Alternative criteria for writing system typology

2.1 Descriptive perspective of existing typologies

No matter how abridged by necessity, any outline of influential typology proposals must acknowledge, at least briefly, Gelb’s (1969 [1952]) seminal classification, which essentially set the trajectory for most subsequent proposals. Gelb’s typology consisted of five categories in total, but it also implemented a fundamental division between two main categories. Thus, although the first category included both (1) pictorial representations and (2) mnemonic devices, which Gelb regarded as forerunners of writing, it was differentiated from a second category of full writing. The full-writing category was further subdivided into the three classes of (3) word-syllabic (mixing logography and syllabography), (4) syllabic, and (5) alphabetic. However, despite the key insight regarding the significance of full writing, Gelb’s typology was undeniably flawed by his misguided zeal to portray the evolution of writing as teleological in nature, holding that all writing systems invariably transform via logography and syllabary to become alphabets in their final stages (Coulmas 1996a; Daniels 1990; 2001).

The next classification proposal that warrants singling out is that advanced by Sampson (2015 [1985]), because, although it largely conforms structurally with Gelb’s (1969 [1952]), even if the similarities are partially obscured by terminological differences, it has also been of special importance for the historical development of writing systems research. Within Sampson’s classification, the first division is between semasiographic (included on the basis of conjecture and, thus, not subdivided further) and glottographic writing systems, with the latter divided into logographic and phonographic writing systems. Even though Sampson acknowledges that writing systems based on polymorphemic units do not exist, also on a conjectural basis, the logographic category is subdivided into polymorphemic and morphemic units. Finally, the phonographic category is divided into the three subcategories of syllabic, segmental and featural. Of these, the first two of syllabic and segmental correspond to Gelb’s (1969 [1952]) categories of syllabic and alphabetic, respectively, with segmental being a preferable term that avoids labelling a category after its exemplar, but Sampson proposed the featural category solely for Korean Hangŭl, based on the correlations between graphic components and phonetic features.

Although the classification proposed and developed by Daniels (1990; 2001; 2018) has, arguably, been the most influential over the last three decades, it too is firmly rooted in the legacy of Gelb (1969 [1952]). Indeed, as Daniels readily acknowledges, the initial motivations for his typology were to resolve the inadequacies that he perceived in Gelb’s tripartite classification, which stemmed directly from Gelb’s fallacious belief in the teleological nature of writing system evolution. Thus, more specifically, while Daniels’ typology has always recognized Gelb’s three categories of full writing (albeit with slightly modified terms)—(1) logosyllabary (morphosyllabary), (2) syllabary, and (3) alphabet (Greek-type script), respectively—it also expanded on that range by including two other categories. Modeled on the term alphabet, Daniels (1990) coined two new terms for these categories. The first was abjad (Semitic-type script) to refer to writing systems where each character stands for a consonant (with the term derived from the first signs of the Arabic script). The second new term was abugida (Sanskrit-type script), where each character stands for a consonant accompanied by a particular vowel with other vowels indicated by additions to the character (with the term derived from the initial signs of Ethiopic). Finally, it should also be noted that prior to Daniels (2018), the classification has also recognized featural as a sixth category, but it now treats Korean Hangŭl as an alphabet.

The last typology proposal to include within the present selective outlines is that recently advanced by Gnanadesikan (2017), because, while it focuses solely on phonemic writing systems, it seeks to categorize them according to more detailed classification descriptions. Significantly, Gnanadesikan treats all phonemic writing systems as segmentary types, defined as “a script all or most of whose signs are used in such a way as to encode individual phonological segments, or phonemes” (Gnanadesikan 2017: 21). Accordingly, the range of typology categories is determined by a small set of factors, including phonological unit size (i. e., syllable, mora, feature), representation of higher-order structure (i. e., syllabically arranged/space, linear) and degree of vowel inclusion (i. e., all, most, some, none). To illustrate the more nuanced descriptions that this typology yields compared to previous proposals, the Phoenician writing system is, for example, categorized as “consonantal linear segmentary”, Devanāgarī as “mostly vowelled āksharik segmentary”, and Hangŭl as “fully vowelled syllabically arranged featural segmentary” (Gnanadesikan 2017: 29).

