The status of Chinese adjective-noun combinations ([A N]) has been debated for decades. 1 Many linguists have argued that Chinese [A N] are compounds, that is, words (Zhu 1956; Zhu 1980; Fan 1958; Sproat and Shih 1988; Sproat and Shih 1991; Lu 1990; Dai 1992; McCawley 1992; Duanmu 1998; Duanmu 2000; Feng 2001; Xu 2006; among others). By contrast, Li and Thompson (1981) claim that some semantically transparent Chinese [A N] such as hong hua ‘red flower’ are phrases without explaining why they cannot be compounds. Paul (2005) argues that some semantically transparent Chinese [A N] can be phrases. Schäfer (2009) argues that there is no evidence for deciding the status of the semantically transparent Chinese [A N].
My article only deals with [A N] whose status, either a word or a phrase, is controversial. I do not discuss exocentric compounds or compounds without an expressed semantic head. These pass all tests for wordhood and hence are uncontroversially words (Chao 1968; Duanmu 2000). For example, xiao-qi (small-air) ‘stingy’ is an exocentric compound. 2
Endocentric [A N] form a continuum according to their semantic transparency. Those at one end of the continuum are semantically opaque and the adjective is bleached semantically. Examples include hei-ban (black-board) ‘blackboard’, da-yi (big-coat) ‘overcoat’, and hong-hua (red-flower) ‘safflower’. 3 Semantically opaque endocentric [A N] must be listed in the lexicon (Feng 2001). At the other end, the meaning of the compound is fully determined by those of its constituents and is semantically transparent. Examples include xiao zhuozi ‘small table’ and huang chenshan ‘yellow shirt’. Semantically transparent endocentric [A N] can be produced online.
There is no clear-cut division at any point along this continuum. For example, congming ren (intelligent person) ‘a person who is intelligent and understands the situation’ has a lexicalized sense ‘to understand the situation’ in addition to the meaning that is deduced from its constituent parts. Hence, congming ren must be listed in the lexicon in this sense. It is less transparent than those endocentric compounds with no lexicalized sense but it is less opaque than those endocentric compounds whose nonhead is bleached semantically. It is widely accepted that the semantically opaque [A N] in which the nonhead is bleached semantically are words. It is only the status of the semantically (more) transparent [A N] that is controversial.
I defend the wordhood of semantically transparent [A N] with new supporting evidence. 4 Sections 2–8 present arguments for the wordhood of Chinese [A N]. The monomorphemic constraint on the adjective of [A N] compounds, and the tests of modification by degree adverbs, XP substitution, and conjunction reduction support the wordhood of Chinese [A N]. Additionally, Chinese [A N] have a potential naming function and are subject to selectional restrictions. Both are evidence for their wordhood. Sections 9 and 10 argue that neither anaphoric accessibility nor the restrictions on the ordering of adjectives are reliable criteria for establishing the status of Chinese [A N]. Section 11 briefly summarizes the discussions in previous sections. Section 12 shows that the various properties of Chinese [A N] are accounted for under the framework of Construction Morphology (Booij 2010). Chinese [A N] are word-level expressions constructed in an autonomous word formation component of the grammar. I conclude in Section 13.
2 The monomorphemic constraint on Chinese [A N] compounds
It has been observed that a majority of Chinese [A N] contain monosyllabic adjectives (Sproat and Shih 1988; Duanmu 1998; Duanmu 2000). Shen (1997) conducted a corpus-based study of the distribution of Chinese adjectives as either predicates or prenominal modifiers. Shen’s corpus consists of around 60,000 words. Half of the corpus is based on published materials and the other half is based on recordings of natural speech. In Shen’s database, there are 121 [A N] in which the adjective does not contain any modifying element (e.g., hong ‘red’, hutu ‘muddle-headed’). Of the 121 instances, 95 (79 %) contain monosyllabic adjectives. The other 26 (21 %) contain disyllabic adjectives. See Table 1 (adapted from Shen 1997).
Additionally, there are 142 nominal expressions with prenominal adjectives that contain modifying elements such as reduplicants or rhetorical elements that express the degree of adjectives. 5 See (1), in which the presence of de is obligatory. 6
‘a queue that is long to a certain degree’
‘a waist as straight as a brush’
(Adapted from Shen [1997: 244])
In 138 of these 142 instances (97 %) with complex adjectival expressions, the adjectives do not directly precede nouns. Instead, we find the [A de N] construction, in which de intervenes between the adjective and the noun. See Table 2 (adapted from Shen [1997: 244]).
Two points in this data are noteworthy. First, it is surprising that a majority of [A N] contain monosyllabic adjectives, given that disyllabic adjectives outnumber monosyllabic ones by more than two to one in Chinese. According to Putonghua sanqian changyongci biao ‘3,000 commonly used words in Standard Chinese’ (1959), there are 450 common adjectives in Chinese. 7 Of them, 140 (31 %) are monosyllabic, 300 (67 %) are disyllabic, and 10 (2 %) are trisyllabic. Few of the 450 adjectives contain modifying elements. See Table 3.
Zhang (2000) calculates the percentage of Chinese adjectives in terms of their length. According to Zhang’s statistics, there are 1,588 adjectives in Chinese. Of them, 145 (9 %) are monosyllabic, 1,329 (84 %) are disyllabic, and 114 (7 %) are trisyllabic. 8 See Table 4.
The proportions of both disyllabic and trisyllabic adjectives in Zhang (2000) are by far higher than those in Putonghua sanqian changyongci biao ‘3,000 commonly used words in Standard Chinese’ (1959). The reason is that Zhang’s database includes many complex adjectives that contain modifying elements such as reduplicants or rhetorical elements that express the degree of adjectives. See (1).
The second point in need of an explanation is that few adjectives with modifying elements directly precede nouns in Chinese, as shown in Table 2. The above two issues can be accounted for by the monomorphemic constraint:
The monomorphemic constraint: The adjective in [A N]N is monomorphemic.
In this article, I follow the traditional definition of morpheme as “the smallest individually meaningful element in the utterances of a language” (Hockett 1958: 123). The monomorphemic constraint rules out compounds in which a noun is directly preceded by complex adjectives that contain modifying elements. By contrast, this constraint does not apply to the [A de N] construction, which is widely considered to be phrasal in nature. See (1). Reduplicated forms like chang-chang (long-long) ‘long to a certain degree’ are morphologically complex, not only because they express more than the meaning of their base forms (Li and Thompson 1981), but also because reduplication in general introduces morphologically complex forms. According to Inkelas and Zoll (2005) and Booij (2010), reduplication is a morphological process and each reduplication pattern carries a construction-specific sense, which varies across languages.
Similar but less general constraints have been proposed by others. Shen (1999) remarks that zhuangtai xingrongci ‘complex adjectives’ such as reduplicated adjectives (1a) and modifier-head compound adjectives (1b) are usually derived from xingzhi xingrongci ‘simplex adjectives’ and do not directly precede nouns. 9 I define the monomorphemic constraint in a more general way so that it forbids any morphologically complex adjectives to directly modify nouns.
This constraint also explains why a majority of Chinese [A N] contain monosyllabic adjectives, although disyllabic adjectives, excluding both reduplicated and modifier-head compound adjectives, outnumber monosyllabic ones by more than two to one in Chinese. Li and Thompson (1981) observe that in Chinese a syllable generally corresponds to a morpheme and vice versa. They note that disyllabic expressions are generally bimorphemic in Chinese. Hence, disyllabic adjectives generally violate the monomorphemic constraint in (2). As a consequence, a majority of Chinese [A N] contain monosyllabic adjectives.
‘a person who is intelligent and understands the situation’
(Shen 1997: 244)
(Shen 1997: 244)
(Shen 1997: 244)
(Duanmu 2000: 171)
Li and Thompson (1981) note that polysyllabic morphemes do exist in Chinese; it is sometimes difficult to determine the morpheme status of a syllable (see also Sproat and Shih 1996). For example, polysyllabic expressions such as putao ‘grape’, hudie ‘butterfly’, boli ‘glass’, and meigui ‘rose’ are clearly monomorphemic because they cannot be decomposed into smaller meaningful units. It is very hard to argue that words like feng.liu (wind.flow) ‘amorous’ are bimorphemic because otherwise the whole meaning of this expression would be unrelated to the meaning of either of its constituent parts and neither part is syntactically active. They claim that, except for clear cases of polysyllabic morphemes, most polysyllabic expressions are polymorphemic in Chinese.
I agree with Li and Thompson (1981) that the disyllabic adjectives in (3a)–(3c), (3 g) are disyllabic morphemes, so that they can directly combine with nouns without violating the monomorphemic constraint. For example, it is not clear what piao means in (3a). The syllable liang corresponds to a Chinese character meaning ‘bright’, but it is still not clear how this meaning is related to ‘beautiful’. It is very hard to decompose the adjectives in (3b)–(3c) into smaller meaningful units in modern Chinese. Additionally, the syllables liang, ming and tu are not pronounced in full and carry a neutral tone, which makes it even more difficult to recognize their meanings.
Each adjective in (3d)–(3f) formally consists of two morphemes, so they violate the monomorphemic constraint. These adjectives are what Duanmu (2007) calls “pseudo-compounds”: each of them is semantically opaque and equals a monomorpheme from a semantic perspective. Semantic opacity correlates with whether all its constituent parts add to the meaning of an adjective. For example, zhong-da (heavy-big) expresses the meaning of one of its constituents. The expression zhong-da de shi (heavy-big DE event) is semantically very close to da shi ‘big event’. Hence, semantic opacity predicts that a complex disyllabic adjective will occur in [A N] compounds to the extent that the complex adjective violates compositionality. The more opaque a complex adjective is, the more likely it is to be found in [A N] compounds. 11 This prediction is borne out. If we reconsider the 26 [A N] with disyllabic adjectives that contain no modifying element in Shen’s (1997) database (Table 1), almost all of these adjectives are either disyllabic morphemes or pseudo-compound adjectives, which are semantically opaque.
The monomorphemic constraint also explains why bimorphemic words like fu-ze (take-responsibility), which are ambiguously adjectives or verbs, lose their adjectival readings when they directly modify nouns (Xu 2006). According to Packard (2000), the word fu-ze has an internal V-O structure. In Chinese, fu-ze can have either a verb meaning ‘be in charge (of something)’ or an adjective meaning ‘responsible, having a sense of responsibility, acting responsibly’. When fu-ze is an adjective, it can be modified by degree words such as feichang ‘very’ (4). 12
‘Zhangsan is very responsible.’
When fu-ze acts as a verb, it means ‘be in charge of (something)’ instead of ‘act responsibly’ (5).
|Zhangsan fuze||yi ge xiangmu.|
|Zhangsan be in charge of||one cl project|
‘Zhangsan is in charge of a project.’ not ‘Zhangsan acts responsibly for a project.’
Additionally, verbal fu-ze cannot be modified by degree words such as feichang ‘very much’ (6).
|*Zhangsan feichang||fuze||yi||ge xiangmu.|
|Zhangsan very much||be in charge of||one||cl project|
When fu-ze occurs in the de-construction, it can be either an adjective or a verb, so that it can express both readings (7).
|fuze||de ren||/ laoshi||/ yisheng||/ guanyuan|
|responsible /||be in charge||DE person||/ teacher||/ doctor||/ official|
‘person/teacher/doctor/official with a sense of responsibility’ or ‘person/teacher/doctor/official in charge’
By contrast, when fu-ze directly modifies nouns, it expresses the verb reading ‘be in charge’ and not the adjective reading ‘responsible, having a sense of responsibility, acting responsibly’. See (8). Hence, fu-ze acts as a verb instead of an adjective in the construction [fu-ze N] ‘N in charge’.
|fuze||ren||/ laoshi||/ yisheng||/ guanyuan|
|be in charge||person||/ teacher||/ doctor||/ official|
‘person/teacher/doctor/official in charge’
not ‘person/teacher/doctor/official with a sense of responsibility’
The bimorphemic adjective fu-ze cannot directly precede nouns, because then it would violate the monomorphemic constraint, which requires the adjective in [A N]N to be monomorphemic.
