Phonological derivation from proximal to distal demonstratives in Chinese

2.1 The database for the present analysis

It is difficult to determine how many dialects Chinese has, because they are remarkably different in terms of phonology, lexicon, and grammar. Even within a dialect family, such as the Wu dialect, people from different areas often cannot communicate using their own dialects. Consequently, there are different views about the divisions of Chinese dialects (Li 2001: 29).^[4] For the purpose of the present study, I adopt the so-called eight major dialects as follows:

the Northern dialect,^[5] the Jin dialect, the Xiang dialect, the Gan dialect, the Wu dialect, the Min dialect,^[6] the Hakka dialect, and the Yue dialect^[7]

I reserve the term ‘dialect’ for each of the eight dialectal families and use ‘subdialect’ for smaller dialectal areas.^[8] In reality, the Chinese language comprises more than 1000 subdialects with remarkable differences in terms of phonology, lexicon, and grammar. This article does not make any further differentiations among them and simply uses the term ‘subdialect’ for them.

To guarantee the accuracy and complete coverage of the present study, I have established a database on the basis of the following investigations.

1. Cao (2008). A well-balanced investigation of all eight dialects, covering 930 subdialects.

2. Huang (1996). A collection of papers and monographs, covering 37 subdialects of the eight dialects.

3. Wu (2000). An investigation of 53 subdialects spoken in Hunan Province, mainly belonging to the Xiang dialect but including the Hakka dialect and the Northern dialect as well.

4. Qiao (2000). An investigation of 21 subdialects of the Jin dialect.

5. Beida (1995).^[9] An investigation of 20 representative subdialects of all eight dialects.

6. Li (2001) and Yuan (2001). Two general surveys of Chinese dialects, covering all eight dialects.

The present analysis is based on the demonstratives of more than 1000 subdialects. Although there is some overlap between the works mentioned above, the details of their descriptions vary greatly, and hence, they are all valuable for my study. Cao (2008) and Beida (1995) are dictionaries that provide only partial forms of demonstratives.

2.2 Lexical sources for chinese demonstratives and their distributions in dialects

A replacement of systems of demonstratives took place in late Medieval Chinese, used from the seventh century to the tenth century. This period gave birth to all the demonstratives that are still used in Contemporary Chinese, that is, zhè, gè and dǐ.^[10] The origin of these demonstratives has long been debated in the field of Chinese linguistics, especially the origins of the proximal and distal demonstratives zhè and nà^[11] (for details, see Jiang and Cao 2005; Lü 1985; Ohta 1987; Wang 1989). Here, I mention only the most convincing analyses without going into any detail. Jiang (1999) postulates that the demonstrative dǐ originated from a locative word meaning ‘bottom’; I would like to leave this issue open. Zhang (2001) employs solid phonological evidence to argue that the demonstrative zhè in the Northern dialects and gè in many southern dialects originated from two general classifiers in Medieval Chinese. It is less debatable that the demonstrative gè was grammaticalized from its classifier usage because the two words have the same phonological form in many dialects (for details, see Lin 2018; Qian 1997, 2014), but it is problematic to claim that the demonstrative zhè was also derived from the classifier zhī, because the two phonological forms are different in Standard Chinese. However, this issue is successfully resolved by Zhang (2001), who provides the following two pieces of phonological evidence.

1. In Qie Yun (edited in 601 AD, a phonological dictionary) and Piao Tong Shi Yan Jie (edited in 1515 AD, a textbook to teach Koreans Chinese) the demonstrative zhè and the classifier zhī have exactly the same phonological form.

2. In at least nine present-day subdialects, the demonstrative zhè and the classifier zhī have the same phonological form. For instance, in the Fuqing subdialect (belonging to the southern Min dialect), both are realized as tsia ²¹ .^[12]

According to Cao (2008), 131 subdialects have the same phonological forms for the proximal demonstrative zhè and the general classifier zhī. The classifier gè was grammaticalized from the noun ‘bamboo’ and the classifier zhī was grammaticalized from the noun ‘bird’ in Medieval Chinese, both of which were among the few earliest classifiers (Wang 1989: 18–41). The earliest examples of the demonstratives gè, zhè (zhī), and dǐ are illustrated in the following examples (Lü 1985: 183–245).

