Do children know whanything? 3-year-olds know the ambiguity of wh-phrases in Mandarin

ABSTRACT

Wh-phrases in Mandarin have an interrogative (like English what) and an indefinite (like English a/some) interpretation. Previous comprehension studies find that children can access both interpretations around 4.5 years old; studies with younger children focus on production and find that children between 2 and 4.5 do not reliably produce the indefinite interpretation in naturalistic speech or in elicited imitation tasks. In this article, we use comprehension tasks to examine 3-year-olds’ interpretation of wh-phrases. We find that they have adult-like interpretations of wh-phrases in two different contexts: in dou -sentences (Experiment 1), where the indefinite interpretation is the only available interpretation and the whole sentence receives a universal reading (roughly equivalent to English any), and in negated sentences (Experiment 2), where the interpretation of wh-phrases depends on prosodic prominence and the indefinite interpretation leads to an existential reading of the sentence.

Acknowledgments

We want to thank the directors, teachers, and parents at Beijing Tangjialing Hongying School, Beijing Xinglinwan Preschool, Beijing Shuangzhuang Science Park Preschool, Beijing Shangzhuang Yiming Preschool, and Xintongxin Preschool. We are also grateful for the suggestions from the audiences at the Acquisition Lab Meeting and LSLT at the University of Maryland, WCCFL2020, Workshop on Theoretical and Experimental Linguistics at Tsinghua University, MAPLL-TCP-TL 2019, where earlier versions of this work were presented.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

¹ The example is taken from the novel Hongyan by Kuang-pin Luo & Yiyan Yang (1961); see Liu & Yang (2021) for more examples.

² If we change the subject of (8) to Zhangsan, the ignorance inference could be satisfied, and yet the indefinite interpretation is not as acceptable:

(i) Zhangsan xihuan shei/shenme ren

Zhangsan like  who/what person

a. ‘Who does Zhangsan like?’

b. ??’Zhangsan likes someone.’

However, when we add a modification to the object wh-phrase, the indefinite interpretation is acceptable again:

(ii) Zhangsan xihuan yuyanxuexi     de shei/shenme ren

Zhangsan like   Linguistics Department poss who/what person

a. ‘Which person from Linguistics Department does Zhangsan like?’

b. ‘Zhangsan likes someone from Linguistics Department.’

It is unclear why manipulating the length of object NP would affect the acceptability of wh-indefinites in affirmative sentences; see Liu & Yang (2021) for more discussion on this puzzle.

³ For some speakers, prosodic prominence on the numeral yi could make the sentence more acceptable, assigning focus on the quantity of people that Xiaoxiao met. The sentence has an interpretation similar to “he didn’t meet ONE person (he met two).” However, intuition varies across the native speakers we consulted.

⁴ As correctly pointed out by a reviewer, this is true except in dou sentences like (7), where wh only has the indefinite interpretation but is associated with prominence. In Experiment 1, we took advantage of this property of dou -sentences to make sure that the sentences with dou where wh is interpreted as an indefinite and sentences without dou where wh is interpretation as an interrogative have similar prosodic contours.

⁵ In fact, Zhou & Crain (2009) examined all instances of meiyou from Mandarin corpora in CHILDES and found no instance of quantificaitonal meiyou co-occurring with wh -phrases. A reviewer correctly points out that this does not mean that children have no exposure to this type of structure. Although it is possible that children have some exposure to this structure, their argument here is that this structure is rare enough that children might not be able to learn the interpretation of wh in this specific configuration from the input.

⁶ As noted by many, the position of wh-phrases relative to dou matters to their interpretation too. In pre-dou positions, wh-phrases are interpreted non-interrogatively, as demonstrated in (18), but when the wh positions to the right of dou as in (i), the sentence is a wh-question. In this article, we focus on pre-dou wh-phrases. (i) Xiaoxiao dou chi-le shenmeXiaoxiao dou eat- asp whata. What all did Xiaoxiao eat?b. (Not possible) ’Xiaoxiao ate everything.’

⁷ Five adult participants chose to use headphones. Using headphones did not influence the adult behavior; all adults behaved the same way in this experiment.

