Scalar implicatures are inferences that strengthen what is sometimes called the “basic meaning” of the sentence:
(1) John ate some of the cookies
(1a) Basic Meaning: that John ate some, possibly all, of the cookies
(1b) Scalar Implicature: that John did not eat all of the cookies
(1c) Strengthened Meaning: that John ate some but not all of the cookies (BM + SI)
This strengthening has been shown to generate various detectable “signatures,” some of which are highlighted in (2):
(2) SI Signatures
(2a) SIs tend to disappear in DE environments (e.g., the restrictor of “every”).
(2b) SIs are detectable, but not very robust, in non-DE environments (e.g., the scope of “every”).
(2c) SIs are processed slow: (1a) is processed faster than (1c) (cf. Bott & Noveck, 2004; and much work since).
(2d) SIs show up late in acquisition: There is a stage of development at which children behave as if they assign (1a) to (1) but do not assign (1c) to (1) (cf. Noveck, 2001; and much work since).
So-called “free-choice” inferences, exemplified in (3), have been shown to also disappear in negative environments. Taking this to be one of the signatures of an SI (cf. (2a)), it has been argued that free-choice inferences should be derived in the cognitive system that computes SIs (e.g.,Kratzer & Shimoyama, 2002; Schulz, 2005; Alonso-Ovalle, 2005).
(3) John may eat the cookies or the pie
(3a) Basic Meaning: that John is allowed to eat one, and possibly both, of the cookies and the pie
(3b) Free-Choice: that John is allowed to eat the cookies and he is allowed to eat the pie
In stark contrast with the SI in (1), however, free-choice (3b) is not processed slower than (3)’s basic meaning (3a) (cf. (2c); Chemla & Bott, 2014), and free-choice (3b) is preferred to the basic meaning (3a) in positive embeddings, such as in the nuclear scope of “every” (cf. (2b); Chemla, 2009).
In this talk, I present evidence that free-choice and SIs also have diverging developmental signatures (cf. (2d)). Specifically, I present evidence that children (3;9-6;4, M = 4;11) compute conjunctive free-choice SIs for disjunctive sentences (reporting on joint work with Ken Wexler, Andrea Astle, Deepthi Kamawar, and Danny Fox). Our finding replicates earlier results showing that children often interpret disjunctions as if they were conjunctions (Paris, 1973; Braine and Rumain, 1981), and extends this to embedding in the scope of “every.” We argue that this conjunctive SI follows from: (i) Katzir’s (2007) theory of alternatives in the steady state, (ii) the assumption that children differ from adults by not accessing the lexicon when generating alternatives, and (iii) Fox’s (2007) mechanism for free-choice computation in the steady state. We further provide evidence that children at this stage of development share the adult preference for free-choice SIs in matrix and embedded positions.
These data raise the challenge of explaining why free-choice and SIs both disappear in negative environments but differ with respect respect to developmental trajectories, embeddability, and processing complexity (see Chemla & Singh, 2014 for generalizations to other scalar items). I will explore strategies for addressing this challenge.