Due to diachronic change and cross-dialectal variations, every demonstrative above has contemporary reflexes with a variety of phonological forms. The consonants in the onsets are the most important clues indicating demonstratives with the same origin. Table 2 shows the diachronic changes and cross-dialectal variations of the three demonstratives with different origins.

The onsets of these three demonstratives are [k], [tʂ] or [t], and their dialectal variations are listed as follows:

1. If the onset consonants of the demonstratives are [tʂ], [ts], [tɕ] or [tʃ], they originated from the general classifier zhī.

2. If the onset consonants of the demonstratives are either [k] or [ɡ], they originated from the general classifier gè.

3. If the onset consonants of the demonstratives are either [t] or [d], they originated from the locative word dǐ.

As for the 930 subdialects in Cao (2008), the distribution of the three demonstratives with different origins is shown in Table 3.

The distinction between the investigated subdialects of the southern dialects, such as the Wu dialect and the Xiang dialect, is much more fine-grained than in the Northern dialects, probably because of different complexities in the relationship between them. In the other 179 subdialects, most of the phonological forms are derived via the same principle, as I will discuss in the subsequent sections.

4.1 The Asymmetry between proximal and distal demonstratives

Within a grammatical opposition such as the one between singular vs. plural, as mentioned above, the singular is more basic and remains unmarked, while the plural is less basic and marked (Corbett 2004: 17). For demonstratives, the proximal ones are more basic and unmarked (default), and the distal ones are derivative and marked (Ariel 1990). This distinction is confirmed by observations concerning human cognition and the frequency of occurrence. Proximal demonstratives indicate a smaller distance between the speaker and the referent, while distal demonstratives indicate a farther distance. The orientation of human cognition is typically from the decitic center to a farther location. Langacker (1991: 242–246) relates the proximal-distal contrast in demonstratives to the present—past distinction in the tense system, where proximal corresponds to present and distal correspond to past. Likewise, Ariel (1990) argues that proximal demonstratives are less marked than distal demonstratives. These assumptions are supported by the different frequencies of the proximal zhè and distal nà throughout the history of Chinese, cf. Table 7.

Clearly, the frequency of the proximal zhè is always higher than that of the distal nà, which shows that the former is more basic than the latter.

There are some exceptions to the derivational direction between proximal and distal demonstratives. In 33 of the 930 subdialects mentioned by Cao (2008) (accounting for 3.5% of the total), the demonstratives that were grammaticalized from the classifier gè were first realized as distal demonstratives, and then gave rise to the corresponding proximal demonstratives through a phonological principle, as illustrated in Table 8. This irregular realization causes interesting derivations, which will be discussed in the subsequent sections.

Some subdialects even have more than one proximal demonstrative, each originating from a different source. For instance, the Shanghai subdialect has two proximal demonstratives, diəʔ ¹² and geʔ ¹² ; the former is the traditional demonstrative, and the latter is the newly emerging one (Huang 1996: 487). Obviously, the proximal demonstrative diəʔ ¹² was grammaticalized from the locative word dǐ, and the proximal geʔ ¹² was grammaticalized from the classifier gè.

Thus far, I have mainly discussed proximal demonstratives without mentioning the corresponding distal demonstratives. As mentioned previously, it has been convincingly argued that two general classifiers plus one possible locative word first developed into proximal demonstratives, but no sources for the distal demonstratives have been found.^[16] The main focus of this article is on where the corresponding distal demonstratives come from and how they emerged. This question is addressed in the following section.

4.2 Phonological forms of twin-syllables and cognate words in chinese

Chinese is a tonal language, and tonal contours function to distinguish meanings. Additionally, Chinese is basically a monosyllabic language, and approximately 90 percent of its morphemes, defined as minimal pairings of meaning and form (usually [smaller than] words), are monosyllabic (Lü 1963). In all dialects, basic demonstratives in Chinese are monosyllabic, regardless of whether they are proximal or distal. Chinese does not allow any consonant clusters, and hence, only a single consonant can occur at the onset or coda position. The glides, [j], [w], and [y] may occur in the initial position of the rhyme. The maximal syllable of Chinese can be represented as follows^[17]:

(5)

When the onset position is empty, the glides (or semivowels) are raised to occupy this position. In this case, the glides may function to distinguish medial from distal demonstratives, as we will see in the next section.

