2015: Logography and the classification of writing systems: a

September 11, 2018 | Author: Anonymous | Category: Documents
Share Embed

Short Description

These features have implications for how Chinese writing is processed in the ... morphosyllabic, classification of writi...


Logography and the classification of writing systems: a response to Unger* Zev Handel

University of Washington, USA

In response to Unger (2014), I argue that Chinese does not merely lie along one end of an undifferentiated continuum of writing systems plotted according to the degree of phonological representation found in its graphs. Rather, two features of Chinese writing make it categorically distinct from even orthographically “deep” alphabetic writing systems like English: (1) the high prevalence of graphs that represent distinct meaningful linguistic units (i.e. morphemes) and (2) the use of graphic components (variously termed significs, determinatives, taxograms, classifiers, radicals) to represent the general semantic domains of those represented morphemes. These features have implications for how Chinese writing is processed in the brain, how it changes over time, and how it has been adapted for the written representation of other languages. For these reasons we should recognize that Chinese writing is distinct from phonographic systems of writing. Any dispute over which term is most appropriate for characterizing Chinese and the other writing systems of its type—logographic, morphographic, morphosyllabic, etc.—is secondary in importance to the recognition of the validity of this categorical distinction. Keywords: Chinese writing, logogram, logographic, morphosyllabic, classification of writing systems, script borrowing, radical

*‌I am grateful to J. Marshall Unger for his thoughtful critique, which began the discussion that is continued in this response, and to the editors of Scripta, who graciously provided me with a venue for this response. I am also indebted to Richard Salomon, Sven Osterkamp, William G. Boltz and Ik-sang Eom for helpful comments and suggestions. Finally, two anonymous reviewers provided invaluable feedback that has enabled me to improve the paper. SCRIPTA, Volume 7 (October 2015):109–150 © 2015 The Hunmin jeongeum Society



0. Background Unger & DeFrancis (1995:45) define logograms as “units of writing that do not provide a visual clue to the pronunciation of the speech strings they represent”. Handel (2013:23) noted that “Chinese writing is usually characterized as logographic”. In a footnote, I elaborated as follows: I use the term logographic in its widely accepted sense, referring to a writing system whose graphic units represent individual morphemes of the spoken language. These units are called logographs or logograms. Unger & DeFrancis (1995) object to this characterization of Chinese writing, but there is a bit of sleight of hand at work here. In making their argument, they change the terms of debate without being explicit about it. Their objection is based on redefining logogram in terms of a graph’s internal structure rather than its referent, namely as a unit of writing that contains no visual clue to its pronunciation (1995:45, 50). While this conception may be useful as a way of thinking about the internal structure and function of graphs, by repurposing an existing technical term they confuse rather than clarify the underlying issues. By redefining the term logographic in a way that precludes any writing system from being logographic, they thereby render it useless as a way of characterizing writing systems like Chinese whose graphic referents are primarily morphemes rather than semantically empty phoneme strings.

Unger 2014 is an elaborated response to this footnote.1 In this article I will respond in turn to Unger’s response. I wish to emphasize at the outset that I agree with Unger in many respects, more than he realizes. In several places in his article Unger attributes to me, explicitly or implicitly, positions that I have not expressed and do not hold. (This can be forgiven as the inevitable result of attempting to extrapolate my views from a single paragraph of a tangential footnote.) As a result, a number of his arguments are directed against “straw man” positions. Below I will make clear where I agree with  I wish to acknowledge an anonymous reviewer of Handel 2013, who suggested that this footnote could be “made the subject of a more extensive paper on the nature of Chinese characters and logographic writing more generally…. This … seems to me to be the beginning of a fruitful and intelligent discussion.” I am pleased that Prof. Unger has provided an opportunity for us to engage in that discussion.


Logography and the classification of writing systems


Unger, and where I believe his criticisms are misdirected. In other areas, however, there is genuine scholarly disagreement, and it is in these areas that I wish to engage substantively with Unger’s claims and arguments. As I see it, the disagreements lie primarily in these domains: 1) Terminology. What does “logographic” mean, and what should it mean? 2) T ‌ ypology. What does it mean to establish a valid categorization of writing systems (or, indeed, of any object of scientific inquiry)? 3) ‌E vidence. How conclusive is the evidence from the nascent fields of psycholinguistics and neuroscience about how written Chinese is processed during reading, in comparison with other writing systems? 4) ‌Domain. What aspects of writing systems structure, function, and history are legitimate objects of scholarly analysis?

I will take these up one at a time in sections 1-4 below. Before proceeding to these points, one other thing deserves to be said. Regarding the use of the term “logographic”, which was the catalyst for this debate, Unger (77) concedes that he and DeFrancis did in fact give it a new definition.2 His justification is that the redefinition of the term was useful for their own purposes of categorization.3 DeFrancis and I did not “simply . . . repurpose an existing technical term.” Rather, we stripped the definition of “logogram” of its circularity and sinocentric exceptionalism so that it can be used to distinguish partial from full writing in all parts of the world objectively.

I certainly have no quarrel with the goal to which Unger and DeFrancis put their redefinition: developing a typology that is capable of distinguishing partial from full writing, and of distinguishing theoretical

 In this article, bare page number references always refer to Unger 2014. Thus “Unger (77)” indicates “Unger (2014:77)”. 2

 If I understand it correctly, this redefinition of “logogram” has two aspects. The first is that it is divorced from the question of the size of the linguistic unit represented by a graphic unit. The second is that it applies only to individual graphs, not to writing systems as a whole. Based on the redefinition, it makes no sense to characterize an entire writing system as logographic; rather, one can only talk about the degree of logography in the writing system, i.e. the proportion of its graphs that are logograms. 3



or artificial systems of writing from actually occurring systems of writing.4 Furthermore, I agree wholeheartedly with Unger and DeFrancis about the importance of those distinctions. I criticized this redefinition of an accepted technical term because proliferating individualized interpretations of technical terms renders scholarly debate about the underlying issues difficult if not impossible. I still maintain that view, but after reading Unger’s critique, I now see that my use of the term “sleight of hand” was unfairly dismissive. My phraseology implied a deliberate attempt to deceive, which is clearly not the case, and for which I offer Unger an apology. To the contrary, Unger and DeFrancis make a substantive argument that is independent of their terminological usage, and it is with that argument that I will engage beginning with section 2 below.

1. Characterizations ‌ of Chinese writing and the definition of “logographic” Unger criticizes my characterization of Chinese as logographic in this way (76): I call [Handel’s] approach essentialist because it purports that every writing system has a “central organizing principle” (Sampson 1994) whereby all instances of all its graphic units (with negligibly few exceptions) represent either morphemes (or units such as phrases) or else phonemes (or units such as syllables). Classifying the world’s writing system[s] on this basis allegedly tells us something deeply important about them.

 I must admit that I do not understand several things about Unger’s statement. First, it is not clear to me why the characterization of a writing system as “logographic” (in the older sense) should be “sinocentric”. The term has been used, with validity in my judgment, to classify the Sumerian cuneiform writing system and the Mayan hieroglyphic writing system, both of which share many features with Chinese character writing. Second, it is unclear to me what is “circular” about the old definition, which does not preclude the identification of morphemes based on analysis of the spoken language alone. Finally, I do not see how the new definition helps to distinguish true writing from partial writing. As discussed below, Unger (79) describes the existence of artificial full writing systems like military codes, in which an arbitrary graph is used for each word. Unger’s revised definition of logogram does not affect the classification of such a system as full or partial writing, so far as I can tell. Under the old or new definition, such a system could justifiably be called both “full” and “logographic”. 4

Logography and the classification of writing systems


Unger’s critique that my characterization of Chinese writing is “essentialist” is based on his assumption that I believe that “every writing system has a central organizing principle” whereby “all instances of all its graphic units … represent either morphemes … or else phonemes”. In other words, he ascribes to me the same position as Sampson (1994). I have never stated such a position and do not believe it to be true, as is clear from my earlier publications (e.g. Handel 2009) and as I will further explain below. I do believe, however, that a non-“essentialist” description of Chinese writing as logographic is possible and appropriate. Whether this tells us something “deeply” important about writing systems or not is, I grant, a matter of judgment. As I will show, I believe that it does say something useful and interesting, providing both explanatory and predictive power, and is therefore a scientifically valid classification. While there is no single universally accepted set of definitions for terms such as alphabetic, syllabic, and logographic that are used to classify writing systems, a survey of experts in the field reveals a high degree of consensus, along with a sophisticated and nuanced recognition of the inexact nature of the terms and the fuzzy edges of the categories.5 These categories are based on a shared recognition that, from a linguistic perspective, it is useful to analyze writing systems in terms of the units of spoken language that are represented by the units of graphic analysis.6 Graphic units typically correlate with basic linguistic units such as segments (phonemes), syllables or moras, morphemes, and words. The terms letter, syllabogram, and logogram/morphogram are applied to graphs that represent segmental  My use here of the terms alphabetic, syllabic, and logographic is not meant to be an exhaustive catalog of terms in writing systems typology, or even to imply that these three are the most important or the most common. It is merely a recognition of their frequency of use in the literature. Other types, such as abugida and abjad, are widely recognized as well, and there remains lively debate about the most effective and meaningful way to categorize writing systems—a debate in which the Unger-DeFrancis typology plays an important part. Daniels (1996:4) points out that “half a dozen fundamentally different types of writing systems have been devised with respect to how symbols relate to the sounds of language (and there’s no reason more types could not be invented)”. 5

6  The linguistic perspective is, of course, not the only one relevant to the analysis or categorization of writing. One can look at various aspects of the mechanics of the writing system (direction of writing, characteristic stroke shapes, and so on) and of the social function and history of the writing system. Categorization based on linguistic analysis is just one method, and is the one that is naturally of concern to linguists.