3 The monomorphemic constraint in other languages
The monomorphemic constraint is found in other languages, notably Dutch and German. According to Booij (2010) and Hüning (2010), the adjective in both Dutch and German [A N] compounds needs to be simplex. See (9)–(10), 13 in which the monomorphemic adjectives can be either monosyllabic or disyllabic, as those in Chinese. 14
Dutch [A N] compound
snel-trein ‘fast train, express train’ (Hüning 2010)
zuur-kool ‘sour cabbage’
edel-metaal ‘noble-metal, precious metal’
bitter-koekje ‘bitter cookie’
German [A N] compound
Schnell-zug ‘fast train, express train’ (Hüning 2010)
Bitter-mandel ‘bitter almond’
Sauer-teig ‘sour dough’
Fertig-gericht (ready food) ‘convenience food’
Locker-schnee (loose-snow) ‘a special type of snow’
By contrast, according to Booij (2010) and Hüning (2010), the monomorphemic constraint does not apply to [A N] phrases in Dutch and German, in which complex adjectives are possible. See (11), which shows that complex adjectives are allowed in [A N] phrases, but not in [A N] compounds in Dutch and German. Agreement takes place in [A N] phrases but not in [A N] compounds in Dutch and German (Booij 2010).
die wissenschaftliche Hilfskraft ‘graduate assistant’
het wetenschappelijk(e) onderwijs ‘scientific training/education’
The monomorphemic constraint distinguishes words from phrases crosslinguistically. It does not apply to the phrasal [A de N] construction in Chinese and cannot be ascribed to phrasal syntax. Hence, Chinese [A N] are words, under which it is normal to observe such types of non-syntactic constraints. See McCawley (1992) for discussion of other non-syntactic constraints on Chinese word formation.
4 Alternative accounts of the majority of Chinese [A N] with monosyllabic adjectives
In this section, I discuss several alternative accounts of the majority of Chinese [A N] with monosyllabic adjectives. These approaches include a prosodic one (Feng 1998; Feng 2001) and two semantic ones. One of the two semantic approaches has been proposed in Paul (2005; 2006) and Schäfer (2009); the other in Huang (2006).
4.1 A prosodic account of Chinese [A N]
Feng (1998; 2001) claims that Chinese [A N] pattern like verb phrases prosodically. For example, both Chinese [A N] and verb phrases allow for the prosodic structure of a monosyllabic constituent followed by a disyllabic one (1+2 hereafter) (e.g., xiao zhuozi ‘small table’, chi shuiguo ‘eat fruits’). Hence, understanding the prosodic pattern of verb phrases may help to explain why a majority of Chinese [A N] contain monosyllabic adjectives.
Closer examination, however, shows that a prosodic explanation must be rejected. Although 2+1 is disfavored by verb phrases, 2+2 is very productive in making verb phrases (Duanmu 2012), showing that verb phrases are prosodically unrelated to [A N] compounds. In fact, as long as a nominal complement is not shorter than its verb head, the resulting verb phrase will be prosodically acceptable, because the head tends not to bear greater phonological prominence in a head-complement relation in many languages (see e.g., Cinque 1993) and phonological prominence is realized by word-length in terms of syllable numbers in Chinese (Duanmu 1990; Duanmu 2000; Duanmu 2012; Lu and Duanmu 1991; Lu and Duanmu 2002). See (12), in which disyllabic verbs occur. The bimorphemic verb in (12a) is a pseudo-compound, in that it expresses the meaning of one of its constituents. According to Packard (2000), the verbs in both (12b)–(12c) are morphologically complex and bear an internal V-O structure.
‘be in charge of projects’
‘invest 20,000 dollars’
If Chinese [A N] patterned like verb phrases prosodically, then (just as with verbs) morphologically complex adjectives such as reduplicated adjectives and modifier-head compound adjectives should directly precede nouns, as long as the second element is not shorter than the first one in [A N]. Additionally, fu-ze (take-responsibility) should both act as an adjective and directly precede a noun, as long as the noun is not shorter than fu-ze in terms of syllable numbers. These predictions fail to be borne out but are explained by the monomorphemic constraint.
4.2 Paul’s (2005) semantic account of Chinese [A N]
I now turn to semantic accounts of Chinese [A N]. Based on Chen (1955), Paul (2005) and Schäfer (2009) argue that Chinese [A N] must always have a naming function and the adjective must express a defining property of the head noun, so that any Chinese [A N] “has to result in a natural, plausible classification” (Paul 2005: 772). By contrast, an [A de N] does not bear a naming function and its adjective does not express a defining property. Paul’s evidence for the naming function of Chinese [A N] comes from Fu (1987), who remarks that in a few diagnostic contexts such as the identification context in (13), only the de-less modification structure is allowed. According to Paul (2005) and Schäfer (2009), these contexts enforce naming, which results in a natural and plausible classification of entities.
This sentence becomes worse if the linker shi ‘be’ is replaced with jiao ‘be called’, further showing that [A de N] cannot name entities, but [A N] can. See (14).
‘This is called “black scarf”.’
Paul (2005) and Schäfer (2009) argue that the semantic constraint that requires a natural and plausible classification rules out [A N] with complex adjectives such as reduplicated adjectives and modifier-head compound adjectives, because they reflect speakers’ subjective evaluation of the property expressed by the adjective rather than solely referring to that property (see Zhu 1980; Tang 1988). Schäfer (2009) argues that because of this general semantic constraint, the fact that complex adjectives do not occur in [A N] does not support either the compound or the phrasal analysis of Chinese [A N].
There are several problems with this semantic constraint. Above all, since Paul (2005) and Schäfer (2009) do not clearly define “natural and plausible classification”, the claim that Chinese [A N] must result in a natural and plausible classification is empirically unfalsifiable and the argumentation is circular. In fact, Paul (2005) and Schäfer (2009) notice this problem and remark that it is often unclear what counts as a natural and plausible classification (see also Bolinger 1967).
Second, the notion “subjective evaluation” is not well defined. It is not clear why the simplex adjective in [A N] such as xiao zhuozi ‘small table’ could not express speakers’ subjective evaluation of a property. Hence, the non-occurrence of [A N] with complex adjectives cannot be ascribed to speakers’ subjective evaluation of a property.
Third and most tellingly, no semantic constraint could explain why a majority of Chinese [A N] contain monosyllabic adjectives, although disyllabic adjectives, even excluding complex adjectives such as reduplicated adjectives and modifier-head compound adjectives, outnumber monosyllabic ones substantially in Chinese.
4.3 Naming and descriptive functions of Chinese [A N]
Contrary to Paul (2005), semantically transparent Chinese [A N] can describe entities. Above all, the semantic difference between [A de N] and [A N] with the same adjective is very subtle and most contexts allow both constructions (Paul 2005). I used pairs like huang chenshan vs. huang de chenshan ‘yellow shirt’, hong wazi vs. hong de wazi ‘red sock’ to test the semantic difference between the two constructions in contexts other than the one in (14). My consultants, five native speakers of Chinese, told me that they could not tell the difference between them semantically. This shows that semantically transparent Chinese [A N] can not only name but also describe entities so that in most contexts they are interchangeable with [A de N]. It is widely accepted that [A de N] only describe entities.
Second, the claim that Chinese [A N] can either name or describe entities captures the range of linguists’ intuitions about the naming or descriptive function of Chinese [A N]. Paul (2005) and Schäfer (2009) claim that Chinese [A N] always bear a naming function, whereas Sproat and Shih (1993; 1996) claim that semantically transparent Chinese [A N] describe entities if their nominal head is free. According to Sproat and Shih, when a bound nominal root acts as a head, it only denotes a kind. For example, huang-shi (yellow-lion) only denotes a kind of lion and hong-yi (red-ant) only denotes a kind of ant, because both shi ‘lion’ and yi ‘ant’ are bound roots. By contrast, huang shizi (yellow lion) means ‘lion that happens to be yellow’ and hong mayi (red ant) means ‘ant that happens to be red’. In other words, huang ‘yellow’ and hong ‘red’ simply describe the properties of shizi ‘lion’ and mayi ‘ant’, both of which are free.
Third, if Chinese [A N] can either name or describe entities, we will get a plausible five-way contrast among [A de N], nonconventionalized Chinese [A N] with either a free or a bound nominal head, and conventionalized Chinese [A N] with either a free or a bound nominal head. 15
Consider Table 5. It is widely accepted that the naming function must always be associated with Chinese [A N] that are conventionalized and/or have a bound nominal head, all of which are bold-faced in the following table. Notice that compared to huang-shi (yellow-lion) ‘the type of lion with the property of being yellow’, mei-nü (beautiful-woman) ‘beautiful woman’ is a conventionalized expression and hence lexicalized, given that it is usually listed in dictionaries. See, for example, Xiandai Hanyu Cidian ‘Modern Chinese Dictionary’ (2000), in which mei-nü is listed. It is typical of words to name entities while it is typical of phrases to describe entities (Erben 2000; Bauer 2003). Conventionalization can make it more likely for [A N] to name entities, because even an [A N] phrase can bear a naming function if it is conventionalized and hence lexicalized (Koefoed 1993; Booij 2002; Booij 2010; Schäfer 2009; Hüning 2010). Accordingly, it is less likely for an [A N] to name an entity if it has not been conventionalized. Moreover, Chinese [A N] with a bound nominal head must always name entities (Sproat and Shih 1993; Sproat and Shih 1996). Accordingly, it is less likely for Chinese [A N] without a bound nominal head to name entities. When the Chinese language has a bound nominal head and a corresponding free form, we will get a contrast. That is, [A N] with the free form will not be used to name entities (Sproat and Shih 1993; Sproat and Shih 1996).
Assuming that there is an opposition between naming and description, nonconventionalized Chinese [A N] with a free nominal head can either name or describe entities. Their word status allows them to name entities. On the other hand, neither are they conventionalized nor do they contain a bound nominal head, which makes it possible for them to describe entities. 16 It is therefore predicted that a semantically transparent Chinese [A N] with a free nominal head will always be ambiguous between naming and description. Hence, in the contexts that enforce naming, such as the one in (14), semantically transparent Chinese [A N] with a free nominal head can occur while [A de N] cannot, because the latter can only describe entities. In other contexts, both constructions can occur because both can describe entities, so it is very hard for Chinese speakers to distinguish them semantically.
Fourth, when the repetition of de is disallowed phonologically, speakers will replace [A de N] with [A N]. 17 In the following sentence, huang chenshan (yellow shirt) ‘yellow shirt’ is preferred to huang de chenshan (yellow DE shirt) ‘yellow shirt’, because otherwise de would be repeated leading to ill-formedness. The first de after Lisi is a possessive marker, whose presence is obligatory.
|As long as||be||Lisi||DE||yellow||DE||shirt,||Zhangsan||all|
‘As long as it is a yellow shirt of Lisi, Zhangsan wants to have it.’
