1 Word order of utterance particles

Utterance particles in Chinese, which are also known as “sentence-final particles”, are functional words that occur in the sentence-final position, expressing some grammatical meanings and pragmatic information, such as the speaker’s attitude. For example, in Mandarin, the interrogative 嗎 ma in 1 and the assertive 吧 ba in 2 are typical utterance particles that convey some pragmatic information. In Cantonese 喎 wo in 3 with the low rising tone ([wɔ13] in IPA) is a hearsay evidential utterance particle.

  1. (1)


    • ta__kan__shu__ma?

    • he__read__book__MA

    • Did he read?

  2. (2)


    • ta__kan__shu__ba.

    • he__read__book__BA

    • He read, I believe.

  3. (3)


    • keoi__tai__syu__wo.

    • he__read__book__WO

    • It is said that he read.

In the literature of generative grammar, it has been widely assumed that utterance particles in Chinese are heads of functional projections in syntax, such as CP (Lee 1986, Cheng 1991, among many others), split CP in a cartographic periphery domain (Li 2006), or some functional phrases beyond CP (Tang 邓思颖 2010, Tang 2015).

There are two ways to derive the sentence-final word order of the utterance particles in Chinese. One way is to claim that the phrase containing the utterance particle is head final, abbreviated as “XP” in 4, in which the utterance particle is the head and the clause is its complement. A theory-internal problem is the inconsistency in setting the value of the head parameter in Chinese, given that Chinese is head initial in the clausal domain (Huang 1982).

  1. (4)

    [XP clause [X particle ]]

Another possibility is to claim that the utterance particle is head initial underlyingly while the clause in the complement undergoes movement to the specifier of the phrase (Tang 1998), as in 5, in the spirit of Kayne (1994), deriving the apparently sentence-final order in Chinese. Although the right word order can be derived, the motivation of moving the clause seems puzzling.

  1. (5)

    [XP clause i [X’ [X particle ] t i ]]

In this paper, it is argued that the apparent head final word order of the utterance particles in Chinese is derived by a generalized syntactic schema that may cover both predicative utterance particles and non-predicative utterance particles. It will be further conjectured that utterance particles and other sentence-final expressions are not heads, contra the analyses in 4 and 5.

2 Predicative utterance particles

There are at least six utterance particles that are used frequently in spoken Mandarin, such as 了 le, 呢 ne, 來著 laizhe, 嗎 ma, 吧 ba, and 啊 a (and their variants) (Zhu 朱德熙 1982). Among these utterance particles, some of them have a verbal origin. For instance, the temporal particles了 le in 6 (conveying some aspectual meaning) and 來著 laizhe in 7 (conveying recent past) were undoubtedly derived from the verbs 了 liao ‘finish’ and 來 lai ‘come’, respectively (Ota 太田辰夫 1987, Cao 曹广顺 1995, among others). The evaluative mood of the utterance particle 吧 ba in 8 is inherited from the verb 罷 ba ‘stop, finish’ (Ota 太田辰夫 1987).

  1. (6)


    • xia__yu__le.

    • fall__rain__LE

    • It has rained.

  2. (7)


    • xia__yu__laizhe.

    • fall__rain__LAIZHE

    • It rained.

  3. (8)


    • ni__qu__ba!

    • you__go__BA

    • It is suggested that you go.

It seems that the predicative function of these particles is still preserved in modern Mandarin. For example, the realization of the event of raining in 6 is signaled by the predicative utterance particle 了 le (cf. “There is an event of raining and this event of raining has been realized”), which inherits the aspectual meaning from the verb 了 liao ‘finish’. The temporal meaning of the event of raining in 7 is associated with 來著 laizhe (cf. “There was an event of raining and this event of raining had come to its existence”), which inherits the existential and directional meaning from the deictic verb 來 lai ‘come’ and the affirmative meaning from 著 zhuo ‘order’. The evaluative mood of 吧 ba in 8 can be interpreted as a solicitation of approval or agreement from the hearer that may settle certain debates or uncertainties (cf. “certain uncertainty of your going can be ceased”), which inherits the meaning of settling down from the verb 罷 ba ‘stop’. It is highly likely that these three particles in Mandarin, namely, 了 le, 來著 laizhe, and 吧 ba, all have a verbal origin, which can be traced back to verbs like 了 liao ‘finish’, 來 lai ‘come’, 著 zhuo ‘order’, and 罷 ba ‘stop’.

The counterparts of the Mandarin utterance particles that have a verbal origin can also be found in Cantonese, such as 喇 laa ([la33]) (likely cognate with Mandarin le) and 嚟 lei ([lei21]) (cognate with Mandarin 來著 laizhe). For the Cantonese utterance particles that cannot be seen in Mandarin, some of them also have a verbal origin and may play a role in predication in Cantonese, such as 添 tim ([tim53]), 啩 gwaa [kwa33]), and 喎 wo.

The utterance particle 添 tim ‘too’ has an emphatic function and was historically derived from the verb 添 tim meaning ‘to add’. In fact, 添 tim ‘add’ can still be used as a verb in modern Cantonese that means “to add”, as in 10. Example 9 with 添 tim could be paraphrased as “There is an event of raining and this event happens additionally/unexpectedly” (out of the expectation of the speaker).

  1. (9)


    • lok__jyu__tim!

    • fall__rain__TIM

    • It is raining unexpectedly.

  2. (10)


    • keoi__tim-zo__jat__wun__faan.

    • he__add-PERF__one-CL__rice

    • He ate one more bowl of rice.

The utterance particle 啩 gwaa expresses the speaker’s uncertainty and was historically derived from the verb 估 gu (or romanized as gwu) ‘guess’ (Chao 1947). In modern Cantonese, 估 gu ‘guess’ is still used as a verb, as in 12. Example 11 with 啩 gwaa could be paraphrased as “That it will rain tomorrow is what I guess”. This predicative utterance particle conveys some modality meaning.

  1. (11)


    • lok__jyu__gwaa.