In terms of auditory phonetics, the concept of ‘sonority’, which is central to this analysis, can be defined from two closely related perspectives: energy and openness. First, it refers to the carrying power of a sound; the more sonorous a sound is, the more articulatory effort is required to produce it (Clark and Yallop 1992: 61). Second, it refers to relative openness of the vocal tract, which directly determines the relative loudness of a sound. The wider open the mouth is, the more sonorous the sound is. For example, among vowels, the low vowels exhibit the highest degree of sonority and the high vowels have the lowest degree of sonority (Zsiga 2013: 333–334). Syllables are structured in accordance with certain principles relating to the sonority of sounds. It has been known for over a century that the construction of syllables is guided by the following principle (Kenstowicz 1994: 254):

Sonority Sequencing Principle: onsets must rise in sonority toward the nucleus and codas must fall in sonority from the nucleus.

Interestingly, crosslinguistic regularity in constructing a syllable is widely found in the phonological derivation between proximal and distal demonstratives in many Chinese dialects. While the Sonority Sequencing Principle (SSP) works in a linear fashion when constructing a syllable across languages, it behaves in a non-linear fashion in the derivation of distal demonstratives from proximal demonstratives in Chinese. Overall, the sonority of vowels is higher than that of consonants, but vowels differ in sonority, too. According to Kiparsky (1979) and Kenstowizc (1994: 254–255), the vowels are ordered in terms of sonority as follows:

Notice that among all vowels, schwa [ǝ] has the lowest sonority and the low vowel [a] has the highest sonority; the other vowels are located in between these extremes. The sonority scales of sounds at the onset positions are as follows (Selkirk 1984: 112):

When no consonant occurs in the onset position, the high vowels [i] and [u], originally the initials of the rhyme, automatically take the onset position and are pronounced as glides (semivowels) [j] and [w].

The SSP operates within syllables in all languages. For example, English has monosyllabic words such as plain [plein] and plant [plænt], but the syllables cannot be structured as plani or plnat. In Chinese, however, this phonological principle is not only found inside a syllable but also within a group of semantically related words in a cross-syllabic fashion. As mentioned above, Chinese is basically a monosyllabic language, which means that a syllable usually stands for a meaningful morpheme that corresponds to a Chinese character. Due to a historical disyllabification tendency, nearly 80% of words in contemporary Chinese are dissyllabic, and their phonological forms are determined by the two meaningful morphemes. For instance, in ɕiao⁵¹ ʂi⁵⁵ (teach + master) ‘teacher’, there is no phonological constraint between the two syllables. Many nouns become disyllabic through the addition of nominal suffixes such as -zi, -er or -tou, whose syllables are also independent. However, there is always a portion of disyllabic words whose two syllables are phonologically constrained. In these cases, neither of the two syllables conveys any semantic content, and they cannot be separated and used independently. These words are called ‘twin-syllable’ words, as illustrated in (8) (Shi 1995).

Between the two syllables of a twin-syllable word, even the tones are closely related to each other. For instance, nearly 100% of Chinese onomatopoeias have a high tone, e.g., p ^h u ⁵⁵ toŋ ⁵⁵ , p ^h iŋ ⁵⁵ p ^h aŋ ⁵⁵ , and huŋ ⁵⁵ loŋ ⁵⁵ .

On the basis of the sonority scales of sounds and the SSP, we can immediately recognize that the construction of twin-syllables in Chinese obeys a principle similar to what is found cross-linguistically within a single syllable. In the following formula, S stands for syllable, O stands for onset, N stands for nucleus, and C stands for coda; the subscripts indicate whether the segments in question belong to the first syllable or the second syllable. The function f _s is defined as returning the sonority value of a segment.

For example, according to the definition of sonority degree, in the onomatopoeia huŋ ⁵⁵ loŋ ⁵⁵ , the onset of the second syllable [l] is more sonorous than that of the first syllable [h], the nucleus of the second syllable [o] is more sonorous than that of the first syllable [u]. The codas of the two syllables are equally sonorous.