phonemes, syllables, and morphemes, respectively. Many writing systems are in turn characterized as alphabetic, syllabic, or logographic/ morphographic depending on whether they primarily make use of letters, syllabograms, or logograms respectively as their most common and basic graphic units. Nearly all scholars categorize Chinese as logographic or morphographic according to this typology.7 If we narrow our focus to the definitions of logographic and logogram, and to scholarly classifications of Chinese as belonging to this type, the primary difference seen among scholars is not in their fundamental analysis but in their terminology. There is general agreement that Chinese writing employs most of its graphs to represent morphemes. Whether logographic is the appropriate term for such a writing system depends on whether logographic writing is defined as primarily representing words (as the etymology of the word logographic, from Greek logos λόγος ‘word’, implies) or morphemes. Because it is judged impossible for a natural and complete writing system to be of the former type, i.e. each graph representing a spoken word (and indeed no natural example of such writing is known), scholars who define logographic in terms of word representation—for purely theoretical reasons—always choose an alternative term (such as logosyllabary or morphosyllabary) for writing systems like Chinese. There is no impediment to retaining the term logographic as a characterization, however, if its definition is modified to encompass morphemes along with, or instead of, words. To take three examples: Coulmas (2003) retains the term logographic but carefully specifies its definition, while Rogers (2005) and Daniels (1996) eschew logographic in favor of another term. But all three agree on the basic definition of whatever term that they use to characterize Chinese writing, namely that it involves a specific relationship between graphic units and linguistic units, that the morpheme is a key linguistic unit, and that this type of writing can be contrasted with types of writing which relate to semantically empty phonological units. Rogers (2005:14) says “Chinese [is] a writing system where the primary relationship of graphemes is to morphemes. Such a system can be called morphographic, and those graphemes can be termed morphograms.  I recognize that an appeal to scholarly consensus does not constitute argumentation, and that Unger takes issue with the typological framework underlying these consensus categorizations. In this section I intend to deal only with terminological issues, reserving substantive argumentation for subsequent sections. 7

Logography and the classification of writing systems


Other authors on writing describe graphemes as related to words, rather than to morphemes, often using the term logogram. I am unaware of systems where the primary relationship of graphemes is to words, as opposed to morphemes. Accordingly, I will use the terms morphogram and morphographic here.” Daniels (1996:4) similarly characterizes a writing system like Chinese as “a logosyllabary, [in which] the characters of a script denote individual words (or morphemes) as well as particular syllables”. Coulmas’s definitions are a bit more complex. He (2003:40) first says: “One way of classifying writing systems is by the level of linguistic analysis to which their basic functional units relate. Writing systems whose basic functional units are interpreted as words are known as ‘logographic’ or ‘word writing’ systems.” But he goes on to note that the term logographic “requires some qualification, as will become apparent in the following discussion of two major logographic writing systems … the Sumerian and the Chinese” (2003:41). After describing the two writing systems, Coulmas revisits the terminology: “‘Logogram’ is … appropriate, but [etymologically] inaccurate, because the term suggests that the word is the prominent unit of writing. This is, however, not the case in either Sumerian or Chinese, not in any event if the word is understood as a well-defined unit of linguistic analysis …. Syllables and morphemes are more relevant as linguistic units” (2003:59–60). In Handel 2013, the definition of logographic that I employed was in reference to “a writing system whose graphic units represent individual morphemes of the spoken language.” This may sound categorical or absolute, but the broader context makes it clear that I did not intend to imply that in a logographic writing system like Chinese all graphs must represent morphemes. This was made explicit in an earlier version of the same article, Handel (2012:161–163), which gives a detailed explanation and several examples of the aspects of Chinese writing that deviate from the pure definition of logography: characters employed purely for sound value as syllabograms; characters employed to write one syllable of a bisyllabic morpheme; characters employed to write monosyllabic twomorpheme words; and one character employed to write a sub-syllabic morpheme.8 Overall, the number of graphs that in the modern Chinese  These details and examples in Handel 2012 were omitted from Handel 2013 because




writing system are not employed as pure logograms is not insignificant, but these are very much in the minority. That most Chinese characters of both the modern and historical Chinese writing system represent morphemes is simply an empirical fact. My working definition of logographic can be revised using the word typically to explicitly acknowledge that morpheme representation is a strong tendency, not an “essentialist” absolute, as follows: (1) Logographic refers to a writing system whose primary graphic units, logograms or logographs, typically represent individual morphemes of spoken language.

I will argue below that it is not only accurate to characterize Chinese writing in this way, but that it is also a useful characterization, in that it captures features of the writing system that differ significantly from features of non-logographic writing systems, in terms of both abstract analysis and real-world usage and evolution. My point here is that the specific terms and definitions are less important than the underlying conceptualization. If one wishes instead to reserve the term logographic for a hypothetical writing system in which graphic units represent not morphemes but whole words of spoken language, then I have no objection to instead characterizing writing systems like Chinese as morphographic, where morphographic has the same definition that I gave to logographic above: (2) Morphographic refers to a writing system whose primary graphic units, morphograms or morphographs, typically represent individual morphemes of spoken language.

If one wishes to also capture the fact that Chinese characters generally represent syllable-long units of spoken language (a fact that is intimately connected with the tendency of modern Chinese morphemes to be of space considerations, and because they were not relevant to main thesis of the article, which was concerned with the implications of the 20th-century character simplification policy in mainland China. An understanding of these deviations is of course an important part of any analysis of the modern Chinese script. While a complete description is beyond the scope of this paper, more details will be provided in section 5 below.

Logography and the classification of writing systems


monosyllabic), then I have no objection to characterizing Chinese as morphosyllabic, as DeFrancis (1984:88) does. Unger (76) says that “Handel treats ‘morphosyllabic’ as merely an alternative to ‘logographic’—the two are virtual synonyms for him”, and when it comes to Chinese I do indeed see this as a terminological distinction without much substantive difference: morphosyllabic is more specific than logographic. It is worth noting in this regard that DeFrancis does not actually reject terms like logographic for characterizing Chinese, he merely says that they “are good as far as they go, but do not go far enough” (1984:88). So when it comes to Chinese, we could replace characterizations (1) and (2) above with (3): (3) Morphosyllabic refers to a writing system whose primary graphic units, morphosyllabograms or morphosyllabographs, typically represent individual monosyllabic morphemes of spoken language.

What all of the characterizations of Chinese described above have in common is that they relate the primary graphic units of the script to a particular linguistic unit of the spoken language: the morpheme.9 As noted earlier, DeFrancis and Unger defined logograms as “units of writing that do not provide a visual clue to the pronunciation of the speech strings they represent”. No natural full writing system—i.e. one that can reproduce all of the utterances of spoken language—lacks phonographic elements (either as graphs or components of graphs), and so by Unger’s definition Chinese, like all natural writing systems in use today, is not even close to being fully logographic.10 Indeed, as has long been recognized, the  It is important to note, however, that a characterization of Chinese writing as morphosyllabic is tenable only for the last 2000 years or so, before the major typological shift of spoken Chinese that took place around the Hàn 漢 dynasty (206 bce-220 ce). There remains considerable dispute about the type of speech units represented by Chinese characters before then, mainly because of ongoing uncertainty about Old Chinese phonology and morphology. Current research in Old Chinese reconstruction suggests that through the first millennium of their use, Chinese characters often represented words that were polymorphemic and could be longer than one syllable. (For example, Baxter & Sagart (2014) reconstruct Old Chinese 鼻 *Cə-bi[t]-s ‘to smell’ as a two-syllable word with three morphemes (prefix, root, suffix) and Old Chinese 飼 *s-m-lək-s ‘to feed’ as having four morphemes (two prefixes, root, suffix) and as many as three syllables.) This would justify calling early Chinese characters logographic even in the etymological sense of the term, i.e. characterizing a writing system in which basic graphs represent complete words of the spoken language. 9


 Unger points out that it is possible to devise a fully logographic (his definition)



vast majority of Chinese characters contain phonetic elements that provide (or historically provided) an approximate indication of the pronunciation of the (morpho-)syllable represented. I am in agreement with Unger on this point. It seems to me, then, that if we set aside the terminological differences and focus on the substantive area of dispute, the question on which Unger and I are currently engaged is whether, as I claim, Chinese writing can be characterized as a writing system whose primary graphic units typically represent individual monosyllabic morphemes of spoken language

and further, whether characterizing writing systems like Chinese in such a way “tells us something important about them”. Note that this characterization is not concerned with the internal structure of the graphs or the manner in which they do or do not encode phonetic information. It is only concerned with the relationship between graphic units and speech units. If I read Unger’s critique correctly, he seems to feel that a characterization on such a basis is both pragmatically pointless and intellectually dishonest, because (1) all actually occurring full writing systems are essentially the same, so there is no value in subcategorizing them; and (2) that any categorization based on represented speech units fails because it would necessarily be approximate rather than absolute. I am in complete agreement that categorizations based on correlations with linguistic units are always approximate. Whatever term and definition we choose for a characterization of Chinese, we must recognize that the characterization is approximate, that is, it captures a generalization. Not all morphemes of Chinese are monosyllabic; not all Chinese characters represent morphemes; and not all Chinese characters represent monosyllables.11 But these caveats need not in and of themselves invalidate writing system, such as a military code, in which each word of a language is assigned a written representation unrelated to the word’s pronunciation. He quite rightly notes the artificial and cumbersome nature of such systems, which have never evolved naturally, and whose value as secret codes lies precisely in how difficult it is to read them.

 Some examples of such characters as used in Modern Standard Written Mandarin are:


兒 -r (nominal suffix; monomorphemic, sub-syllabic); 珊 shān (part of shānhú ‘coral’; submorphemic, monosyllabic); 廿 èrshí (‘twenty’; bimorphemic, bisyllabic). One important

Logography and the classification of writing systems


the usefulness and accuracy of the generalization. There is another matter of some importance that I will take up below as well, and that is the internal structure of graphs in morphographic writing systems. In practical terms, morphographic writing systems must employ some graphic mechanism for distinguishing homophonous and near-homophonous morphemes. This is accomplished by means of graphic elements known as determinatives, significs, semantic components, classifiers, or taxograms, and which in the case of Chinese are popularly known as radicals. As I will demonstrate below, the presence of these elements has important ramifications for the cognitive processing of scripts like Chinese and for their historical development. They are categorically different from the inconsistency in spelling of, say, written English, which lends English what is sometimes called its “logographic” character, in which distinct homophonous morphemes often have different written forms (e.g. to, too, two).

