Against a unified treatment of obligatory trigger effects: evidence from acquisition

Introduction: It has been known since at least Heim (1991) that presupposition triggers are sometimes obligatory in contexts where the presuppositions are satisfied (1a); failure to use the relevant trigger results in an inference, an anti-presupposition, that its presupposition does not hold (1b). To account for such cases, Heim (1991) proposed the pragmatic principle Maximize Presupposition (MP), which has since been recruited to account for a range of phenomena involving competition between two lexical items that differ only in their semantic presuppositions (Percus 2006, Chemla 2008, Sauerland 2008, a.o). Beyond cases like (1), we find obligatory trigger effects in sentences like (2), in which the triggers do not have an obvious lexical competitor. One line of analysis claims that the cases in (2) follow from the same principle of MP, taking elements like too as competing with a null morpheme (Chemla 2008). Another line of analysis takes (1) and (2) to be fundamentally different, with too being inserted in (2) to counteract unwanted exhaustivity implicatures that would be generated otherwise (Krifka 1999, Saebo 2004, Bade 2016). For instance, Bade (2016) argues that too is inserted to block the otherwise obligatory insertion of an EXH-operator (c f . Magri 2011), which interprets the second sentence in (2) exhaustively wrt. the QUD and create a contradiction. (1) a. John broke both of his legs. b. # John broke all of his legs. (2) a. John came to the party. Bill did, too. b. # John came to the party. Bill did. The MP-analysis is complicated by the fact that unlike the more straightforward cases, obligatoriness of too disappears in certain embedded contexts (3). (3) a. John came to the party. It’s not the case that Bill did {3too, 3 / 0}. b. John broke both of his legs. It’s not the case that Bill broke {3both, 7all} of his legs. On the other hand, the unwanted implicatures approach has trouble making sense of the fact that the effect unexpectedly persists in certain embedded contexts. In (4), for instance, the trigger appears in a DE-environment, where the insertion of an EXH-operator is not usually expected (Chierchia et al. 2012). Moreover, the trigger is part of a conjunction, which independently marks additivity, raising the question of why too’s putative contribution of blocking exhaustive interpretations is not redundant in this case. (4) If John came to the party and Bill did {too, # / 0}, it would have been a good time. The resulting state of affairs is one in which it seems difficult to adjudicate between the two theories on the basis of the intuitive judgments discussed in the literature. Our goal in this paper is to offer evidence of a different sort, namely from child language, that contributes novel arguments to this theoretical debate. The two theories make diverging predictions about the developmental trajectory of sensitivity to obligatory trigger effects in (1) vs. (2). If (1) and (2) are governed by the same underlying principle, e.g. MP, all else being equal, we expect children’s abilities with both environment to emerge at the same time. On the other hand, if the effects have independent sources, we might see reflexes of this distinction in development. Present Study: We compare children’s sensitivity to obligatory trigger effects with both vs. too (a between-subjects factor) in a Felicity Judgment Task. Sentences with and without the trigger served as alternative descriptions of a pictured event, as illustrated in Figure (1). Afterwards, two characters described what happened and the child had to judge which one described it better. In the critical