2.2 From descriptive to evaluative considerations

While both select and brief in nature, the short outlines of influential typologies in Section 2.1 vividly underscore one key point; all typology proposals to date have adopted the notion of representational mapping as either their sole or primary criterion for classifying writing systems. Indeed, the very distinctions between writing system types (abstract sense) hinge on identifying the dominant level of representational mapping. Such descriptive classifications certainly highlight the crucial relations between a writing system (or, rather, what are assumed to be its basic, or “default”, units) and the language represented (or, rather, its linguistic units). They are, however, highly reductive in nature. Firstly, by focusing only on the dominant level of representational mapping, such as the grapheme-phoneme-mappings of segmentaries (Gnanadesikan 2017), other levels of representation (for example, morphography) are either neglected or completely discounted. This leads to the situation where typologies generally fail to account for the fact that most writing systems are mixed in nature (Günther 1988: 43). As languages and scripts both undergo changes over time, the consistency of the basic mapping relations between them can become greatly attenuated, particularly in terms of lexical consistency (Sampson 2018). The mixed nature of writing systems is discussed further in Section 3.1 with respect to both the GWS and the JWS. A second reason why the exclusive focus of typologies on representational mappings is problematic is that it fails to acknowledge features of writing systems that are unrelated to representation, such as those that emerge when using writing systems for writing and reading. Section 3.2 considers how different approaches to marking visual boundaries influence the psychological processes of visual word recognition within reading.

As with general linguistics, description is foundational for grapholinguistics, but evaluation has, for quite some time, also assumed a prominent role within the relevant literature. Unlike languages, which are seldom compared in terms of “quality”, writing systems have often been the targets of various kinds of evaluative assessments, as epitomized by Rogers’ (1995: 31) claim that “some writing systems are better than others”. Indeed, the notions of assessing and contrasting the ‘quality’ of diverse writing systems are blatantly manifest in a number of criteria lists, ranging from teleological assumptions (Gelb 1969 [1952]) to unequivocally evaluative labels, such as “ideal” (Bauernschmidt 1980), “optimal” (Rogers 1995), “practical” (Venezky 1977) and even “perfect” (Venezky 2004). Moreover, although some criteria lists have emerged within the context of literacy development to explicitly guide the creation of writing systems for yet unwritten languages (i. e., Smalley 1964; Bauernschmidt 1980; Cahill 2014), others seek to explain how different characteristics of writing systems influence processing, especially reading (i. e., Daniels and Share 2018). Highly illustrative of the second motivation, Daniels and Share (2018: 101) enumerate ten “dimensions of complexity” within writing systems that they identify as posing particular processing challenges, especially for dyslexics. The dimensions are (1) linguistic distance, (2) nonlinearity, (3) visual complexity, (4) historical change, (5) spelling constancy despite morphophonemic alternation, (6) omission of phonological elements, (7) allography, (8) dual purpose letters, (9) ligaturing, and (10) inventory size. However, the list’s major flaw is its unsystematic nature, in failing to distinguish between criteria that differ categorically. Consequently, synchronic features, such as nonlinearity, are mixed with diachronic ones, such as historical change. Moreover, some graphetic features (i. e., pertaining to formal and material aspects), such as visual complexity, are mixed with graphematic ones (i. e., pertaining to the relation between graphic and linguistic units), such as omitting phonological elements, while other features are simultaneously relevant at both levels, such as allography and inventory size, which can be interpreted both graphetically and graphematically. Naturally, no list claims to be exhaustive, which is undoubtedly true. For instance, consistent with its focus on processing issues, criteria that are elsewhere treated as “sociolinguistic” (Bauernschmidt 1980) and “cultural” (Rogers 1995) attributes are completely absent from Daniels and Share’s (2018) dimension list.