2. Classification: categories and continua Our substantive disagreement centers in part on the value of subcategorizing writing systems. One aspect of this disagreement has a general theoretical dimension: What is the function of a categorization? In other words, what value is there (if any) to saying that two things are alike and belong to the same class, while two other things are different and belong to separate classes? The question “What is a classification?” is naturally quite broad, touching on aspects of philosophy, ontology, cognitive science, and other fields. Rather than attempting a comprehensive response to the question, I will simply venture a few thoughts that are relevant to the present discussion. It is natural for human beings to classify. It is a basic cognitive trait that is essential to the functioning of our species. The ability to generalize— that is, to predicate features associated with an individual entity of all entities in the same category—is crucial for both basic survival and higherpoint that is sometimes overlooked is that many of the characters that write single syllables of bisyllabic morphemes, like 珊 shān, are in fact morphographic in nature even if they are not, properly speaking, morphograms. This seeming contradiction will be explained in section 5 below.



order cognition. Nevertheless, most categorizations are imperfect. It is seldom possible to posit precise sets of necessary and sufficient conditions that permit all entities to be unambiguously classed into one and only one category. Another way to say this is to acknowledge that categories have fuzzy edges and that some entities have uncertain categorical identities. (See Lakoff 1987.) Consider two examples, one cultural and one scholarly. English speakers broadly categorize edible plant matter into two types, fruits and vegetables. These are cultural categories, with fuzzy boundaries.12 The categories overlap to some extent (as with tomatoes), and they also fail to encompass all produce (edible flowers and nuts fall outside the categories). Despite these logical problems, we find these categories to be culturally useful, and we have no difficulty identifying prototypical members of each (say, spinach as a vegetable and apple as a fruit). Within the field of Chinese dialectology, certain dialects are classified as of the Mandarin type, and certain dialects are classified as of the Wú 吳 type. Despite the fact that prototypical examples of each are indisputable (Běijīng 北京 dialect is Mandarin; Sūzhōu 蘇州 dialect is Wú), there remains ongoing debate about the proper criteria for defining the two types, and there exist numerous “fuzzy” cases of dialects that present a challenge for classification (such as that of the city of Hángzhōu 杭州, which has characteristics of both Mandarin and Wú). Nevertheless, the categories themselves remain useful as reflections of diachronic patterns of development and of synchronic constellations of features. Despite the problems, imprecisions, and imperfections found with these categorizations, they remain valid and useful ways to talk about salient physical and cultural similarities and differences between entities. More importantly, categorizations like that of the Chinese dialects have scientific validity, in that they allow researchers to posit and test generalizations about phenomena. Many real-world phenomena exist on a smooth continuum. Yet we often find it natural, practical, and valuable to divide continua up into discrete categories or classes. To take a very simple example, consider the electromagnetic spectrum (Figure 1). It ranges smoothly from a low frequency of 3 hertz (Hz) to extremely high frequencies above 10 exahertz  They overlap only to some extent with botanical definitions. For example, eggplant and cucumbers are culturally vegetables, although a plant biologist would categorize both as fruits.


Logography and the classification of writing systems


Figure 1. The electromagnetic spectrum13

(Ehz = 1018 Hz). Despite the fact that this is a true continuum, we find it useful to divide the spectrum into categories and sub-categories based on properties that we deem relevant, such as radio waves, microwaves, infrared waves, visible light waves, ultraviolet waves, X-rays, and gamma rays. These categories are related to physical properties and/or practical applications of these electromagnetic waves. Although the continuum itself is smooth, the categorical differences are real: radio waves do not behave like gamma rays. Within this vast spectrum, let us consider the position of visible light. It makes up a tiny fraction of the spectrum as a whole, yet it is obviously a meaningful category for human beings. It is interesting to consider whether there is any value in further sub-categorizing this tiny portion of the electromagnetic spectrum. One could well argue that given the incredible difference between waves found at each pole of the spectrum, and the enormous range of properties found across the spectrum, there is little or no point talking about different kinds or types of visible light. Compared to the difference between radio waves and gamma rays, the internal differentiation of the visible-light part of the spectrum is essentially nonexistent. Yet distinctions between colors in the visible spectrum are important to humans. One could still argue that, even though these differences seem meaningful, it is nevertheless pointless to categorize them  Image from http://hyperphysics.phy-astr.gsu.edu/hbase/ems1.html, accessed September 1, 2014.




because the boundary lines of the categories would inevitably be arbitrary. Where should one draw the boundary lines between “red”, “orange”, “yellow”, “blue”, and so on? Yet human beings do make color categories, do assign prototypical color values to those categories, and do find the distinctions useful and important, even though the categories are nonuniversal and fuzzy-edged.14 One other example, drawn from linguistics, is in order. Syntacticians often point out that, of all the logically possible ways that grammatical structures might be organized in human languages, only a small subset are actually attested. For example, in no human languages are nouns and their determiners separated by intervening predicates. Many linguists believe that these commonalities are attributable to specialized areas of the brain engaged in cognitive processes that are universal. Yet it does not follow that we should not bother to take note of distinctions across languages and create classifications based on them. The lesson here is that we cannot a priori dismiss the value of a classification or categorization because (1) its categories fall within a narrow range of all possibilities along a spectrum; (2) it imposes discreteness on a continuum. The value of a classification or categorization should instead be based on its usefulness: do the different categories capture similarities and distinctions that reflect something meaningful about the world or about human experience? In terms of scholarly and scientific analysis, do those categories allow us to make inferences and predictions that correlate with categorical membership? It is natural, then, to ask whether there are salient similarities and differences among human writing systems that make it useful, practical, or interesting to classify them; and if so, to ask which criteria can be used for making such a classification. As we will see below, Unger makes a strong argument that extant human writing systems fall along a narrow band in the center of a larger continuum encompassing theoretically possible writing systems and, more broadly, graphical communication systems. But his conclusion—that a subcategorization of writing systems is invalid—does not logically follow. To decide whether it is meaningful or pointless to talk about writing systems  I am grateful to Professor Jin-ho Park for suggesting the example of the visible light spectrum.


Logography and the classification of writing systems


Figure 2. “DeFrancis-Unger axial typology of writing systems” from Unger (2014:88, Figure 4))

being alphabetic, syllabic, morphographic, etc. we must determine if there are significant distinctions across these categories and significant commonalities within them, ones that are relevant to empirical questions about the function, use, and history of writing systems. The DeFrancis-Unger typology defines two theoretical extremes along the continuum of full writing systems.15 At one end, the phonographic end, is a system that represents only phonetic or phonological aspects of human speech, and encodes no other information. Something approaching this type would be the phonetic notation used by a field linguist.16 At the other end, the logographic end (according to their novel definition, see section 2 above), only morphemic or lexical identity information is encoded, with no overt indication of any aspect of pronunciation. Something approaching this type would be a code that uses an arbitrary graphic symbol for each word of a language. Figure 2 reproduces Figure 4 from Unger (88).  Unger (78) defines full writing as a system that can transcribe all of “the openended set of utterances possible in a natural language”. Partial writing systems, in contrast, are systems “in which visual marks are used as signals for various purposes by prearrangement among cooperating individuals”, but lack the same capability as full writing systems (77–78). Full writing, then, is equivalent to what others have called “true writing” or “glottographic writing”.


 Although it is my understanding that Unger would not place any actual writing systems at this end of the continuum, it seems to me arguable that orthographically shallow writing systems like Finnish, Serbo-Croatian, and Italian are in fact located very close to there. Although not as detailed as narrow phonetic transcription, they almost completely fit the ideal of one grapheme per phoneme, with no graphic distinctions among homophones. If one takes the phoneme rather than the phone as the basic sound unit of spoken language, then these writing systems really are situated at one end of the spectrum.




Unger (79) says that “Since English and Chinese texts look very different graphically, the naïve observer may be pardoned for jumping to the conclusion that their respective writing systems must operate according to completely different principles (allegedly ‘phonographic’ and ‘logographic’), but English and Chinese are not even close to genuine cases of, respectively, extreme phonography or extreme logography.” I have no quarrel with this general analysis. I agree that in the context of the full spectrum of theoretically possible systems of writing, actual human writing systems occupy a relatively small space. I also agree that English and Chinese writing do not “operate according to completely different principles”; they share those many features inherent in all true writing systems that people actually use. (That these features must exist in a full writing system is in turn the result of certain characteristics and limitations of human brain architecture and cognition, balancing the need for learnability and the need for usability.) But this figure and the accompanying explanation do not constitute evidence against the validity of a classification that puts English writing and Chinese writing into different categories. Unger does not provide any metric for the x-axis, or any objective justification for the positioning of writing systems along it. On what basis is Korean placed to the left rather than to the right of English? There is a slightly larger gap between Arabic/ Hebrew and Chinese than between the other adjacent systems, but what is its measure and significance? Is it large or small? And regardless of the gap’s size, the figure itself is silent about whether crossing this gap leads to any categorically significant effects. Unger’s is not an argument made from hard data.

3. Evidence from psycholinguistic research and other sources Unger (80), quoting DeFrancis & Unger (1994:551–552), says “[u]nless one has hard data (not just classificatory theories) showing that the processing of, say, English texts and Chinese texts in the brain proceeds along different pathways, the hypothesis of a single mechanism [underlying all writing systems] must be preferred.” I read this as a claim by Unger that the method of cognitive processing of a script during the act of reading is the only valid criterion for constructing a typology of writing systems.

Logography and the classification of writing systems


We can then ask two key questions in response to Unger’s arguments. First, is the pathway of mental processing in the brain during the activity of reading a text the only useful and interesting criterion by which it is valid to classify writing systems? This is the criterion question. I will address it in section 4. Second, even if for the sake of argument we suppose that the answer to the criterion question is yes, we can still ask: Is Unger correct to assert that the evidence to date shows no distinction between the way English and Chinese (or, more broadly, between phonographic and morphographic writing systems) are processed in the brain when reading?17 This is the cognitive question. I will address this question now. Unger (80) cites Dehaene (2009) in support of his assertion that “there are hard data proving that the brain has different networks of connections for dealing with phoneme and morpheme recognition. But the data also show that both networks work together, in parallel; are activated only after retinal image processing, invariant letter recognition, and hierarchical assembly of larger graphic units have begun; and are the same for all full writing systems.” Yet a close reading of Dehaene shows precisely the opposite: that despite the similarities in the way writing systems are processed in contrast to other kinds of visual data, they are not the same; there are crucial distinctions between them. These distinctions are downplayed by Dehaene and ignored by Unger. But they are indicative of exactly the kind of meaningful differences that could justify a categorical distinction. As Unger notes, Dehaene’s 2009 book makes a strong claim about the commonalities in brain specialization and processing that are found across all human cultures in relation to reading.18 And indeed, the evidence that Dehaene assembles from a large number of recent psycholinguistic and neurological studies (including his own research) largely supports his overall thesis. But if one reads Dehaene carefully, it is also clear that there are significant differences in how, on the one hand, Chinese characters are  Here we employ phonographic as a cover term for those writing systems whose basic graphic units represent semantically empty phonological strings, including what are traditionally termed alphabetic and syllabic scripts.