Before using the sonority principle to explain the derivation between proximal and distal demonstratives, it should be emphasized that there is a language-specific way of coining new words in Chinese. Phonemes within a syllable are linearly adjacent to each other, so it is not surprising that there are certain phonological constraints between them. However, proximal and distal demonstratives are independent of each other and rarely occur adjacently. How can they show a phonological correlation? The answer lies in the way in which speakers of Chinese create phonological forms for new words. In his pioneering study, Wang (1978[2000]) discovered that a pair of antonyms or a group of semantically related words can be assigned a new phonological form by changing the tones, consonants at onset, or vowels at the nucleus of certain etymological roots. There are thousands of words whose phonological forms have been created in this way (for details, see Wang 1997). Some of them are illustrated in (10).

While the English verbs buy and sell bear no phonological similarity, their corresponding verbs in Chinese share the same syllable and are differentiated by tone only. The special process of creating phonological forms in Chinese mentioned above is crucial for understanding the phonological derivation between proximal and distal demonstratives.

In short, phonological correlations and constraints work in a cross-syllabic fashion in Chinese, including in the creation of phonological formations for words that are semantically related. Since proximal and distal demonstratives are deictically contrastive, the phonological correlations between them are perfectly natural in Chinese.

4.3 Phonological rule for the derivation from proximal to distal demonstratives

The phonological rule governing the derivation from proximal to distal demonstratives is similar to the SSP for constructing syllables. Proximal and distal demonstratives are semantically contrastive and thus can be viewed as pairs of antonyms or semantically related words. As noted previously, classifiers and locative words were the first elements to be grammaticalized into proximal demonstratives.^[18] Instead of recruiting another lexical source for their corresponding distal demonstrative, almost all dialects employ a phonological rule to derive the corresponding distal demonstratives. Specifically, the sonority values of sounds at the onset, nucleus, and coda are remarkably contrastive between the phonological forms of proximal and distal demonstratives. In the following formulas, f _s is defined as a function returning a degree of sonority; O_d, N_d and C_d stand for the onset, nucleus and coda of distal demonstratives; and O_p, N_p and C_p stand for the onset, nucleus and coda of proximal demonstratives.

These three formulas do not operate in the same fashion in differentiating proximal and distal demonstratives in different subdialects. Another phonological principle for constructing a syllable, namely, the Sonority Dispersion Principle (SDP), states that every language prefers to maximize the sonority slope from onset to nucleus and to minimize this value from nucleus to coda (Clements 1990: 291). Similarly, the sonority slopes between the two onsets of proximal and distal demonstratives tend to be maximized and are rarely equal to each other; the sonority slopes between the two nuclei and codas of the two demonstratives tend to be minimized and are often equal to each other, a point that I will discuss in Section 4.4.

According to the database consisting of more than 1000 subdialects, 97% of subdialects can be explained by means of formula (11). First, let us use the demonstratives of Standard Chinese (Mandarin) to test the principle.

According to the sonority scale, f _s ([n]) ≥ f _s ([tʂ]) and f _s ([ɑ]) ≥ f _s ([ɤ]), and both demonstratives share the falling tone. As mentioned previously, this pair of demonstratives in Standard Chinese emerged in late Medieval Chinese. In the literature, the source of the distal demonstrative nà remains unknown, though there are some speculations about its origin. For instance, Ohta (1987: 118) and Wang (1989: 65) speculate that nà might have originated from the two demonstratives of Archaic Chinese, ruò or ěr. This view is problematic for three reasons. First, these two demonstratives ceased to be used around the first century, but nà did not arise until the eighth century. When nà emerged, the dominating distal demonstrative was bǐ, whose corresponding proximal demonstrative was cǐ. Second, their phonological forms were remarkably different, as shown in shown in Table 9.^[19]

Clearly, it is unlikely that the two archaic demonstratives developed into nà according to the evolution of the phonological system of Chinese.

Third, and more importantly, the functions of the demonstratives differ remarkably. For instance, ruò and ér always behaved like adjectival demonstratives; in other words, they could never be used without a nominal head. By contrast, the syntactic use of nà is entirely parallel to that of its corresponding proximal zhè. For instance, both demonstratives are often combined with a classifier to substitute an entire NP, omitting the noun heads. Additionally, both zhè and nà can be used as genitive markers, preceded by a possessor, as illustrated in (13) and (14).

Clearly, the distal nà could not have developed out of any other distal demonstratives in Archaic Chinese. The present analysis suggests that this distal demonstrative might be derived from the proximal demonstrative zhè via a phonological rule. There may be other possibilities of the emergence of the distal demonstratives, however, so I would leave this issue open for future research.