One might propose a rule of de-deletion, which involves no category change and derives [A N] from [A de N]. Such a rule predicts that [A de N] with complex adjectives such as reduplicated adjectives and modifier-head compound adjectives will also undergo de-deletion. By contrast, the substitution of [A N] for [A de N] predicts that [A N] with complex adjectives will be ruled out by the monomorphemic constraint on [A N] compounds. This prediction is indeed borne out. See (16), in which xue-bai de zhi (snow-white DE paper) ‘snow-white paper’ cannot be replaced by *xue-bai zhi (snow-white paper). 18
|As long as||be||Lisi||DE||snow-white||DE||paper||Zhangsan||all|
‘As long as it is Lisi’s snow-white paper, Zhangsan wants to have it.’
Finally, because Chinese [A N] can either describe or name entities while [A de N] only describe entities, Chinese [A N] should occur more frequently than [A de N] with the same adjective. In Shen’s (1997) corpora, there are 104 cases in which monosyllabic adjectives occur in either [A N] or [A de N]. Of them, 95 (91 %) are [A N] and 9 (9 %) are [A de N]. See Table 6 (adapted from Shen 1997).
4.4 Huang’s (2006) semantic type-matching constraint
A second semantic account of Chinese [A N] has been proposed in Huang (2006). Huang proposes a semantic type-matching constraint that requires a modifier to be of the same semantic type as its modifiee. Chinese complex adjectives such as reduplicated adjectives and modifier-head compound adjectives only occur in predicate position, so they are of the semantic type <e, t>. By contrast, a noun head is of type <e>. Hence, this semantic constraint forbids these complex adjectives to directly combine with nouns.
There are several problems with this constraint. First, as with all semantic constraints, it does not explain why a majority of Chinese [A N] contain monosyllabic adjectives, even though disyllabic adjectives greatly outnumber monosyllabic ones in Chinese. According to Huang, monosyllabic adjectives cannot occur in predicate position without functional elements such as hen, which is not stressed and hence can be bleached semantically (see also Li and Thompson 1981). See (17), in which hen is not stressed.
‘Zhangsan is tall.’
(Adapted from Huang [2006: 345])
When hen is stressed, it usually acts as a degree adverb meaning ‘very’. Chinese monosyllabic adjectives are of type <e> and the non-stressed semantically bleached functional element hen has “the function of a type-lifter” (Huang 2006: 353) and denotes “a function of type <e,<e, t≫” (Huang 2006: 353), i.e., hen converts an element of type <e> into one of type <e, t>. Since Chinese monosyllabic adjectives are of type <e>, the semantic type-matching constraint allows them to directly combine with nouns.
But disyllabic adjectives, excluding those containing a modifying element, can also be of type <e> because almost all of them can co-occur with the non-stressed hen. See (18a), in which hen is not stressed. The disyllabic adjective jiao-ao (arrogant-haughty) ‘arrogant’ can co-occur with the non-stressed hen, so it can be of type <e>. Hence, the semantic type-matching constraint predicts that jiao-ao can directly modify the noun xuesheng ‘student’. This prediction, however, fails to be borne out (18b).
‘This student is arrogant.’
‘an arrogant student’
Additionally, as noted above, the bimorphemic word fu-ze (take-responsibility) can act as either a verb meaning ‘be in charge (of something)’ or an adjective meaning ‘responsible, having a sense of responsibility, acting responsibly’. When it acts as an adjective, it can co-occur with the non-stressed hen, showing that it can be of type <e>. See (19). The semantic type-matching constraint predicts that fu-ze can both act as an adjective and directly precede nouns. This prediction, however, fails to be borne out. When fu-ze directly precedes nouns, it can only act as a verb.
‘This doctor is responsible (or has a sense of responsibility).’
Moreover, because Chinese complex adjectives such as reduplicated adjectives and modifier-head compound adjectives describe entities in the [A de N] construction (1) and the adjective in [A N] can be descriptive, it is questionable to refer to a semantic constraint to prevent these complex adjectives from occurring in [A N]. 19
In short, Huang’s semantic account of complex adjectives cannot provide a unified explanation of the failure of any multisyllabic adjective to occur in direct prenominal position.
5 Lexical gaps and selectional restrictions on Chinese [A N]
Lexical gaps often occur in Chinese [A N], but not in [A de N] (Zhu 1980; Duanmu 2000). These gaps arise because of various selectional restrictions that are typical of word formation, besides the monomorphemic constraint. This further suggests that Chinese [A N] are words while [A de N] are phrases.
See (20) and (21) for a contrast between acceptable and unacceptable Chinese [A N] with the same adjective.
(Duanmu 2000: 111)
(Duanmu 2000: 111)
(Duanmu 2000: 111)
(Duanmu 2000: 111)
(Li and Thompson 1981: 123)
‘tall adult man’
(Li and Thompson 1981: 123)
(McCawley [1992: 235], citing Jiang Zixin [personal communication])
(Duanmu 2000: 112)
(Duanmu 2000: 112)
(Duanmu 2000: 112)
(Li and Thompson 1981: 123)
‘good Shaoxing yellow-wine’
(McCawley [1992: 235], citing Jiang Zixin [personal communication])
The unacceptable examples in (21) become acceptable, once de is inserted. See (22).
‘good Shaoxing yellow-wine’
These gaps cannot be ascribed to phrasal syntax, given that they all disappear in the phrasal [A de N] construction. Rather, the gaps in (21) arise because of selectional restrictions. 20 Since there are hundreds of adjectives in Chinese, it is impossible to go through each Chinese adjective in terms of the restrictions on its combination with nouns in a single article. I will analyze the lexical gaps in (21) to illustrate restrictions on Chinese [A N]. For a way to handle these selectional restrictions within a Construction Morphology account, see my discussion in Section 12.
According to Xu (2006), when the adjective shen ‘deep’ directly modifies nouns, it only expresses physical depth. See (23).
It must occur in the [A de N] construction if it is to express abstract profoundness. See (24). The ungrammatical expression *shen shu (deep book), for example, is ruled out by the restriction that requires shen to express physical depth in [A N].
When the adjective duan ‘short’ directly modifies a concrete noun, the noun must express a small object described in terms of length. See (25).
When it modifies a noun that expresses a large object, it must occur in the [A de N] construction. See (26).
The nouns that can be directly modified by the adjective huaji ‘funny’ must be semantically related to ‘performance’. See (27). The expression *huaji ren ‘funny person’ is odd, because ren ‘person’ is not necessarily related to ‘performance’, given that a person may not perform in his lifetime. Hence, de must be inserted between huaji ‘funny’ and ren ‘person’.
‘actor who plays farces’
The nouns that can be directly modified by the adjective congming ‘intelligent’ usually express a human referent, e.g., congming haizi ‘intelligent child’, congming ren ‘the type of person who is intelligent and understands the situation’. Hence, *congming dongwu ‘intelligent animal’ is ruled out by this selectional constraint and de must be inserted between the adjective and noun.
According to Li and Thompson (1981), literary style also plays a role in determining whether or not an adjective may directly modify a noun. “If the head noun is more literary and has a closer affinity to written Classical Chinese, then a modifying adjective is more likely to be able to occur without [de].” (Li and Thompson 1981: 122) Hence, *zhongyao ren ‘important person’ (21e) is not acceptable while zhongyao renwu ‘important person’ (20e) is, because renwu ‘personage’ is a literary term that has its origin in Classical Chinese writings. Additionally, gao han ‘tall adult man’ (20f) is acceptable because han is a literary term from Classical Chinese. By contrast, “gao must take de when it modifies nan-haizi ‘male-child=boy’ … gao must occur with de when it modifies most animate nouns: boy, girl, man, woman, people, and so forth.” (Li and Thompson 1981: 123)
According to McCawley (1992), who cites Jiang Zixin (personal communication), the length of the head noun in [A N] is constrained. That is, the longer the head noun is, the more unacceptable the [A N] will be. See (28)–(29). The adjective hao ‘good’ modifies a head noun. When the length of the head noun increases, the acceptability of the [A N] decreases.
‘a glass of good liquor’
‘a glass of good beer’
‘a glass of good (grape) wine’
|one||cl||good||Shaoxing (place name)||yellow-wine|
‘a cup of good Shaoxing yellow-wine’
(Adapted from McCawley [1992: 235], citing Jiang Zixin [personal communication])
‘a good cup’
‘a good glass’
‘a good glass’
(Adapted from McCawley [1992: 235], citing Jiang Zixin [personal communication])
“If the phonological or morphological form of the element with which hao combines with is the main factor determining the acceptability of semantically coherent combinations, that argues that the rule licensing such combinations is one of word formation rather than of syntax and thus that hao[-N] is indeed a compound rather than a phrasal unit.” 21 (McCawley 1992: 235). See Zwicky (1969), Zwicky and Pullum (1986), and Miller et al. (1997) for discussion of the Principle of Phonology-Free Syntax, which states that “(i)n the grammar of natural language, rules of syntax make no reference to phonology.” (Miller et al. 1997: 68)
Schäfer (2009) claims that lexical gaps do not support the wordhood of Chinese [A N] because they pattern like German lexicalized phrases, whose productivity is also severely restricted. See (30).
‘cohabitation’=not officially sanctioned marriage-like way of living together
(Schäfer 2009: 277)
According to Schäfer, wilde Ehe ‘cohabitation’ must be a phrase because the adjective wild-e carries an inflectional ending. This phrase needs to be listed in the lexicon because its meaning is opaque.
Lexicalized phrases have also been found in other languages (Koefoed 1993; Booij 2002; Booij 2010; Hüning 2010). See the Dutch examples in (31). According to Booij, the adjectives are inflected and agree with the head noun in gender, number, and (in)definiteness. These [A N] are not words in the morphological sense, because agreement is a syntactic operation. “In Dutch, a prenominal adjective ends in -e, unless the NP is indefinite and the head noun is singular and neuter.” (Booij 2010: 176) Booij argues that the Dutch phrasal [A N] in (31) are conventionalized expressions for certain concepts and hence are listed in the lexicon. 22
dikke darm ‘thick intestine, large intestine’
dood spoor ‘dead trail, deadlock’
hoge hoed ‘high hat, top hat’
magere yoghurt ‘lean yoghurt, fat-free yoghurt’
open haard ‘open hearth, fireplace’
rode kaart ‘red card’
vaste benoeming ‘fixed appointment, tenure’
zure regen ‘acid rain’
vrije trap ‘free kick’
witte was ‘white laundry’
zwarte doos ‘black box’
(Booij 2010: 175–176)
Chinese, unlike Germanic languages, does not exhibit indisputable examples of phrasal [A N]. Nor does it allow lexicalized phrasal [A N]: [A de N] cannot be lexicalized. Were there lexicalized phrasal [A N] in Chinese, why couldn’t [A de N] also be lexicalized? Additionally, Booij (2010: 170) remarks that “[p]hrases used as names are often conventionalized expressions, and hence lexical units.” By contrast, semantically transparent Chinese [A N] such as huang chenshan ‘yellow shirt’ are usually produced online, and therefore are not conventionalized.
