Sentence Processing

representations that capture the similarities between these sentences. Production priming has only recently been used to study the nature of children’s linguistic abstractions. Some researchers have found evidence for abstract structural priming in threeand four-year-old children (Huttenlocher et al. 2004, Song & Fisher 2004). Others have not (Gamez et al. 2005, Savage et al. 2003). Recently we developed a novel paradigm that combines structural priming and eye-gaze analyses to investigate priming during online comprehension. Since production tasks are often more difficult for children than comprehension tasks (Hirsh-Pasek & Golinkoff 1996), this may provide a more sensitive measure of children’s linguistic knowledge. Because eye-gaze paradigms provide information about how an interpretation changes during processing, this method allows us to explore the locus of the priming effect and rule out alternate explanations that have been proposed for production priming (e.g. priming of the preposition to). Critically, this technique allows us to explore the representations that children use when understanding sentences with verbs that they already know. If children have item-specific representations (as assumed under the verb island hypothesis, see Ch. 5) then we would expect priming within verbs but not between verbs. In contrast if children have abstract syntactic or semantic categories, then we would expect to see betweenverb priming. The critical sentences in these studies used dative verbs. Dative verbs, such as give, bring or send, typically appear with three arguments: an agent, a recipient and a theme. In English there are two ways in which these arguments can be expressed, as shown in (7). In the prepositional object construction (7a) the theme appears as the direct object while the recipient is expressed by the prepositional phrasemarked by to. In the double object construction (7b) the recipient is the direct object while the theme is expressed as a second noun phrase. (7) a. Tim gave a half-eaten pomegranate to Chris. b. Tim gave Chris a half-eaten pomegranate. Datives are well-suited for developmental studies of priming. The two dative constructions have the same basic meaning and differ only in how the semantic roles get mapped onto syntactic elements. Thus, priming using datives offers a reasonably clear case of structural priming independent of semantics. In addition, both dative constructions are acquired quite early; children appear to comprehend and produce both forms by age three (Campbell & Tomasello 2001, Gropen, et al. 1989). Children were given sets of trials which consisted of filler sentences, followed by two prime sentences, and then a target sentence. The primes were either direct object or prepositional object datives and the final target sentence was also a direct object or prepositional object dative. Our goal Sentence processing 333 //FS2/CUP/3-PAGINATION/CHEL/2-PROOFS/3B2/9780521883375C18.3D 334 [321–338] 30.10.2008 2:52PM was to determine whether direct object and prepositional object datives would prime the interpretation of subsequent utterances that used a different verb and had no common content words. For example, would hearing Send the frog the gift facilitate comprehension of Show the horse the book? To link this priming to eye movements we made use of a well-studied phenomenon in word recognition, the cohort effect (Marslen-Wilson & Welsh 1978). As a spoken word unfolds, listeners activate the lexical items that share phonemes with the portion of the word that they have heard. In the visual world paradigm, this process results in fixations to the referents of words that share phonemes with the target word (Allopenna et al. 1998). These effects are particularly strong at the beginning of a word, when all of the phonological information is consistent with multiple words (the members of this cohort). In our studies we used priming as a top-down constraint whichmight modulate the activation of different members of a phonological cohort. The target trials were either double object (8a) or prepositional datives (8b). (8) a. Bring the monkey the hat. b. Bring the money to the bear. The set of toys that accompanied the utterance contained two items that were phonological matches to the initial part of the direct object noun. One was animate and hence a potential recipient (e.g. a monkey) while the other was inanimate and hence a more likely theme (e.g. some money). Thus the overlap in word onsets (e.g. mon ...) created a lexical ambiguity which was tightly linked to a short-lived ambiguity in the argument structure of the verb. We expected that priming of the direct object dative would lead the participants to interpret the first noun as a recipient, resulting in more looks to the animate match, while priming of the prepositional object dative structure would lead them to interpret it as a theme, resulting in more looks to the inanimate match. To validate our paradigm, we began by examining priming between utterances which shared the same verb (within-verb priming). Since both item-based grammars and abstract grammars posit shared structure between utterances with the same verb, within-verb priming would be consistent with either theory. We found that young four year olds showed robust within-verb priming during the ambiguous region. Young three year olds were slower in interpreting the target sentences, but when we expanded the analysis window to include the whole sentence, we found a reliable priming effect. Childrenwho had heard double object primeswere more likely to look at the potential recipient (the monkey) than children who had heard the prepositional object primes. To examine the nature of the structures that children use, we conducted parallel experiments in which the prime and target utterances had no 334 J E S S E S N E D E K E R //FS2/CUP/3-PAGINATION/CHEL/2-PROOFS/3B2/9780521883375C18.3D 335 [321–338] 30.10.2008 2:52PM content words in common (between-verb priming). Under these circumstances the abstract grammars predict priming, while item-based grammars do not (see figure 18.4). We found that both young four year olds and young three year olds showed between-verb priming. In the three year olds the effect of between verb-priming was almost as large as the effect of withinverb priming, indicating that there was no benefit gained when the two utterances shared a verb. This would suggest that abstract representations play a dominant role in online comprehension in this age group. (a) Abstract structural representations Show the lion the ball Verb Recipient Theme V NP NP Verb Theme Recipient V NP PP Give the horse the ... prime