 Some of the content in Dehaene 2009, including chapter summaries and color versions of the figures, can be found on the World Wide Web at the URL http://readinginthebrain.pagesperso-orange.fr/figures.htm (accessed July 21, 2015).




processed by Chinese readers and kanji are processed by Japanese readers, and how, on the other hand, phonographic writing systems (alphabets, syllabaries, abjads, abugidas) are processed by their readers. At every turn Dehaene employs a rhetorical strategy of hedging or downplaying these differences as insignificant in the face of the commonalities. But it is precisely these differences that may be indicative of a meaningful categorical distinction.19 A key aspect of Dehaene’s argument concerns the left occipito Although not directly relevant to the evidence presented by Dehaene or Unger’s citing of it, it is worth pointing out that Dehaene says a number of things that are simply incorrect about Chinese language and writing. Six of them are discussed below. While these misunderstandings and misrepresentations do not bear directly on Dehaene’s arguments about the universality of reading processes, they do raise doubts about whether he is well read in the psycholinguistic literature concerning Chinese writing and whether he fully understands the import of relevant studies. (i) He says (2009:35) that “a purely phonetic writing system would be useless for Chinese”—completely ignoring the fact that speech is possible in Chinese, and that pinyin writing—even toneless!— is generally a quite effective medium of communication for those practiced in it. (ii) He (2009:36) unwittingly presents, in Figure 1.3, Y. R. Chao’s famous “story in syllable shi”, with no mention at all of the fact that it is a text written not in modern Chinese but in Classical Chinese. The very purpose of the text is to demonstrate that although impossible to understand if read aloud in Mandarin pronunciation, it would have been intelligible to speakers of the classical language, for whom these words were not homophonous. It is moreover simply not true, as Dehaene claims, that “Any Chinese reader can understand this text”, as only readers trained in the grammar and vocabulary of Classical Chinese, i.e. having expertise beyond ordinary literacy, are able to read it. The first sentence of the story would be entirely different if translated into spoken Mandarin and in turn quite understandable if presented in romanized transcription: Yǒu yī wèi zhùzài shíshìlǐ de shīrén jiào Shī shì, xǐhuan chī shīzi, juéxīn yào chī shí zhī shīzi 有 一位住在石室裡的詩人叫施氏, 喜歡吃獅子, 決心要吃十隻獅子 ‘There was a poet by the name of Shī who lived in a stone grotto and was fond of eating lions; he resolved to eat ten lions”. (iii) Dehaene (2009:97) oddly speaks of pinyin as using “the 26 letters of our Latin alphabet, plus a number of graphemes such as ‘zh’, ‘ch’, and ‘ang’.” This is muddled at best; among other problems with the statement, there is no sense in which ‘ang’ can be considered a grapheme. (iv) In his figure 4.1 (2009:175) it is stated that “all writing systems share numerous visual features [including] … an average number of about three strokes per character.” The average number of strokes in a Chinese character is closer to 12. (v) He (2009:176) further notes: “each character is made up of only two, three, or four basic shapes”. It is hard to see how this could possibly be a correct statement, although he never makes clear what he means by “basic shape”. (vi) Dehaene (2009:203) incorrectly states that because they lack alphabetic writing, Chinese readers cannot consciously manipulate phonemes. Yet the existence of what are sometimes called fǎnqiè 反切 languages, Chinese “Pig Latin”-type children’s games in which segmental phonemes are moved and/or transformed, demonstrates the ability of native speakers to manipulate sub-syllabic units of the spoken language. 19

Logography and the classification of writing systems


temporal area of developed literate readers’ brains, which he refers to as the “letterbox area”. It is highly specialized for the recognition of visual elements of writing systems in the process of reading. Located in the same brain area in readers the world over, it responds automatically to written words. In less than one-fifth of a second, a time span too brief for conscious perception, it extracts the identity of a letter string regardless of superficial changes in letter size, shape, or position. It then transmits this information to two major sets of brain areas, distributed in the temporal and frontal lobes, that respectively encode sound pattern and meaning. (Dehaene 2009:53)

Of the location of this area, he (2009:70) says: …the location of the letterbox area turns out to be remarkably similar, both across individuals and even across experimental laboratories, with differences of no more than roughly five millimeters. Even the brains of Chinese and Japanese readers, as we shall see later, are equipped with a brain area specific to reading, and its position is approximately the same as ours. [emphasis added]

One can read the above in two ways: as evidence of remarkable similarity between Chinese and Japanese readers and the “rest of us” (Dehaene takes a notably Anglocentric view in his writing, using phrases such as “our alphabet” and “our language” in reference to English), or as evidence of a crucial distinction in brain organization related to different types of scripts. In discussing the fact that the letterbox area is employed almost exclusively in reading, Dehaene (2009:72) goes on to say: … in most cases speech does not stimulate this [letterbox] region, which thus appears exclusively reserved for the written word. There is one exception, however, which occurs when people are encouraged to spell a word in their mind’s eye …. In such cases, the letterbox area is mildly activated by spoken words—only because participants imagine the written word. This also takes place when Japanese subjects have to think of writing a word. [emphasis added]

Here again we have a notable difference between brain function in Japanese readers and others. (The statement is based on a research study



of Japanese; no information is given about what happens with Chinese readers.) Beginning on page 97, Dehaene has a section titled “The Brains of Chinese Readers”. In this section he says: … the left occipito-temporal letterbox area plays a prominent role in all readers, with only minimal differences linked to the shape and internal structure of characters. … at a spot essentially identical to the letterbox area … brain activations related to word recognition in Chinese readers lie only a few millimeters away from those of English readers.

Once again, Dehaene’s rhetorical moves (“only minimal differences”, “only a few millimeters”) make it sound like these distinctions are too trivial to bear any weight. But is the dismissive attitude warranted? In the brain a distance of a few millimeters can be significant. One might ask instead: Why are there differences between the brains of Chinese readers and readers of alphabets? Why is the letterbox area not located in precisely the same place in the brain? The most plausible—perhaps the only plausible explanation—is that it is the result of an inherent difference between the structures of the writing systems involved. If this is the only difference among human writing systems that produces a measurable and consistent difference in brain structure, then surely it is a difference of some significance, and points to a meaningful categorical distinction. A few pages later, Dehaene is forced to acknowledge differences in the way Japanese readers process morphographic kanji and phonographic kana, the two scripts employed in the Japanese mixed-script writing system, although he again downplays the distinctions, noting that “only subtle differences are seen”.20 But when these “subtle differences” are described in detail, they hardly seem subtle. Indeed, they are obviously of real significance. careful scrutiny reveals a few differences … words written in Kanji produce  Kanji is simply the Japanese term for Chinese characters as used morphographically in the Japanese writing system. Kana encompasses two scripts that are usually described as syllabic, although more precisely they represent moras, not syllables, of spoken Japanese. The kana scripts are phonographic. 20

Logography and the classification of writing systems


more activation in a small bilateral region of the ventral temporal lobe— perhaps because they call for a slightly more global or “holistic” perception mode. Lesion studies also confirm that the brain’s reading networks are not identical for Kanji and Kana … a stroke patient may retain the ability to read in one script but not in the other. (Dehaene 2009:100)21

The evidence presented by Dehaene shows that there is a consistent distinction between the way Chinese characters (whether in written Chinese or as kanji in written Japanese) are processed and the way alphabetic writing—even orthographically deep alphabetic writing like English—is processed. The different types of aphasias in Japanese readers certainly constitute strong evidence for this. So why should this not be a valid basis for a categorical distinction?22 On pages 244–245, Dehaene cites a study that shows differences between Chinese dyslexics and dyslexics who use alphabetic writing systems. Because this evidence presents a challenge to Dehaene’s broader thesis about commonalities in brain function across writing systems, he plays down these differences, going so far as to speculate—utterly without supporting evidence—that one should be able to find an alternative hypothesis to explain the data. It is clear that the jury is still out in terms of psycholinguistic evidence about reading functions in the brain. No doubt we will learn more in coming years as the field matures. But it is also clear that, based on the evidence presented by Dehaene, there are differences in the brain structures of Chinese and Japanese readers as compared to readers of other writing systems. In my view, these differences alone answer the cognitive question: they refute Unger’s contention that we cannot make categorical distinctions among writing systems because they all function the same way  Unger (88) criticizes Dehaene for treating kanji and kana distinctly, saying this “makes no sense except in experimental conditions” because the two always occur together in modern Japanese writing. Yet one can hardly claim that a stroke patient who can no longer read one of the two Japanese scripts is operating in an artificial experimental setting. 21

 Granted, these morphological and functional differences in the brain may not directly correlate to the distinction between phonographic and morphographic scripts. It is possible that some other physical or structural aspect of Chinese characters or of cultural interaction with characters accounts for them. But in the absence of further study, we can hardly conclude as Unger does that our current state of knowledge provides no basis for a categorical distinction. 22



in the brain. But we can go further than this in our conclusions. There is a much wider array of psycholinguistic literature on the reading of Chinese than is cited by Dehaene. Unger (91) notes that I already cited some of this literature in Handel 2013, but opines that I did so selectively, “not mentioning those … that demonstrate how phonological recoding may determine lexical access even in an allegedly logographic writing system.” It is true that I did not cite such research, because nobody is disputing the role that phonological components play in the Chinese writing system or the role that phonological recoding plays in the reading of Chinese. This is settled scholarship and an area in which Unger and I are already in agreement. The research I cited shows interesting differences in the processing of Chinese. Unger ignores the intriguing implications of this research. Here I will elaborate on these implications by presenting the results of some recent studies in psycholinguistics and neuro-imaging that were not mentioned in Handel 2013. Written by specialists in Chinese psycholinguistics, they present a useful counterpoint to Dehaene’s rather naïve view of Chinese writing. Tan & Siok (2006), in a survey of neuroimaging studies of Chinese readers, demonstrate that the left middle frontal gyrus plays a special role in the reading of Chinese, and grows larger over time in Chinese readers as compared to readers of English (or, crucially, of readers of Chinese written in pinyin romanization). This brain region is used by all humans to hold a “limited amount of spatial information in an active state” (2006:361) and the authors surmise that it is used to “mediate visuo-orthographic analysis of written Chinese” (2006:362). They conclude that “[a]reas of functional differences in reading Chinese and English by native readers have been identified by a rapidly growing body of imaging studies” (2006:370). Su & Law (in press), also surveying the literature on recent imaging studies, point out the following about “Chinese-specific” regions of the brain: A review of studies of word reading in alphabetic scripts, Japanese kana (syllabaries) and kanji (Chinese characters used in Japanese writing) and Chinese characters has identified “Chinese-specific” regions, including the left dorsal inferior parietal lobe, left dorsal lateral frontal region (BA9), and right in addition to left inferior occipital and posterior fusiform areas (Bolger et al. 2005). The left BA9 or left middle frontal gyrus (LMFrG) has been hypothesized to underlie visual-spatial analysis of characters and

Logography and the classification of writing systems


mapping between characters and syllables based on a review of Chinese studies employing homophone and rime judgments of characters (Tan et al. 2005). Its significance seems to go beyond its role in adult Chinese reading, as the LMFrG has been argued to be a biological marker of developmental dyslexia in Chinese, differing from the left temporoparietal regions in reading disorders in alphabetic systems (Siok et al. 2004). Chinese children with reading impairment showed abnormally low activation in the LMFrG during homophone judgment and character decision tasks. Moreover, gray matter volume in the LMFrG was found to be positively correlated with reading achievement when both normal and poor young readers were considered (Siok et al. 2008). However, a more recent study involving normal and developmental dyslexic readers in Chinese and English in making decisions about the semantic relationships between written words showed that while normal reading in the two scripts differentially activated the left inferior frontal sulcus (stronger for Chinese) and left posterior superior temporal sulcus (stronger for English), dyslexic readers of English and Chinese demonstrated below-normal activation in the same neural regions, including the left angular gyrus, LMFrG, and the posterior temporal and occipitotemporal regions (Hu et al. 2010).

