Revisiting the processing of English wh-dependencies – an EEG study
Poster D29 in Poster Session D with Social Hour, Friday, October 7, 5:30 - 7:15 pm EDT, Millennium Hall
Keng-Yu Lin1, Edith Kaan1; 1University of Florida
English wh-dependency sentences contain a wh-word that must be integrated with a thematic-role assigner downstream (e.g., Who[i] did the clerk help ___[i]?). Memory/resource-based approaches claim that the wh-word is stored in memory before it can be integrated, with longer storage being more cognitively costly (e.g., Just & Carpenter, 1992; King & Just, 1991). While integration of the wh-word has been associated with a P600 (Kaan et al., 2000), it is debated whether the extra cognitive/memory costs of the dependency distance are reflected in the sustained anterior negativity (SAN) and/or (left) anterior negativity ((L)AN, Fiebach et al., 2002; Phillips et al., 2005). We therefore revisited this issue, hoping to replicate the P600 difference in wh-dependency vs. non-wh-dependency and to examine whether extra cognitive costs of the dependency distance could be observed in SAN and/or (L)AN. We conducted a 2x2 design experiment, with the factors wh-word (whether, who) and distance (short, long), as exemplified in “Sam wondered whether the clerk (at the counter) had helped the customer in the shoe aisle” vs. “Sam wondered who[k] the clerk (at the counter) had helped ___[k] in the shoe aisle.” The long-distance condition contained a prepositional phrase (at the counter) between the wh-word and the thematic-role assigning verb (helped) while the short-distance condition did not. A total of 144 experimental sentences (36 per condition) and 144 distractor sentences were latin-squared and were visually presented word-by-word (SOA=500 ms, ISI=200 ms). EEG data from 29 monolingual English speakers (age 18-27) were preprocessed (EEGLAB/ERPLAB) and were analyzed using linear mixed effects models (R/RStudio). Models included wh-word, dependency distance, and their interactions as fixed effects, and by-participant and by-item random intercepts. We started with a maximum random effect structure, but reduced the number of random slopes until the model converged without warnings. The SAN, (L)AN, and P600 were examined. The SAN was time-locked to the onset of the wh-word, spanning multiple words until the target verb (0-2500 ms for the short-distance condition, 0-4000 ms for the long-distance condition). The LAN and P600 were time-locked to the verb onset and were examined between 300-500 ms and 500-900 ms, respectively. No main effects of wh-word and dependency distance and no interaction effects were found in SAN, P600, and (L)AN. Our results failed to replicate findings of previous studies in which P600 showed sensitivity to dependency vs. non-dependency processing (Kaan et al., 2000; Phillips et al., 2005), though the amplitude of the who condition (0.14 µV) was numerically larger than the whether condition (0.12 µV) in our study. Our SAN and (L)AN did not show sensitivity to the dependency difference, as is consistent with Phillips et al. (2005). Based on the results, it is unclear whether the wh-dependency and the long dependency distance imposed extra cognitive costs on processing in our study. We plan on conducting time-frequency analyses to further examine if extra processing/cognitive costs can be observed in different frequency bands.
Topic Areas: Syntax, Methods