    • fall__rain__GWAA

    • It is raining. I suppose.

  2. (12)


    • ngo__gu__tingjat__wui__lok__jyu.

    • I__guess__tomorrow__will__fall__rain

    • I guess it will rain tomorrow.

wo in 3 with the low rising tone is a hearsay evidential utterance particle (Tang 2015) and was historically derived from the verb 話 waa [wa22] ‘say’ (Chao 1947, Leung 2006). In modern Cantonese, 話 waa ‘say’ can be used as a verb, as in 14. Example 13 with 喎 wo could be paraphrased as “It will rain tomorrow and this piece of information is told you”. This particular utterance particle that was derived from the verb of saying conveys the speech act being performed by the speaker.

  1. (13)


    • lok__jyu__wo.

    • fall__rain__WO

    • It is heard that it is raining.

  2. (14)


    • keoi__waa__tingjat__wui__lok__jyu.

    • he__say__tomorrow__will__fall__rain

    • He said it will rain tomorrow.

3 Predicative core of polysyllabic utterance particles

In the literature, the focus of the study of utterance particles in Chinese is solely on the monosyllabic ones, such as those in 1 to 3 (Hu 胡明扬 1981, Li and Thompson 1981, Zhu 朱德熙 1982, Li 2006, among others in Mandarin, and Kwok 1984, Luke 1990, Law 1990, Fung 2000, among others in Cantonese). On the contrary, polysyllabic ones are almost neglected.

As a matter of fact, utterance particles in Chinese are not all monosyllabic. Some of them are apparently polysyllabic, for example, Mandarin 來著 laizhe in 7, which can be decomposed into 來 lai ‘come’ and 著 zhe. The verbal origin of 來 lai ‘come’ is not debatable as it is still being used as a verb in modern Chinese. For 著 zhe, it was used as a verb 著 zhuo meaning ‘to order’ and has been grammaticalized as an affirmative utterance particle that was added to 來 lai in the Qing Dynasty to reinforce the affirmative meaning of the clause (Sun 孙锡信 1999)a.

Besides 來著 laizhe that has been discussed in the literature, other polysyllabic utterance particles seem to receive less attention. In terms of syllables, polysyllabic utterance particles in Mandarin can be classified into disyllabic ones, such as 也好 yehao, 也罷 yeba, 好了 haole, and 罷了 bale, and trisyllabic ones, such as 就是了 jiushile, as documented in Zhang et al. 张斌主编 (2010: 250)b. In 15, 也好 yehao is formed by two morphemes, namely 也 ye ‘also’ and 好 hao ‘be good’, which is used to indicate that the speaker agrees with the suggestion denoted by the clause. In 16, 也罷 yeba is formed by two morphemes, namely 也 ye ‘also’ and 罷 ba ‘stop, finish’, which is used to express forbearance of the speaker. In 17, 好了 haole is formed by two morphemes, namely 好 hao ‘be good’ and a temporal particle 了 le, which is used to express agreement and judgment of the speaker. In 18, 罷了 bale is formed by two morphemes, namely 罷 ba ‘stop, finish’ and the temporal particle 了 le, which means “that’s all, nothing else”. In 19, 就是了 jiushile is the only trisyllabic utterance particle documented in Zhang et al. 张斌主编 (2010), which is formed by three morphemes, namely the adverb 就 jiu ‘exactly, precisely’, 是 shi ‘be’, and the temporal particle 了 le. This trisyllabic utterance particle 就是了 jiushile is used to give force to the statement denoted by the clause and may mean “that’s all, that’s right, just”. In addition to Zhang et al. 张斌主编 (2010), these five words are also explicitly classified as “particles” (i.e. zhuci in Chinese) in the sixth edition of 现代汉语词典 Dictionary of Modern Chinese, an “authoritative” Chinese dictionary.

  1. (15)


    • shuo-shuo__yehao.

    • say-say__YEHAO

    • It is not a bad idea to talk.

  2. (16)


    • ni__bu__xiang__qu__yeba.

    • you__not__want__go__YEBA

    • It is fine if you do not want to go.

  3. (17)


    • ni__qu__haole.

    • you__go__HAOLE

    • You had better go.

  4. (18)


    • wo__zhishi__shuo-shuo__bale.

    • I__only__say.say__BALE

    • I say it casually only.

  5. (19)


    • ni__fangxin__jiushile.

    • you__take.it.easy__JIUSHILE

    • Don’t worry. Take it easy.

Among the polysyllabic utterance particles we have examined, 好 hao ‘be good’, 罷 ba ‘stop, finish’, and 是 shi ‘be’ are all verbal and can be used as “real” predicates. In 20, 好 hao ‘be good’ is used as a stative predicate, predicated of the subject 他 ta ‘he’. In modern Mandarin, 罷 ba ‘stop, finish’ is a bound morpheme and may form a compound by taking a noun, such as 罷工 ba-gong ‘strike’ (lit. stop-work) in 21, in which the noun 工 gong ‘work’ is regarded as the object of the verb 罷 ba ‘stop, finish’, forming a VO compound. The verb 是 shi ‘be’ is a copula in Mandarin, as in 22.

  1. (20)


    • ta__hen__hao.

    • he__very__good

    • He is very good.

  2. (21)


    • gongren__ba-gong.

    • worker__stop-work

    • Workers are striking.

  3. (22)


    • ta__shi__xuesheng.

    • he__be__student

    • He is a student.

It is proposed that so-called polysyllabic utterance particles in Mandarin can be decomposed into more than one morpheme, one of which is regarded as the predicative core. For example, 好了 haole is formed by a predicative core 好 hao ‘good’ followed by another particle 了 le that also has a predicative origin while 就是了 jiushile can be decomposed into three morphemes “就 jiu + 是 shi + 了 le” with a predicative core 是 shi ‘be’ modified by 就 jiu ‘then’ and followed by another particle 了 le.