My proposal (2013:33) that we “introduce the concept of semantic orthographic depth”, which is criticized by Unger (91) without serious engagement, is based on a wide range of psycholinguistic studies showing that various aspects of Chinese-character cognitive processing are influenced by the semantic components (“radicals”) that appear in over 90% of Chinese characters. Feldman & Siok (1999a:572) observe [T]he semantic attributes of radicals provide another source of activation in the present task. This claim was motivated by the way in which semantic transparency of the radical influenced target recognition. When prime and target were presented in immediate succession (SOA 243 ms) and the mapping between form and meaning of radical in prime and in target was not consistent, inhibition was observed. When the mapping was consistent, facilitation was observed. Our emphasis on the semantic transparency of the radical and the critical role it plays in Chinese character identification should not be interpreted to mean that the orthographic attributes have no role. In fact, we have reported that, under the appropriate temporal constraints, both the semantic and the orthographic characteristics of radicals produce significant



and differentiable effects (Feldman & Siok, 1999[b] ; Siok & Feldman, unpublished manuscript). … What is evident at present is that semantic analysis of radicals must play a role in the visual recognition of Chinese characters. [emphasis added]

Williams & Bever (2010:603) conclude that semantic mediation plays an important role in the processing of Chinese characters: Chinese script is significantly different from alphabetic scripts as semantic information is specifically embedded into characters, thus making a semantic route to meaning much more viable than in alphabetic languages. The existence of twin routes of lexical access—phonetic and semantic— is hypothesized to be a universal constant across languages and scripts, but some have argued that the dominance of phonetic interpretation is a language universal, and not merely restricted to alphabets and syllabaries which, understandably, predispose readers to using phonetic strategies of access over semantic ones, and thus, one would argue, Chinese lexical access must be principally phonologically mediated (Perfetti & Tan, 1998; Zhou et al., 1999). Indeed, there is strong evidence of phonological mediation in Chinese lexical access, but, as asserted by Shen and Forster (1999), such is likely to be task-dependent. The fact that these studies claiming stronger effects for phonological activation are largely naming studies should have raised some flags. Here, it has been demonstrated that evidence for both semantic and phonological processing strategies can be found if one is looking specifically for such, but in a task devoid of any particular advantage for either processing scheme, subject’s error rates spiked significantly when their semantic input was impaired.

These and other psycholinguistic studies show that learners of the Chinese script make use of the semantic elements of Chinese characters in ways that are simply not possible in phonographic writing systems, no matter how orthographically deep (or, in Unger’s terminology, no matter how logographic) they are.23 There is every reason to believe that this semantic component of Chinese writing is a feature of morphographic scripts 23  This is consistent with the argument advanced by Sproat (2000:80–81) that the “logographic” characteristics of English—exemplified by the distinct spellings of homophonous morphemes as seen in examples (1)-(4) below at the end of section 4—are different in kind from those found in Chinese. Sproat refers to the former as “haphazard ‘pseudologographic’ properties”.

Logography and the classification of writing systems


in general, and is not peculiar to Chinese. In an intelligently argued study, Goldwasser (2002) has shown that the semantic determinatives (“classifiers”) of ancient Egyptian hieroglyphic writing correlate not with specific words, but with cognitive semantic categories.24 Hirshorn & Fiez (2014:71–72), in a report that references both psycholinguistic and neurological studies, point out that writing system typology affects the nature of reading disorders such as dyslexia: [T]he core deficit in dyslexic readers of alphabetic languages has to do with phonological awareness …. In Chinese, the primary deficits in dyslexia are thought to reflect visual/orthographic (Ho, Chan, Lee, Tsang, & Luan, 2004; Ho, Chan, Tsang, & Lee, 2002) and morphological (Shu, McBrideChang, Wu, & Liu, 2006) risk factors, rather than a fundamental deficit in phonological awareness. In addition, writing deficits (Tan, Spinks, Eden, Perfetti, & Siok, 2005) are thought to play a larger role in Chinese reading impairments.

They also (2014:82) note measurable effects of native script type on second-language literacy acquisition: Research has shown that the cognitive and neural architecture of your L1 can affect how your reading your [sic] L2 (Tan et al., 2003). For example, Chinese-L1/English-L2 and English-L1/Chinese-L2 bilinguals were compared reading both English and Chinese, two languages that vary on many dimensions such as grain size/mapping principle, phonological consistency/ semantic transparency, and visual complexity (Nelson et al., 2009). It was found that Chinese-L1/English-L2 engaged a similar network while reading English as Chinese, but English-L1/ Chinese-L2 readers engaged additional areas while reading Chinese compared to English. This suggests that Chinese 24  For a related, preliminary investigation of the relationship between semantic components and active cognitive categories in Chinese, see Handel (2014). As far as I am aware there are no naturally-occurring morphographic scripts that do not rely on semantic elements. Such elements are found, in various manifestations, in Sumerian, Egyptian, Mayan, Chinese, Japanese kanji, Vietnamese Chữ Nôm, Zhuang sawndip, etc. This is not surprising given the need to differentiate the written forms of (near-) homophonous morphemes. If one makes use of the same phonetic component for such morphemes, then semantic determinatives are the most effective means of graphic differentiation. (For example: 紡 fǎng ‘to spin (thread)’ contains the “silk radical” 糸, 訪 fǎng ‘to visit’ contains the “speech radical” 言; both have phonetic element 方 fāng ‘square’.) The alternative would be to employ an arbitrary set of distinguishing marks, which would presumably place too great a burden on the memory of script learners.



readers can use a ‘Chinese’ strategy and neural substrates for reading English, but English readers cannot use an ‘English’ strategy to read Chinese (Perfetti et al., 2007). This research highlights the idea that not only do the dimensions of a writing system itself affect the way you read a second language, but so too do the design principles of one’s first language and how similar they are to their second language.

Psycholinguistic and neurological imaging studies remain limited in what they can tell us about real-world processing of writing systems, and many results are controversial. Keeping these caveats in mind, a broad survey of the literature on Chinese language processing provides strong evidence for numerous differences in the way the brain deals with written Chinese when compared to other writing systems. In any serious review of the implications of psycholinguistic research for the categorization of writing systems, Dehaene’s claims must be balanced against this evidence.

4. Valid criteria for the classification of writing systems Let us now set aside the cognitive question and return to the criterion question. For it is not at all a given that our classification of writing systems should depend solely on how the brain processes them during reading. Many other classification criteria might be equally or more valid, depending on what features of writing systems we seek to investigate. This is not merely an abstract exercise in classificatory theory, for there is good reason to believe that distinctions among types of writing systems affect such things as methods of devising written representations for neologisms, borrowing pathways when scripts are adopted to write other languages, and cognitive processes other than those associated with the reading of connected texts by accomplished script users. An appropriate categorization of writing systems can help to explain and predict these types of developments and behaviors. Among the many other criteria that seem potentially useful in writingsystems classification are internal physical structure of graphs, methods of writing new words, correlation of graphs to units of speech, differences in the cognition of writing, differences in the cognition of learning to read, differences in lexicographic practice, differences in pathways of script borrowing, differences in methods of computer encoding, etc. DeFrancis

Logography and the classification of writing systems


and Unger disparage “classificatory theories”, as opposed to “hard data”. But these criteria can be tested for effectiveness by determining whether categories based on them have explanatory or predictive power. I make no attempt here to evaluate all of these criteria, to definitively establish a single diagnostic method for writing systems categorization, or to present a comprehensive typological classification.25 I will only offer a few examples of the real-world importance of the distinction between phonographic and morphographic/logographic writing systems, which I believe must remain a key distinction in any meaningful typology. First, there are real differences in how new written forms are created when morphemes enter a language. A morphographic script like Chinese presents possibilities that are simply not available in an alphabetic script like English. Standard orthographic forms for English are invariably created by using the conventional phonemic associations of the letters of the Roman alphabet to represent the pronunciation of the word. In Chinese there are other possibilities that have no analog in alphabetic writing. For example, when the character for the word jiān ‘sharp’ was created, it was formed from the conjunction of the characters writing the morphemes for ‘small’ (xiǎo 小) and ‘large’ (dà 大), based on their conventional meaning associations without regard to pronunciation: 尖.26 Other examples are wāi 歪 ‘oblique, crooked’, formed by the juxtaposition of the characters writing the morphemes meaning ‘not’ (bù 不) and ‘straight’ (zhèng 正), and the recently created graph ài 鑀 ‘Einsteinium’, formed by the juxtaposition of the semantic element representing metal (金) and the phonetic element 愛 (which separately functions to write the morpheme ài ‘love’).27  In my view, a comprehensive typological classification remains beyond our reach at this point in time. Establishing one or more complete typologies of writing systems will require further discussion and collaboration among scholars as well as additional research and evidence from a variety of academic fields. But I do believe that all of the criteria I have listed remain in play as potentially useful for classification; their suitability as such deserves further evaluation. 25

 For convenience, characters are labeled with their Modern Standard Mandarin pronunciations in pinyin romanization and with an abbreviated gloss reflecting Modern Standard Mandarin and/or Classical Chinese usage. The actual historical pronunciations at the time of character creation were different. 26

 Einsteinium is element #99 of the periodic table, discovered in 1952 and named after Albert Einstein. The character given here is the one used in Taiwan as part of the “traditional script”. The form used in mainland China as part of the “simplified script” is 锿, pronounced āi. It was created through an analogous process, but making use of 27



Let us take another example of the structural distinctions that are seen between English and Chinese writing. It is not possible to find a homophonous triplet in modern English whose standard written forms do not have shared visual representations of shared phonemes.28 Examples of such triplets are: (1) /fɔr/: four, for, fore (2) /tu/: two, too, to (3) /ðer/: their, there, they’re (4) /per/: pair, pare, pear

shared letter graphs: , , shared letter graphs: , shared letter graphs: , , shared letter graphs: , ,

On the other hand, it is quite easy to do this for modern Standard Written Chinese (using modern Mandarin pronunciation). Indeed, one can easily find four, five, six, or even more homophonous morphemes whose written forms share no graphic similarities whatsoever. (5) táng: 堂 ‘hall’, 糖 ‘sugar’, 踼 ‘fall flat’ shared phonetic components: none (6) wǒ: 我 ‘I’, 婐 ‘maid’, 婑 ‘beautiful’ shared phonetic components: none (7) shí: 十 ‘ten’, 時 ‘time’, 食 ‘eat’, 石 ‘stone’, 實 ‘solid’ shared phonetic components: none

The characters in (7) are all high-frequency characters writing common morphemes of spoken Mandarin. One could certainly respond to this data with a shrug of dismissal. After all, these English and Chinese written graphs are processed through basically the same cognitive pathway during reading: the lexical route. But such a dismissal ignores important consequences of what I maintain is a real categorical distinction. For it would be extremely difficult, if not different components.