6 The test of modification by degree adverbs
The non-modifiability of Chinese [A N] by degree adverbs shows that Chinese [A N] are words (Fan 1958; Dai 1992; McCawley 1992; Duanmu 1998; Duanmu 2000; Feng 2001). It is well known that Chinese [A de N] can be modified by degree adverbs while Chinese [A N] in general cannot. See the contrast between (32) and (33) in terms of modifiability by degree adverbs.
|[hen xin]/[geng xin]/[zui xin]/[zheme xin] de shu|
|[very new]/[more new][most new]/[such new] DE book|
‘[very new]/[newer]/[newest]/[such new] books’
(Duanmu 2000: 109)
|*[hen xin]/*[geng xin]/*[zui xin]/*[zheme xin] shu|
|[very new]/[more new]/[most new]/[such new] book|
‘[very new]/[newer]/[newest]/[such new] books’
(Duanmu 2000: 109)
Adverbial modification is a phrase-level phenomenon. It does not apply to Chinese [A N]. This suggests that Chinese [A N] are words, and that degree adverb modification conforms to the Lexical Integrity Hypothesis (Jackendoff 1972; Lapointe 1979; Selkirk 1982; Selkirk 1984; Huang 1984; Anderson 1992; Booij 2009; among many others). Huang (1984: 64) suggests that most differences between a word and a phrase in Chinese can be ascribed to the Lexical Integrity Hypothesis, stated as follows:
The Lexical Integrity Hypothesis (Huang 1984: 60)
No phrase-level rule may affect a proper subpart of a word. 23
Paul (2005) claims that the modification test does not support the wordhood of Chinese [A N]. Instead, she argues for a phrasal analysis of Chinese [A N]. Paul assumes that a degree adverb plus an adjective forms a phrase in Chinese. She claims that a noun cannot be modified by a phrase consisting of a degree adverb and an adjective because of a universal syntactic constraint on the size of a prenominal modifier: the prenominal modifier can only be an adjectival head (see Sadler and Arnold 1994).
This syntactic constraint, however, incorrectly predicts that complex adjectives such as hong-tongtong (red-modifying suffix) ‘reddish’ can directly modify nouns, since they are simple adjectival heads from a syntactic point of view. But cases like *hong-tongtong lian ‘reddish face’ are unacceptable in Chinese. No such incorrect predictions are made under a word analysis of Chinese [A N] with the monomorphemic constraint.
Paul’s (2005) argument for a phrasal analysis of Chinese [A N] is an example of the logical fallacy of affirming the consequent, which works as follows:
If A is true, then B is true.
B is true.
Therefore, A is true.
This is a fallacy because B can be true even if A is not. Paul says that a syntactic account predicts that some Chinese [A N] will have certain properties such as the non-modifiability by degree words. She shows that these [A N] do have these properties. Then she concludes that they must be syntactic. But this is not true. There may be many reasons for them to have these properties that have nothing to do with syntax.
Both Duanmu (2000) and Paul (2005) observe that the superlative degree adverb zui ‘the most’ can sometimes modify [A N] (e.g., zui xin chanpin ‘the latest product’). However, many [A N] cannot be modified by zui (e.g., xin shu ‘new book’ vs. *zui xin shu ‘the latest book’, piaoliang nüren ‘beautiful woman’ vs. *zui piaoliang nüren ‘the most beautiful woman’). Additionally, many [zui A N] do not have their corresponding [A N] (e.g., zui gao qiwen ‘the highest temperature’ vs. *gao qiwen ‘high temperature’, zui huai dasuan ‘the worst outcome’ vs. *huai dasuan ‘bad outcome’, zui jia xuanshou ‘the best player’ vs. *jia xuanshou ‘good player’). These discrepancies point to the conclusion that [zui A N] and [A N] are two constructions independent of each other, so that we cannot use [zui A N] to tell whether the corresponding [A N] is a word or phrase. Paul (2005) also argues that zui should be treated separately from other degree adverbs because ding ‘utmost, extreme’, an equivalent of zui used only in the spoken language, only modifies [A de N]. The [zui A N] construction is exempt from the monomorphemic constraint because it is distinct from the general [A N] construction. The distinction can easily be modeled within Construction Morphology, which we discuss later.
Paul (2005) and Schäfer (2009) assume that a Chinese [A N] must always co-occur with a naming function, and that naming prevents it from being modified by a degree adverb. Schäfer (2009), for example, claims that Chinese [A N] pattern like German lexicalized [A N] phrases, so that they always bear a naming function and therefore lose their modifiability by degree words. See the German examples in (36). The expression grüner Tee can be a lexicalized [A N] phrase and express a particular type of tea. Under this interpretation, it cannot be modified by degree adverbs. On the other hand, when grüner Tee acts as a non-lexicalized phrase and means ‘tea that looks green’, it can be modified by degree adverbs as shown in (36b). See also Rainer and Varela (1992); Booij (2002; 2010) and Hüning (2010), who argue that lexicalized phrases pattern like compounds in terms of various syntactic behaviors such as the non-modifiability by degree adverbs.
=1. the subtype of tea [lexicalized phrase]
2. tea that looks green [standard intersective interpretation]
‘very green tea’=tea that looks very green
(Schäfer 2009: 282)
However, there is no lexicalized [A N] phrase in Chinese and Chinese [A N] can describe entities in addition to naming. Hence, we should expect Chinese [A N] to be modifiable by degree adverbs when they describe entities just like (36b). Therefore, the fact that Chinese [A N] cannot ever be modified by degree adverbs must be ascribed to their wordhood instead of semantic constraints.
There are two independent reasons for the non-modifiability of Chinese [A N] by degree adverbs, both of which follow from the wordhood of Chinese [A N]. First, since a Chinese [A N] is a noun, it cannot be modified by adverbs, given that adverbs generally do not modify nouns in any language. 24 Second, it has been shown that an adverb plus an adjective forms a phrase in Chinese (Dai 1992; Duanmu 2000; Paul 2005). Hence, such phrases cannot occur inside a noun because of Lexical Integrity.
7 XP substitution
Another test based on Lexical Integrity is the “XP substitution” test (Fan 1958; Duanmu 1998; Duanmu 2000). Fan (1958: 214) notes that N in [A de N] can be replaced by [X N], where X is a [Numeral-Classifier N] unit or a [Demonstrative N] unit, but N in [A N] cannot (37).
[A de N] → [A de XP]
[A N] → *[A XP]
[A de XP]
|xin||de [san ben shu]|
|new||DE three cl book|
‘three new books’
|xin||de [nei ben shu]|
|new||DE that cl book|
‘that new book’
(Duanmu 2000: 110)
|*xin||[san ben shu]|
|new||three cl book|
|*xin||[nei ben shu]|
|new||that cl book|
(Duanmu 2000: 110)
The failure of a Chinese [A N] to undergo XP substitution is often used as evidence to support its wordhood. Because of Lexical Integrity, a word cannot be manipulated by phrase-level rules. In (39a), *xin san ben shu is unacceptable because the phrase san ben shu ‘three books’ occurs inside a word. By contrast, san ben shu ‘three books’ can occur in the phrasal de-construction (38a).
Schäfer (2009) claims that no XP substitution is involved in the above cases. Instead, these cases show that we cannot change the order of san ben (three cl) and xin ‘new’ in san ben xin shu ‘three new books’. By contrast, we can change the order of san ben (three cl) and xin de (new DE) in san ben xin de shu ‘three new books’, so that we derive xin de san ben shu ‘three new books’.
Schäfer’s claim makes sense only if the two expressions that differ in word order are identical in terms of semantic scope. Otherwise they must be constructed independently of each other. In fact, the two differ in semantic scope. Consider (40). To address a question like A, we can reply with B1, which means ‘there are new books here and the number of them is three’. We cannot reply with B2, which means ‘there are three books here and the quality of them is new’.
‘How many new books there are here?’
|You||san||ben xin||de shu|
|have||three||cl new||DE book|
‘There are three new books (here).’
|*You xin||de san||ben shu|
|have new||DE three||cl book|
Hence, expressions like [san ben [xin de shu]] and [xin de [san ben shu]] are constructed independently of each other, given that their structural difference is accounted for by semantic scope, defined as follows: “[Scope] concerns semantic compositionality. In particular, given three items X, Y, and Z, items X and Y combine with each other and then combine as a unit with Z. The semantics of Z is added to that of X and Y as a unit.” (Rice 2000: 24)
Since expressions like san ben xin de shu and xin de san ben shu differ in semantic scope, no simple word order change can be involved in either (38) or (39). Hence, the ungrammaticality of the cases in (39) must be ascribed to the failure of XP substitution, which supports the wordhood of Chinese [A N].
8 Conjunction reduction
Conjunction reduction is another test based on Lexical Integrity that supports the wordhood of Chinese [A N]. Fan (1958) observes that conjunction reduction can be applied to [A de N], but not to [A N]. The two [A de N] are coordinated in (41a) and the repeated element xin de is deleted in the second conjunct in (41b). 25
|[xin||de||xie] he [xin de||shoubiao]|
|new||DE||shoe and new DE||watch|
‘new shoes and new watches’
|xin||de||[xie he shoubiao]|
|new||DE||shoe and watch|
‘new [shoes and watches]’
By contrast, the two [A N] are coordinated in (42a) while the repeated element xin ‘new’ cannot be deleted in the second conjunct, as shown in (42b).
|[xin xie] he||[xin shoubiao]|
|new shoe and||new watch|
‘new shoes and new watches’
|*xin||[xie he shoubiao]|
|new||shoe and watch|
‘new [shoes and watches]’
If conjunction reduction is syntactic, it supports the wordhood of Chinese [A N], given that Lexical Integrity forbids words to be manipulated by syntactic rules (Huang 1984; Duanmu 1998; Duanmu 2000).
Booij (1985) argues that in Dutch and German conjunction reduction is phonological in nature because such reduction resorts to phonological units such as phonological word and its projection. In Dutch, for example, *[[blauw]A
ig]A en [[rod]A ig]A ‘bluish and reddish’ will be ungrammatical if the suffix -ig is deleted because -ig is not an independent phonological word (Booij 1985: 149). By contrast, [[storm]N achtig]A en [[regen]N achtig]A ‘stormy and rainy’ will be grammatical if the suffix -achtig is deleted because -achtig can form an independent phonological word (Booij 1985: 151).
Similar cases can be found in German. For example, *[[Salz]N
ig]A und [[Mehl]N ig]A ‘salty and mealy’ will be ungrammatical if the suffix -ig is deleted (Booij 1985: 149). By contrast, [[Freund]N schaft]N oder [[Feind]N schaft]N ‘friendship or hostility’ can undergo conjunction reduction because -schaft can form an independent phonological word (Booij 1985: 152).
Based on Booij (1985), Schäfer (2009) remarks that conjunction reduction does not lend support to the wordhood of Chinese [A N] because it can be phonological in nature and insensitive to morphosyntactic categories such as word and phrase. Additionally, Schäfer (2009: 284) remarks that “Mandarin conjunctions like he ‘and’ and gen ‘and’ can only be used with nominal expressions and not with verbs, verb phrases, or clauses”. Hence, *hong
chenshan he huang chenshan ‘red (shirts) and yellow shirts’ will be ungrammatical if reduction takes place, because hong ‘red’ is an adjective so that it cannot be conjoined by he ‘and’.
Contra Schäfer (2009), conjunction reduction supports the wordhood of Chinese [A N]. In Dutch and German, phonological reduction can take place even when two conjuncts belong to different morphosyntactic categories. In (43) and (44), one conjunct is a compound and the other is a phrase.