⁸ The experiment was run on PsychoPy3.0.0 (Peirce et al. 2019).

⁹ During practice, if the child still would not give any responses, the experimenter would provide additional prompts: either “Let’s help her. Is Xiaoxiao right?” if Xiaoxiao’s utterance was a statement, or repeat the question if Xiaoxiao’s utterance was a question. The experimenter stopped giving the additional prompts during the test phase.

¹⁰ Child participants were asked to give Xiaoxiao a stamp if they agreed with Xiaoxiao’s guess about the winner to keep them engaged in the game. When we were piloting this experiment, some adults were very reluctant to participate in stamping. Considering that introducing stamps stretched the length of the experiment, and that adults do not need this extra step to stay attentive, we did not include the stamp for adult participants.

¹¹ We ran a model with both participants and test items as random factors, but the effect of items was extremely close to zero, so we excluded items as a random factor in the final model.

¹² Same as yes/no-responses, we did run a model with both participants and items as random factors, but there was close to zero variance for items, and thus the final model did not include items as a random factor.

¹³ We thank an anonymous reviewer for pointing this out to us.

¹⁴ For some adults, the sentence (32) elicits a vague “not much” interpretation instead of the clear-cut “nothing” interpretation: Little Lamb didn’t pack much in the box. This is especially the case if the contexts allow for a contrast between significant items versus insignificant items (Huang 2013). For example, if Little Lamb only packs one piece of candy when she should have packed food to survive, the candy is negligible. In this context, the sentence Little Lamb didn’t pack shenme would be judged as true, because although “Little Lamb didn’t pack anything” is false, the candy that she packs is sufficiently insignificant to pass as “not much.” In our task, since there was a requirement to pack all three items, each of these items was made significant, and thus we can avoid this ambiguity. Moreover, even if some participants still assign the “not much” inference, we have established that the agent has packed the majority of items required, so the “didn’t pack much” interpretation is still false.

¹⁵ In the model with both participants and items as random factors, the factor items had close to zero variance and were excluded from the final model.

¹⁶ An anonymous reviewer suggests that Figure 20 gives the appearance that there should be a statistically significant age effect. When we consider all trials as independent data, we do indeed find one ($\beta =-0.70,p<.05$), but it disappears when we control for participant identity across trials by using it as a random factor. This indicates that the supposed “age effect” is a difference between individuals, not between adults and children. So the statistics are as expected if there is indeed no effect between age groups. We do want to acknowledge that there is a question about whether a larger sample size would have been appropriate. We, however, cannot do a posthoc power analysis here, as a power analysis with obtained effect size does not tell us anything that the $p$ value does not already tell us (namely, that there is no significant effect in the sample, see Hoenig & Heisey (2001) for discussion on the inadequacy of power analysis with obtained effect size). Unfortunately, we could not obtain a priori power from a previous study of similar design (Zhou 2015), but our number of participants per condition is comparable to Zhou (2015).

¹⁷ A reviewer points out that the percentage of “no” responses is roughly equivalent to the percentage of “yes” and “packed/unpacked-item” responses combined in this way and argues this means that we can’t conclude that children had an adult-like indefinite interpretation. However, we don’t think “yes” and “packed/unpacked-item” responses can be combined. As discussed in Section 4.2, saying “yes,” like saying “no,” clearly indicates that the speaker is responding to a statement, while the packed/unpacked-item responses could be responding to either a statement or a wh -question. Thus, it doesn’t make sense to collapse the “yes” responses with packed/unpacked-item responses and contrast them with the “no” responses in that way.

¹⁸ One possible explanation for the “yes” responses that adults and children gave is that they had the “not much” interpretation mentioned in footnote 13, despite our attempt to discourage this inference (that is, they are agreeing with Xiao Xiao that Little Lamb didn’t pack much because the most important item was not packed). In future work, we plan on eliminating this complication, for instance by having a condition where Little Lamb has all of the items in her box.

¹⁹ We thank an anonymous reviewer for this suggestion.

Additional information

Funding

This material is based upon work supported by the National Science Foundation under Grant No. 1449815.