Plurality in Mandarin Chinese and typicality effects
Authors
Claire Rong
Affiliation: École Normale Supérieure (Paris)
Category: Linguistics
Keywords: Plurality, Pragmatics, Typicality, Chinese
Schedule & Location
Date: Tuesday 2nd of September
Time: 15:30
Location: GSSR Plenary Hall (268)
View the full session: Quantifiers, Plurals & Numbers
Abstract
Plurality in Mandarin Chinese and typicality effects
Context
Chinese is an articleless language in which bare nouns (BNs) are widely used and are number-neutral, making no singular/plural distinction ([8]). Post-verbal BNs are indefinite unless anaphoric or unique definites. Nouns can be preceded by a singular indefinite marking, [one + cL], or a plural indefinite marking, [one + xie]. xie is a plural classifier which replaces the singular classifier (abbr. cL) usually associated with the noun. In formal speech, both classifiers are preceded by the numeral glossed as “one”, though cL can also follow other numerals. Despite containing “one", [one + cL] usually functions like the English “a”, except in explicit counting or when carrying focus:
(1) wǒ shì yī gè xué-shēng I be one CL student 'I am a student.'
[one + cL] and [one + xie] are positive polarity items (PPIs). An example showing it for [one + xie]:
(2) tā méi yǒu kàn guò yī xiē shū he NEG PFV read PFV one xie book 'He hasn't read some books.' (only the wide scope reading is accessible)
I present two puzzles: the first concerns the truth conditions of [one + cL], [one + xie], and BNs; the second involves gradient effects in production and perception experiments. I will show how both puzzles can be explained through pragmatic competition and typicality effects.
1st puzzle
By embedding [one + cL] and [one + xie] in a DE environment (not negation, because they are PPIs) as in (3), I show that [one + cL] triggers a pragmatic uniqueness inference and that [one + xie] triggers a pragmatic plurality inference. The following example uses [one + xie], but [one + cL] yields the same conclusion.
(3) měi dāng wǒ jiàn dào yī xiē yăn-yuán wǒ dōu huì gāo-xìng each when I meet PFV one xie actor I DOU will happy ‘Each time I see some actors, I am happy.' ✓ I am happy if I see 1 actor. ✓ I am happy if I see >1 actors.
Both forms have a literal meaning of ≥ 1, though [one + cL] is generally singular (like a NP in English) and [one + xie] plural (like the English bare plural). BNs, by contrast, are fully number-neutral, having the same meaning of ≥ 1 and triggering no number inference in positive or negative contexts. The plurality inference of [one + xie] arises as a higher-order implicature, driven by competition with the pragmatically strengthened (exactly one) reading of [one + cL] [6, 9].
It appears that [one + cL], [one + xie] and the BN all have the same truth conditions of ≥ 1. However, this contradicts the judgments observed in universally quantified sentences. First note that in English, sentences (4a) and (4b) both can truthfully describe situations where each box contains exactly one book, situations where each contains several books, and crucially, mixed situations where some contain one book, others several. Experimental evidence from [3] confirms that (4b) is judged true in mixed situations.
(4) a. Each box contains a book. b. Each box contains books.
If [one + cL/xie] have a weak denotation, mixed situations should also be acceptable in the Mandarin sentences corresponding to (4). However, in mixed situations, most informants find [one + cL/xie] hard to accept, unless singular or plural situations dominate across boxes. In contrast, BNs are perfectly acceptable, consistent with their number-neutrality. To confirm that the number inferences of [one + cL/xie] are pragmatic, a simple argument shows that embedding in DE environments is a more conclusive test than assessing mixed situations: if [one + cL/xie] had strong denotations, then the observed inferences would be stronger than (i.e. cannot be strengthened to) the meaning of [one + cL/xie] in the restrictor of a DE operator. Therefore, we need to reconcile a weak denotation with this apparent contradiction.
2nd puzzle: production
Inspired by [2], we recruited 277 native English speakers on Prolific and presented stimuli varying in the number of boxes (0–10) containing a unique object, each condition instantiated 11 times with different objects (Fig.1a). As shown in Fig.1b, higher proportions of unique-object boxes increased a NP use, while bare plural use declined. A logistic regression predicting singular use significantly outperformed a null model (full: χ²(1) \= 29.92, p \< 0.001; excluding extreme conditions: χ²(1) \= 9.20, p \= 0.002). These gradient effects challenge existing accounts of bare plural interpretation ([5, 6, 9, 4]). While a Chinese production study is forthcoming, our English results support a role for typicality-based effects, aligning with [1].
2nd puzzle: perception
120 English and 120 Chinese speakers rated “Each box contains trees”/(5) on a continuous scale from "bad description" to "good description" while viewing images of 10 boxes, with 0, 5, or 10 containing a single tree (randomized conditions). As shown in Fig.2b, English participants show no Bayesian alignment between production and perception. While English and Chinese patterns are qualitatively similar, they differ in amplitude. Unlike our informant-based results, this perception task used a continuous scale rather than binary judgments.
(5) měi gè hé-zi lĩ dōu yǒu yī xiē shù each CL box in DOU is one xie tree
Discussion
Existing theories do not account for gradience, suggesting that cognitive mechanisms like typicality-based effects influence plural and quantified statement interpretation. Typicality reflects how well a situation exemplifies a proposition. Two sources are considered: (i) an independent factor [7], and (ii) a function of truth conditions, based on distances to closest true and false scenarios [1]. Typicality helps resolve our first puzzle and explains gradient effects in the second, but comparing these approaches sheds light on English-Chinese differences.
Van Tiel [7] models typicality in quantification through individual exemplars: a scenario best fits “Each box contains [one + xie] books” when each box holds a typical number, i.e. > 1. Similarly, one + cL book is best exemplified by exactly one book. By contrast, Chemla & Spector [1] define a sentence's typicality as a weighted sum of readings: literal (L), weak (W) (adding the global implicature “at least one box contains several books"), and strong (S) (strengthened to "each box contains several books" through local implicature), with typicality given by T(p) \= α + β × L(p) + γ × W(p) + δ × S(p). English-Chinese differences may result from varying weights. Alternatively, van Tiel's model projects typicality via a harmonic mean, where boxes with fewer books lower the sentence's overall typicality. This account predicts cross-linguistic differences only if book quantities per box receive distinct typicality values in each language.
Figures
(a) Production experiment stimulus
(b) Production results in English
Figure 1: Production experiment and results
(a) Perception results in Chinese
(b) Perception results in English
Figure 2: Perception experiment results in Chinese and English