(Booij 1985: 143–144)
(Booij 1985: 144)
Unlike Dutch and German, Chinese does not allow the type of reduction in (43) and (44). In (45a), for example, hong de he huang chenshan does not mean ‘red shirts and yellow shirts’, which shows that conjunction reduction cannot apply here. In (45d), mutou he piaoliang de zhuozi means ‘wood and beautiful tables’ instead of ‘wooden tables and beautiful tables’, which shows that conjunction reduction cannot apply here, either. Notice that except for huang ‘yellow’ in (45b), each individual conjunct in (45), either reduced or non-reduced, can uncontroversially be conjoined by he ‘and’. Hence, the failure of the examples in (45) to undergo conjunction reduction cannot be ascribed to the property of he ‘and’.
|*hong||de ||he||huang chenshan|
|red||DE shirt||and||yellow shirt|
‘red shirts and yellow shirts’
|*huang ||he||hong||de chenshan|
|yellow shirt||and||red||DE shirt|
‘yellow shirts and red shirts’
‘beautiful tables and wooden tables’
‘wooden tables and beautiful tables’
It is not clear what phonological rules or principles might prevent the examples in (45) from undergoing conjunction reduction. If every case of conjunction reduction were phonological in Chinese, these examples should be as good as those in Dutch and German, where conjunction reduction disregards whether a conjunct is a word or phrase. Hence, their failure to undergo conjunction reduction must be explained by syntax.
Like Chinese, English has cases in which conjunction reduction does not apply to two conjuncts of different morphosyntactic categories. For example, neither *ice
cream and heavy cream nor *heavy cream and ice cream can undergo conjunction reduction (Mark Aronoff [personal communication]).
Xu and Liu (1999) discuss cases in which several monosyllabic attributive modifiers precede a noun. See (46). At first glance, these cases seem to be derived from coordinated expressions. For example, jia da kong hua (fake big empty talk) seems to be derived from jia hua, da hua, he kong hua ‘fake talk, bragging talk, and empty talk’.
‘fake bragging empty talk’ or ‘fake talk, bragging talk, and empty talk’
‘true good beautiful thing’ or ‘the true, the good, and the beautiful’
‘good and bad strains’
(Xu and Liu 1999: 99–100)
The examples in (46) have nothing to do with conjunction reduction. Above all, many such cases do not have corresponding non-reduced coordinated expressions. For example, you lie pinzhong (good bad strain) cannot be derived from *you pinzhong he lie pinzhong (good strain and bad strain), in which neither *you pinzhong ‘good strain’ nor *lie pinzhong ‘bad strain’ is acceptable. See Fan (1958) for similar reasoning.
Second, many coordinated expressions cannot undergo conjunction reduction to bring about the construction in which several attributive modifiers precede a noun. For example, huang chenshan he hong chenshan ‘yellow shirt and red shirt’ cannot undergo conjunction reduction to derive *huang hong chengshan (yellow red shirt). As Fan (1958) points out, cases like jia da kong hua (fake big empty talk) are not coordinated expressions but generated by a productive construction in written Classical Chinese, which also generated Classical Chinese examples such as dong xi han (east west Han) ‘the Eastern and Western Han Dynasties’ and nan bei chao (south north dynasty) ‘the Southern and Northern Dynasties’.
Third, the order of attributive modifiers in each example in (46) is fixed. For example, jia da kong hua (fake big empty talk) cannot be changed into *da kong jia hua (big empty fake talk) and zhen shan mei shiwu (true good beautiful thing) cannot be changed into *shan mei zhen shiwu (good beautiful true thing). By contrast, the order of conjuncts in non-reduced coordinated expressions can be changed. For example, jia hua, da hua he kong hua ‘fake talk, bragging talk, and empty talk’ can be changed into da hua, kong hua, he jia hua ‘bragging talk, empty talk, and fake talk’. This contrast shows that the constructions with several attributive modifiers in (46) are lexicalized and bear no relation to conjunction reduction.
9 Anaphoric accessibility
Paul (2005) argues that some Chinese [A N] can be phrases because their noun head is anaphorically accessible, as shown in (47).
|I||think||yellow||DE||shirt||compared to||red||DE Ø|
‘I think that yellow shirts are prettier than red ones.’
(Adapted from Paul 2005: 763)
The noun head of huang chenshan ‘yellow shirts’ can be referred to by a zero anaphor, symbolized by Ø. Anaphoric accessibility also applies to the phrasal [A de N] construction (47).
By contrast, those [A N] whose heads cannot be anaphorically accessed are considered compounds in Paul (2005). See (48), in which the head of hong-hua ‘safflower’ is not anaphorically accessible.
|*Amei||bu xiang||chi||hong-hua,||huang||de Ø||hai keyi.|
|Amei||not want||eat||red-flower=||safflower yellow||DE Ø||still ok|
‘Amei doesn’t want to take safflower, yellow ones are still ok’.
(Adapted from Paul 2005: 763)
Examples (47) and (48) involve what Postal (1969) terms outbound anaphora, under which a part of a word is anaphorically inaccessible. See (49). The underlined English expressions are words, yet an anaphor cannot refer to their internal parts.
*He took the teapot and poured it into the cup. (it=tea) (Spencer 1991: 42)
*The best wombatmeat comes from young ones. (ones=wombats) (Postal 1969: 226)
Paul’s phrasal analysis of some Chinese [A N] based on anaphoric accessibility is another example of affirming the consequent. According to Ward et al. (1991), the felicity of outbound anaphora has nothing to do with either syntax or morphology. See also Sproat (1993). Instead, it should be ascribed to pragmatic factors. Ward et al. (1991: 439) claim that outbound anaphora is “in fact fully grammatical and governed by independently motivated pragmatic principles” because outbound anaphora is possible in certain contexts. Ward et al. (1991: 450) assume that “a genuinely ungrammatical construction is ungrammatical in all (nonmetalinguistic) contexts, and cannot be ‘amnestied’ by pragmatic or discourse factors.” Consider (50), in which the underlined expressions are words but their internal elements are anaphorically accessible. The expression cocaine use is a compound with the main stress falling on cocaine.
McCarthyites are now puzzled by him. (him=McCarthy) (Lieber 1984: 188)
Although casual cocaine use is down, the number of people using it routinely has increased. (it=cocaine) (Ward et al. 1991: 454)
Apart from English, Schäfer (2009) shows that German [A N] compounds are visible to outbound anaphora. In (51), Grünglas ‘green glass (as a subcategory of glass in the context of recycling)’ is a compound with “the main accent falling on the adjective and the adjective itself being realized in its shortform, i.e., without any inflectional ending.” (Schäfer 2009: 286) The noun head of the compound Grünglas is anaphorically accessible.
|Ich||bin das Grünglas losgeworden,||das weiβe||liegt||noch im|
|I||am the green-glass got.rid.off||the white-nom.sg||lies||still in.the|
‘I got rid of the green glass, the white glass is still in the car.’
(Schäfer 2009: 286)
I present evidence that outbound anaphora cannot distinguish words from phrases in Chinese, either. In (52), mei-nü ‘beautiful woman’, da-han ‘hefty man’, bai-shu ‘white rat’, and qing-jiao ‘green pepper’ are all compounds because nü ‘woman’, han ‘man’, shu ‘rat’, and jiao ‘pepper’ are bound roots. See, for example, Bresnan and Mchombo (1995) and Harris (2006) for similar reasoning. Each of these bound roots, which is part of a compound, can be referred to by a zero anaphor. This shows that Chinese words are visible to outbound anaphora, which therefore is not a phrase-level rule. Because outbound anaphora cannot distinguish words from phrases either in Chinese or crosslinguistically, there is no compelling evidence for treating Chinese [A N] as phrases.
|Zhangsan zhi||xiang qu||mei-nü.||Chou||de Ø|
|Zhangsan only||want marry||beautiful-woman||ugly||DEØ|
|ta bu yao.|
|he not want|
‘Zhangsan only wants to marry a beautiful woman. He does not want an ugly one.’
|Amei zhi xiang jia||da-han.||Shencai xiao||de Ø||ta bu yao.|
|Amei only want marry||big-man||stature small||DE Ø||she not want|
‘Amei only wants to marry a hefty man. She does not want one small in
|Women||zhi||mai bai-shu,||bu mai hei||de Ø.|
|We||only||buy white-rat||not buy black||DE Ø|
‘We only buy white rats, not black (ones).’
‘We only buy green peppers, not red (ones).’
Ward et al. (1991: 454) argue that “[a] key factor in determining the felicity of outbound anaphora is the semantic transparency of the word containing the antecedent of the anaphor (see Lieber 1984). The containing word must be sufficiently transparent for the word-internal morpheme to successfully evoke an accessible discourse entity[, which is referred to by an anaphor].” According to Ward et al. (1991: 454), “semantic opacity generally inhibits outbound anaphora.” The expression huang (de) chenshan ‘yellow shirts’ in (47) is visible to outbound anaphora while hong-hua ‘safflower’ in (48) is not, because the latter is semantically opaque so that it can no longer be straightforwardly interpreted based on its component parts. Accordingly, a hearer will access the meaning of the semantically opaque compound directly, i.e., without morphologically decomposing it into smaller parts, one of which can be anaphorically accessible.
Additionally, in (47) huang (de) chenshan ‘yellow shirts’ stands in contrast to hong de Ø ‘red ones’ in terms of color. Hence, a connection can be established between an expression containing an antecedent and one containing a zero anaphor, which facilitates outbound anaphora. By contrast, because of semantic opacity, in (48) hong-hua ‘safflower’ cannot stand in contrast to huang de Ø ‘yellow ones’ in terms of color. Such a connection cannot be established, which inhibits outbound anaphora.
10 Adjectival ordering restrictions
Sproat and Shih (1991) propose adjectival ordering restrictions (AOR) that require an adjective expressing the property COLOR to be closer to its modified noun than one expressing the property SIZE because the former adjective expresses a more inherent property of the noun than the latter one (see Whorf 1945).
Feng (2001) observes that some Chinese [A N] obey AOR while others do not. In (53), da panzi ‘big plate’ obeys AOR. The adjective bai ‘white’ must be closer to the noun than da ‘big’.
‘a big plate’
‘a big white plate’
(Feng 2001: 168)
By contrast, in (54), da-guar (big-gown) ‘an unlined long gown’ does not obey AOR. The adjective bai ‘white’ must be outside da-guar.
‘an unlined long gown’
‘a white unlined long gown’
(Feng 2001: 168)
AOR have been assumed to be syntactic constraints (Feng 2001; Paul 2005; Schäfer 2009). Accordingly, it has been claimed that Chinese [A N] that obey AOR are phrases and those that do not are words (Paul 2005; Schäfer 2009). If this is true, then Chinese [A N] like da panzi ‘a big plate’ must be phrases and Chinese [A N] like da-guar ‘an unlined long gown’ must be words.
A phrasal analysis of Chinese [A N] based on AOR is another case of affirming the consequent because AOR cannot distinguish words from phrases. AOR are constraints on semantic compositionality and order adjectives based on their senses (Sproat and Shih 1991). AOR are spiritually very similar to the semantic relevance principle proposed in Bybee (1985), which has been applied to inflectional morphology. Bybee (1985: 13) remarks that “[a] meaning element is relevant to another meaning element if the semantic content of the first directly affects or modifies the semantic content of the second.” Bybee (1985) uses the semantic relevance principle to account for the well-known fact that case markers are farther away from the nominal stem than number markers (Greenberg 1963). She remarks that:
[T]he expression of number occurs closer to the noun base because it is more relevant to the meaning of the noun. Number has a direct effect on the entity or entities referred to by the noun. Case, on the other hand, has no effect on what entity is being referred to, but rather only changes the relation of that same entity to the other elements in the clause. (Bybee 1985: 34)
Like AOR, the scope constraint is semantically based and has been applied to both the orderings of words in a phrase and affixes in a word (see Rice 2000; Paster 2005; Aronoff and Xu 2010). Aronoff and Xu (2010: 389) define the scope constraint in morphology as follows: “Given two scope-bearing features f1 and f2, if f1 scopes over f2, then I2, an exponent of f2 cannot be farther away from the same stem than I1, an exponent of f1.”
