681). Phonologically, [d̪ʱ] and [d͡ʑʱ] are fairly distinct in their places and manners of articulation. Hence, claiming that they would be written with one and the same graph merits further scrutiny. Pandey (2011b, p. 13) hypothesises that certain handwritten instances of |𑈝| [d͡ʑʱ(ə)] are effectively graphetic allographs of |𑈐| [d͡ʑ(ə)]. Offering support to this hypothesis is the lack of graphematic distinction between unaspirated and aspirated voiced φ-plosives observed in Khojki-script texts, and in unstandardised Landa-script texts more broadly. Graphematic underdifferentiation between [d͡ʑʱ(ə) d͡ʑʱ(ə)] is also attested in Stack’s chart (Table 10.1), which lists near-identical graphs for the two.

Other alternations such as the use of |𑈘 𑈖| for one or more of [ɖ ɽ ɳ (ɽ̃) (ɽʱ)] seems plausible when viewed against the fact that [ɖ ɽ] are not graphematically distinguished in many Indic-script-based writing systems (but see footnote 84). In systems that do have separate graphs for [ɖ ɽ], such as 𝚜𝚍-𝙳𝚎𝚟𝚊 |ड ड़|, the graphs themselves are rasmically identical and differ only subsegmentally. Moreover, given that [ɳ] is often realised intervocalically as [ɽ̃] in several Indo-Aryan languages (§4.3.1), the phonological distribution in question lies well within the realms of possibility.

Certain graphs in the 𝚖𝚞𝚕-𝙺𝚑𝚘𝚓-𝚡-𝚝𝚛𝚊𝚍 inventory bear graphetic similarities to their canonical equivalents or near-equivalents in other inventories. Thus, 𝚖𝚞𝚕-𝙺𝚑𝚘𝚓 |𑈉| [kʰ(ə)] corresponds graphetically and phonologically to 𝚜𝚍-𝚂𝚒𝚗𝚍 |𑊻| [kʰ(ə)], with both corresponding graphetically — but not phonologically — to 𝚜𝚊-𝙳𝚎𝚟𝚊 |ष| [ʂa] (Table 8.1).

Stack’s chart (1849a, p. 7) lacks a distinct graph for [j(ə)], listing instead the graph |𑈐| against it. This means that, in 𝚖𝚞𝚕-𝙺𝚑𝚘𝚓-𝚡-𝚝𝚛𝚊𝚍, |𑈐| could denote one or more of [d͡ʑ(ə) d͡ʑʱ(ə) j(ə)]. However, this mapping is consistent with that of 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝚝𝚛𝚊𝚍 |𑋘|, which could denote one or more of [d͡ʑ(ə) d͡ʑʱ(ə) j(ə)] (Table 8.1). In later 𝚖𝚞𝚕-𝙺𝚑𝚘𝚓-𝚡-𝚝𝚛𝚊𝚍 texts, [j(ə)] is denoted by |𑈥|, which likely emerged as a spin-off of the free γ-vowel |𑈂| [ɪ i] (Asani, 1991, p. 61).

Much like 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝟷𝟾𝟼𝟿 and contemporary 𝚜𝚍-𝙳𝚎𝚟𝚊 practice, 𝚖𝚞𝚕-𝙺𝚑𝚘𝚓-𝚡-𝚝𝚛𝚊𝚍 uses the same bound graph to denote the nasalisation suprasegmental on a φ-vowel as well as a nasal φ-consonant that forms a cluster with a homorganic oral φ-consonant. However, the use of this graph is erratic, especially in early 𝚖𝚞𝚕-𝙺𝚑𝚘𝚓-𝚡-𝚝𝚛𝚊𝚍. Graphetically, it may appear as a dot or as a circle (Shackle & Moir, 1992, pp. 38–40), mirroring similar practice in 𝚐𝚞-𝙶𝚞𝚓𝚛 at the time (Grierson, 1908, p. 338). There was no dedicated virama in early 𝚖𝚞𝚕-𝙺𝚑𝚘𝚓-𝚡-𝚝𝚛𝚊𝚍. Some early writers added a dot-like graph above a γ-consonant to function as a makeshift virama, which made it liable to misinterpretation as an anusvara (Asani, 1991, p. 63; Shackle & Moir, 1992, p. 37). Later manuscripts and early print texts often used |⁠𑈭 | with the function of a virama (see (27)). Thus, 𝚖𝚞𝚕-𝙺𝚑𝚘𝚓-𝚡-𝚝𝚛𝚊𝚍 |𑈟𑈭𑈦𑈰𑈤| may be decoded as [pᶦɾem] or [pɾem] (Asani, 1991, p. 63), and |𑈈𑈭𑈥𑈯𑈴| as [kᶦjũ] or [kjũ] (Shackle & Moir, 1992, p. 41). Phonologically, this practice reflects the free alternation of [Ø] with epenthetic [ɪ] in many spoken Sindhi varieties (§4.3.1). This practice is also consistent with the use of 𝚜𝚍-𝙰𝚛𝚊𝚋 | ⁠◌ِ |, 𝚜𝚍-𝙳𝚎𝚟𝚊 |⁠ि|, 𝚜𝚍-𝚂𝚒𝚗𝚍-𝚡-𝟷𝟾𝟼𝟿 |⁠𑋡| and 𝚜𝚍-𝙶𝚞𝚛𝚞 |⁠ਿ|, all of which canonically denote |ɪ|, in near-free variation with their respective graphs for [Ø] (jazm or virama).

Later manuscripts also featured bound forms of the γ-consonants |𑈥 𑈦| [j(ə) ɾ(ə)] when denoting the second element of a φ-consonant cluster, as in |𑈈𑈵𑈥| [kj(ə)] and |𑈩𑈵𑈦| [sɾ(ə)] (Asani, 1992; Pandey, 2011b). Depending on the extent of graphetic integration, the bound γ-consonant may be classified as a ligature.

Numerals in 𝚖𝚞𝚕-𝙺𝚑𝚘𝚓, |૧ ૨ ૩ ૪ ૫ ૬ ૭ ૮ ૯ ૦|, may be considered identical to those in 𝚐𝚞-𝙶𝚞𝚓𝚛 (Moir, Shackle, & Mitha, n.d., p. 27; Pandey, 2011b, p. 10), which are themselves near-identical to numerals in other Landa inventories (see Example (21)) and in Devanagari (see Example (14)). Manuscripts feature expected graphetic variation in numeral shapes. Manuscripts in 𝚖𝚞𝚕-𝙺𝚑𝚘𝚓-𝚡-𝚝𝚛𝚊𝚍 were also rich in indigenous punctuation practices attested in other Indic-script-based writing systems. These included the use of |𑈸 𑈹| to mark the end of sentences and verses, respectively (Pandey, 2011b, p. 9). At the same time, 𝚖𝚞𝚕-𝙺𝚑𝚘𝚓-𝚡-𝚝𝚛𝚊𝚍 diverged from its contemporaries in using the colon-like graph |𑈺| as a word separator (Asani, 1992, p. 44). While graphetically similar to the visarga in several Indic-script inventories, 𝚖𝚞𝚕-𝙺𝚑𝚘𝚓-𝚡-𝚝𝚛𝚊𝚍’s lack of a dedicated visarga ruled out any potential conflation (Pandey, 2011b, p. 9). Manuscripts also featured |𑈻 𑈼|, lengthened as needed, to fill the space between the end of a sentence or paragraph and the page margin (Pandey, 2011b, p. 9).