28  Dehaene (2009:31ff) talks at some length about the advantages inherent in the “irregular” spellings of English (“The Hidden Logic of Our Spelling System”), but seems to be unaware of or uninterested in the historical sources of this irregularity. His description implies that English script users somehow colluded to create distinct spellings for homophonous words in order to reduce ambiguity and increase the ability to recognize words using the lexical (rather than phonological) pathway during reading. The fact that English spelling distinctions represented real differences in pronunciation, historical or dialectal, and thus that English began (as all phonographic writing systems do) as an orthographically shallow system, deserves a place in any analysis of these “logographic” (as Unger & DeFrancis use the term) aspects of English writing.

Logography and the classification of writing systems


impossible, for speakers of another language to borrow the English writing system morphographically. Could speakers of an unwritten language, who are literate in written English, employ to write their word meaning ‘four’, for their word meaning ‘front’, and for their benefactive case marker? As far as I know there are no unambiguous historical examples of this kind of borrowing.29 Yet Chinese characters can be and have been borrowed in just this way in the history of written Japanese, written Korean, and written Vietnamese, among other languages (DeFrancis 1987; Handel 2009). This is a direct consequence of the fact that the basic units of written English (i.e. letters) represent semantically empty phonological units or strings, while the basic units of Chinese (i.e. characters) represent morphemes, units of language that inherently contain meaning.30 This difference is independent of the fact that letter sequences are recognized “logographically” through the lexical route by fluent readers, and has had tremendous consequences for the development of

 A possible exception is to be found in the so-called “heterograms” of Middle Persian and Parthian: orthographic Aramaic forms (written phonographically with an abjad or consonantary) presumably used to represent synonymous but phonologically unrelated words. For example, /šāhān šāh/ ‘king of kings’ might be represented graphically as , where is the written form of Aramaic ‘king’. There remains controversy over the linguistic interpretation of these ancient written forms, and much remains unknown about scribal practice. Even if these heterograms really functioned as described, they would constitute a singular event, in contrast with what is the norm when a morphographic script is borrowed. See Durkin-Meisterernst 2004/2012. (I am grateful to Sven Osterkamp for alerting me to this phenomenon.) 29

 Unger (80) objects to treating letters and Chinese characters as comparable units of writing systems: “The functional units of a writing system are typically not its smallest identifiable graphic units …. The orthographic word of English, the syllabic block of Korean, or the individual character of Chinese are all … basic [functional] units.” But what constitutes a “functional unit” depends on what functionality one is interested in. When the Roman script is borrowed to write another language, it is the individual letters that are manipulated by people and so are the functional units of borrowing— not the orthographic words. When the Korean script is adopted to write other languages (as it was for the Cia-Cia language of Indonesia), it is again the individual letters that are the functional units of borrowing, not syllable blocks. The same is true when native script users of English or Korean decide how best to transcribe newly encountered foreign words: they manipulate individual letters. This kind of functional saliency should be taken into account in any typology of writing systems. (An unrelated but interesting question raised by Unger’s statement is why it must be the syllable block, rather than the space-delimited orthographic word, that is identified as the “functional unit” of Korean writing.) 30



writing systems throughout East Asia.31 I contend, therefore, that the tendencies of Chinese characters in the modern Chinese script to represent morphemes and to contain semantic elements are important and valid criteria for the classification of Chinese writing into a distinct category. But is it in fact true that most Chinese characters represent morphemes? This is another point on which I differ with Unger.

5. Is Chinese writing morphographic? We can now return to the question we began with: Is it reasonable to characterize the modern Chinese writing system as a script in which the vast majority of graphs represent morphemes (or, more generally, meaningbearing linguistic units)? Unger never provides any statistical basis for refuting the claim. He does, however, provide some examples of script usage that most observers take to be morphographic but which he argues are in fact phonographic. Here he is not talking about the small number of graphs that function purely phonographically in the transcription of foreign loanwords,32 but about graphs that are typically considered to represent morphemes, such as 大 which writes the word dà ‘large’: For example, in the usual way of writing the Chinese word dàxué ‘university’

大学, neither character is a logogram in the strong sense Handel prefers

because, synchronically, this word is just a single two-syllable morpheme. The characters could perhaps be described as logographic etymologically, but etymology is not meaning, and speakers unaware of a word’s etymology can still learn and use it properly. (81)

Unger’s example is poorly chosen. I of course agree that etymological meaning is usually irrelevant to synchronic function or meaning in the minds of language and script users. But Unger provides no evidence that the syllables dà and xué are functioning as he claims, or that dàxué is “just  Indeed, one might say the same about the Akkadian borrowing of the morphographic Sumerian script. 31

 An example is 卡, pronounced kǎ, which is used to render the first syllable of such borrowed English words as card, cartoon, and calorie.


Logography and the classification of writing systems


a single two-syllable morpheme”. We are asked to accept his claim on faith. But one need only juxtapose the members of the morphemically minimal triplet below to make a strong argument that the dà and xué of dàxué are both full-fledged morphemes in modern spoken Chinese: xiǎoxué ‘elementary school’ (‘small’ + ‘study’) zhōngxué ‘junior/high school’ (‘middle’ + ‘study’) dàxué ‘college/university’ (‘large’ + ‘study’)

This is elementary morphological analysis, and taken together with the written forms of the obviously related free morphemes xiǎo ‘small’ (小), zhōng ‘middle, mid-sized’ (中), dà ‘large’ (大), and xué ‘study’ (学), it provides strong evidence that the graphs 大 and 学 do indeed represent morphemes when writing the word dàxué ‘university’.33 Indeed, it is difficult to see how one could argue that dàxué is a single morpheme instead of a bimorphemic compound without similarly claiming that the English words “high school” and “middle school” are each but a single morpheme. Unger (6) also points out that the word húdié ‘butterfly’ is a single morpheme, though written with two Chinese characters 蝴蝶. This is of course correct.34 Húdié is one of a significant minority of Chinese bisyllabic monomorphemic words, most of which are nativized foreign borrowings, ideophones (including, but not limited to, onomatopoeia), or names of plants, insects and vermin.35 In the written forms of these words the syllabic aspect of the morphosyllabic writing system trumps the morphemic aspect. In other words, the one-character-per-syllable feature of the writing system  Here, following Unger, I use the “simplified” form 学 which is in current use in mainland China, rather than the equivalent “traditional” form 學 in current use in Taiwan and Hong Kong (Handel 2013). Elsewhere in this paper I consistently use traditional forms.


 Kennedy (1955), in a famous essay, emphatically made this point in refuting the “monosyllabic myth” of Chinese. 34

 Sproat (2000:143, 154–155) estimates that there are about one hundred such words; he provides a list of 74. Examples include hàndàn 菡萏 ‘lotus’, biānfú 蝙蝠 ‘bat’, chángyáng 徜徉 ‘roam leisurely’. It seems possible that, aside from borrowings (not all of which are easily identified as such), all of these monosyllabic two-syllable words ultimately are ideophonic in origin, as the words for insects, vermin, and plants could derive from mimetic expressions related to the appearance or movement of these entities. This hypothesis is supported by the fact that the Old Chinese pronunciations of many of these words show consonance, assonance, or rhyme across the two syllables. 35



takes precedence over the one-character-per-morpheme feature.36 But such words do not lend support to Unger’s refutation of the morphographic nature of the writing system. Hú in húdié may be a meaningless syllable by itself, but 蝴 only writes this meaningless syllable when it is part of the morpheme meaning ‘butterfly’. When the meaningless syllable hú occurs in other bisyllabic morphemes, it has a different written representation. Examples (8), (9), and (10) are all monomorphemic words.37 (8) húdié (9) húlu (10) shānhú

蝴蝶 葫蘆 珊瑚

‘butterfly’ ‘gourd’ ‘coral’

Indeed, it is almost always the case that in the written representation of such two-syllable morphemes, both graphs share the same semantic element, which can be thought of as a semantic element for the morpheme that is graphically distributed across the two characters. In the case of húdié 蝴蝶 ‘butterfly’ it is 虫, referencing the semantic domain of insects and similar creatures.38 These written forms are therefore clearly morphographic, not phonographic, in nature, because their use is restricted to specific morphemes, and their graphic structure makes this restriction clear.39  I know of only one case in the modern standard writing system where the reverse is true, i.e. where morphological representation takes precedence over syllabic representation: the sub-syllabic suffix -r is written with a separate character 兒, as in the bimorphemic monosyllabic word huàr 畫兒 ‘painting’. 36

 Note that all three characters representing hú syllables contain a phonetic element, namely 胡. Thus according to Unger’s definition, none of the three is a logogram.


38  See Sproat (2000:144–154) and commentary on Sproat’s analysis by Boltz (in press). Early bisyllabic borrowings, which were nativized as single morphemes, follow the same orthographic pattern of having a “distributed” semantic element, e.g. bōlí 玻璃 ‘glass’ (from Pali) and pútáo 葡萄 ‘grape’ (from Persian). This orthographic phenomenon persisted into the 19th and early 20th centuries, e.g. níngméng 檸檬 ‘lemon’ and kāfēi 咖啡 ‘coffee’ (both words entering Chinese from European sources). More recent borrowings, however, are written with existing Chinese characters used as phonograms, the same method employed since ancient times for foreign transcription. As a result they lack a distributed semantic element, e.g. tǎnkè 坦克 ‘military tank’, qiǎokèlì 巧克力 ‘chocolate’. As Chinese speakers become increasingly skilled in English, and with computer standardization hindering the casual creation of new Chinese characters, this trend will likely continue, with the possibility of transforming the nature of the Chinese writing system significantly in ensuing decades or centuries.