Although it is well-known in the literature that Cantonese is a language that has rich particles, much attention was paid to those monosyllabic utterance particles only. In Cheung 張洪年 (1972), two disyllabic utterance particles in Cantonese are identified, namely 罷啦 baalaa ([pa35 la55]) in 23 and 係啦 hailaa ([hɐi35 la55]) in 24, and are analyzed as “compound particles”. According to Cheung 張洪年 (1972), “compound particles” are formed by more than one morpheme as an inseparable unit (see also Wong 黃卓琳 2014). In 23, 罷啦 baalaa is formed by two morphemes 罷 baa ‘stop, finish’ and a temporal particle 啦 laa. Its morphology is somewhat similar to 罷了 bale in Mandarin. Unlike Mandarin 罷了 bale, Cantonese 罷啦 baalaa is mainly used to form an imperative (Tang 鄧思穎 2009), having a function of advocating or advising. Cantonese 係啦 hailaa in 24 seems to be the counterpart of Mandarin 就是了 jiushile, which is also used to give force to the statement denoted by the clause.

  1. (23)


    • nei__heoi__baalaa.

    • you__go__BAALAA

    • You had better go.

  2. (24)


    • ngo__heoi__hailaa.

    • I__go__HAILAA

    • All right, I will go.

There are two more examples, namely 定啦 dinglaa and 得 dakgaa, that can be analyzed as polysyllabic utterance particles in Cantonese (Wong 黃卓琳 2014)c. In 25 定啦 dinglaa ([tɪŋ35 la55]) denotes an epistemic meaning, which can be decomposed into a predicative core ding ‘stable’ and the temporal particle 啦 laa. In 26 得 dakgaa ([tɐk55 ka35]) contributes some conditional meaning and can be decomposed into a predicative core 得 dak ‘possible, able’ and an another utterance particle gaa that is likely to be formed by merging 嘅 ge ([kɛ33]) and 啊 aa ([a35]). A characteristic that these two polysyllabic utterance particles share is the existence of a predicative core.

  1. (25)


    • keoi__wui__heoi__dinglaa.

    • he__will__go__DINGLAA

    • It is likely that he will go.

  2. (26)


    • keoi__hang__heoi__dakgaa.

    • he__willing__go__DAKGAA

    • Something would be possible only if he is willing to go.

Interestingly, a predicative element can be found in all these polysyllabic utterance particles in Mandarin and Cantonese. Considering the facts in Mandarin and Cantonese, we may conclude that the polysyllabic utterance particles in Chinese are all formed by a predicative core, which may be either preceded by an adverbial element or followed by another utterance particle that may also have a predicative origin. The predicative core in the examples of polysyllabic utterance particles include 好 hao ‘be good’ in Mandarin, 罷 ba ‘stop, finish’ in Mandarin and its counterpart 罷 baa ‘stop, finish’ in Cantonese, 是 shi ‘be’ and its counterpart 係 hai ‘be’ in Cantonese, 定 ding ‘stable’ in Cantonese, and 得 dak ‘possible, able’ in Cantonese.

4 Syntax of predicative utterance particles

The utterance particles that were historically derived from verbs or predicative elements are still predicative in modern Chinese in some sense, serving as “grammaticalized predicates”. This “grammaticalized predicate” and its preceding clause may form a “serial verb” construction in a broad sense (cf. Li and Thompson 1981), having a syntactic structure like 27.


In 27, F is regarded as a conjunction that takes two conjuncts, XP and YP. XP, the internal conjunct, is a position reserved for the utterance particles while YP, the external conjunct, is a position for the main clause that precedes the utterance particles. If the utterance particle is predicative, XP could be regarded as a secondary predicate and F functions as a coordinating conjunction that links the main clause and the secondary predicate, somewhat like and in Englishd. In terms of their function, some secondary predicates are predicated of the main clause in YPe, which may express judgments and evaluative meanings, and some secondary predicates indicate the temporal sequence of events, the resultant state of the event of the main clause, or the speech act being performed by the speaker, from which the aspectual, temporal, and modality meanings and other pragmatic functions are derived.

Predicative utterance particles should be different from the canonical lexical predicates although both of them seem to be predicates. The predicative element, i.e. the root of the predicative utterance particles, is more “bleached”, losing some lexical meanings. For example, the evaluative meaning of 吧 ba could have inherited the meaning of settle down from the verb 罷 ba ‘stop’. The lexical meaning of 罷 ba ‘stop’ has been “bleached” and it only expresses the judgment of the speaker. Some of the predicative utterance particles seem to be still quite lexical, for instance, 好 hao (originally derived from 好 hao ‘good’) in 好了 haole, which could be analyzed as a “defective” predicate, still straddling the line between lexical and functional.

In 8, repeated in 28, the utterance particle 吧 ba is regarded as a secondary predicate in the complement of F, i.e. in the XP position, and the main clause is the specifier of FP, i.e. in the YP position, as indicated in 29, which could mean something like “You should go and this is what I suggest to settle certain uncertainty”. The function of 吧 ba in XP is associated with the evaluative meaning expressed by the second conjunct “this is what I suggest to settle certain uncertainty”. Other irrelevant details are omitted.

  1. (28)


    • ni__qu__ba!

    • you__go__BA

    • It is suggested that you go.


Polysyllabic utterance particles, however, can be decomposed. For instance, the Mandarin disyllabic utterance particle 好了 haole in 17, repeated in 30, is decomposed into two morphemes, namely 好 hao ‘good’ and 了 le. For the derivation, F1 is first merged with 好 hao ‘good’ and second merged with the main clause 你去 ni qu ‘you go’, forming the structure in 31. This is the first step.

  1. (30)


    • ni__qu__haole.

    • you__go__HAOLE

    • You had better go.


For the second step, a different functional category F2 is chosen and first merged with 了 le and second merged with FP1, as in 32, deriving the word order “clause + 好 hao + 了 le”. FP1 “clause + 好 hao” is embedded in FP2. The same analysis can apply to the disyllabic utterance particle 罷啦 baalaa in Cantonese.