Under the scope constraint, case markers are farther away from the nominal stem than number markers because “case expresses the relation of an entity or a number of entities to other elements in a clause.” (Aronoff and Xu 2010: 390) If AOR are semantically based like these others, and not phrase-level, they provide no evidence for the phrasehood of Chinese [A N].
For various reasons, the first element of Chinese [A N] like da-guar (big-gown) ‘an unlined long gown’ cannot be separated from the noun head. Above all, the first element of many such [A N] is bleached semantically so that they do not obey AOR. In the example at hand, da in da-guar ‘an unlined long gown’ is bleached semantically and does not express the property SIZE. Hence, bai da-guar ‘a white unlined long gown’ vacuously satisfies AOR, which are constraints on semantic compositionality and order semantically meaningful adjectives.
In fact, almost all [A N] in which the adjective cannot be separated from the noun are conventionalized terms for particular types of entities, so they must be listed in the lexicon with this fixed form. The conventionalized or lexicalized sense of such [A N] is maintained only if the first element immediately precedes the noun head. For example, xiao-dou (small-bean) is a conventionalized term for ‘azuki bean’. Its conventionalized sense will be lost if xiao does not immediately precede the noun head. For example, xiao hong-dou (small red-bean) no longer expresses ‘azuki bean’. Instead, it means ‘small Indian Licorice’ and hong-dou (red-bean) refers to a different plant from xiao-dou (small-bean).
AOR, like all principles of compositionality, are overridden by the constraint to maintain a conventionalized or lexicalized sense of a fixed expression. Similarly, AOR are overridden by conventionalized collocations. For example, hong-xiao-bing (red-small-soldier) ‘young soldiers who fought for the Cultural Revolution’ is acceptable while *xiao-hong-bing (small-red-soldier) is unacceptable because hong ‘red’ cannot directly combine with bing ‘soldier’ given that *hong-bing (red-soldier) ‘soldier who is red’ is semantically infelicitous.
11 Brief summary
I have argued that there is compelling evidence for the word status of Chinese [A N], contra Paul (2005) and Schäfer (2009). My findings are briefly summarized in Table 7, in which I list the previously discussed criteria that have been used to judge the status of Chinese [A N] in terms of word or phrase. The symbol ? indicates that a criterion in its left is incapable of determining the status of [A N]. Every decisive test supports wordhood. When applied properly, the naming test clearly distinguishes words from phrases. See (14). It is much more usual for words to name. Unlike Dutch and German, Chinese does not have lexicalized [A N] phrases, which have a naming function.
12 A Construction Morphology account of Chinese [A N]
The various properties of Chinese [A N] can be expressed and formalized under the framework of Construction Morphology (CM) (Booij 2010). CM “assumes that there are specifically morphological generalizations or rules that cannot be reduced to either syntax or phonology.” (Booij 2010: 3) That is, CM “takes the lexicalist position that the grammars of natural languages have a relatively autonomous morphological sub-gramamar.” (Booij 2010: 3) (see Lapointe 1980; Selkirk 1982; Di Sciullo and Williams 1987; Chomsky 1995). “The word remains an essential unit for stating regularities. Both words and phrasal constructions are domains over which certain generalizations can be stated, and hence the domains of “word” and “phrase” are both essential for the analysis of natural languages” (Booij 2009: 97) (see Blevins 2006). Under CM, we can distinguish Chinese [A N]N from [A de N]NP as word and phrase, based on their distinct properties that have been discussed in previous sections. The properties of Chinese [A N] can be accounted for in the word formation component of the grammar in CM.
CM is built upon the notion construction, defined as “a pairing of form and meaning.” (Booij 2010: 11) An individual construction is characterized by a holistic property that cannot be derived from the properties of its constituent parts. Chinese [A N] are word-level constructions with the following holistic properties that cannot be derived from their constituent parts.
Selectional restrictions on Chinese [A N].
The monomorphemic constraint on the adjective of Chinese [A N].
First, Chinese [A N] with different adjectives are subject to different selectional restrictions. For example, in [A N] the adjective shen ‘deep’ usually combines with nouns that express physical depth instead of abstract profoundness (Xu 2006). For example, both shen shui ‘deep water’ and shen de shu (deep DE book) ‘deep book’ are fine while *shen shu ‘deep book’ is bad, because shen ‘deep’ cannot directly modify shu ‘book’, which would otherwise express abstract profoundness.
These selectional restrictions are associated with constructional idioms (Langacker 1987; Fillmore et al. 1988; Goldberg 1995; Goldberg 2006; Kay and Fillmore 1999; Pitt and Katz 2000; Jackendoff 1997; Jackendoff 2002; Booij 2010; among others). Constructional idioms are morphological or syntactic schemas with both lexically specified positions and open slots that are represented by variables.
The selectional restriction that requires shen ‘deep’ to directly combine with nouns expressing physical depth is associated with the constructional schema in (55).
[[shen]A [x]Ni]Nj ↔ [deep SEMi]j (where SEMi=objects that can express physical depth)
Following Booij (2010), the variable x stands for a phonological form. The symbol ↔ stands for correspondence. To the left of the double arrow stand both the phonological form shen x and the morphosyntactic form [A N]N, which are conflated into one representation for ease of exposition. To the right of the double arrow stands the corresponding semantic interpretation. SEM stands for semantic interpretation. Co-indexation via the lower-case variables i, j states the correspondence between the phonological, morphosyntactic, and semantic properties of words. The naming or descriptive function of the constructional schema in (55) and of any word construction depends on factors such as whether the nominal head is free and/or whether its instantiation has been conventionalized. In any lexicalist theory, the (potential) naming function of any noun follows directly from the fact that it is a word, so there is no need to stipulate in a rule whether a noun or adjective in a compound names or describes.
Second, the monomorphemic constraint requires the adjective of a Chinese [A N] to be monomorphemic. It is a default property of the Chinese [A N] construction. With both inheritance hierarchy and default inheritance, CM allows not only for the expression of this default property but also for some overrides. 26 Under CM, the lexicon is made up of hierarchical arrays of constructions. CM adopts the notion of inheritance hierarchy (Flickinger 1987; Krieger and Nerbonne 1993; Riehemann 1998; Riehemann 2001; König 1999; Hippisley 2001; Sag et al. 2003; Kilbury et al. 2006). According to their common properties, expressions are classified into a general category or meta-construction, which stands on top of a hierarchical tree. This general category or meta-construction is subdivided into or dominates more specific categories or constructions. The properties of the general category or meta-construction are inherited by the specific categories or constructions that it dominates. Each individual word or specific instantiation of a construction sits at the bottom of the hierarchical tree.
In Chinese, an overwhelming majority of [A N] obey the monomorphemic constraint but there are some [A N] with pseudo-adjectival compounds, which are morphologically complex but semantically simple (see Section 2). This can be described via the inheritance hierarchy in (56), represented by an inheritance tree. The schemas in this hierarchy are classified in terms of the monomorphemic constraint. The general schema, which the monomorphemic constraint is associated with, stands on top of the inheritance tree and dominates other schemas. Its default property which requires the adjective to be monomorphemic is inherited by its dominated schemas except [A N]N (A=µµ ↔ SEMµ), which overrides this default property and allows for bimorphemic pseudo-adjectival compounds. (I follow Anderson 1992 and use the symbol µ to represent a morpheme.)
Additionally, the [zui A N] ‘the most A N’ construction is exempt from the monomorphemic constraint because it is distinct from the general [A N] construction (see Section 6). This is normal in a default system and can easily be modeled in (57).
Under CM, the lexicon is made up of various types of inheritance hierarchies. Each individual word may form the end node of a number of such hierarchies. For example, a classification in terms of semantic interpretation may cross-classify with a classification in terms of morphological restrictions. For example, the construction [[a]Ak [b]Ni]Nj ↔[SEMi with relation R to SEMk]j can stand on top of an inheritance tree and dominate the words under it. The symbols a and b each stand for a phonological form. “The nature of R is not specified but is determined for each individual compound on the basis of the meaning of the compound constituents, and encyclopedic and contextual knowledge (Downing 1977; Jackendoff 2009).” (Booij 2010: 17) In (58) the adjectives are bleached semantically and both compounds have a lexicalized sense. In (59) the adjectives are not bleached semantically and both compounds have a lexicalized sense that incorporates or adds to the literal sense. In (60) the [A N] are semantically transparent and their naming function emerges in a context that enforces naming. In (61) the adjective is semantically redundant and its presence is optional.
hei-ban (black-board) ‘blackboard’
hong-hua (red-flower) ‘safflower’
congming ren (intelligent person) ‘a person who is intelligent and understands the situation’
xiao yifu (small clothes) ‘the type of clothes that bear the property SMALL and are designed for kids’
xiao zhuozi (small table) ‘(the type of) table that is small’
hong yifu (red clothes) ‘(the type of) clothes that are red’
da-suan (big-garlic) ‘garlic’
da-xiang (big-elephant) ‘elephant’
Chinese [A N] are word-level expressions constructed in an autonomous word formation component of the grammar. They exhibit holistic properties that cannot be ascribed to either their constituent parts or the syntax. I have provided a CM account because it is helpful in understanding the data. 27 Other accounts are possible. I am not arguing against them.
I have examined several views about the status of Chinese [A N] as either words or phrases. I have reviewed various criteria for judging their status and concluded that Chinese [A N] are words. I have provided new evidence for the wordhood of Chinese [A N] in the form of a newly proposed monomorphemic constraint on the adjective of [A N] and the tests of modification by degree adverbs, conjunction reduction, and XP substitution. Chinese [A N] have a (potential) naming function and are subject to selectional restrictions. Both are evidence for their wordhood. Anaphoric accessibility is not a reliable criterion for judging the status of Chinese [A N] because it is semantically and pragmatically determined and independent of phrasal syntax. The restrictions on the ordering of adjectives cannot judge the status of Chinese [A N], either, because they are constraints on semantic compositionality and therefore independent of phrasal syntax, too.
I have used the framework of Construction Morphology (Booij 2010) to encode the various properties of Chinese [A N], which are word-level constructions with holistic properties that cannot be derived from either their constituent parts or the syntax. I have shown that Chinese [A N] are constructed in an autonomous word formation component of the grammar, which is compatible with Jackendoff’s (2009: 114) conclusion that “compounding is only barely syntactic.”
This work was supported by grant 15BYY004 to the author from the National Social Science Foundation of China. Portions of this work were presented at the Joint Conference of the 15th International Association of Chinese Linguistics (IACL-15) Conference and the 19th North American Conference on Chinese Linguistics (NACCL-19) at Columbia University, the 25th North American Conference on Chinese Linguistics (NACCL-25) at University of Michigan, and a colloquium held by the Department of East Asian Languages and Literatures in University of Hawai‘i at Mānoa. I am grateful to those audiences for useful discussion. I also thank Giorgio Francesco Arcodia, Zhiming Bao, Geert Booij, San Duanmu, Martin Schäfer, and two anonymous reviewers for their very helpful comments. Special thanks are given to Mark Aronoff, who spent countless time discussing this work and provided me with many valuable comments, criticisms, and suggestions. All errors remain my own.