 Boltz (in press) makes this same point, adding that the two characters used to write


Logography and the classification of writing systems


One could certainly choose better examples in support of Unger’s position. For example, the Chinese word for ‘police’ is jǐngchá. Unlike the case of dàxué ‘university’, which is part of a morphologically minimal set with xiǎoxué and zhōngxué, and which is transparently related both phonologically and semantically to xué ‘to study’, jǐngchá does not obviously break down into morphemes related to the meaning ‘police’. The written form, 警察, reveals its etymological origins, as jǐng 警 means ‘to warn, to guard’ (as in jǐnggào ‘to warn’) and chá 察 means ‘to inspect, to scrutinize’ (as in guānchá ‘to observe’). It would certainly be plausible to claim that synchronically jǐngchá is a single morpheme, written with two Chinese characters that have purely phonographic function (and are conventionally associated with the word, so that other homophonous characters are never used). However, it would be equally plausible to argue that because (a) literate Chinese speakers learn from an early age to write jǐngchá as 警察; (b) they know the meanings of the morphemes normally represented by these two characters; and (c) the meanings ‘to guard’ and ‘to inspect’ are plausibly related to the meaning ‘police’; therefore the etymological morphological structure is still synchronically alive for these speakers. The morphology of jǐngchá is reinforced for such speakers through their daily use of the writing system, and is reflected in the lexicographic structure of reference works such as dictionaries.40 This sort of thing simply doesn’t happen with the same degree of prevalence in English, despite the retention of many etymological spellings.41 Ask a literate native speaker if police and policy share a meaningful element, and you will not receive a confident response. But ask a literate native speaker of Chinese if jǐngchá and jǐnggào share a meaningful element, and the húdié “by virtue of the semantic classifiers used, do have, or at least imply, a meaning even when written in isolation”. (emphasis added)

 Our two hypotheses about the degree of morphological transparency of words like dàxué and jǐngchá are testable. For example, a well designed psycholinguistic experiment could determine if exposure to the spoken word jǐngchá ‘police’ primes other spoken words containing the morpheme jǐng ‘to warn’ while failing to prime words containing unrelated homophonous morphemes, such as jǐng ‘scenery’ (written 景). I do not know if such experiments have been carried out or reported in the literature. 40

 One might suppose that folk etymologies (e.g. the belief that history derives from his story) are analogous, but in fact they are not: the Chinese situation just described causes etymological morphological structure to persist, i.e. keeps synchronic morphology aligned with etymological morphology, while folk etymology does the opposite. 41



response will be definitive.42 One can even plausibly argue that for literate native speakers, the hú and the dié of húdié ‘butterfly’ are meaningful elements, i.e. they have acquired morphemic status. Here Unger’s argument is turned on its head: etymologically húdié may have been one morpheme, but in the modern language it has become two. Consider the following words in which dié appears, isolated from hú, with the meaning ‘butterfly’: (11) fèngdié (12) diéyǒng (13) diéxínggé (14) diégǔ

鳳蝶 蝶蛹 蝶型閥 蝶骨

‘swallowtail butterfly’ ‘chrysalis’ ‘butterfly valve’ (used in gas piping) ‘sphenoid bone’ (due to its butterfly-like shape)

Moreover, this use of dié is productive: new compounds can be formed from it. In all cases it is written with 蝶, never with other characters pronounced dié, thus preserving the morphographic link between written and spoken form that is so pervasive in Chinese writing. It is highly probable that this morphological development within the spoken language was partly catalyzed by features of the writing system as a whole and by the morphographic nature of the character 蝶 in particular, leading to a reinforcing cycle. When newly coined lexical items (like diégǔ 蝶骨) are introduced to a wider user base through the written medium, the graphic form 蝶 signals the meaning of the underlying morpheme precisely because it is morphographic to begin with. This makes the new word easier to process and understand and more likely to enter the spoken language, strengthening the independent morphemic status of dié and making it more likely to combine in a versatile way in the creation of new compounds. Chao (1968:139) had this kind of development in mind when he said “The  I am grateful to an anonymous reviewer for pointing out that the high frequency of the word jǐngchá ‘police’ has led to the formation of a new morpheme, jǐng ‘police’. (Alternatively, we can think of this as an extended meaning of the earlier morpheme jǐng ‘to warn, to guard’.) This is the result of the use of jǐng as an abbreviated form of jǐngchá to form compounds such as jǐngchē 警車 ‘police car’, which cannot be interpreted as meaning ‘guard car’ or ‘warning car’. This development, in which jǐng has acquired the meaning of the compound jǐngchá, shows parallels to the creation of the morpheme dié ‘butterfly’ described below. It can be broadly described as an instance of the phenomenon in historical semantics termed displacement. (An example of displacement from the history of English is the word daily acquiring the meaning of ‘daily newspaper’.) 42

Logography and the classification of writing systems


predominance of monosyllabism [of morphemes in spoken Chinese] is so great that a speaker or writer tends to read meaning into single syllables when there was none originally…. Which all goes to show that completely meaningless syllables in Chinese are always felt as something of an anomaly.” Where I noted (2013:24) that for script users, “… each Chinese character also has an associated pronunciation and meaning, namely the pronunciation and meaning of the morpheme with which it is or has been conventionally associated. For native users of the script, these linguistic features are thought of as inhering within the written graph itself,” Unger (77) replies that such observations are meaningless as a basis for script analysis, because “as every anthropologist is aware, self-reported beliefs— especially ones reinforced by schooling—are not the whole story.”43 I fully agree. I do not believe that self-reported beliefs have any reliability when it comes to cognitive structures and processing. But although they are not “the whole story”, they are indeed part of the story, and in the case of writing systems, an important part. Setting aside the pedagogical and lexicographic tradition within China itself, literacy instruction in premodern Japan, Korea, and Vietnam—where Classical Chinese was the high-prestige formal written language—was predicated on the recognition that Chinese characters have associated meanings, i.e. are morphographic. These meanings were made explicit through memorizing Chinese-character pronunciations in association with native-language semantic glosses: the kun くん of Japanese and hun 훈 of Korean. Perhaps, as Unger would have it, this is all a fantasy or mass delusion about Chinese writing. But if so, it is a fantasy that has acquired linguistic substance. The development and use of entire writing systems—Korean idu 이두, Vietnamese Chữ Nôm, and Japanese kanji-kana majiribun 漢字仮名混じり文 (mixed-script writing), to name the three best known— is due to it. It was only through the recognition of the meanings of the Chinese morphemes represented by Chinese characters that early scribes in Japan, Korea, and Vietnam were able to repurpose those graphs to write synonymous noun and verb roots in their native languages (Handel 2009). These developments are far removed from how written words on the page 43  In quoting me, Unger (77) omits the important word “associated” before “pronunciation and meaning”. Since he gives no explicit indication that he has done so, this omission was likely an oversight. Associations can, of course, be broken or lost.



are processed in the brain, but they are certainly within the domain of linguistic inquiry, and a classification of writing systems which sheds light on such historical developments is clearly useful, if not essential.

6. Conclusion: towards a new typology I am not aware of any knowledgeable scholar who would dispute the role of phonetic elements in Chinese writing, and indeed psycholinguistic research confirms that graphs in Chinese writing are sometimes processed via the phonological route when reading. The important contributions of both DeFrancis and Unger to this understanding, in an era before sophisticated psycholinguistic studies were possible, should not be underemphasized. But by adopting a novel definition of “logographic” in order to facilitate their typology of writing systems, Unger and DeFrancis give these phonographic features of Chinese writing exclusive primacy, and therefore erase all useful distinctions between the structure and function of Chinese writing and of other writing systems. It is equally (if not more valid) to turn the issue around, and note that despite the fact that Chinese writing shares certain universal properties with other human writing systems, it is also distinct from phonographic writing systems in important and interesting ways. The traditional definition of logographic (or of its equivalent morphographic) is useful in capturing these differences. I would like to stress that I do not claim that these categories are “essential” or “absolute”, contra Unger’s claims. They are not essential in that the function of Chinese characters (whether individual ones or the script taken as a whole) can change over time or in different linguistic contexts (Handel 2009, see especially page 92 notes 4 and 5); they are not absolute in that alphabetic writing systems contain non-alphabetic features, syllabic writing systems contain non-syllabic features, and logographic writing systems contain non-logographic features. Nevertheless, the differences among these systems are real and significant, and are no less so for being clear-cut tendencies rather than absolute categorizations. The tendency of English written forms to correspond to the phonemic segments of words (with some visual distinctions across homophonous morphemes) and of Chinese characters to correspond to morphemes

Logography and the classification of writing systems


may not be “organizing principles” (the phrase used by Sampson, which DeFrancis & Unger object to), but to recognize them as significant tendencies, i.e. as real patterns in the data, is simply to acknowledge what are empirical facts. To label the one alphabetic and the other morphographic is to acknowledge these facts and to recognize their significance. Such recognition need not imply a claim that these features are inherent, essential, or immutable, that Chinese characters cannot be repurposed to be used purely phonographically, that their role cannot change over time. I therefore reject Unger’s accusation that a characterization of Chinese as logographic/morphographic is essentialist. I agree with the Unger-DeFrancis typology of writing systems to the extent that I believe it is useful to distinguish full from partial writing systems, to precisely define the endpoints of the continuum of theoretically possible writing systems, and to identify structural constraints on actually occurring human writing systems within that continuum. Their typology seems to capture something essential about the way that human writing systems work.44 But I disagree that this is the only useful way of categorizing writing systems. Many strands of evidence support the idea that writing systems lying at different ends of this continuum are categorically different. One need only broaden one’s field of inquiry to see why a categorization that recognizes these differences is also of immense value. In conclusion, I believe the evidence shows that there is real value in sub-categorizing writing systems, even while recognizing that the categories have fuzzy boundaries. There is nothing essentialist about such a categorization, because it makes no assumptions about inherent, immutable properties or “organizing principles”; it is simply an empirically-based set of distinctions. Nor is it a pointless “classificatory exercise”, because the sub-categorization reflects real-world differences and predicts realworld consequences. Once this is recognized, then whether one wishes to 44  That said, one wonders where the extinct Tangut writing system would fit into their typology. All indications are that among the thousands of Tangut graphs, very few contained overt phonological components. Tangut then, may be an example of a writing system that approaches the “logographic” end of their continuum, beyond the expected range for actually occurring writing systems. On the other hand, Tangut was created in a short period of time, rather than developing organically, and although a large corpus of texts was produced in Tangut, we do not know to what degree the script could be read by literate speakers of the language without the use of lexicographic reference aids.