The Mandarin disyllabic utterance particle 好了 haole can be followed by another utterance particle, such as 吧 ba in 33f. To derive the word order “clause + 好 hao + 了 le + 吧 ba” in 33, I assume that an additional functional category F3 is introduced and merged with 吧 ba and FP2 accordingly, as in 34. In this configuration, FP2 is further embedded. According to the analysis presented here, 33 could roughly mean something like “you should go and this option is the best and this message has been conveyed and this is what I suggest to the hearer to settle certain uncertainty”.

  1. (33)


    • ni__qu__haole__ba.

    • you__go__HAOLE__BA

    • Perhaps you should go.


For the syntax of the Mandarin trisyllabic utterance particle 就是了 jiushile in 19, as repeated in 35, I assume that it should be decomposed into a predicative core 是 shi ‘be’ followed by another predicative utterance particle 了 le and preceded by an adverbial 就 jiu ‘then’. The adverbial 就 jiu ‘then’ is adjoined to the verb 是 shi ‘be’. F1 is then merged with XP headed by 是 shi ‘be’. After the derivation of FP1 is completed, F2 is introduced and merged with 了 le, deriving the word order “clause + 就 jiu + 是 shi + 了 le” on the surfaceg.

  1. (35)


    • ni__fangxin__jiushile.

    • you__take.it.easy__JIUSHILE

    • Don’t worry. Take it easy.


Along the lines of the discussion in this paper, the syntax of polysyllabic utterance particles is treated on a par with that of the “particle cluster”, i.e. stacking of utterance particles in a specific order, such as “了 le + 吧 ba” in 37. As shown in the configuration in 38, the clause with 了 le is embedded in FP2, deriving the word order “了 le + 吧 ba” on the surface. The utterance particles 了 le and 吧 ba are complements of F1 and F2, respectively, reminiscent of the status of 好 hao and 了 le in the disyllabic utterance particle 好了 haole in 32.

  1. (37)


    • xia__yu__le__ba.

    • fall__rain__LE__BA

    • Perhaps it has rained.


The above considerations lead to the conclusion that the polysyllabic utterance particles are in fact a subgroup of the particle cluster and should be treated as “compounds in disguise”h. In other words, describing these utterance particles as “polysyllabic” would be quite misleading.

5 Syntax of non-predicative utterance particles

Examples 39, 40, and 41 are yes-no questions, also known as “VP-Neg questions”i. 39 is from Mandarin. The negation 沒有 meiyou ‘not’ negates the completion of an event (Li and Thompson 1981). 不 bu ‘not’ provides a neutral negation (Li and Thompson 1981) and may form a VP-Neg question, such as 40, in some northern dialects. 41 is a Cantonese example. The negation 未 mei ([mei22]) ‘not yet’ may also form a VP-Neg questionj.

  1. (39)


    • ta__kan-le__shu__meiyou?

    • he__read-PERF__book__not

    • Did he read?

  2. (40)


    • ta__xianzai__shenti__hao__bu?

    • he__now__body__good__not

    • Is his health fine?

  3. (41)


    • keoi__tai-zo__syu__mei?

    • he__read-PERF__book__not

    • Has he read yet?

The syntactic schema in 27 also applies to yes-no questions, for example, 42. The functional category F connects two conjuncts, i.e. the main clause in the external conjunct YP and the negation with the identical VP in the internal conjunct XP. Deletion takes place and the identical VP in the internal conjunct is elided, leaving the negation on the surface.


In Mandarin, the utterance particle 嗎 ma in 1, as repeated in 43, functions as an interrogative marker that forms a yes-no question.

  1. (43)


    • ta__kan__shu__ma?

    • he__read__book__MA

    • Did he read?

Note that although both 嗎 ma in 43 and 沒有 meiyou ‘not’ in 39 can form yes-no questions and look quite similar on the surface, they behave differently. Firstly, the adverb 到底 daodi ‘really’ expresses the speaker’s impatience or irritation in not getting the answer to a question (Law 2008), which can go with 沒有 meiyou in 44, but not 嗎 ma ‘not’ in 45k.

  1. (44)


    • daodi__ta__lai-le__meiyou?

    • really__he__come-PERF__not

    • Did he really come?

  2. (45)


    • daodi__ta__lai-le__ma?

    • really__he__come-PERF__MA

Secondly, the discrepancy between 46 and 47 shows that ma is compatible with a negative clause while meiyou ‘not’ is not (Shi 石定栩 2007)l.

  1. (46)


    • ta__mei__qu__Beijing__ma?

    • he__not__go__Beijing__MA

    • Isn’t it that he did not go to Beijing?

  2. (47)


    • ta__mei__qu__Beijing__meiyou?

    • he__not__go__Beijing__not

Although 嗎 ma was from a negation historically (Ota 太田辰夫 1987) and may have a similar syntactic structure as what the interrogatives with 沒有 meiyou ‘not’ have, it is assumed that VP-ellipsis was a historical process that took place diachronically in interrogatives with 嗎 ma. In modern Chinese, 嗎 ma has been fully grammaticalized to be an utterance particle that purely and simply conveys the speech act being performed by the speaker. The complement of F in interrogatives with 嗎 ma contains nothing but the grammaticalized particle 嗎 ma underlyingly, as in 48. In other words, the so-called elided VP does not exist in the numeration in the first place. Historically 嗎 ma was derived from a verbal negation 無 wu ‘not’ (Ota 太田辰夫 1987) and its verbal properties may still be preserved in modern Chinese, functioning as a secondary predicate in the internal conjunct in 48.

  1. (48)

    [[Subject VP] F [ma]]

On the contrary, VP-ellipsis in VP-Neg questions is a synchronic process, which perhaps takes place in the phonological component, i.e. PF, and the elided VP is in the derivation from the numeration all the way to the semantic component, i.e. LF, in synchronic syntax. What F conjoins is the positive clause in the external conjunct and the negative clause in the internal conjunct. 49 is the underlying structure for VP-Neg questions. According to 49, the underlying structure for 39 would be 50. F connects the positive clause and the negative clause, which would mean something like “Did he read or not read?”.

