It has been claimed that there are two main approaches in studies of language acquisition; the Constructionist Approach (Goldberg 2003, Tomasello 2000, 2003 among others), and the Generative Approach. This study reports some experimental evidence that supports the latter rather than the former.

In recent work, Tomasello proposes an input-matching ("usage based") model of language acquisition. According to the model, preschool children use "analogy" to learn their first language, based on the input (i.e., this is a 'what you see is what you get' approach).  The model hypothesizes strictly local combinations of expressions and conservative learning model. Later children form abstract "constructions", which are built from the item-based schemas. Assuming this model, we can predict that children would "overgenerate" in the interpretation of the constructions that cannot be grasped with such local strategies.

We challenge this model, using principles of formal semantics to adjudicate between the two alternative accounts of children's interpretation of the universal quantifier. Two experiments were designed to see whether children are sensitive to the semantic property of downward entailment, which distinguishes between the first argument and the second argument of the universal quantifier every in adult languages. In each study, 'extra' objects were made highly salient in the experimental contexts, to evoke non–adult responses if these were permitted in child grammars. In addition, the contexts were constructed to satisfy the felicity conditions of the task, as prescribed by Crain et al. (1996).

The first experiment investigated children's awareness that the first argument of the universal quantifier is downward entailing (licensing inferences from a set to its subsets), while its second argument is upward entailing. Consequently, (1b) can be true even if (1a) is false, because the underlined NPs appear in the first argument of every. The opposite pattern of inferences obtains when the NPs appear in the second argument; here, (2a) can be true even if (2b) is false.

(1)   a. Every troll has a potato chip.              (False)

       b. Every purple troll has a potato chip.    (True)

(2)   a. Every smurf caught a bug.                  (True)

       b. Every smurf caught a blue bug.           (False)

The situations depicted in (1) and (2) were incorporated into a Truth Value Judgment task (Crain and McKee 1985; Crain and Thornton 1998). The experiment involved 20 children (mean age 4;10). For sentences like (1), children rejected the (a) examples 100% of the time but the same children accepted the (b) examples 90% of the time. Just the reverse pattern was manifested by children for sentences like (2); they accepted the (a) examples 95% of the time and rejected the (b) examples 100% of the time. Moreover, children never pointed to the salient 'extra' objects to justify their rejections in either of the test conditions.

The second experiment was based on the observation that the disjunction operator or generates different entailments depending on the properties of the environment it occurs in (e.g., Partee et al. 1990). Thus, when the disjunction operator or appears in the second argument, as in (3), the truth conditions are those associated with the exclusive-or reading of disjunction, but only a 'conjunctive' reading is possible when or appears in the first argument, as illustrated in (4).

(3)  Every child picked a tiger or a dinosaur        

                  picked a dinosaur.

(4)  Every lady who bought an egg or a banana got

      a basket.

      --> Every lady who bought an egg got a basket and

         every lady who bought a banana got a basket.

To assess children's knowledge of this distinction in the interpretation of disjunction, we conducted an experiment, again using the Truth Value Judgment task, with twenty English-speaking children (age: 3;11-5;8) In the first condition, children were presented with sentences like (3) in a situation in which one child had picked a tiger, one had picked a dinosaur and one had picked both. If children know that the second argument of every is not downward entailing, they should accept the target sentences. By contrast, if children do not know the entailment properties of the second argument of every, they should reject them. In the second condition, children were presented with sentences like (4) in a context in which only those girls who had bought an egg received a basket. If children know that the first argument of every is downward entailing, they should reject the target sentences. By contrast, if children do not know the entailment properties of the first argument of every, then they should accept them. The finding was that children consistently accessed the conjunctive interpretation of disjunction for sentences like (4), but almost never for ones like (3). Again, no child ever pointed to the 'extra' objects as a reason for rejecting any of the test sentences. 

In conclusion, we revealed that preschool children are aware of the asymmetric entailment relations of every, and of the environment-sensitive nature of the entailments it generates. Therefore, the data support the Generative Assumption, according to which children have access to the same linguistic principles that govern adult languages.