label the category containing Egyptian hieroglyphs, Mayan hieroglyphs, Sumerian cuneiform, Japanese kanji and Chinese characters as logographic or something else simply becomes a matter of terminological choice. I am happy to sidestep terminological disputes by referring to this category as morphographic, and to more specifically characterize Chinese writing of the last two millennia as morphosyllabic. Extending this categorization of Chinese writing into a comprehensive typology of writing systems will depend in part on taking evidence from both psycholinguistic studies and the history of script development and adaptation. A categorization arrived at through such means would constitute a broad revision of both traditional classifications and the UngerDeFrancis typology. The present study might be considered a first step in this direction, as it establishes one crucial distinction that is supported by both types of evidence.45 Unger & DeFrancis (1995:53–54) consider that a hypothetical reader might object to their typology by saying “that writing systems should nevertheless be classified into predominantly logographic and predominantly phonographic types, according to the proportion of logograms to phonograms they employ, even though the difference between these types pales into insignificance when either is compared with a true code [i.e. a writing system completely lacking phonographic elements]. The question, however, is not whether such a typology is possible but whether it is useful. Are there real-world correlates of this distinction between ‘largely logographic’ and ‘largely phonographic’ writing systems, or is drawing such a distinction simply a facile, deadend exercise of no consequence?”. I hope that I have shown here that this rhetorical question can indeed be answered in the affirmative: there are real-world correlates of this distinction, and they do indeed justify the classification of writing systems as phonographic or logographic according to the traditional definitions of those terms.  An important question that still remains unanswered has to do with the causal factor for this crucial distinction. We have seen that morphographic writing systems and phonographic writing systems create categorically different effects. But what is it about morphographic writing that leads to these effects? Is it that the majority of graphic units represent meaningful units of speech, or is it that the majority of graphic units contain semantic elements? In other words, is it their representational function or their internal structure that is definitive, or is it the combination of both factors? 45

Logography and the classification of writing systems


References Baxter, William & Laurent Sagart (2014). Old Chinese: a new reconstruction. Oxford: Oxford University Press. Bolger, Donald J., Charles A. Perfetti & Walter Schneider (2005). Crosscultural effect on the brain revisited: universal structures plus writing system variation. Human Brain Mapping 25:1. 92–104. Boltz, William G. (in press). Orthographic monosyllabicity. In Benjamin Tsou, Bit-chee Kwok & Andy Chin (eds.) Commemorative essays on the 120th birthday of Professor Y. R. Chao. Hong Kong: The Commercial Press. Chao, Yuen-Ren (1968). A grammar of spoken Chinese. Berkeley: University of California Press. Coulmas, Florian (2003). Writing systems: an introduction to their linguistic analysis. Cambridge: Cambridge University Press. Daniels, Peter T. (1996). The study of writing systems. In Peter T. Daniels & William Bright (eds.) The world’s writing systems. New York: Oxford University Press. 3–12. DeFrancis, John & J. Marshall Unger (1994). Rejoinder to Geoffrey Sampson, “Chinese script and the diversity of writing systems”. Linguistics 32. 549–554. DeFrancis, John (1984). The Chinese language: fact and fantasy. Honolulu: University of Hawai’i Press. Dehaene, Stanislas (2009). Reading in the brain: the new science of how we read. New York: Penguin. Durkin-Meisterernst, D. (2004). Huzwāreš. In Ehsan Yarshater (ed.) Encyclopædia Iranica Vol. XII, Fasc. 6. New York: Encyclopædia Iranica Foundation. 585–588. [An updated version is available online at http://www.iranicaonline.org/articles/huzwares (updated on 23 March 2012).] Feldman, Laurie Beth & Witina W. T. Siok (1999a). Semantic radicals contribute to the visual identification of Chinese characters. Journal of Memory and Language 40. 559–576. Feldman, Laurie B. & Witina W. T. Siok (1999b). Semantic radicals in phonetic compounds: implications for visual character recognition in Chinese. In Jian Wang, Albrecht W. Inhof & Hsuan-Chih Chen (eds.)



Reading Chinese script: a cognitive analysis. Mahwah, NJ: Lawrence Erlbaum Associates. 19–35. Goldwasser, Orly (2002). Prophets, lovers, and giraffes: wor(l)d classification in ancient Egypt. Wiesbaden: Harrassowitz Verlag. Handel, Zev (2009). Towards a comparative study of sinographic writing strategies in Korean, Japanese, and Vietnamese. Scripta 1. 89–125. Handel, Zev (2012). Can a logographic script be simplified and remain logographic? Lessons from the 20th century Chinese writing reform informed by recent psycholinguistic research. Interactions of writing systems (Proceedings of Scripta 2012, 8–10 October). Seoul: Hunmin jeongeum Society. 157–223. Handel, Zev (2013). Can a logographic script be simplified? Lessons from the 20th century Chinese writing reform informed by recent psycholinguistic research. Scripta 5. 21–66. Handel, Zev (2014). The role of semantic classifiers in modern Chinese writing: an investigation of categories reflected in neography. Paper presented at the Idea of Writing Conference, Leiden. Hirshorn, Elizabeth A. & Julie A. Fiez (2014). Using artificial orthographies for studying cross-linguistic differences in the cognitive and neural profiles of reading. Journal of Neurolinguistics 31. 69–85. Ho, Connie Suk-Han, David Wai-Ock Chan, Suk-Han Lee, Suk-Man Tsang & Vivian Hui Luan (2004). Cognitive profiling and preliminary subtyping in Chinese developmental dyslexia. Cognition 91:1. 43–75. Ho, Connie Suk-Han, David Wai-Ock Chan, Suk-Man Tsang & Suk-Han Lee (2002). The cognitive profile and multiple-deficit hypothesis in Chinese developmental dyslexia. Developmental Psychology 38:4. 543– 553. Hu, Wei, Hwee Ling Lee, Qiang Zhang, Tao Liu, Li Bo Geng, Mohamed L. Seghier, Clare Shakeshaft, Tae Twomey, David W. Green, Yi Ming Yang & Cathy J. Price (2010). Developmental dyslexia in Chinese and English populations: dissociating the effect of dyslexia from language differences. Brain 133:6. 1694–1706. Kennedy, George A. (1955). The butterfly case (part I). Wennti 8. 1–47. (Reprinted in Tien-yi Li (ed.) (1964). Selected Works of George A. Kennedy. New Haven: Far Eastern Publications, Yale University. 274– 322.) Lakoff, George (1987). Women, fire, and dangerous things: what categories

Logography and the classification of writing systems


reveal about the mind. Chicago: University of Chicago Press. Nelson, Jessica R., Ying Liu, Julie Fiez & Charles A. Perfetti (2009). Assimilation and accommodation patterns in ventral occipitotemporal cortex in learning a second writing system. Human Brain Mapping 30:3. 810–820. Perfetti, Charles A. & Li Hai Tan (1998). The time course of graphic, phonological, and semantic activation in Chinese character identification. Journal of Experimental Psychology. Learning, Memory, and Cognition 24:1. 101–118. Perfetti, Charles A., Ying Liu, Julie Fiez, Jessica Nelson, Donald J. Bolger & Li-Hai Tan (2007). Reading in two writing systems: accommodation and assimilation of the brain’s reading network. Bilingualism: Language and Cognition. 10:2. 131–146. Rogers, Henry (2005). Writing systems: a linguistic approach. Oxford: Blackwell. Sampson, Geoffrey (1994). Chinese script and the diversity of writing systems. Linguistics 32:1. 117–32. Shen, Di & Kenneth I. Forster (1999). Masked phonological priming in reading Chinese words depends on the task. Language and Cognitive Processes 14:5-6. 429–459. Shu, Hua, Catherine McBride-Chang, Sina Wu & Hongyun Liu (2006). Understanding Chinese developmental dyslexia: morphological awareness as a core cognitive construct. Journal of Educational Psychology 98:1. 122–133. Siok, Wai Ting & Laurie B. Feldman (unpublished manuscript). Activation of the form and meaning of semantic radicals in Chinese character recognition. Siok, Wai Ting, Charles A. Perfetti, Zhen Jin & Li Hai Tan (2004). Biological abnormality of impaired reading is constrained by culture. Nature 431. 71–76. Siok, Wai Ting, Zhendong Niu, Zhen Jin, Charles A. Perfetti & Li Hai Tan (2008). A structural–functional basis for dyslexia in the cortex of Chinese readers. Proceedings of the National Academy of Sciences of the United States of America 105:14. 5561–5566. Sproat, Richard (2000). A computational theory of writing systems. Cambridge: Cambridge University Press. Su, I-Fan & Sam-Po Law (in press). Neural correlates of character



recognition and phonological access. In Rint Sybesma (ed.) Encyclopedia of Chinese language and linguistics. Leiden: Brill. Tan, Li Hai & Wai Ting Siok (2006). How the brain reads the Chinese language: recent neuroimaging findings. In Ping Li, Li Hai Tan, Elizabeth Bates & Ovid J. L. Tzeng (eds.) The handbook of East Asian psycholinguistics, vol. 1: Chinese. Cambridge, New York: Cambridge University Press. 358–371. Tan, Li Hai, Angela R. Laird, Karl Li & Peter T. Fox (2005). Neuroanatomical correlates of phonological processing of Chinese characters and alphabetic words: a meta-analysis. Human Brain Mapping 25:1. 83–91. Tan, Li Hai, John A. Spinks, Ching-Mei Feng, Wai Ting Siok, Charles A. Perfetti, Jinhu Xiong, Peter T. Fox & Jia-Hong Gao (2003). Neural systems of second language reading are shaped by native language. Human Brain Mapping 18:3. 158–166. Tan, Li Hai, John A. Spinks, Guinevere F. Eden, Charles A. Perfetti & Wai Ting Siok (2005). Reading depends on writing, in Chinese. Proceedings of the National Academy of Sciences of the United States of America 102:24. 8781–8785. Unger, J. Marshall & John DeFrancis (1995). Logographic and semasiographic writing systems: a critique of Sampson’s classification. In Insup Taylor & David Olson (eds.) Scripts and literacy: reading and learning to read alphabets, syllabaries and characters. Dordrecht: Kluwer. 45–58. Unger, J. Marshall (2014). Empirical evidence and the typology of writing systems: a response to Handel. Scripta 6. 75–95. Williams, Clay & Thomas Bever (2010). Chinese character decoding: a semantic bias? Reading and Writing 23:5. 589–605. Zhou, Xiaolin, William Marslen-Wilson, Marcus Taft & Hua Shu (1999). Morphology, orthography, and phonology in reading Chinese compound words. Language and Cognitive Processes 14:5-6. 525–565. Zev Handel Department of Asian Languages and Literature University of Washington, USA

[Received 13 March 2015; revision received 23 July 2015; accepted 31 July 2015]

View more...


Copyright © 2017 EDOC Inc.