3.1 Linguistic fit: Mixed representational levels

As Frost (2012: 266) provocatively claims, “every language gets the writing system it deserves”. Naturally, by its very formulation, the comment is overtly evaluative. What does it mean to claim that a language “deserves” a particular writing system? Most succinctly, it expresses the conviction that a writing system should adequately accommodate the specific features of a language, which is precisely what linguistic fit attempts to encapsulate. As a visually-mediated semiotic system, a writing system must be appropriate to the specific language. Obviously, as the sections on both processing and sociocultural fits underscore below, linguistic fit is not the only basis for evaluating writing systems, and is possibly not even the most important.

As a broad category, linguistic fit encompasses the criteria, or parameters, that pertain to the relations between a writing system and ‘its’ language (Meletis 2018). One criterion that has been examined extensively is transparency. Graphematic units (ranging from individual graphemes up to word-length strings) are considered to be transparent if they relate to a single linguistic unit. For instance, the grapheme combination <gh> in English is not transparent, because, at the phonographic level, it can represent a range of phonemes (e. g., /f/ in <enough> and /g/ in <ghostly>). Neef and Balestra (2011) have proposed a value of graphematic transparency (gt-value) as an index of phonographic transparency. In addition to outlining how to assess grapheme transparency for a particular writing system, they also demonstrate the notion’s application by comparing the German and Italian writing systems. However, as gt-values are only applicable to segmental phonographic writing systems, resonating directly with the issues outlined above, evaluating transparency is only possible in terms of the dominant level of representational mapping. For alphabets, such as German and Italian, the mapping is primarily of segmental graphemes onto segmental phonemes. Thus, once more, this approach to assessing transparency effectively neglects the extent to which a writing system may relate to other linguistic units, whether at the syllabic or morphemic levels. Moreover, similar limitations also apply to Katz and Frost’s (1992) influential orthographic depth hypothesis, which only treats writing systems that are phonographically transparent as being orthographically shallow (i. e., transparent). Accordingly, the Chinese writing system is evaluated as being orthographically deep; an assessment that completely ignores the fact that at the morphographic level, which is highly relevant for reading, it is predominately transparent. Measurements of transparency must be “multidimensional” (echoing the sentiments of Daniels and Share 2018). Such measures must acknowledge that (a) writing systems can be transparent with respect to multiple linguistic levels and that (b) different kinds of transparency are mutually exclusive. These points can be illustrated by contrasting the GWS and the JWS.

It seems quite uncontroversial to classify the GWS as an “alphabet”, i. e. a segmental phonographic writing system with separate graphemes for both consonant and vowel phonemes. It is, thus, not surprising that analyses have focused on the correspondences between graphemes and phonemes. However, most descriptions of the GWS also acknowledge its so-called morphological principle (Dürscheid 2016: 143–146; Karg 2015: 58–60). One pertinent phenomenon is the graphematic non-representation of final (obstruent) devoicing in German. For example, even though the phonological representation of the word Hund ‘dog’ is /hʊnt/, it is written as <Hund> with a final <d> (which graphematically relates to /d/), rather than *<Hunt>, which would be the graphematically transparent spelling. This is because all the other forms across the lexeme’s inflectional paradigm, such as the genitive singular Hunds ‘dog.gen’ or the nominative plural Hunde ‘dogs’, include the phoneme /d/ and are transparently written with <d>. Thus, a preference for morphographic consistency across the paradigm is stronger than a propensity for phonographic transparency, highlighting how the two are (often) in conflict. Recently, as the autonomous paradigm within grapholinguistics has gained greater credence, more serious scrutiny has been given to the notions that phonographic representational mappings are central and that morphography should be regarded as either a secondary “intrusion” (if spun negatively) or a “surplus” (if its benefits are stressed), which are effectively reinforced by crude typological classifications such as “alphabet” (Schmidt 2018; Berg 2019). The crux of these criticisms lies in the realization that polysegmental correspondences, rather than segmental phonographic correspondences, should be the basis of graphematic analyses, and that the most important correspondences are those that exist between the written and spoken forms of words. This view holds that morphology and syntax are both modality-neutral aspects of language (Berg 2019: 5), because words can be realized as either speech or writing (or, in the case of sign languages, as signing). Taking the argument further, segmental phonographic correspondences should not be regarded as being primary but rather as being epiphenomenal; the reason why the German word klar ‘clear’ (/klaːɐ̯/) is graphematically represented as <klar> (i. e., includes <r>) is simply because klare ‘clear.nom.f.sg’ /ˈklaːʁə/ (with /ʁ/) is written <klare>. As Schmidt (2018: 78, our translation) points out, “the reference system for the interpretation of individual letters and their correspondences is knowledge about [...] pairs of graphematic and phonological words” (cf. also Berg 2019: 7). It also bears mention that blanket classifications, such as “alphabet”, fail to capture the syllab(ograph)ic level, which also plays a role in many segmental phonographic writing systems, as further discussions of the GWS in Section 3.2 observe.