  1. (49)

    [[Subject VP] F [not ]]

  2. (50)

    [他看了書] F [沒有看書]

    • [ta__kan-le__shu ] F [ meiyou__kan__shu]

    • he__read-PERF__book__F__not__read__book

    • Did he read?

Such VP-Neg questions are regarded as disjunctive questions and F is interpreted as a disjunctive conjunction, just like or in Englishm. After deletion applies, the identical VP in the internal conjunct is elided and the negation remains, deriving the word order of yes-no questions “clause + negation”. VP-Neg questions have two logically possible answers, one positive and one negative, which can be denoted unambiguously by syntax. With the adverb 到底 daodi ‘really’, the speaker of 44 impatiently is asking for a yes or no answer to the question by choosing either the positive answer or the negative answer offered by the configuration in 49. On the other hand, 45 is incompatible with the adverb 到底 daodi ‘really’ because no such choices, i.e. positive or negative answers, can be offered by the configuration in 48.

In VP-Neg questions, if the external conjunct is a negative clause, the negative clause would be repeated and negated in the second conjunct, as in 51. As the negative clause cannot be negated again, the ungrammaticality of 47 is expected, exhibiting the phenomenon of reversed polarity from the preceding clause.

  1. (51)

    *[他沒去北京] F [沒有沒去北京]

    • [ta__mei__qu__Beijing] F [meiyou__mei__qu__Beijing]

    • he__not__go__Beijing__F__not__not__go__Beijing

The acceptability of 46, however, suggests that yes-no questions with the utterance particle 嗎 ma should have a different syntactic structure, as in 52. The negative clause is not copied and will be not negated. The utterance particle 嗎 ma is then reinterpreted as a secondary predicate to express the speech act in modern Chinese.

  1. (52)

    [ta__mei__qu__Beijing] F [ma]

    • he__not__go__Beijing__F__MA

    • Isn’t it that he did not go to Beijing?

The syntactic schema may be extended to sentences with the focus utterance particle 呢 ne in Mandarin, as in 53. For the historical source of 呢 ne, it could be from 聻 ni ‘that’ in pre-modern Chinese (Ota 太田辰夫 1987), preserving a demonstrative function, or from 裏/里 li ‘place’ (Ota 太田辰夫 1987), conveying a locative meaning. Although the origin is still debatable, a similarity that both 聻 ni ‘that’ and 裏/里 li ‘place’ share is anaphoricity, which 呢 ne in modern Chinese inherits. The anaphoric function of 呢 ne in 53 is still preserved, which is used to call the hearer’s attention to the information denoted by the clause 他很高 ta hen gao ‘he is very tall’.

  1. (53)


    • ta__hen__gao__ne!

    • he__very__tall__NE

    • Listen, he is very tall.

The anaphoric function of 呢 ne is quite similar to 這樣子 zheyangzi ‘this appearance’ in the sentence-final position in 54, a popular colloquial expression frequently used by the young generation, particularly in colloquial Mandarin spoken in Taiwan. This expression could be paraphrased as “or something like that” in English. In some occasion 這樣子 zheyangzi ‘this appearance’ could be shortened as 醬子 jiangzi (literally meaning ‘sauce, paste’) in the spoken language and on the internet.

  1. (54)


    • wo__dagai__liu__fenzhong__jiu__shuizhao__le__zheyangzi/jiangzi.

    • I__probably__six__minute__then__sleep__LE__this.appearance

    • Probably I will fall asleep in six minutes (or something like that).

The anaphoric expressions, such as 呢 ne and 這樣子 zheyangzi, are kind of repetition of the main clause that precedes them and they are used to reinforce the meaning denoted by the main clause. I assume that the main clause and such anaphoric expressions are conjoined by F, as in 55, and are in YP and in XP, respectively.


The syntax of both predicative and non-predicative utterance particles in Chinese can be captured by the same syntactic schema in 27 with the functional category F as the head. F could be interpreted either as a coordinating conjunction or as a disjunctive conjunction. The utterance particles are in the internal conjunct, i.e. the complement of F. Under this hypothesis, utterance particles are not heads that select the main clause. Instead, the highest project is FP, a conjunction structure, and it is the head F that conjoins the main clause and the utterance particle. Neither the main clause nor the phrase headed by the utterance particle would be the root. It is the conjunction structure FP that should be regarded as the root. Along these lines, all sentences would be dominated by a conjunction structure headed by a conjunction F and ended with an internal conjunct, a position reserved for expressing aspectuality, modality, and speech act.

6 A generalized schema: consequences

The proposed conjunction configuration in 27 for both predicative utterance particles and non-predicative utterance particles in Chinese may have some interesting consequences.

The generalized schema FP in 27 can be stacked (or embedded) recursively in syntax. The configuration proposed in 34 is a good example, in which FP is projected three times with different purposes and functions. The (relative) hierarchical order of different FPs is likely to be constrained by semantics, particularly in the syntax-semantics interface, in the sense of the cartographic approach (Cinque 1999, among others). The FPs associated with speech act should be the highest while those associated with aspectual properties would be the lowest.

Let us examine the following examples in Cantonese. The utterance particles 喇 laa ([la33]) denotes aspectuality and 咩 me ([mɛ55]) denotes interrogative. 56 shows that the word order should be “喇 laa + 咩 me”, not “*咩 me + 喇 laa”, which implies that the aspectual FP that contains 喇 laa should be lower than the interrogative FP that contains me.

  1. (56)


    • keoi__zau-zo__laa__me/__*me__laa?

    • he__leave-PERF__LAA__ME__ME__LAA

    • He has left, hasn’t he?

Among the utterance particles in the speech act domain, for instance, Cantonese 咩 me and 嗬 ho ([hɔ35]) in 57, 咩 me encodes the epistemic state of the speaker while ho (also known as an interjection) discusses that of the addressee (Lam 2014). 57 shows that the word order should be “咩 me + 嗬 ho”, which suggests that the FP with the addressee orientation should be syntactically higher than the FP with the speaker orientation.

