Reading‐Time Evidence for Intermediate Linguistic Structure in Long‐Distance Dependencies

Abstract

Most linguistic theories since Chomsky (1973) have hypothesized that longdistance dependencies crossing multiple clauses are mediated by intermediate structures. This paper provides a new source of evidence for the existence of such intermediate structures: reading times during online sentence comprehension. The experiment presented here compares reading times for two structures involving the long-distance extraction of a wh-filler: (a) a structure in which a clause intervenes between the endpoints of the extraction, and (b) a structure in which a nominalization of the clause intervenes. The logic of the experiment relies on two hypotheses: first, that intermediate structures mediate the relationship between a wh-filler and its h-roleassigning verb when a clause intervenes between them but not when a nominalization intervenes; and second, that reading times for a word increase as linear distance increases between the word and the position on which it is dependent in the partial structure for the input (Gibson 1998(Gibson , 2000Grodner, Watson & Gibson 2000). In combination, these hypotheses predict that reading times at the region in which the verb assigns a h-role to the wh-filler will be faster in the clausal conditions than in the nominalized conditions, because in the clausal conditions intermediate structure mediate the wh-filler verb dependency and cause it to be more local. This prediction was confirmed.* We would like to thank Susanne Tunstall for her help in running the experiment reported here. We would like to thank the following people for useful discussions of earlier presentations of this work:

Agreement and movement: A syntactic analysis of attraction

“…Syntactic theory assumes the existence of an intermediate representation between these two configurations, in which the object passes via a position that intervenes on AGREE. The finding that this intermediate position of the object triggered attraction is, to our knowledge, the first piece of experimental evidence for the role of intermediate representations in language production (for related evidence in language comprehension see Gibson & Warren, 2004).…”

Object interference in subject–verb agreement: The role of intermediate traces of movement

“…Hence, one cannot consider a moved element to remain active in an unqualified way throughout the whole structural stretch. Second, the experimental results by Gibson and Warren (2004) showing a facilitating effect of intermediate traces 7 Hartsuiker and colleagues (2001) found weaker interference from the object (whether phrasal or pronominal) than from the prepositional phrase modifier in Dutch embedded SOV structures. This result is consistent with our approach under standard assumptions on Dutch word order.…”

“…However, these studies manipulated traces of thematic positions of moved elements, not intermediate traces, as we suggest here, which only play a role within syntax. To our knowledge, the only piece of evidence for the role of intermediate traces in language comprehension is that recently provided by Gibson and Warren (2004). The authors reported shorter reading times at the verb when the moved NP could transit via an intermediate position left empty (e.g., The manager who i the consultant claimed that the new proposal had pleased t i will hire five workers tomorrow) than when it could not because of the nature of the structural representation (e.g., The manager who i the consultant's claim about the new proposal had pleased t i will hire five workers tomorrow).…”

“…Gibson and Warren (2004) suggested that intermediate structures facilitate processing by reducing the distance of integration between a moved element and its integration site. That is, locality in formal syntax reflects processing constraints by which elements are regularly reactivated in precise positions within the syntactic structure.…”

An fMRI study on the processing of long-distance wh-movement in a second language

“…When these types of sentences are parsed, the filler is loaded into working memory until the appropriate gap is found, in order for the dependency to be resolved (Active Filler Hypothesis; . This leads to increased working memory (WM) demands (Gibson 1998), which can be expressed with longer reading times for sentences with long-distance dependencies, compared to control sentences of equal length but without wh-dependencies, such as (3) and (4) (Gibson & Warren 2004). Although both (1) and (2) contain a long-distance dependency, in (1) the filler crosses an embedded verb phrase (VP), introduced by that.…”