The phonological forms of demonstratives in Standard Chinese are based on Pekingese.^[20] Indeed, there are many phonological variants of the distal demonstratives in the Northern dialects that are used by approximately 70% of the population. The proximal demonstratives that originated from the classifier zhī are realized as [tʂɤ], [tsɤ], [tʂi], [tsɪ], [tʂei], [ʦei], [ʦit], and [tɕiaʔ] in different subdialects. For the sake of simplicity, Table 10 shows only the distal demonstratives whose corresponding proximal demonstratives are [tʂɤ].

There are 21 different phonological forms of the distal demonstratives that correspond to the proximal demonstrative [tʂɤ]. Even within a given subdialect, there are typically two or more phonological forms for distal demonstratives. These facts show that it is likely that these distal demonstratives did not originate from any independent lexical sources. Of the 205 subdialects summarized in Table 10, only five seem to be exceptions to the phonological principle (the onset consonants are [k] and zero). In other words, with respect to the emergence of their distal demonstratives, the phonological rule can account for 97% of the subdialects. In phonological theory, [k] is regarded as the least sonorous plosive and has a lower degree of sonority than the affricative [tʂ]. There are possible explanations for this. The phonological properties of the distal and proximal demonstratives suggest that this group of subdialects might have developed the classifier gè into their distal demonstratives, and the classifier zhī into their proximal demonstratives. Indeed, some subdialects obtained their distal demonstratives from the classifier gè, which in turn caused a reversal of the derivation process between the two demonstratives. The Huanning subdialect chose the low vowel [a] with the highest sonority as its distal demonstrative, which shows that it is not a real exception because there is still a remarkable increase in the sonority of the nucleus.

4.4 The sonority asymmetry between the segments of proximal and distal demonstratives

According to the phonological principle in (11), the degrees of sonority of the onsets and nuclei of distal demonstratives must be either higher than or equal to the corresponding segments of proximal demonstratives.^[21] However, this requirement varies from one segment to another, as stated in the following.

1. For the segments at the onset, the sonority of distal demonstratives is usually higher than but rarely equal to that of proximal demonstratives.

2. For the segments at the nucleus or coda, it is equally possible that the sonority of distal demonstratives is either higher than or equal to that of proximal demonstratives. This pattern differs across dialects.

The above asymmetry is manifested by the SDP, which says that a syllable tends to maximize the sonority slope from onset to nucleus, and to minimize the slope from nucleus to coda (Clements 1990: 291), as mentioned above. This principle can be revised to maximize the sonority slope from the onset of the proximal demonstratives to that of the distal demonstratives, and to minimize this value from the nucleus/coda of proximal demonstratives to those of distal demonstratives.

Within the 320 subdialects of the Northern dialect,^[22] 311 subdialects (97%) have increased the sonority values from the onsets of proximal demonstratives to those of distal demonstratives. Since the proximal demonstratives mostly originated from the classifier zhī, the onsets of these subdialects are [tʂ] or [ts]. The onsets of the distal demonstratives are mostly [n] and sometimes [m], [l] or [w], see Table 11.

For the nine subdialects exhibiting equal degrees of sonority in the two onsets, the consonants of the onsets of their proximal demonstratives are either nasals or liquids, which have the highest sonority values among all consonants, cf. Table 12. Thus, it is impossible to increase the sonority from the first onset to the second.

The Baoji subdialect deserves special attention. It is the only subdialect in which the onset of the proximal demonstratives is [k]. This indicates that it originated from the classifier gè, which is typically found in the Southern dialects, making this the only exception in the Northern dialect.

Within this group of 320 subdialects, 59 exhibit a sonority increase from the first nucleus to the second, as shown in Table 13. The vowels of the first nucleus are either [i] or its variant [ɪ].

In the remaining 261 subdialects, the distal demonstratives always have an alternative form with the same vowel as that of the corresponding proximal demonstratives. This means that the sonority scales of the two nuclei can be equal, cf. Table 14. Standard Chinese has two pairs of phonological forms for its proximal and distal demonstratives: [tʂɤ]/[na] and [tʂei]/[nei]. The first pair shows a sonority increase between the two nuclei, but the second pair exhibits equal sonority values in the two nuclei.

