Do Children Interpret 'or' Conjunctively?

Preschoolers often struggle to compute scalar implicatures (SI) involving disjunction (or), in which they are required to strengthen an utterance by negating stronger alternatives, e.g., to infer that, “The girl has an apple or an orange” likely means she doesn’t have both. However, recent reports surprisingly find that a substantial subset of children interpret disjunction as conjunction, concluding instead that the girl must have both fruits. According to these studies, children arrive at conjunctive readings not because they have a non-adult-like semantics, but because they lack access to the stronger scalar alternative and, and employ doubly exhaustified disjuncts when computing implicatures. Using stimuli modeled on previous studies, we test English-speaking preschoolers and replicate the finding that many children interpret or conjunctively. However, we speculate that conditions which replicate this finding may be pragmatically infelicitous, such that results do not offer a valid test of children’s semantic competence. We show that when disjunctive statements are uttered in contexts that render the speaker’s intended question more transparent, conjunctive readings disappear almost entirely. INTERPRETATION OF DISJUNCTION 4 Older preschoolers and kindergarteners often struggle to compute scalar implicatures (SIs) involving quantifiers like some (Smith, 1980; Noveck, 2001; Papafragou & Musolino, 2003) and logical operators like disjunction (Braine & Rumain, 1981; Chierchia et al., 2001). For example, when presented with an utterance containing some like the one in (1a), many children fail to derive the implicature in (1c). Similarly, when presented with a sentence like the one in (2a) many children fail to derive an exclusive interpretation of or characterized by (2c): (1) a. The boy took some of the bananas. b. The boy took all of the bananas. c. The boy took some, but not all of the bananas. (2) a. The girl has an apple or an orange. b. The girl has an apple and an orange. c. The girl has an apple or an orange, but not both. On most accounts, deriving a scalar implicature involves accessing a stronger alternative statement that is generated by replacing one scalar term (e.g., some, or) with another, stronger, one (e.g., all, and). For example, to derive an implicature for the sentence containing some in (1a), children must access and negate the stronger alternative in (1b). Likewise, to derive the implicature for the sentence containing or in (2a), children must access and negate the stronger alternative in (2b). On standard accounts of disjunction, a failure to do so should result in an inclusive interpretation, wherein the listener concludes that the utterance in (2a) is acceptable even when the girl has both an apple and an orange (Crain & Khlentzos, 2010; Gazdar, 1979; McCawly, 1993; Pelletier, 1977). INTERPRETATION OF DISJUNCTION 5 Children’s difficulties with SI have been variously attributed to general processing limitations (Chierchia et al., 2001; Reinhart, 2004; Pouscoulous et al., 2007), difficulty in understanding the communicative goals of the experimenters (Musolino & Papafragou, 2003; Papafragou & Tantalou, 2004), and a tendency to be more tolerant of pragmatic infelicity (Katsos & Bishop, 2011). More recently, evidence has begun to accumulate in support of an account focusing on children’s access to relevant linguistic alternatives that are necessary to derive an SI (Barner et al., 2011; Barner & Bachrach, 2010; Chierchia et al., 2001; Foppolo et al., 2012; Hochstein et al., 2014; Tieu et al., 2016; Skordos & Papafragou, 2016). These accounts argue that difficulties in accessing necessary linguistic alternatives might explain why children fail to derive implicatures, whether such difficulties are due to a lack of associations between scale mates, a failure to detect which alternatives are contextually relevant, or a lack of working memory capacity to compute the contribution of alternatives while simultaneously considering a sentence’s basic meaning. While most studies in support of the “access to alternatives” view have focused on conditions under which children derive weak vs. strong interpretations of utterances, both of which are available to adults, two recent studies have argued for this account by pointing to an entirely different form of evidence, namely the interpretation of disjunction as conjunction. Specifically, these studies report that, given an utterance like (2a) above, some children conclude that the girl must have both types of fruit. In one study documenting this phenomenon, Singh et al. (2016) tested 4and 5-year-old English-speaking children (N=31) using a modified Truth Value Judgment task (Crain & Thornton, 1998). Surprisingly, they found that children accepted disjunctive statements (e.g., The boy has an apple or a banana) only about 35% of the time when one of the disjuncts was true (henceforth 1-Disjunct-True trials), but 76% of the time