Consistent with the common secondary sense of writing system (Section 1), the JWS refers to the complete set of graphic symbols used to represent written Japanese. As the usual Japanese term <漢字仮名交じり文> /kan-ji-ka-na-ma.jiri.bun/^[1] ‘mixed kanji and kana writing’ patently signifies, the basic graphematic convention, as outlined in Section 3.2, is to simultaneously employ the component scripts in complementary ways (Joyce and Masuda 2018; 2019). Interestingly from the perspective of writing systems research, the JWS constitutes a unique approach to mixing representational mapping principles as materialized by its blend of four scripts. The core script is morphographic kanji (Joyce 2011; 2016), but the morphography of kanji is rather complex, as kanji are associated with both Native-Japanese (NJ) and Sino-Japanese (SJ) morphemes. For example, <山> means ‘mountain’ and is associated with both the NJ word /yama/ and the SJ morpheme /san/. Word-formation processes involve both lexical strata (Kageyama and Saito 2016), such as NJ <山登り> /yama-nobo.ri/ and SJ <登山> /to-zan/ which both mean ‘mountain climbing’, although compounding is particularly productive with SJ morphemes (Kobayashi et al. 2016). Thus, two-kanji compound words (bimorphemic words) are particularly significant within the Japanese lexicon, both as independent words, like <関係> /kan-kei/ ‘relation’, and as the components of longer compounds, such as the three-kanji compound word <可能性> /ka-nō-sei/ ‘potentiality’ (two-kanji compound plus suffix) and the four-kanji compound <自分自身> /ji-bun-ji-shin/ ‘oneself’ (combining two two-kanji compounds) (Joyce and Masuda 2021).

Despite their separate historical developments, the two syllabographic scripts of <ひらがな> /hiragana/ and <カタカナ> /katakana/ are equivalent in terms of their potential to graphematically represent Japanese syllables (mora). Both kana sets have five vowel symbols, such as <あ> and <ア> /a/ respectively, but most kana symbols are consonant-vowel mora, such as the unvoiced <か> and <カ> /ka/ and voiced <が> and <ガ> /ga/, respectively. As the most recently integrated script (basically from the mid-twentieth century when first taught at elementary school), phonemic <ローマ字> /rōma.ji/ ‘Roman letters’ are more marginal in terms of general usage but they are common in popular media contexts. Moreover, while highly context-dependent, kanji and Arabic numerals are both commonly used to represent numbers (Joyce and Masuda 2018).

Clearly, typologies of writing systems need to adequately account for the complex interactions between different levels of representational mapping. As both the perspective shift towards the GWS and the constitutive mixing of the JWS highlight, the classification of writing systems based solely on the dominant level of representational mapping is essentially untenable.

