Isomorphism for All (but Not Both): Floating as a Means to Investigate Scope

Abstract

This article investigates the so-called isomorphism effect (Musolino 1998; Musolino, Crain & Thornton 2000) in the comprehension of scopally ambiguous sentences containing negation and floating quantifiers. Given that floating quantifiers can appear in up to three surface positions relative to negation, I propose that they provide us with the ideal methodological tool to test for scope ambiguity resolution while holding constant various factors, including: (i) the associated noun phrase, and thereby the relevant thematic roles in the test stories; (ii) the syntactic position of the associated noun phrase, and thereby the relevant interpretive mechanism for achieving either a wide or narrow scope construal; (iii) the discourse contexts in which the test sentences are presented. Using a truth value judgment task, I show that both 4-year-olds and adults display isomorphic preferences in their interpretation of ambiguous sentences containing the floating quantifier all, no matter its surface position. In the case of both, children and adults display a preference for isomorphism only when both precedes negation. Crucially, for both quantifiers, children and adults display the same interpretive preferences, lending further support to the general view that children and adults do not differ in their grammatical representations of such scopally ambiguous sentences (Musolino & Lidz 2003, 2006; Gualmini 2004; Conroy, Lidz & Musolino 2009).

Notes

¹ Note that I do not argue for the isomorphism effect as a constant across all quantifiers; rather my goal is to show that when various factors are held constant, children and adults essentially pattern alike—whether their baseline preferences reflect isomorphic readings or not.

² As pointed out by an anonymous reviewer, many models of children’s resolution of scope ambiguities have been proposed since the original observation of the isomorphism effect (see, for example, Crain, Ni & Conway’s [1994] and Notley et al.’s [2012] Semantic Subset Principle, as well as Gualmini et al.’s [2008] Question Answer Requirement model). The focus of the present article will not be to provide evidence for or against these models. Rather, the goal is simply to show that when we test the same quantifier while keeping constant various unrelated factors, children and adults are alike in their interpretive preferences of scopally ambiguous sentences. The question of how these baseline preferences are derived (e.g., under competing models of scope ambiguity resolution) is beyond the scope of this article.

³ There has not been a great deal of previous research on children’s acquisition of the floating quantifiers both and all. Labelle & Valois (2001), for instance, examined the French floating quantifier chacun ‘each,’ but their study was concerned with whether children correctly allowed the floating quantifier to quantify over the subject but not the object. Our test sentences will all involve quantification of both/all over the subject noun phrase, and children’s responses, in particular their justifications, will reveal that they have no problems quantifying over the subject NP. As another example, Roeper et al. (2007) discussed English all and German alles, but specifically in the context of triggering exhaustive interpretations of wh-questions; thus their findings are not directly relevant to the present study. The fact that there is little previous work on children’s knowledge of both and all makes it all the more important to have floating quantifier control trials; as we will see in Experiments 1 and 2, these provide a basic measure of children’s ability to interpret the floating quantifiers, particularly without any interfering interactions with negation. Success on these control trials, i.e., targetlike interpretation of sentences such as All the girls fed the pandas and The girls all fed the pandas, will provide a general measure of children’s knowledge of the floating quantifier and give us an objective criterion for including the children’s data in the analysis.

⁴ Prosody likely has the potential to bias toward one reading over another (I thank Kamil Ud Deen for raising this point). In particular, whether or not the floating quantifier is stressed, and whether there is a small pause following the noun phrase, both appear to influence one’s judgments about which reading is more plausible. This is likely tied to issues of focus, which are orthogonal to the goals of the present study. To avoid these potential complications, the researcher who recorded the puppet’s lines was instructed to do so using as neutral (but natural) an intonation as possible, in particular without stressing the floating quantifier, and without any marked pauses within the sentences.

FIGURE 1 Final image accompanying the puppet’s utterance of the test sentence All the boys didn’t climb the tree/The boys all didn’t climb the tree/The boys didn’t all climb the tree.

⁵ An anonymous reviewer points out that the age range of children tested was relatively large (3;10–5;08). In particular there were nine children under the age of 4 years, 10 children between 4;00–4;05, 14 children between 4;06–4;11, and five children over the age of 5 years. A two-way ANOVA on the children’s data revealed a significant main effect of quantifier position, , , no effect of age, , , and no significant interaction, , .

FIGURE 2 Proportion of no-responses by position of the quantifier all. Note that no-responses corresponded to isomorphic readings in the prenominal and floated prenegation conditions, and to nonisomorphic readings in the floated postnegation condition.

⁶ Twenty-one of the 42 children also participated in Experiment 1. There was a temporal overlap in the two experiments, such that some children participated in the both experiment first, while others participated in the all experiment first. To minimize any contaminating effects between the two experiments, the two sessions were always spaced apart by at least five weeks (the average time elapsing between the two experiments was 9.5 weeks).

⁷ Here it was slightly less clear that follow-up justifications would allow us to unambiguously tease apart the two readings. We expected follow-up justifications for no-responses (WS) to make reference to the character who had carried out the action (e.g., it’s false that both of them failed to carry out the action because one of them did carry out the action). On the other hand, justifications for yes-responses (NS) might also make reference to the character who had carried out the action but highlight the fact that he was the only one who had done so (e.g., it’s true that it’s not the case that both of them carried out the action because only one of them did so).

⁸ An anonymous reviewer points out the relatively large age range of the children tested (3;10–6;08). In particular, there were seven children under the age of 4 years, 12 children between 4;00–4;05, 14 children between 4;06–4;11, and 9 children over the age of 5 years. A two-way ANOVA on the children’s data revealed no effect of quantifier position, , , no effect of age, , , and no significant interaction, , .

⁹ An anonymous reviewer points out that given that contextual manipulations have been shown to influence children’s and adults’ interpretations of scopally ambiguous sentences, perhaps we might be able to reverse the preferences observed here for both. I agree with this speculation that with the right contextual manipulations, we might be able to prompt children and adults to show a preference for the nonisomorphic reading in the prenominal and floated prenegation conditions and for the isomorphic reading in the postnegation condition.