Sentence and text processing

Saturday, October 01st,   2011 [17:20 - 19:20]

PS_2.107 - Eye tracking evidence for pronoun resolution in students with intellectual disability

Tavares, G. , Fajardo, I. , Ávila, V. , Ferrer, A. & Salmerón, L.

Department of Developmental and Educational Psychology. University of Valencia. Valencia, Spain.

This study investigates how students with intellectual disability and low reading comprehension levels process and resolve anaphoric pronouns. The influence of the order of mention of the antecedent and its syntactic function in the resolution of ambiguous pronouns in Spanish texts were tested by means of the analyses of readers’ eye movements. Eighteen participants with intellectual disability read 32 counterbalanced mini-stories of two sentences in a self-paced reading task. The first sentence contained 2 proper names. The second sentence began with a subject pronoun (the anaphor) referring to either the subject or the object of the first sentence. The arguments immediately following the pronoun disambiguated it. The study followed a 2 (antecedent position: first mentioned name vs. second mentioned name) x 2 (antecedent function: subject vs. object) within-participant design. For several pre-defined areas of interests (object and subject of the first sentence and disambiguation area in the second sentence), the number of regressions, gaze durations, first fixation, regression path and total fixations were calculated. The congruence of the preliminary analysis of these measures with the use of a less costly general cognitive strategy for pronoun referent assignment (first mentioned account) versus a grammatical analysis (subject preference account) is discussed.

PS_2.108 - Can cognates modulate language switching costs in sentence context?

Bultena, S. ¹ , Dijkstra, T. ¹ & Van Hell, J. ^{2, 3}

1 Donders Institute for Brain, Cognition and Behaviour. Radboud University. Nijmegen, The Netherlands.
2 Department of Psychology, The Pennsylvania State University. University Park, PA, USA.
3 Behavioural Science Institute. Radboud University. Nijmegen, The Netherlands.

Switching between languages in production or recognition is associated with a measurable cost. Yet, numerous studies have shown that bilinguals access lexical representations from both languages in parallel. This is especially clear for cognates, which are translation equivalents with form-overlap, such as the Dutch-English word FILM. Cognates co-activate representations in two languages and therefore provide an interesting test ground for language switching. According to Clyne’s (2003) trigger hypothesis, cognates may facilitate codeswitching in the speech of habitual codeswitchers. This is corroborated by recent evidence that noun cognates reduce switch costs in reading. We examined the trigger hypothesis using an unexplored word class, verb cognates, and studied how verb cognates affect language switching in sentences. In a shadowing task, Dutch-English bilinguals were presented with sentences that could start in L1 Dutch or in L2 English; the verb prior to the switch was manipulated for cognate status. Although switching from L1 to L2 showed no effect of the cognate, latencies of slow shadowers indicated that switching from L2 to L1 was easier when the switch was preceded by a cognate verb rather than a noncognate control verb, suggesting that verb cognates can to some extent modulate switch costs in sentence context.

PS_2.109 - Stereotypical and grammatical gender cues in pronoun resolution: Evidence from German

Esaulova, Y. , Reali, C. & Irmen, L.

Department of Cognitive and Theoretical Psychology. University of Heidelberg. Heidelberg, Germany.

In this eye-tracking study we investigated the processing of grammatical and stereotypical gender cues in pronoun resolution. Materials contained a gender typical role name (e.g., electrician, beautician) in masculine or feminine grammatical gender as an antecedent in the first clause and a pronoun (he or she) as an anaphor in the second. In both early and late measures, masculine pronouns were fixated longer after an incongruent feminine as compared to a masculine antecedent. No reliable effect of antecedent gender was found for feminine pronouns. Typicality did not affect pronoun resolution but role name processing. Fixation times were shorter when the typicality and grammatical gender of the role name were congruent as compared to the incongruent case. This effect was reliable in early stages of processing for typically male role names. Results show that pronoun resolution is mainly guided by grammatical features, such as grammatical gender, whereas role name processing also comprises conceptual information from semantic memory, such as gender typicality. However, masculine and feminine grammatical gender constrain pronoun resolution to a different extent.

PS_2.110 - Emotional content of words in sentences modulates syntactic processing

Fernández Hernández, A. ¹ , Martín-Loeches, M. ^{1, 2} , Casado Martínez, P. ^{1, 2} , Jiménez-Ortega, L. ^{1, 2} & Fondevila, S. ¹

1 Center for Human Evolution and Behavior,UCM-ISCIII, Madrid, Spain
2 Psychobiology Department, Complutense University of Madrid, Madrid, Spain

In the present study we investigate the influence of the emotional valence of words in the syntactic and semantic processing of a sentence by means of event-related brain potentials (ERP). Whereas the possibility is open that emotional content may affect semantic processes due to their heuristic and general computation nature, syntactic processes would be unaffected, as they are often considered algorithmic and encapsulated. The ERP were time-locked to correct or incorrect (50%) adjectives (target word) in sentences with the structure Det-N-Adj-V while subjects performed a correctness judgement. The adjective could also be positive, negative, or neutral, but there was no mention about the emotional content of the words. Results showed a strong effect of emotional content in the words only in the syntactic condition. Words with negative valence elicited larger amplitudes in both early (Left Anterior Negativity) and late (P600) components that emerged to syntactic violations. Semantic processes (N400), in contrast, appeared unaffected. Our results suggest that syntactic processing may share resources with other processes, i.e., that it is not as encapsulated as reported before. The relevance of emotional information in language processing is also supported, adding to recent, yet scarce evidences in this line.

PS_2.111 - Affectedness as a factor at the semantics/syntax interface in sentence processing: ERP data

Rausch, P. . ^{1, 2, 3} , Krifka, M. ^{2, 4} & Sommer, W. ³

1 Berlin School of Mind and Brain. Humboldt-Universität zu Berlin. Berlin, Germany.
2 Department of German Language and Linguistics. Humboldt-Universität zu Berlin. Berlin, Germany.
3 Department of Psychology. Humboldt-Universität zu Berlin. Berlin, Germany.
4 Centre for General Linguistics. Berlin, Germany.

From a sentence like 'The doctor cured Mary', we can infer that Mary necessarily underwent a change of state in the course of the event expressed by the verb 'cured', while the same is not true in a sentence like 'The doctor treated Mary'. Verbs thus imply different degrees of 'affectedness' for their object arguments and this factor is one of the semantic key determinants for verb-argument linking at the semantics/syntax interface. To assess the impact of affectedness for this linking process during online sentence processing, we conducted a self-paced-reading experiment and an ERP study. To minimize the influence of sentence internal syntactic cues, we used German deverbal event nominalizations of verbs implying different affectedness levels (e.g. 'admiration-treatment-assassination') embedded in sentence contexts and followed by genitives that either referred to agents or patients of an event introduced in a context sentence. ERP data were analyzed using a wavelet-based functional mixed model. First ERP analyses converge with reading time and acceptability patterns and suggest a prominent role of a frontal P600-like component, while no effect on the N400 is found. We discuss the findings in light of recent processing studies on compositional semantics and the roles of the N400/P600 component families.