3.2 Processing fit: Demarcation of boundaries

Admittedly, it may seem quite trivial to claim that writing systems must be suitable for their users’ processing needs, but as one of the primary functions of writing is communication, clearly successful communication will be highly dependent on the degree to which a writing system affords efficient processing. Accordingly, processing fit is a broad category that embraces a number of physiological and cognitive aspects, which have become the focus of greater attention since the processing turn within grapholinguistics. One hallmark of that perspective shift has been its deeper appreciation of the mutual constraints between mind and tool that have shaped the evolution of writing systems. As Dehaene (2009: 150) astutely points out, over time writing systems have had to become sufficiently “easy to acquire”, with the natural corollary of also becoming “easy to use”. Consistently, the category of processing fit aptly accentuates how, as products of the human mind, writing systems are dynamic artifacts.

One obvious feature of some writing systems that greatly influences processing, even though it is conspicuously absent from Daniels and Share’s (2018) list, despite its focus on (processing) complexity, is word spacing. Most of the world’s writing systems insert spaces between “words” (even if that linguistic category is inherently problematic). Word spacing initially developed out of changes in reading practices; specifically, the shift from reading aloud to silent reading and the different processing demands imposed on readers (Parkes 1992; Saenger 1997). Moreover, as most writing systems have been adapted from existing systems (or have, at least, been highly influenced by them), once word spacing was invented, it was adopted by many systems. Reading studies conducted with alphabet users have observed disrupted processing when word spaces are removed, with negative effects for both eye-movement control and word identification (Morris et al. 1990; Sheridan et al. 2016). However, not every writing system has word spacing, with notable exceptions being the Chinese, Japanese and Thai writing systems. Yet, it would be grossly inaccurate to claim that such writing systems are bereft of information that signals word or other boundaries, such as between syllables and morphemes. Accordingly, a binary typology distinction between “spaced vs. unspaced” writing systems alone would also be untenable, as it would be too imprecise and reductionist in nature, similar to existing typologies based solely on the dominant level of representational mapping.

The GWS employs word spacing in the form of blank spaces inserted between units that most often correspond to morphosyntactic words. The notion of word spacing is also fundamental to a proposal for the so-called graphematic word (Fuhrhop 2008).^[2] However, while it is uncontestably true that morphosyntactic units are salient processing units and that blank spaces are highly visible boundary cues, visual cues also signal graphematic syllables (Fuhrhop and Buchmann 2009; Fuhrhop et al. 2011), even though, admittedly, that concept is not without some controversy and the signaling is not always consistent. Specifically, graphematic syllable boundaries are preferentially occupied by “long” basic shapes, referring to those with either an ascender, such as |d|, or a descender, as in |p|. In contrast, the nuclei of graphematic syllables are occupied by “compact” basic shapes, such as |a|, |e|, and |u|, which are predominantly used to materialize vowel graphemes. Thus, a length hierarchy is certainly discernable, where the shapes at syllable boundaries exhibit length and central shapes are compact. That noted, however, the visual demarcation of graphematic syllables is not as salient as that for graphematic words, given that some consonants that occur at syllable boundaries, such as |s|, violate the notion of length hierarchy constituting a form of visual cueing. However, it is worth stressing that evidence supports the diachronic emergence and evolution of the graphematic syllable in German; an example being the gradual disappearance of long shapes, such as |y|, from the syllable nucleus position (Fuhrhop and Schmidt 2014). Consistent with Dehaene’s (2009) insight noted earlier, cognitive factors have fostered a gradual strengthening of graphematic syllable structures that has been synchronically described for German.

In contrast to the focus on the JWS’s mixture of representational mappings and associated scripts in Section 3.1, the emphasis here is on how the JWS’s graphematic conventions effectively serve to visually differentiate both between content and grammatical words and, to a lesser degree, between lexical strata. To that aim, Figure 1 presents Joyce and Masuda’s (2018: 183) example of an authentic Japanese sentence. Taken from a Wikipedia entry for the Japanese Industrial Standard’s (JIS) main character set, it contains examples of all component script.

Figure 1

Example of a Japanese sentence, with phonological gloss and translation (Joyce and Masuda 2018: 183).