As mentioned in Section 3, in the Min dialect, with the rhymes (i.e., nucleus and coda) remaining the same, the shift between [ts] and [h] refers to the proximal and distal demonstratives, respectively. Similarly, in the Jin dialect, some subdialects can have as many as six phonological forms for each of the proximal and distal demonstratives, and the rhymes remain unchanged; different consonants in the onset thus identify different demonstratives, cf. Table 15.

In short, in distinguishing proximal from distal demonstratives, the sonority scales are maximized from the first onset to the second onset and minimized from the first nucleus to the second nucleus.

4.5 The reversal of the derivational process between proximal and distal demonstratives

Approximately 30 of the 930 subdialects documented in Cao (2008) (accounting for 3% of the total) first developed the classifiers into distal demonstratives and then derived the corresponding proximal demonstratives by means of the same phonological rule. In other words, the derivational process proceeded in reverse, from distal to proximal demonstratives, showing some very interesting properties. The reverse derivation is mainly found in Cantonese (the Yue dialect), cf. Table 16. A successful explanation for this phenomenon will lend further support to the hypothesis that there is a phonological rule guiding the derivation between distal and proximal demonstratives.

First, let us focus on the consonants at the onsets. The classifier gè has the onset consonant [k], which has the lowest degree of sonority among all sounds. Once it is realized as the distal demonstrative in a particular subdialect, there are only two strategies left to satisfy the formula f _s (O_d) ≥ f _s (O_p): (a) Find another plosive consonant for the onset of the proximal demonstrative, e.g., [t] in Liannan and Fogang; in this case, the formula becomes f _s (O_d) = f _s (O_p); (b) evacuate the onset of the proximal demonstratives, as in Conghua and Longmen; in this case, the formula becomes f _s (O_d) > f _s (O_p).

However, quite a few subdialects, such as those spoken in Hong Kong, Zhongshan and Boluo, violate the formula by reversing the sonority increase from the onset of the distal demonstratives to that of the proximal demonstratives. An appropriate explanation is that these subdialects have different rankings of the rules: the optimal rule is that there must be a sonority increase between the two onsets of proximal and distal demonstratives, regardless of the direction. As a result, nasals and liquids are used as the onsets of proximal demonstratives in Hong Kong, Zhongshan and Boluo.

In the Yue dialect, the phonological form of the classifier gè is [kɔ], with a central vowel as its nucleus. According to the sonority hierarchy, [ɔ] is more sonorous than the high vowels [i] and [ǝ]. Thus, to satisfy the formula f _s (N_d) ≥ f _s (N_p), it is possible to increase the sonority value of the nucleus of a proximal demonstrative to that of a distal demonstrative. All vowels of the proximal demonstratives are either [i] or [ǝ], as shown in Table 16, which satisfies the phonological principle.

The generalization made above can perfectly explain the group of subdialects illustrated in Table 17. The demonstrative gè is realized as both proximal and distal, differentiated by the sonority of the vowels. The nuclei of the proximal demonstratives are either [i] or [ǝ], which are the least sonorous vowels.

Furthermore, in the Xiang family, there are three ways to derive proximal demonstratives where the classifier gè is first realized as a distal demonstrative instead, as shown in Table 18.

First, like the Hong Kong subdialect in the Yue family, the Pingjiang subdialect shows that there must be a sonority increase between the two onsets of proximal and distal demonstratives, regardless of the direction. As a result, the liquid [l] occupies the onset position of the proximal demonstrative. Second, like the Liannan subdialect in the Yue family, the Yizhang subdialect finds another plosive consonant [t] for the onset of its proximal demonstrative; in this case, the formula becomes f _s (O_d) = f _s (O_p). Third, like the Conghua subdialect in the Yue family, the Miluo and Yanling subdialects evacuate the onset of proximal demonstratives; in this case, the formula becomes f _s (O_d) > f _s (O_p).

In summary, the phonological principle works in reverse when the classifier gè is first realized as a distal demonstrative in those subdialects. Once again, this phenomenon reveals that the derivation between proximal and distal demonstratives is consistently governed by the phonological rule.