  1. (57)


    • daai__seng__zau__dak__gaa__laa__me__ho/__*ho__me?

    • big__voice__then__okay__GAA__LAA__ME__HO__HO__ME

    • What, can one get by just by being loud? I assume you’d agree it’s a valid question, right? (Lam 2014: 64)

The status of 嗬 ho in Cantonese is somewhat like the interjection eh in English (mainly spoken in North America), as in 58. The addressee-oriented particles or interjections are the internal conjunct of the highest FP, i.e. FP2 in 59, while other utterance particles or speech act oriented expressions are embedded in the external conjunct, i.e. FP1. Hence, the main clause and the utterance particles (or interjections) are in a conjunction relation forming FP, which is projected in the highest syntactic position, i.e. the root, of the sentence.

  1. (58)

    What’s he talking about, eh?


In addition to utterance particles, the complement of F can also be reserved for other kinds of elements to express some pragmatic meanings, which can be labeled as “sentence-final expressions” or more specifically, “utterance tags” (Yap et al. 2014).

Recall that the focus utterance particle 呢 ne in Mandarin is analyzed as an anaphoric expression that is used to reinforce the meaning denoted by the main clause. Interestingly, tags in English tag statements, such as 60 and 61 (Crystal 2008), are reminiscent of the reinforcing function of 呢 ne in Mandarin. For example, that was and is John are tags that partially duplicate the main clause to reinforce the meaning denoted by the main clause. Although these phoneticians in 62 is not analyzed as a tag, Crystal (2008) points out that there should be some connection between 62 and tag statements. Utterance particles and other sentence-final expressions like tags in 60 and 61 and the expression in 62 should be treated on a par syntactically although tags and tag-like sentence-final expressions, but not utterance particles, are prosodically separated from the rest of the clause. The presence of an intonational break distinguishes tags and tag-like sentence-final expressions from utterance particles (cf. Sailor 2012).

  1. (60)

    That was a lovely drink, that was.

  2. (61)

    He’s a nice man, is John.

  3. (62)

    They’re all the same, these phoneticians!

The proposed configuration in 55 should be able to apply to the expressions like tags in English, in which the tags are in the complement position of the conjunction F. For example, 63 would be the underlying structure for 60, in which the clause is repeated in the second conjunct and F is interpreted as a coordinating conjunction, somewhat like and. To avoid repetition, a lovely drink is elided, leaving the subject that and the copula was.

  1. (63)

    [that was a lovely drink] F [that was ]

In spoken Cantonese, some part of the clause could be repeated and put in the final position following the utterance particles, forming apparent tag statementsn. In 64, the subject 我 ngo ‘I’ and the nominal adverbial 聽日 tingjat ‘tomorrow’ are repeated and put after the utterance particle 啊 aa. In 65, the subject 佢 keoi ‘he’ and the adverb 都 dou ‘also’ are repeated and put after the utterance particle 喎 wo. The repetition of these expressions seems to reinforce the information denoted by the main clause, somewhat similar to the function of tags in English. If the proposed configuration in 55 applies to 64 and 65, the repeated expressions would be analyzed as the internal conjunct of F while the main clause with the utterance particles 啊 aa and 喎 wo would be embedded in the external conjunct. The repeated predicate is elided, leaving the subject, for instance, 我 ngo ‘I’ in 64 and 佢 keoi ‘he’ in 65, and the adverbials, for instance, 聽日 tingjat ‘tomorrow’ in 64 and 都 dou ‘also’ in 65 on the surface, as represented in 66.

  1. (64)

    我聽日交文啊, 我聽日

    • ngo__tingjat__gaau__man__aa, __ngo__tingjat.

    • I__tomorrow__submit__paper__AA__I__tomorrow

    • I will submit a paper tomorrow indeed.

  2. (65)

    佢都幾好喎, 佢都

    • keoi__dou__gei__hou__wo, __keoi__dou.

    • he__also__quite__good__WO__he__also

    • He is quite nice indeed.

  3. (66)

    [Subject Adverb VP] F [Subject Adverb ]

The proposed generalized syntactic schema can also be extended to tags in tag questions. Isn’t he in 67 is a tag in English, which should be derived from a similar structure of VP-Neg questions in Chinese by negating the repeated positive clause followed by VP-ellipsis in the second conjunct, as in 68o. F is interpreted as a disjunctive conjunction.

  1. (67)

    He is coming, isn’t he?

  2. (68)

    [he is coming] F [isn’t he ]

The internal conjunct is not necessarily a negative clause, for instance, 69. The internal conjunct is a positive clause when the external conjunct is a negative one. What the configuration in 70 offers is the positive and negative answers to be chosen by the hearer.

  1. (69)

    He is not coming, is he?

  2. (70)

    [he is not coming] F [is he ]

There is another type of yes-no questions in English, for example, 71, in which or not is following the interrogative clausep. The same syntactic schema can cover 71, in which the functional category F is overtly realized as the distinctive conjunction or. What or conjoins is a positive interrogative in the external conjunct and the repeated predicate negated by not, as in 72. The repeated predicate is deleted by VP-ellipsis on the surface, somewhat like the VP-Neg questions in Chinese in 49. The analysis of 71 shows that the functional category F can be overt (as realized as or), not necessarily emptyq.

  1. (71)

    Is he going or not?

  2. (72)

    [is he going] or [not ]

Among various constructions in Chinese and English we have examined so far, the internal conjunct contains some empty elements. Some of them may be derived by deletion, for instance, VP-ellipsis, and some may be base-generated silent elements, in the sense of Kayne (2015). Although the functional category F plays a crucial role in conjunction, the relation between the two conjuncts can also be brought together by co-reference between the empty elements in the internal conjunct and their antecedent in the external conjunct in some cases.