Anderson, Stephen R. 1992. A-morphous morphology. Cambridge: Cambridge University Press. Google Scholar
Arcodia, Giorgio Francesco. 2012. Lexical derivation in Mandarin Chinese. Taiwan: Crane Publishing. Google Scholar
Baker, Mark. 2003. Lexical categories: Verbs, nouns, and adjectives. Cambridge: Cambridge University Press. Google Scholar
Bauer, Laurie. 2003. Introducing linguistic morphology, 2nd edn. Edinburgh: Edinburgh University Press. Google Scholar
Bhat, D. N. S. 1994. The adjectival category: Criteria for differentiation and identification. Amsterdam & Philadelphia: John Benjamins. Google Scholar
Booij, Geert. 1985. Coordination reduction in complex words: A case for prosodic phonology. In Harry Van Der Hulst & Norval Smith (eds.), Advances in non-linear phonology, 143–160. Dordrecht: Foris. Google Scholar
Booij, Geert. 2002. Constructional idioms, morphology, and the Dutch lexicon. Journal of Germanic Linguistics 14(3). 301–329. Google Scholar
Booij, Geert. 2009. Lexical integrity as a formal universal: A constructionist view. In Sergio Scalise, Elisabetta Magni & Antonietta Bisetto (eds.), Universals of language today, 83–100. Dordrecht: Springer. Google Scholar
Booij, Geert. 2010. Construction Morphology. Oxford: Oxford University Press. Google Scholar
Brown, Dunstan & Andrew Hippisley. 2012. Network Morphology: A defaults-based theory of word structure. Cambridge: Cambridge University Press. Google Scholar
Bybee, Joan L. 1985. Morphology: A study of the relation between meaning and form. Amsterdam & Philadelphia: John Benjamins. Google Scholar
Chao, Yuen-Ren. 1968. A grammar of spoken Chinese. Berkeley & Los Angeles: University of California Press. Google Scholar
Chen, Qiongzan. 1955. Xiushiyu he mingci zhi jian de ‘de’ zi de yanjiu [An investigation of de between modifier and noun]. Zhongguo Yuwen 10. 22–27. Google Scholar
Chomsky, Noam. 1995. The Minimalist Program. Cambridge, MA: MIT Press. Google Scholar
Cinque, Guglielmo. 1993. A null theory of phrase and compound stress. Linguistic Inquiry 24. 239–297. Google Scholar
Croft, William. 1991. Syntactic categories and grammatical relations. Chicago: University of Chicago Press. Google Scholar
Dai, John Xiangling. 1992. Chinese morphology and its interface with the syntax. Columbus, OH: Ohio State University dissertation. Google Scholar
Di Sciullo, Anna-Maria & Edwin Williams. 1987. On the definition of word. Cambridge, MA: MIT Press. Google Scholar
Dixon, R. M. W. 2004. Adjective classes in typological perspective. In R. M. W. Dixon & Alexandra Y. Aikhenvald (eds.), Adjective classes: A cross-linguistic typology, 1–49. Oxford: Oxford University Press. Google Scholar
Duanmu, San. 1990. A formal study of syllable, tone, stress, and domain in Chinese languages. Cambridge, MA: MIT dissertation. Google Scholar
Duanmu, San. 1998. Wordhood in Chinese. In Jerome Packard (ed.), New approaches to Chinese word-formation: Morphology, phonology and the lexicon in modern and ancient Chinese, 135–196. Berlin & New York: Mouton de Gruyter. Google Scholar
Duanmu, San. 2000. The phonology of Standard Chinese. Oxford: Oxford University Press. Google Scholar
Duanmu, San. 2007. The phonology of Standard Chinese, 2nd edn. Oxford: Oxford University Press. Google Scholar
Erben, Johannes. 2000. Einführung in die deutsche Wortbildungslehre. Berlin: Erich Schmidt Verlag. Google Scholar
Fan, Jiyan. 1958. Xing-ming zuhe jian de zi de yufa zuoyong [The grammatical function of de in adjective-noun constructions]. Zhongguo Yuwen 1958(5). 213–217. Google Scholar
Feng, Shengli. 1998. Lun hanyu de “ziran yinbu” [On “natural feet” in Chinese]. Zhongguo Yuwen 1. 40–47. Google Scholar
Feng, Shengli. 2001. Lun hanyu ci de duo-wei-xing [The multidimensional properties of word in Chinese]. Contemporary Linguistics 3(3). 161–174. Google Scholar
Flickinger, Daniel Paul. 1987. Lexical rules in the hierarchical lexicon. Stanford, CA: Stanford University dissertation. Google Scholar
Fu, Jingqi. 1987. La structure du syntagme nominal en chinois. University of Paris dissertation. Google Scholar
Goldberg, Adele. 1995. Constructions: A Construction Grammar approach to argument structure. Chicago: Chicago University Press. Google Scholar
Goldberg, Adele. 2006. Constructions at work: The nature of generalization in language. Oxford: Oxford University Press. Google Scholar
Greenberg, Joseph. 1963. Some universals of grammar with particular reference to the order of meaningful elements. In Joseph Greenberg (ed.), Universals of language, 73–113. Cambridge, MA: MIT Press. Google Scholar
Hippisley, Andrew. 2001. Word formation rules in a default inheritance framework: A Network Morphology account of Russian personal names. In Geert Booij & Jaap Van Marle (eds.), Yearbook of Morphology 2000, 221–261. Dordrecht: Kluwer. Google Scholar
Hockett, Charles F. 1958. A course in modern linguistics. New York: Macmillan. Google Scholar
Huang, James C.-T. 1984. Phrase structure, lexical integrity, and Chinese compounds. Journal of the Chinese Language Teachers Association 19(2). 53–78. Google Scholar
Hüning, Matthias. 2010. Adjective + Noun constructions between syntax and word formation in Dutch and German. In Alexander Onysko & Sascha Michel (eds.), Cognitive perspectives on word formation, 195–215. Berlin & New York: De Gruyter Mouton. Google Scholar
Inkelas, Sharon & Cheryll Zoll. 2005. Reduplication: Doubling in morphology. Cambridge: Cambridge University Press. Google Scholar
Jackendoff, Ray. 1972. Semantic interpretation in generative grammar. Cambridge, MA: MIT Press. Google Scholar
Jackendoff, Ray. 1997. The architecture of the language faculty. Cambridge, MA: MIT Press. Google Scholar
Jackendoff, Ray. 2002. Foundations of language. Oxford: Oxford University Press. Google Scholar
Jackendoff, Ray. 2009. Compounding in the parallel architecture and conceptual semantics. In Rochelle Lieber & Pavol Štekauer (eds.), The Oxford handbook of compounding, 105–129. Oxford: Oxford University Press. Google Scholar
Kilbury, James, Wiebke Petersen & Christoph Rumpf. 2006. Inheritance-based models of the lexicon. In Dieter Wunderlich (ed.), Advances in the theory of the lexicon, 429–480. Berlin & New York: Mouton de Gruyter. Google Scholar
Koefoed, Geert. 1993. Benoemen: Een beschouwing over de faculté du langage. Amsterdam: P. J. Meertens-Instituut. Google Scholar
König, Jean-Pierre. 1999. Lexical relations. Stanford, CA: CSLI. Google Scholar
Krieger, Hans-Ulrich & John Nerbonne. 1993. Feature-based inheritance networks for computational lexicons. In Ted Briscoe, Valeria De Paiva & Ann Copestake (eds.), Inheritance, defaults, and the lexicon, 90–136. Cambridge: Cambridge University Press. Google Scholar
Langacker, Ronald. 1987. Foundations of Cognitive Grammar, Vol. 1. Theoretical prerequisites. Stanford: Stanford University Press. Google Scholar
Lapointe, Steven G. 1979. A lexical analysis of the English auxiliary verb system. GLOT 2. 215–254. Google Scholar
Lapointe, Steven G. 1980. A theory of grammatical agreement. Amherst, MA: University of Massachusetts dissertation. Google Scholar
Larson, Richard & Gabriel Segal. 1995. Knowledge of meaning: An introduction to semantic theory. Cambridge, MA: MIT Press. Google Scholar
Leben, William. 1973. Suprasegmental phonology. Cambridge, MA: MIT dissertation. Google Scholar
Li, Charles N. & Sandra A. Thompson. 1981. Mandarin Chinese: A functional reference grammar. Berkeley & Los Angeles: University of California Press. Google Scholar
Lieber, Rochelle. 1984. Grammatical rules and sublexical elements. In David Testen, Veena Mishra & Joseph Drogo (eds.), Papers from the parasession on lexical semantics, 187–199. Chicago: Chicago Linguistic Society. Google Scholar
Lu, Bingfu. 1990. The structure of Chinese nominal phrases. Storrs, CT: University of Connecticut master thesis. (Published in Comparative studies on the structure of noun phrases, Research Report (PI: Mamoru Saito), Department of Linguistics, University of Connecticut, Storrs, 1–50.). Google Scholar
Lu, Bingfu & San Duanmu. 1991. A case study of the relation between rhythm and syntax in Chinese. Paper presented at the Third North American Conference on Chinese Linguistics. Cornell University. Google Scholar
Lu, Bingfu & San Duanmu. 2002. Rhythm and syntax in Chinese: A case study. Journal of the Chinese Language Teachers Association 37(2). 123–136. Google Scholar
Lu, Jianming. 1985. Duo he shao zuo dingyu [Duo ‘many’ and shao ‘few’ as modifiers]. Zhongguo Yuwen 1. 71–75. Google Scholar
McCarthy, John J. & Alan Prince. 1993. Prosodic Morphology I: Constraint interaction and satisfaction. Ms., University of Massachusetts, Amherst & Rutgers University. Google Scholar
McCawley, James D. 1992. Justifying part-of-speech assignments in Mandarin Chinese. Journal of Chinese Linguistics 20(2). 211–246. Google Scholar
Miller, Philip H., Geoffrey K. Pullum & Arnold M. Zwicky. 1997. The principle of phonology-free syntax: Four apparent counterexamples in French. Journal of Linguistics 33. 67–90. CrossrefGoogle Scholar
Milsark, Gary. 1988. Singl-ing. Linguistic Inquiry 19(4). 611–634. Google Scholar
Packard, Jerome L. 2000. The morphology of Chinese. Cambridge: Cambridge University Press. Google Scholar
Paster, Mary. 2005. Pulaar verbal extensions and phonologically driven affix order. In Geert Booij & Jaap Van Marle (eds.), Yearbook of Morphology 2005, 155–199. Dordrecht: Springer. Google Scholar
Paul, Waltraud. 2005. Adjectival modification in Mandarin Chinese and related issues. Linguistics 43(4). 757–793. Google Scholar
Paul, Waltraud. 2006. Zhu Dexi’s two classes of adjectives revisited. In Christoph Anderl & Halvor Eifring (eds.), Studies in Chinese language and culture. Festschrift in honour of Christoph Harbsmeier on the occasion of his 60th birthday, 303–315. Oslo: Hermes Academic Publishing. Google Scholar
Paul, Waltraud. 2010. Adjectives in Mandarin Chinese: The rehabilitation of a much ostracized category. In Patricia Cabredo-Hofherr & Ora Matushansky (eds.), Adjectives: Formal analysis in syntax and semantics, 115–151. Amsterdam & Philadelphia: John Benjamins. Google Scholar
Postal, Paul M. 1969. Anaphoric islands. In Robert Binnick, et al. (eds.), Papers from the fifth regional meeting of the Chicago Linguistic Society, 209–239. Chicago: Chicago Linguistic Society. Google Scholar
Prince, Alan & Paul Smolensky. 1993/2004. Optimality Theory: Constraint interaction in generative grammar. Malden, MA: Blackwell. [Revision of 1993 technical report, Rutgers University Center for Cognitive Science. Available on Rutgers Optimality Archive, ROA-537.]. Google Scholar
