Presentation
Search Abstracts | Symposia | Slide Sessions | Poster Sessions | Lightning Talks
Rapid use of argument roles in verb prediction: Evidence from ERPs using a comprehension-production interleaved trials design
Poster B12 in Poster Session B, Tuesday, October 24, 3:30 - 5:15 pm CEST, Espace Vieux-Port
Eun-Kyoung Rosa Lee1, Colin Phillips1; 1University of Maryland
Previous findings in comprehension paradigms suggest that expectations about upcoming verbs are insensitive to argument roles (i.e., comparable N400 amplitudes to verbs in role-appropriate and role-reversed contexts)[1,2,3]. However, production studies using a speeded cloze paradigm suggest the opposite, with verb continuations showing clear sensitivity to argument roles[4]. To test whether this conflict is due to different tasks or to different measures, we conducted an interleaved comprehension-production study. Results place the blame on task differences. When we recorded ERPs in a setting where comprehenders were ready to speak, we found argument role sensitivity that has rarely been seen in prior ERP studies. A total of 24 English-speaking adults participated in the experiment. Materials were taken from a previous study that did not find the N400 effect to role-reversals [Role-appropriate: The restaurant owner forgot which customer the waitress had served…; Role-reversed: The restaurant owner forgot which waitress the customer had served...; Control: Abby brushed her teeth after every meal (plausible) / game (implausible)]. Sentences were truncated before the verb for production trials. Participants read sentences word-by-word and either judged the plausibility of sentences or produced cloze completions when a probe was given. Crucially, the comprehension and production trials were interleaved and randomized such that participants did not know in advance which task was required, until they either saw the probe eliciting a cloze response or reached the end of the sentence, motivating them to be prepared to generate a continuation at any time as a sentence unfolded. The N400 responses to target verbs in the comprehension trials and rate of role-reversed verb completions in the production trials were analyzed. The production results replicated earlier findings; role-reversed verbs were rarely produced as cloze completions (6%), and when produced, they had slower onset latencies than when produced in the role-appropriate contexts (+250 ms). These findings suggest that argument role information was actively used to constrain productions to role-appropriate verbs. Importantly, the comprehension trials revealed a significant N400 effect to role-reversals; in the 350-500 ms time window, cluster-based permutation tests revealed significant negative clusters between the role-appropriate and role-inappropriate conditions (cluster t-statistic = -544, p = .001; cluster t-statistic = -107, p = .03). There were no positive clusters found in the P600 time window. The control conditions elicited a significant N400 effect (cluster t-statistic = -2626, p < .001) as expected. The results together suggest that engaging in a speeded cloze production task which requires rapid generation of candidates for upcoming words and selection of a response can result in the rapid use of argument roles to constrain initial predictions to role-appropriate verbs. This may occur through increased activation of role-appropriate candidates or stronger inhibition of role-inappropriate candidates that were initially activated based on lexical associations. The findings highlight the effect of task-related goals and strategies on prediction during real-time sentence processing[5]. References: [1] Chow et al. (2016). LCN. [2] Kuperberg et al. (2003). CBR. [3] Kim & Osterhout (2005). JML. [4] Lee et al. (2022). 35th HSP. [5] Brothers et al. (2017). JML.
Topic Areas: Language Production, Syntax and Combinatorial Semantics