As already noted, kanji are used to graphematically represent both SJ and NJ content words, such as SJ <地名> /chimei/ ‘place names’, and the stems of NJ verbs, adjectives, and some adverbs, such as <含む> /fuku.mu/ ‘include (plain-present)’. Hiragana are used primarily to represent functional words and inflections, such as the <含む> inflection, <である> /dearu/ (copula, plain-present) and the grammatical marker <の> /no/ ‘possessive; nominalization’. Katakana are used to represent Foreign-Japanese (FJ) words (loanwords not from Chinese), foreign and species names, onomatopoeia, emphases and glosses, such as FJ <バイト> /baito/ ‘byte’. Rōmaji are usually used to represent names and foreign words, particularly within advertising and mass media contexts, such as the JIS part of the reference code. Moreover, consistent with the example’s technical nature and its horizontal orientation, Arabic numerals are used for the number <6,879>, rather than the kanji equivalent <六千八百七十九>, which requires more space.

As even the short example in Figure 1 plainly attests, the JWS’s graphematic conventions effectively serve to vividly signal word boundaries, where a function word in hiragana invariably follows a content word, typically represented by kanji but also by katakana (Taylor and Park 1995; Joyce et al. 2014). Although Japanese texts use both commas (、) and periods (。), as there are no conventions for hyphenation, words simply wrap to the following line, as the compound <符号化文字集合> /fu-gō-ka-mo-ji-shū-gō/ ‘encoded character set’ indicates.

In contrasting the GWS and the JWS in terms of their approaches to marking syllable and word boundaries, this section has sought to illustrate how it would be overly simplistic to assume that strategies for visually differentiating linguistic units, such as word spacing, might be equally effective across writing systems. However, in light of psycholinguistic evidence that forms of visual demarcation do facilitate reading processes, clearly, further research is warranted into how diverse writing systems deploy various demarcation strategies.

3.3 Sociocultural fit: Orthographic regulation and reforms

Sociocultural fit represents the third pillar of the organizational framework, which are all equally important. Indeed, as Cahill (2014) states:

People accept or reject an orthography based on sociolinguistic factors. If a group doesn’t want to use an orthography, it doesn’t matter how linguistically sound it is – they won’t use it. So ‘what the people want’ is not just one more factor; it is the most critical factor in acceptance of an orthography.

(Cahill 2014: 16, emphasis in original)

Although the criteria under sociocultural fit can be analytically more elusive in nature, they are absolutely crucial in terms of actually using writing systems. In light of the stubborn term confusions within writing systems research, as acknowledged in Section 1, it should be stressed that Cahill’s use of “orthography” corresponds to the secondary sense of writing system. As explained earlier, it is preferable to reserve orthography to refer to the standardizing regulations that typically fall under the province of external stakeholders, such as authorities for language polices. From the perspective of grapholinguistics, descriptions of how writing systems function in terms of actually implementing their principles of representational mapping should be located at the graphematic level. More specifically, the objective of graphematic analysis should be to elucidate a particular writing system’s graphematic solution space (Neef 2015).

According to Neef’s (2015) conceptualization of the graphematic solution space, in German, for example, the word Fuchs ‘fox’ could be written *<Fuks>, *<Fux>, *<fuchs>, *<Vuchs>, *<vux>, and <Fuchs>. These variants (and others) are all licensed at the graphematic level of the GWS, but, consistent with the asterisk convention of marking unacceptable forms, only <Fuchs> is regarded as orthographically correct. As stressed earlier, the orthographic level is always prescriptive in nature. Of course, that is not to imply that regulating authorities always exist to determine what is “correct”, for although the Rat für deutsche Rechtschreibung (Council for German Orthography) and various Japanese Ministries, such as the Bunkachō (Agency for Cultural Affairs) have such roles for the GWS and the JWS, respectively, there is no such authority for English, for example, which is why English is often called a “self-organizing” system (Berg and Aronoff 2017).^[3] Our interpretations of orthographic regulations are often highly Eurocentric in focusing on certain aspects, such as grapheme-phoneme-correspondences, word divisions, hyphenation, capitalization, and the spelling of loan words (Coulmas 1996a: 379); issues that are not necessarily relevant for many writing systems.