4.6 The glottal sounds at the onset of the distal demonstrative

The crosslinguistic distribution of glottal sounds lends further support to the phonological principle proposed in this article, guiding the derivation from proximal to distal demonstratives in Chinese. In the literature, there are two opposite views on the sonority value of glottals. One view is that glottals are highly sonorous (Chomsky and Halle 1968: 301; Levin 1985; Pike 1954). By contrast, other researchers have argued that glottals have the lowest degree of sonority, basically identical to that of obstruents (Heffner 1950; Lass 1976; Lombardi 1999: 3.2; Zec 1988). Clements (1990: 322) claims that glottals behave arbitrarily in terms of their position on the sonority scale, effectively having no sonority value.

Empirical evidence suggests that in sonority distance restrictions, glottals usually behave like highly sonorous elements. For example, Gujarati allows only glides, liquids, and [h] as the second members of onset clusters: [kjal] ‘opinion’, [krupa] ‘kindness’, [kleʃ] ‘fatigue’, and [khǝrǝc] ‘cost’ (Cardona 1965: 31). In contrast, it is rare for [h] to have the same distribution as other fricatives, as is evident, for example, in a comparison of the English words [slɪt] ‘slit’, [flɪt] ‘flit’, and *[hlɪt]. Likewise, in Chinese dialects, the two glottals [ʔ] and [h] behave like nasals and liquids, which are highly sonorous. In the Jin dialect, for instance, only [ʔ], [n], [m], and [ŋ] can occur in the coda position. In the derivation process from proximal to distal demonstratives, the glottal [h] behaves like a nasal or liquid in the Min dialect, as discussed below.

The consonants in the onset position of distal demonstratives are mainly nasals ([n], [m], and [ŋ]) and the liquid [l]. According to Cao (2008), however, 111 subdialects use the uvular [x] as the onset for their distal demonstratives. In this work, no onsets are transcribed as glottal [h]. Both [x] and [h] are fricatives. Distal demonstratives of this type are mainly found in the subdialects of the Min family. However, according to Beida (1995: 557) and Yuan (2001: 97), in the phonological system of the Xiamen subdialect (i.e. the representative area of the Min dialect), only [h] exists, and [x] is absent. Among the 20 representative subdialects in Beida (1995), there are four subdialects with [h] as the onset of the distal demonstrative and only one with [x] as the onset.

As Table 19 shows, [h] is used as the onset of the distal demonstrative more frequently than [x]. In addition, the sound [x] of demonstratives is recorded as [h] in Yuan (2001: 235–283). I therefore believe that the transcriptions of Cao (2008) are inaccurate and that most instances of [x] in the onset are actually instances of the glottal fricative [h]. Consequently, a majority of the 111 subdialects should have [h] onsets in their distal demonstratives. Accordingly, in the Min dialect, [h] has a sonority value comparable to that of a nasal or liquid, which implies an increase in sonority from the onset of the proximal demonstratives to that of the distal demonstratives. This can explain why a shift between [ts] and [h] is used to distinguish proximal from distal demonstratives in the Min dialect, which Yuan (2001) regards as a kind of inflection.

4.7 The different specifications between onset and nucleus

A remarkable feature of demonstratives is that their phonological forms are variable in almost all Chinese dialects. In Cao (2008), only one of the 930 subdialects has a single form for its proximal and distal demonstratives. The other subdialects have two or more phonological forms for each of their demonstratives. Such variability is not found in any other grammatical morphemes. For instance, the associative marker de, the most frequently used morpheme, has only a single phonological form within specific subdialects, e.g., [tǝ] in Standard Chinese, [kɤʔ] in Suzhou, and [kɔ] in Yangjiang (see Beida 1995 for details).

In Standard Chinese, the proximal demonstrative zhè has two phonological forms, [tʂɤ⁵¹] and [tʂei⁵¹],^[23] and the distal demonstrative also has two corresponding forms, [nɤ⁵¹] and [nei⁵¹]. The situation in dialects is much more complex. Some subdialects have as many as six different forms for each of their proximal and distal demonstratives. This means that a single subdialect can have more than 10 distinctive forms for its demonstratives, as shown in Table 20 (Qiao 2000: 127).

Clearly, the consonants in the onset positions are fixed for proximal and distal demonstratives, but the rhymes are variable. In this case, there is a rigorous correspondence between the rhymes of proximal and distal demonstratives, including medial demonstratives if there are any.

In many other subdialects, the consonants at the onset position are the same, and in these cases, the vowels in the nucleus must be distinctive, distinguishing proximal from distal demonstratives; typically, the sonority increases from the proximal nucleus to the distal nucleus, see Table 21.