73 is spoken in colloquial English and has an interpretation close to that of a tag question. Kayne (2015) assumes that 73 is related to 74. A difference between these two sentences is that 73 contains silent ISN’T THAT, as represented in 75, while ISN’T THAT are audible in 74r. 75 could be rephrased in 76 under the generalized schema proposed in this paper, in which F is a coordinating conjunction that conjoins the external conjunct we’re on the list and the internal conjunct ISN’T THAT right. The antecedent of the silent pronominal THAT in the internal conjunct is the clause in the external conjunct. Considering subject-auxiliary inversion in the tags in 68 and 70, the internal conjunct should have a full-fledge clausal structure like CP (or a structure larger than CP, such as FP. See footnote 18). Apparently truncated expressions such as those in 72 and 73 should also have a full-fledge clausal structure that contains a syntactic position for the negation in 72 and the silent subject in 76, for example, a function project like TP, and a position for the raised silent auxiliary in 76, for example, CP.

  1. (73)

    We’re on the list, right?

  2. (74)

    We’re on the list, isn’t that right?

  3. (75)

    we’re on the list ISN’T THAT right

  4. (76)

    [we’re on the list] F [ISN’T THAT right]

A similar analysis can be extended to sentence-final 對 dui ‘right’ in Mandarin, as in 77, somewhat like tag questions in English. Mandarin 對 dui ‘right’ is a verb. It can form a yes-no question in the form of ‘A-not-A’ or followed by the interrogative particle 嗎 ma in 77s. 78 is the simplified syntactic representation of 77, in which F is a coordinating conjunction that conjoins the main clause in the external conjunct and the ‘tag’ in the internal conjunct. The verb 對 dui ‘right’ in the internal conjunct is predicated of a silent subject, which could be either an empty pronominal pro (Zhang and Tang 张和友, 邓思颖 2011), given that Chinese is a pro-drop language, or silent THAT, in the sense of Kayne (2015).

  1. (77)

    我們都被邀請, 對不對/對嗎?

    • women__dou__bei__yaoqing, __dui-bu-dui/__dui ma?

    • we__all__BEI__invite__right-not-right__right__MA

    • We are all invited, right?

  2. (78)

    [we are all invited] F [THAT right]

Given that 79 could be paraphrased as 80, Kayne (2015) assumes that 79 is in fact derived by 81 with silent TELL, silent ME, and the adverb again associated with silent TELL. The syntax of 81 can also be captured by the generalized syntactic schema, as represented in 82. The functional category F is a coordinating conjunction that conjoins two conjuncts. TELL ME are silent elements that are base-generated in the internal conjunct and TELL takes a clausal object CP which is the repeated question referring to the external conjunct. After the repeated object CP is elided, the adverb again remains.

  1. (79)

    Where do they live, again?

  2. (80)

    ?Where do they live, tell me again?

  3. (81)

    where do they live TELL ME again

  4. (82)

    [where do they live] F [TELL ME again]

It seems that the interpretation of 79 is rather similar to the echo questions marked with 話 waa in Cantonese (Tang 1998), as in 83. The interrogative particle 話 waa has the high rising tone ([wa35]), which was originally derived from the verb of saying 話 waa with the low level tone ([wa22])(cf. 14). Following Kayne (2015) analysis of again, I assume that 83 can be analyzed as 84. In the internal conjunct, AGAIN and ME are silent and the clausal object CP of the verb of saying is elided, leaving the verb of saying alone to denote the speech act of the speakert. The verb of saying in the internal conjunct is then realized as 話 waa with the high rising tone.

  1. (83)


    • keoidei__zyu__hai__bindou__waa?

    • they__live__in__where__WAA

    • Where do they live, again?

  2. (84)

    [where do they live] F [AGAIN tell ME ]

In Cantonese silent TELL ME could be modified by the adverb 先 sin ‘first’. In 85 the adverb 先 sin ‘first’ is used to reinforce the interrogative meaning of the question, which may express some reservation (Tang 邓思颖 2006). I assume that 85 may have a syntactic representation in 86, in which F is a coordinating conjunction that conjoins the interrogative in the external conjunct and the expression associated with the speech act in the internal conjunct. In the internal conjunct, the adverb sin ‘first’ is an overt element that modifies the predicate headed by silent TELL with two empty objects, namely silent ME and the elided clausal object CP, giving rise to the meaning of “Tell me who is the prettiest first”. After VP-ellipsis takes place, sin ‘first’ will be left alone on the surface, on a par with again in 79u.

  1. (85)


    • bingo__zeoi__leng__sin?

    • who__most__pretty__SIN

    • (Frankly tell me) who is the prettiest?

  2. (86)

    [who is the prettiest] F [first TELL ME ]

Recent studies have shown that utterance particles and some specific intonation are in complementary distribution (Wakefield 2011, Zhang 2014). If there is indeed such a correlation between the existence of utterance particles and intonation, it is speculated that intonation could have the same syntactic status as utterance particles, which could also be in the complement of F underlyingly, i.e. in the internal conjunct XP in 87. In terms of semantics and function, utterance particles and intonation are the same, while in terms of phonetics, they are different. The former is segmental and the latter suprasegmental. The speculation in 87 awaits supporting evidence particularly from large scale typological studies along these lines, cf. Cheng (1991) clausal typing hypothesis and Feng’s 冯胜利 (2015) diachronic study of Chinese particles, which I leave open in this paper.

  1. (87)

    [YP main clause] F [XP intonation]

Our discussion of the generalized syntactic schema of utterance particles in Chinese and its consequences in sentence-final expressions in English, such as tags, may support Kayne (2015) conjecture that sentence-final particles are not heads, which can be rephrased as in 88v.

  1. (88)

    Sentence-final expressions are not heads.

If the conjecture in 88 is proved to be correct, sentence-final expressions, including utterance particles, tags, and intonation, are all phrases in the complement of the conjunction F, i.e. in the internal conjunct. The root clause in all languages is then interpreted as a conjunction structure. In other words, all sentences should be dominated by a conjunction structure cross-linguistically, which would be regarded as the ultimate highest phase in syntax. The conjunction analysis of the root clause may bring us to a new horizon of viewing and understanding the deep nature of human language. Needless to say, more research along these lines will be fruitful in further exploring the syntactic structure of utterance particles in other Chinese dialects as well as in other languages that have utterance particles and other sentence-final expressions. It is hoped that the generalized schema hypothesized in this paper could shed light on the study of the syntax of the periphery and comparative grammar.