Japanese debate concerning the difficulties of the JWS, known as <国字問題> /koku-ji-mon-dai/ ‘problems of the national script’, can be traced back to the Meiji period (1868–1912) when the written language was seen as an obstacle to Japan’s modernization (Seeley 1991; Twine 1991). As it has been estimated that there were more than ten thousand kanji in daily use at the start of the Meiji period (Twine 1991), understandably, one focus has been on restricting the number of kanji. Thus, since the mid-twentieth century, the Japanese government has issued a series of guidelines, including the current <常用漢字表> /jō-yō-kan-ji-hyō/ list of 2,136 kanji, as revised in 2010 (Bunkachō 2010). There is also a subset of 1,026 <教育漢字> /kyō-iku-kan-ji/ ‘education kanji’, which are carefully introduced across the elementary-school years, with the remaining 1,110 kanji covered at high school. While it is true that the guideline revisions have tended to slightly increase the numbers of official kanji, the revisions can be regarded as periodic fine-tuning, rather than signaling a fundamental shift in policy. For example, additions made in 2010 resolved some inconsistencies with proper nouns, such as <藤> NJ /fuji/ and SJ /tō/ ‘wisteria (flower)’, which is part of many common family names, like <藤原> /fuji-wara/ ‘Fujiwara’ and <佐藤> /sa-tō/ ‘Satō’. On the other hand, because the jōyō kanji list only has guideline status, although newspapers and official publications generally adhere, it is certainly not prescriptive for all written Japanese language. In the era of electronic information, a more realistic imposition on modern kanji usage is the JIS X 0208 character-encoding standard, referred to in Figure 1, which specifies 6,355 kanji.

Moreover, under the <現代仮名遣い> /gen-dai-ka-na-zuka.i/ ‘modern kana usage’ conventions amended in 1986 (Ministry of Education, Culture, Sports, Science and Technology-Japan 1986), kana-syllable correspondences are generally quite consistent, even though some historical influences remain. Essentially, there are three key exceptions, when <を>, <は>, and <へ> represent the grammatical particles for object /o/, topic /wa/ and destination /e/, respectively, rather than their conventional correspondences to /wo/ (now obsolete), /ha/ and /he/, respectively. Some retained conventions relate to preserving the morphological transparency of compound words. For example, the kana representation of <鼻血> /hana-ji/ ‘nose-bleed’ [nose + blood] is <はなぢ> /hanadi/ rather than <はなじ> /hanaji/, because the NJ word for ‘blood’ is /chi/, which is graphematically represented as either <血> or <ち>.

Despite the existence of various official guidelines, which largely codify the lexical strata-script associations, outlined in Section 3.2, as orthographic conventions, there are, unquestionably, profound consequences of the JWS’s distinctive mixture of mapping principles and scripts. The most striking is the highly fungible nature of the contemporary JWS, where graphematic variants are ubiquitous (Joyce et al. 2012; Joyce and Masuda 2019). As Joyce and Masuda (2019) seek to elucidate, the notion of the graphematic solution space in the case of the JWS shifts considerably. For example, the graphematic solution space for /watashi/ ‘I’ (polite first-person pronoun used by both genders but more commonly and less formally by females) consists of <私>, <わたし>, <ワタシ>, and <watashi>. However, selecting among them depends on numerous interrelated factors of intentionality, such as script sensibilities.

As yet, no typologies of writing systems have attempted to consider the extent to which writing systems can be “meddled with” orthographically by external regulators, such as the Council for German Orthography, but perhaps Neef’s (2015) notion of the graphematic solution space could be suitably expanded in that context. Even though proposed expressly for phonographic writing systems, the basic concept could be applied to potentially analyzing other levels of graphematic representation (Meletis 2020). Unquestionably, writing systems differ in terms of their latitudes towards representational variation, reflecting fundamental differences related to freedoms of choice and creativity that writing systems afford. In the case of the GWS, the degree of graphematic freedom is extremely limited, but the situation is radically different in the case of the JWS.