when both INTERPRETATION OF DISJUNCTION 6 disjuncts were true (henceforth 2-Disjunct-True-trials). In addition, children displayed a similar behavior with disjunction embedded under a universal quantifier (e.g., Every boy has an apple or a banana), accepting 1-Disjunct-True trials about 45% of the time and 2-Disjunct-True trials, 75% of the time. This pattern of responses is particularly surprising when seen from the standard perspective that children often do not derive implicatures at all; if that were the case, they should generally be inclusive and be equally likely to accept disjunction in 1-Disjunct true and 2Disjunct-True trials. Critically, Singh et al. classified children according to their individual response patterns: Four children performed in an adult-like manner, accepting 1-Disjunct-True utterances while rejecting 2-Disjunct-True utterances (henceforth “Exclusive”). Another four children showed the pattern of failure often reported in previous studies, accepting both 1Disjunct-True and 2-Disjunct-True trials (henceforth “Inclusive”). Finally, 21 out of 31 children (or 67%) responded as if or was and, rejecting 1-Disjunct-True utterances while accepting 2Disjunct-True trials (henceforth “Conjunctive”). Adopting the “access to alternatives” account, Singh et al. argue that children in their study arrive at conjunctive interpretations because they lack access to stronger alternative statements 1 Against this hypothesis that children simply think or always means and, Singh and colleagues (2016) point out that children do not derive conjunctive meanings in downward entailing environments, such as when disjunction is embedded under the scope of negation (Gualmini & Crain, 2002; Goro & Akiba, 2004; Jing, Crain & Hsu, 2005), when it occurs in the antecedent of conditionals (Su, 2014), and when it is in the nuclear scope of only, before, not every, etc. (Goro, Minai & Crain, 2005; Jing, Crain & Hsu, 2005; Gualmini & Crain, 2002; Notley, Zhou, Jensen & Crain, 2012; cf. Tieu et al., 2017; Singh et al., 2016, for discussion). INTERPRETATION OF DISJUNCTION 7 derived by replacing or with and. Consequently, they are restricted to deriving implicatures from alternatives that are contained within the original assertion – i.e., the individual disjuncts. Also, and critical to their account, Singh et al. adopt the grammatical view of implicature (Fox, 2007; Chierchia, et al., 2009), and propose that children’s disjunctive alternatives are exhaustified before being negated, resulting in multiple exhaustification (for additional discussion of this account, see Crnič, Chemla, & Fox, 2015; Fox & Katzir, 2011; Franke, 2011; Katzir, 2007; Spector, 2016). Specifically, when children hear a disjunctive statement like (3a), they compute alternatives by exhaustifying each separate disjunct in (3b) and (3c), resulting in the statements in (3d) and (3e): (3) a. The girl has an apple or an orange. b. The girl has an apple. c. The girl has an orange. d. The girl has an apple but not an orange (i.e., only an apple). e. The girl has an orange but not an apple (i.e., only an orange). The original disjunctive statement is then strengthened by negating these exhaustified alternatives, resulting in the statements in (4a) and (4b), which, when combined with the original statement result in a conjunctive meaning, in (4c). (4) a. It’s not the case that the girl has only an apple. b. It’s not the case that the girl has only an orange. c. The girl has an apple or an orange (6a), but doesn’t have only an apple, and doesn’t have only an orange (7a & 7b) (i.e., the girl has both an apple and an orange). According to this theory, adults do not derive the implication in (4c) because they have access to the alternative with and – i.e., “The girl has an apple and an orange.” The inclusion of INTERPRETATION OF DISJUNCTION 8 this alternative, since it is stronger than the plain disjunctive statement, leads to its negation, and the corresponding inference that the girl does not have both. This inference directly contradicts the inference in (4c). Since contradictory inferences are not permitted by the exhaustification operator, the derivation of (4c) is blocked (see Fox’s 2007 discussion of innocent exclusion for details). Thus, on Singh et al.’s analysis, their study simultaneously supports the idea that access to alternatives limits children’s implicatures, as well as the idea that children’s implicatures can involve multiple iterations of exhaustification. As they note, such an analysis integrates work on acquisition (e.g., Paris 1973; Braine & Rumain 1981; Tieu et al. 2017) with a broader account of multiple exhaustification, which previous studies have used to explain other forms of implicature, including so-called “free choice” inferences (see Fox, 2007; Chierchia et al. 2009; see also Geurts, 2010; Franke 2011; Meyer, forthcoming, for accounts that do not employ multiple exhaustification). The power of this unifying account depends on the reliability of the deve