7 Concluding remarks

A novel analysis has been proposed for the utterance particles in Chinese. It has been argued that there are two types of utterance particles in terms of their diachronic origin, namely predicative and non-predicative, both of which can be derived by a generalized syntactic schema. The main clause and the utterance particles are conjoined by the functional category F, forming a conjunction structure. It has also been argued that polysyllabic utterance particles in disguise are in fact a subgroup of “particle cluster”. Both polysyllabic utterance particles in disguise and the particle cluster are derived by the same schema that is stacked (or embedded) recursively. It has also been shown that the generalized syntactic schema can be applied to other sentence-final expressions in Chinese and English, leading to a conjecture that sentence-final expressions are not heads.

8 Endnotes

aThere is a possibility that cannot be denied that 來 lai and 著 zhe in 來著 laizhe may be reanalyzed as a compound in modern Mandarin due to some morphological reason although they should be decomposed into two independent predicative utterance particles historically.

bZhang et al. 张斌主编 (2010: 250) also document these polysyllabic utterance particles: 似的 side, 的话 dehua, 着呢 zhene, 而已 eryi, and 便了 bianle. 似的 side ‘like’ is usually preceded by the verb 像 xiang ‘resemble’ forming the “像…似的 xiangside” sequence, which is not a real utterance particle. 的话 dehua ‘if’ is a marker of conditional subordinate clauses and should be different from utterance particles. 着呢 zhene is always attached to adjectival predicates and could be analyzed as an adjectival suffix, not an utterance particle. 而已 eryi could also be from a predicative element historically, which calls for future research on its diachronic syntax. Among all these elements, 便了 bianle may be qualified to be polysyllabic utterance particles, falling within the scope of my analysis although it is rarely used in contemporary spoken Mandarin.

cThese two examples are not discussed in Cheung 張洪年 (1972).

dIf the internal conjunct XP in 27 is a predicative, F could also be regarded as a “relator” that mediates the relationship between XP and its subject YP, in the sense of den Dikken (2006).

eThe secondary predicate may be predicated of an empty subject in the second conjunct, along the lines in Huang (1992), and the empty subject refers to the main clause in the first conjunct. The empty subject could be an empty pronominal in the sense of Huang (1992) or a silent expression in the sense of Kayne (2015). See the discussion in section 6 of this paper.

fExample 33 is due to a reviewer of this paper.

gCantonese 就係啦 zau hailaa seems to be the counterpart of the trisyllabic utterance particle 就是了 jiushile in Mandarin. However, there are two differences between Cantonese and Mandarin. First, Mandarin allows 就是 jiushi without 了 le while Cantonese does not (*就係 zauhai). Second, Cantonese allows 係啦 hailaa without the adverbial 就 zau ‘exactly’ while Mandarin does not (*是了 shile).

hSee Wong 黃卓琳 (2014) for more evidence and further elaboration along these lines.

iSee also Cheng et al. (1996) for an alternative syntactic analysis of VP-Neg questions in Chinese.

jThe Cantonese counterpart of Mandarin 沒有 meiyou ‘not’ is 冇 mou ‘not’. In modern Cantonese, 冇 mou ‘not’ cannot form a VP-Neg question though it was acceptable in old Cantonese (Yue 2004). The Mandarin counterpart of Cantonese 未 mei ‘not yet’ is 還沒有 hai meiyou ‘not yet’. See Law (2014) for the derivation of Cantonese 冇 mou ‘not’.

kLaw (2008) observes that 到底 daodi ‘really’ is incompatible with 嗎 ma, but he does not discuss 沒有 meiyou ‘not’.

lIn Shanghainese the interrogative utterance particle (口伐) va is incompatible with negative clauses, on a par with VP-Neg questions with 沒有 meiyou in Mandarin (Shi 石定栩 2007).

mI am grateful to Jo-wang Lin (personal communication) for drawing my attention to the disjunctive nature of F in 50.

nThere seem to be some similarities between apparent tag statements like 64 and 65 and the “dislocation focus construction” like (i) though Cheung’s (2008) movement analysis of dislocation deviates from the deletion analysis of tags proposed in this paper. I am grateful to Kwun-kin Chan (personal communication) for his input of Cantonese examples like 64 and 65.

  1. (i)

    買一部手機啊, 佢會淨係

    • maai__jat-bou__saugei__aa,__keoi__wui__zinghai

    • buy__one-CL__cellphone__AA__he__will__only

    • He will only buy a cellphone. (Cheung 2008:19)

oSee, for example, Sailor (2012) for the discussion of distributional and behavioral similarities of tags VP-ellipsis in English.

pAs pointed out by John Wakefield (personal communication), the fronted auxiliary in 70 may be deleted: He going or not?

qThe existence of or in 70 suggests that the head of FP can be overtly realized, contra Kayne (2015) conjecture that the high sentential heads are all silent in all languages.

rCapitals are used to indicate non-pronunciation in Kayne (2015).

sThere should be an embedded FP in the internal conjunct in which the embedded F conjoins 對 dui ‘right’ and the utterance particle 嗎 ma. Irrelevant details are omitted.

tThe word order of AGAIN in 83 differs from again in 81, given that adverbials (and adjuncts in general) should precede the predicate in Chinese.

uThe analysis presented here is also reminiscent of Liu 劉丹青 (2013) who proposes that it is the speech act verb that 先 sin ‘first’ in 84 modifies.

vKayne (2015) examines right and again in English, 講 kong ‘tell’ in Taiwanese, and tu in French and his original statement is “All pronounced elements that could have been taken to be sentence-final particles in one language or another actually share the non-sentence-final-particle status that right, again, kong and tu share”. Some version of 87 had been proposed independently in an early stage of my work before I came across Kayne (2015).