When OR is assigned a conjunctive inference in child language

ABSTRACT

It has been proposed that children differ from adults in that children license a conjunctive inference to disjunctive sentences that lack any licensing expression. The proposal is that children infer “A and B” from sentences of the form “A or B.” Although children’s conjunctive interpretations of disjunction have been reported in some studies, they have not been observed in other studies. This study investigates one possible source of the different findings, which is the presence or absence of objects that are contextually introduced, beyond those mentioned in the test sentences. Using a Truth Value Judgment Task, we conducted three experiments with preschool Mandarin-speaking children and a control group of adults. Test sentences in Experiment 1 included disjunction and a deontic modal verb, and the contexts only included objects that were mentioned in the test sentences. In Experiment 2, the deontic modal verb was omitted from the test sentences. Experiment 3 presented the same sentences as Experiment 2, but the contexts included additional objects beyond those mentioned in the test sentences. Both children and adults assigned a conjunctive inference to the test sentences in Experiment 1. In Experiment 2, with no deontic verbs and no contextually introduced objects, a subset of children computed conjunctive inferences, whereas none of the adults did. In Experiment 3, neither group made a conjunctive inference. The findings reveal that entities in nonlinguistic contexts influence children’s understanding of logical expressions. Moreover, preschool children already have linguistic knowledge that modal verbs license conjunctive inferences for sentences with disjunction.

Acknowledgments

For helpful comments on the account we propose in this article, we wish to thank Gennaro Chierchia, Uli Sauerland, Alexandre Cremers, Rosalind Thornton, Drew Khlentzos, William Snyder, David Barner, Alan Bale, Raj Singh, Peng Zhou, Lyn Tieu, Yimei Xiang, Cory Bill, and the audience at the 7th Generative Approaches to Language Acquisition North America Conference. We are also grateful to Jeffrey Lidz, Kristen Syrett, and four anonymous reviewers for helpful comments on an earlier draft. Finally, we are thankful to the kindergarten affiliated with Hubei University of Technology, Wuhan, China.

Compliance with ethical statements

This study has also received the Ethics Approval from Faculty of Human Science-Human Research Ethics Sub-Committee, Macquarie University, Australia; the ethics reference number is 5201500028.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

¹ The finding that children accept sentences with disjunction when both disjuncts are true also argues against the proposal that children initially assign an exclusive-or interpretation of disjunction (e.g., Morris 2008).

² Conjunctive inferences are also licensed in certain sentences with existential quantifiers and in some generic sentences (see, e.g., Fox 2007; Nickel 2011). To illustrate, consider examples (i) and (ii).

(i) a. (This job is very hard.) Some workers take three days to complete it or do not finish it at all.

b. ⇝ Some workers take three days and some workers do not finish it at all.

(ii) a. Lions live in China or in Vietnam.

b. ⇝ Lions live in China and lions live in Vietnam.

³ In sentences in which disjunction appears in the scope of a deontic modal verb, the conjunctive inference is compatible with the exclusivity inference that John is not allowed to eat both candy and ice cream. This inference is derived at the first step of the algorithm.

⁴ The responses of 10 child participants were excluded from the analysis because they responded at chance levels in at least one of the test conditions.

⁵ Children’s egalitarian interpretation was also reported in a study by Hirohisa & Thornton (2016), who investigated English-speaking children’s interpretation of pseudocleft sentences such as (i).

(i) A shell or a plant is what every diver brought back.

⁶ An anonymous reviewer inquired why Experiment 1 included control sentences but no filler sentences, whereas Experiment 2 contained filler sentences but no control sentences. We wish to point out that these two types of sentences have the same function of counterbalancing the Yes and No responses. As compared with Experiment 1, Experiment 2 used the prediction mode of the TVJT. In such a paradigm, a filler sentence is more conducive to fulfill the aforementioned function than a control sentence.

Additional information

Funding

This work was supported by Macquarie University. Specifically, it was supported by an International Macquarie University Research Excellence Scholarship to the first author (No. 2014016) and by the Macquarie University Postgraduate Research Fund (2016) to the first author.