Pullum, Geoffrey & Arnold Zwicky. 1991. A misconceived approach to morphology. In Dawn Bates (ed.), Proceedings of the Tenth West Coast Conference on Formal Linguistics, 387–398. Stanford, CA: CSLI Publications. Google Scholar
Rainer, Franz & Soledad Varela. 1992. Compounding in Spanish. Rivista di Linguistica 4. 117–142. Google Scholar
Rice, Keren. 2000. Morpheme order and semantic scope: Word formation in the Athapaskan verb. Cambridge: Cambridge University Press. Google Scholar
Riehemann, Suzanne Z. 2001. A constructional approach to idioms and word formation. Stanford, CA: Stanford University dissertation. Google Scholar
Ross, John Robert. 1972. Doubl-ing. Linguistic Inquiry 3. 61–86. Google Scholar
Sag, Ivan, Thomas Wasow & Emily M. Bender. 2003. Syntactic theory: A formal introduction. Stanford, CA: CSLI. Google Scholar
Schlücker, Barbara & Matthias Hüning. 2009. Compounds and phrases. A functional comparison between German A + N compounds and corresponding phrases. Italian Journal of Linguistics/Rivista di Linguistica 21(1). 209–234. Google Scholar
Selkirk, Elisabeth. 1982. The syntax of words. Cambridge, MA: MIT Press. Google Scholar
Selkirk, Elisabeth. 1984. Phonology and syntax: The relation between sound and structure. Cambridge, MA: MIT Press. Google Scholar
Shen, Jiaxuan. 1997. Xingrongci jufa gongneng de biaoji moshi [The marking model of the syntactic functions of adjectives]. Zhongguo Yuwen 4. 242–250. Google Scholar
Shen, Jiaxuan. 1999. Buduicheng he biaojilun [Asymmetry and the markedness theory]. Jiangxi: Jiangxi Education Press. Google Scholar
Spencer, Andrew. 1991. Morphological theory: An introduction to word structure in generative grammar. Oxford: Blackwell. Google Scholar
Sproat, Richard. 1993. Looking into words. In Sharon Hargus & Ellen Kaisse (eds.), Studies in lexical phonology, 173–195. San Diego: Academic Press. Google Scholar
Sproat, Richard & Chilin Shih. 1988. Prenominal adjectival ordering in English and Mandarin. In James Blevins & Juli Carter (eds.), Proceedings of the North East Linguistic Society 18, vol. 2, 465–489. Amherst, MA: GLSA. Google Scholar
Sproat, Richard & Chilin Shih. 1991. The cross-linguistic distribution of adjective ordering restrictions. In Carol Georgopoulos & Roberta Ishihara (eds.), Interdisciplinary approaches to language: Essays in honor of S.-Y. Kuroda, 565–593. Dordrecht: Kluwer Academic Publishers. Google Scholar
Sproat, Richard & Chilin Shih. 1993. Why Mandarin morphology is not stratum-ordered. In Geert Booij & Jaap Van Marle (eds.), Yearbook of Morphology 1992, 185–217. Dordrecht: Kluwer Academic Publishers. Google Scholar
Tang, Ting-Chi. 1988. Guoyu xingrongci de chongdie guilü [Reduplication rules for adjectives in Mandarin Chinese]. In Ting-Chi Tang (eds.), Hanyu cifa jufa lunji [Studies on Chinese morphology and syntax], 29–57. Taipei: Student Book Company. Google Scholar
Xiandai Hanyu Cidian [Modern Chinese Dictionary] (2000) The Linguistics Institute of the Chinese Social Sciences Academy (eds.). Beijing: Shangwu yinshuguan. Google Scholar
Xing, Fuyi. 1997. “hen shu-nü” zhi lei shuofa yuyan wenhua beijing de sikao [Reflections on hen shu-nü from linguistic and cultural perspectives]. Yuyan yanjiu 2. 1–10. Google Scholar
Xu, Jianhua & Fuhua Liu. 1999. Danyin xingrongci dingyu de he-zhi xi-zhi yu yuxu wenti [Monosyllabic attributive adjectives and the problem of word order]. Yuyan jiaoxue yu yanjiu 3. 97–106. Google Scholar
Xu, Zheng. 2006. On the formation of adjective-noun combinations in Mandarin Chinese. In Qian Gao (ed.), Proceedings of the Seventeenth North American Conference on Chinese Linguistics (NACCL-17), 463–482. Los Angeles: GSIL Publications. Google Scholar
Yip, Moira. 1998. Identity avoidance in phonology and morphology. In Steven G. Lapointe, Diane K. Brentari & Patrick M. Farrell (eds.), Morphology and its relation to phonology and syntax, 216–246. Stanford, CA: CSLI. Google Scholar
Yu, Alan C. L. 2003. The morphology and phonology of infixation. Berkeley, CA: University of California dissertation. Google Scholar
Yu, Alan C. L. 2007. A natural history of infixation. Oxford: Oxford University Press. Google Scholar
Zhang, Guoxian. 2000. Xiandai hanyu xingrongci de dianxing tezheng [The characteristics of adjectives in modern Chinese]. Zhongguo Yuwen 5. 447–458. Google Scholar
Zhongguo Wenzi Gaige Weiyuanhui Yanjiu Tuiguang Chu [Chinese Language Reform Committee Research and Popularization Office]. 1959. Putonghua sanqian changyongci biao [3,000 commonly used words in Standard Chinese]. Beijing: Wenzi Gaige Chubanshe. Google Scholar
Zhu, Dexi. 1956. Xiandai hanyu xingrongci yanjiu [A study of modern Chinese adjectives]. Yuyan yanjiu 1. 83–112. Google Scholar
Zhu, Dexi. 1980. Xiandai hanyu yufa yanjiu [Research on modern Chinese grammar]. Beijing: Shangwu yinshuguan. Google Scholar
Zwicky, Arnold M. & Geoffrey K. Pullum. 1986. The principle of phonology-free syntax: Introductory remarks. Working Papers in Linguistics 32. 63–91. Columbus, OH: Ohio State University. Google Scholar
I do not aim at giving a crosslinguistically valid definition of adjectives. Dixon (2004) remarks that adjectives can be recognized for every human language while the definition or behavior of adjectives varies among languages. Hence, I follow the traditional and widely-accepted definition of Chinese adjectives in Zhu (1956; 1980). For further discussion of adjectives, see Croft (1991); Bhat (1994); Larson and Segal (1995); Baker (2003); and Paul (2010).
In Shen’s (1997) database there are 177 nominal expressions with adjectives that contain modifying elements. Of the 177 instances, 35 contain quantifiers such as duo ‘many’ and shao ‘few’ as prenominal modifiers. Shen notes that these quantifiers differ from adjectives in several ways when used as modifiers. See also Zhu (1956) and Lu (1985) for relevant discussion. Hence, I do not take the 35 cases into consideration so that there remain 142.
Putonghua sanqian changyongci biao ‘3,000 commonly used words in Standard Chinese’ (1959) was compiled by Zhongguo Wenzi Gaige Weiyuanhui Yanjiu Tuiguang Chu ‘Chinese Language Reform Committee Research and Popularization Office’. It is one of the most authoritative sources for commonly used words in Standard Chinese. I refer to it, following Duanmu (2000; 2007).
Zhang (2000) does not provide any information about the size of his corpus.
Duanmu (2000) suggests that [A N] with disyllabic adjectives may be derived from [A de N] via a rule of de-deletion. The questions that remain unanswered include, for example, why only some [A de N] with disyllabic adjectives can undergo de-deletion, and why [A de N] with adjectives that contain modifying elements cannot undergo de-deletion.
The examples in (4) – (8) are from Xu (2006).
Booij (2010: 176) remarks that the adjective in Dutch [A N] compounds is strictly simplex. German also imposes the monomorphemic constraint, but with some exceptions such as adjectives in –al (National-staat ‘nation state’), -iv (Suggestive-frage ‘suggestive question’), and –ig (Niedrig-wasser ‘low tide’). See also Schlücker and Hüning (2009: 212–213), who remark that German adjectives with foreign suffixes like -al or -iv are possible in [A N] compounds.
An expression is conventionalized if it has been chosen in a language community to denote a particular concept (Booij 2010).
If we only consider the criterion based on naming and description, it is also possible that a nonconventionalized Chinese [A N] with a free nominal head stands in between a word and a phrase, assuming that there is a continuum between word and phrase. Based on the monomorphemic constraint, lexical gaps, the tests of modification by degree adverbs, XP substitution, and conjunction reduction, I argue that it is a word.
See Yip (1998) for a comprehensive discussion of avoidance of repetition of identical or similar forms in various languages. In English, it is well known that sequences of words ending in -ing are disliked (Ross 1972; Milsark 1988; Pullum and Zwicky 1991; Yip 1998), e.g., *John was starting reading the book (Yip 1998: 232).
This case shows that the monomorphemic constraint on [A N] compounds overrides the prosodic constraint that requires avoidance of repetition of identical forms. The monomorphemic constraint rules out [A N] with complex adjectives and the prosodic constraint forbids de to be repeated. This can be accounted for under the ranking M(orphology) ≫ P(rosody) in Optimality Theory (Prince and Smolensky 1993/2004; McCarthy and Prince 1993). See Sproat and Shih (1996) and Yu (2003; 2007) for detailed discussions of this ranking in both Chinese and other languages.
See Paul (2010) for other criticisms of Huang’s semantic type-matching constraint.
Zhu (1980) remarks that [A N] like *bai shou ‘white hand’ and *zang tang ‘dirty candy’ are ungrammatical while bai zhi ‘white paper’ and zang yifu ‘dirty clothes’ are fine. However, Shen (1997) argues that expressions like bai shou ‘white hand’ can be grammatical in the right context. Additionally, it is not clear why food names such as tang ‘candy’ cannot be modified by zang ‘dirty’ because zang xigua ‘dirty water melon’ and zang pingguo ‘dirty apple’ are fine. *Zang tang ‘dirty candy’ sounds odd probably because zang ‘dirty’ and tang ‘candy’ share a similar rhyme so that *zang tang ‘dirty candy’ violates the universal constraint that forbids identical or similar forms to be adjacent (Leben 1973; Yip 1998).
See Xu (2006) for more restrictions on Chinese [A N] that are characteristic of word formation.
See Booij (2010: 178–179) for a list of lexicalized [A N] phrasal constructions in various European languages.
Booij (2009) argues that the Lexical Integrity Principle should be formulated in such a way that syntactic rules are allowed to see/access the word-internal structure. That is, in such an understanding of Lexical Integrity, parts of a word can be anaphorically accessible. See the discussion in Section 9.
Xing (1997) discusses some Chinese denominal words that are modifiable by degree adverbs, e.g., hen shu-nü (very noble-lady) ‘very much like a noble lady’. These denominal words usually express a metaphoric sense ‘being like x, showing the properties of x’. It is beyond the scope of this article to discuss in detail the contexts where nouns have such metaphoric senses and are modifiable by degree adverbs.
See Chaves (2008) for arguments that the Chinese examples in question undergo conjunction reduction rather than simple coordination of either heads or nonheads.
Many other morphological frameworks such as Network Morphology also adopt inheritance hierarchy and default inheritance (see e.g., Brown and Hippisley 2012).