Tsinghua University – University of Amsterdam Joint Research Centre for Logic

Invited Talk – Haihua Pan

Revisiting Chinese donkey sentences

(joint work with Hang Kuang)

Cheng and Huang’s (1996) classic theory of Chinese donkey sentences proposes that two interpretational strategies are needed for a satisfactory account of Chinese donkey sentences, namely the unselective binding mechanism of classic DRT and the E-type strategy. Unselective binding applies to the so-called bare conditionals (1a) in which identical wh-phrases appear in pairs and are bound by a covert necessity operator, and the E-type pronoun strategy applies to ruguo-conditionals (1b) and dou-conditionals (1c). The empirical basis and theoretical analysis of Cheng and Huang (1996) have been challenged, notably by Pan and Jiang (2015), who object to the claim that a strict correspondence exists between the type of conditionals and the interpretational strategy. Cheng and Huang (2020) addresses some of the issues raised by Pan and Jiang, and argue that their original analysis can be maintained. In this paper, we review in detail the responses from Cheng and Huang (2020), and conclude that their arguments cannot hold. We also discuss the question-based approach to bare conditionals (e.g. Liu 2017, Li 2021) and point out some potential problems.

(1) a. Shei xian lai,    shei  xian chi.

          who first come who first  eat 

          ‘Whoever comes first eats first.’

b. Ruguo ni   kanjian shei, jiao ta  lai      jian wo.

          if         you see       who  ask he come  see  me

    ‘If you see someone, ask him to come see me.’

c. Ni   jiao shei  jinlai, wo dou jian ta.

    you ask  who enter    I    all  see  him

    ‘Whoever you ask to come in, I’ll see him.’

According to Pan and Jiang (2015), it is not the case that, as argued in Cheng and Huang (1996), the wh-phrase in the consequent clause of bare conditionals cannot be replaced by a pronoun (2a and 2b). It is also incorrect to say that the pronoun in the consequent clause of ruguo-conditionals and dou-conditionals cannot be replaced by a wh-phrase (2c and 2d). Based on these observations, Pan and Jiang argue that there is no strict correlation between sentence types and interpretational strategies. What Cheng and Huang observe is just the default patterns. Deviations from the default patterns are possible but require extra contextual support. 

(2) a. Shei bu  dui,  wo jiu  shuo ta.

          who not right I   then say  he

         ‘Whoever is not right, I will say he is not right.’ 

b. Shei  yao   zhe po-chang,          rang gei   ta    hao   le.

          who  want this broken-factory  give  to   who good sfp

          ‘Whoever wants this broken-factory, give it to him then.’

      c. Ruguo shei yao    zhe po-chang,          jiu  rang shei dao bangongshi lai     zhai wo.    

           if        who want  this broken-factory then let   who  to   office         come find me

          ‘Whoever wants this broken factory, let him come to my office to see me.’

d. Amei shuo: “Gei  shei kan,  shei  dou hui  shuo wo shi haoxinhaoyi.”

          Amei say     give who look  who  all   will say    I    be  good-will

  ‘Amei said: ‘Whomever you give to look at (it), he will say that I meant well.’’

Cheng and Huang (2020) argue that examples in (2) are not problematic for their analysis. According to them, the anaphoric wh-phrase in bare conditionals can be replaced by a pronoun only if the connective jiu ‘then’ is present, which is indicative of a covert ruguo ‘if’, though they admit that a ruguo-conditional does not have to use the E-type strategy: when the anaphoric element is a wh-phrase, as in (2b), the unselective binding strategy has to be adopted. 

    If one follows this line of argument, then all that determines what interpretational strategy to be adopted is what the anaphoric element is, which simply amounts to a stipulation: irrespective of what type of conditionals they occur in, wh-phrases are presumed to be interpreted as unselectively bound variables and pronouns via the E-type strategy, which is different from the claims in Cheng and Huang (1996). Cheng and Huang’s new suggestion that the unselective binding strategy is not excluded for ruguo-conditionals is just what Pan and Jiang argue for. Furthermore, some bare conditionals with jiu do not allow presence of ruguo (3a vs. 3b), suggesting that the presence of jiu does not necessarily indicate the presence of a covert ruguo

(3) a. Ni zenme gen   wo shuo de, wo jiu  zenme gen   ta    shuo de.

          you how   with me say   de    I   be  how    with him  say   de

  ‘I said it to him in the way that you said it to me.’

      b. *Ruguo ni zenme gen wo shuo de, wo jiushi zenme gen ta shuo de.

Cheng and Huang further argue that those bare conditionals without jiu which also allow wh-pronoun alternation in the consequent, as shown in (2b), should be reanalyzed as composed of two independent sentences: a rhetorical question followed by a suggestion. Their evidence comes from the intonation difference between a true bare conditional and (2b): (2b) can be pronounced with a pause or a pause particle between the two clauses, while a true bare conditional sounds more like a single compact unit. However, the evidence based on intonation is not reliable, as a true bare conditional (i.e. whwh conditional) may also be pronounced with a pause particle (4). The presence or absence of a pause is not an indication of the interpretive strategy used. Finally, Chen and Huang attribute a universal reading to the second wh-phrase in dou-conditionals like (2d). For them, (2d) means “No matter who you show (it) to, everyone will say that I meant well.” This interpretation is clearly incorrect; the sentence actually means “No matter who you show (it) to, the person who you show it to will say that I meant well.” Even if one allows their interpretation, our interpretation is still a salient interpretation and cannot be excluded. Consequently, the examples in (2) remain problematic for Cheng and Huang.

(4) Shei xian dao     jiaoshi        ne, wo  jiu    jiangli  shei.

      who first arrive classroom   prt  I    then reward  who

‘I will give a reward to whoever arrives at the classroom first.’

Some recent accounts of Chinese donkey sentences targeting specifically bare conditionals (also called wh-conditionals in these accounts) propose that the two wh-clauses in these conditionals denote questions. This kind of analysis faces a serious problem concerning the distribution of pronouns in the consequent clause. Analyzing wh-conditionals as saying essentially that the short answer to the first wh-question is also the short answer to the second wh-question (e.g. Liu 2016) does not predict the possibility of replacing the second wh-phrase with a pronoun. Even if the question-based account can be made to allow pronouns in the consequent (e.g. Li 2021), it is still left unaccounted for the question why the occurrence of pronouns is rather restricted. The limited distribution of pronouns receives an explanation from Pan and Jiang (2015), as the use of pronouns in bare conditionals deviates from the default patterns and requires extra contextual support, such as the imperative flavor in (2b). This question-based account also has nothing to say about similar patterns using reflexives like ziji ‘self’ and empty pronouns.


  1. Cheng, Lisa Lai-Shen, and C.-T. James Huang. 1996. Two types of donkey sentences. Natural Language Semantics, 4, 121-163.
  2. Cheng, Lisa Lai-Shen, and C.-T. James Huang. 2020. Revisiting donkey anaphora in Mandarin Chinese: A reply to Pan and Jiang (2015). International Journal of Chinese Linguistics, 7(2), 167-186.
  3. Li, Haoze. 2021. Mandarin wh-conditionals: A dynamic question approach. Natural Language Semantics, 29, 401-451.
  4. Liu, Mingming. 2016. Varieties of alternatives. Ph.D. thesis, Rutgers University.
  5. Pan, Haihua, and Yan Jiang. 2015. The bound variable hierarchy and donkey anaphora in Mandarin Chinese. International Journal of Chinese Linguistics, 2(2), 159-192.