Academic journal article Canadian Journal of Experimental Psychology

Thematic Roles in Sentence Parsing

Academic journal article Canadian Journal of Experimental Psychology

Thematic Roles in Sentence Parsing

Article excerpt

Abstract Two eyetracking experiments examined the reading of sentences like "While the police/truck stopped the Datsun disappeared into the night." A paper by L. Stowe (Thematic structures and sentence comprehension. In G. N. Carlson and M. K. Tanenhaus (Eds.), Linguistic structure in language

processing, 1989) indicated that an inanimate subject ("truck" in "After the truck stopped the Dastun disappeared...") is taken as theme of the ergative verb ("stop"), preventing the assignment of the postverbal noun phrase ("the Datsun") as direct object. This eliminates the disruption of reading that is normally observed on the disambiguating verb ("disappear"). The present experiments found the pattern of results reported by Stowe when looking at the disambiguating region of a sentence. However, the results for earlier regions suggest that the postverbal noun phrase was initially taken as direct object of an ergative verb even when the subject was inanimate. It appears that the inanimacy of the subject may not have guided the initial syntactic analysis, but rather facilitated the revision of an initial misanalysis.

Resume Deux experiences permettant de capter les mouvements oculaires ont porte sur la lecture de phrases telles « While the police/truck stopped the Datsun disappeared into the night ». Dans un document de L. Stowe (Thematic structures and sentence comprehension. G.N. Carlson and M.K. Tanenhaus (Eds.), Linguistic structure in language processing, 1989), il est indique qu'un sujet inanime (« truck » dans « After the truck stopped the Datsun disappeared (...) »), est considere comme le theme du verbe ergatif (« stop »), empechant ainsi d'attribuer au syntagme nominal postverbal (« the Datsun ») le role d'objet direct. Cette approche elimine la perturbation de la lecture qu'on observe normalement lorsqu'intervient le mot desambiguisant (« diseappear »). Les experiences que nous avons menees ont conduit aux resultats signales par Stowe pour ce qui est du segment desambiguisant de la phrase. Par contre, les resultats se rapportant aux premiers segments nous portent a croire que le syntagme nominal postverbal etait d'abord vu comme le complement d'objet direct d'un verbe ergatif meme quand le sujet etait inanime. Il semble donc que le caractere inanime du sujet n'ait peut - etre pas guide l'analyse syntaxique initiale, mais qu'il ait plutot facilite la revision de l'analyse initiale erronee.

Two of the principal questions that have driven research on sentence comprehension during the past decade concern (a) the role that discourse information, meaning, and world knowledge play in assigning a structural analysis to a sentence, and (b) the role that lexically specific information plays in such structural analysis. These questions have seemed particularly important to researchers who are interested in questions of modularity and interactive processing (e.g., Fodor, 1983; Marslen - Wilson, 1987). Some have argued for a strongly interactive system that makes decisions, efficiently and without delay, on the basis of any relevant information as soon as it becomes available (Marslen - Wilson, 1987). Others have argued for a more weakly interactive system that, for instance, initially builds several possible analyses on the basis of purely syntactic information but quickly uses discourse information to select among them (Crain & Steedman, 1985; Altmann & Steedman, 1988). Still others have argued for a highly modular system in which initial decisions about what structures to build are based solely on grammatical information, and in which other sources of information are used to evaluate these decisions and to guide reanalysis (Frazier, 1987).

The role of extragrammatical information (including discourse information, meaning, and world knowledge) in sentence comprehension bears on these distinctions. Any language processing system must have some way of using all relevant information if it is to be an adequate model of the human system. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.