If the onsets remain the same, the vowels must be different. In either case, the consonants at the onset position are specified.

4.8 Cliticization of demonstratives

In some dialects, only one phoneme is used to express demonstratives. These phonemes, typically [a], [i], [u], [n], and [ŋ], all have a high degree of sonority, carrying the potential to be syllabic, as shown in Table 22.

Since the focus of this article is on the phonological derivation between proximal and distal demonstratives, I have not discussed their syntax. The derivation process may be obscured by phonological reduction from cliticization or fusion with other adjacent elements, which is related to syntax. In this section, I will address the phonological reduction of demonstratives and several special cases for which the derivation processes cannot be identified at present.

In many dialects, demonstratives cannot immediately precede a noun and must be connected by a classifier (for details, see Huang 1996). To replace an entire NP, a demonstrative has to combine with a classifier. In addition, demonstratives are often combined with locative words to indicate places or adverbial suffixes to modify a VP. These factors may cause demonstratives to cliticize to, or fuse with, the following elements. Their high frequency of occurrence with classifiers may trigger the fusion of demonstratives with the following classifiers or locatives. Ariel (1990) proposed an Accessibility Marking Scale, a cline showing the tendency toward phonological reduction:

It is cross-linguistically common that pronominal elements cliticize to an element in their environment. Likewise, demonstratives are often phonologically reduced to a single phoneme. For instance, Lezgian has two demonstrative roots, i ‘proximal’ and a ‘distal’, which must be combined with other elements to be used (Haspelmath 1993: 259).

Since they are frequently combined with other elements, demonstratives are subject to cliticization, which in turn causes their phonological reduction. This reduction may occur in the onset or nucleus. In the Liuyang subdialect (Li 2001), for instance, the distal demonstrative n cannot be used alone and must co-occur with the classifier ke, as illustrated in (16):

When used as an adverb to modify APs or VPs, the distal demonstratives in Liuyang have two forms – koŋ and n – indicating that the reduced form n should be phonologically reduced from koŋ.

Likewise, in the Jin dialect, the distal demonstrative u is a reduced form and is usually combined with other elements. For instance, u ²¹³ .ɛ in the Yangqu subdialect is a fusion between a distal demonstrative and a classifier. The full form in the Yuanping subdialect, its neighboring subdialect, is uai ⁵⁴ (Qiao 2000: 126–127).

Of the 24 subdialects of the Jin family (Qiao 2000), six subdialects use [u] or [ɔ] to encode distal demonstratives only, including Pinglu, Wanrong, Linyi, Xiangfen, Huozhou, and Pinyao. Within this dialectal family, many subdialects use u alone as the onset of their distal demonstratives. There is robust evidence that this u is a reduced form, as shown in Table 23.

Here, let us focus on the single phoneme [u] in the Yuanping subdialect in Table 23, which constitutes a distal demonstrative. There is a neat correspondence between each form of the proximal and distal demonstratives in the two subdialects. With the same rhymes, the shift between [tʂ] and [u] at the onset distinguishes proximal from distal demonstratives. Apparently, the distal [u] in the Yuanping subdialect is actually a reduction of [ui].

Of the 93 subdialects of the Xiang family, 13 use one of the following phonemes to encode their proximal demonstratives.

These single-phoneme demonstratives are also the result of phonological reduction due to cliticization. Within the same dialectal family, these vowels often occur at the nucleus of demonstratives, as shown in Table 24.

Among the 930 subdialects of Cao (2008), there are three exceptions to the phonological rule. All of them have [pi⁴⁴] as their distal demonstratives, where the onset is [p], the consonants with the lowest degree of sonority, and the nucleus is [i], the vowel with the lowest sonority value. Their corresponding sounds in proximal syllables are more sonorous, cf. Table 25.

These distal demonstratives have a similar phonological form as bǐ [pi²¹³], which emerged in Archaic Chinese and was replaced around the tenth century. It is likely that these three subdialects still retain the relics of the old distal demonstratives.

It thus seems that the phonological rule posited in this article worked even in Archaic Chinese. The old pair of demonstratives cǐ [tsi²¹³] and bǐ [pi²¹³] has been reconstructed as [tshǐe²] and [pǐai²] (Li and Zhou 1999: 56), with an increase in sonority from the first nucleus [i] to the second [a].