Poster Slam Session E
Saturday, August 18, 2:45 – 3:00 pm, Room 2000C, Chair: Patti Adank
Interaction of morphological and long-distance dependency processing: MEG evidence
Suhail Matar1,2, Ryan King1, Wesley Leong1, Liina Pylkkänen1,2;1New York University, 2NYU Abu Dhabi Research Institute
INTRODUCTION Comprehending an entire sentence requires composition at multiple levels, from morphemes to words, phrases to larger phrases, and the resolution of many types of dependencies. The speed at which all this happens suggests that these mechanisms co-occur and overlap. But while the brain bases of the various levels have all been studied individually, their interaction is not understood. METHOD We addressed how long-distance dependency formation interacts with morphological complexity. Specifically, semantically transparent compounds, opaque compounds and monomorphemic nouns were contrasted in the filler site of so-called right-node-raising constructions, a long-distance dependency in which the filler follows the gap, as in (1) (Leiken et al., 2015). (1) The executive board ridiculed/congregated and the manager criticized the teamwork/chairman/campaign during the meeting. When the verb of the first conjunct was transitive (but not when it was intransitive), the object of the second conjunct had to be linked back to the object position in the first conjunct. MEG measurements (N = 22) on the comprehension of such expressions (46 sets) allowed us to test the spatial and temporal organization of the requisite morphological and syntactic processes, both separately and together. Our analyses covered the entire sentence but here we report only on our results for the target nouns, for space reasons. ROIs included regions previously sensitive to long distance dependencies (LIFG: BA44, BA45, BA47), compound composition (LATL: BA38), and lexical access and syntactic composition (left BA21, BA22, BA37). RESULTS All our effects occurred within the 350-450ms time window, indicating that this time-window is the main temporal locus of a variety of syntactic and morphological operations. Morphological complexity: As in prior work (Brooks & Cid de Garcia, 2015; Flick et al., 2016, subm.), the LATL showed larger amplitudes for transparent compounds compared to opaque compounds or monomorphemic targets at ~350-400ms. A similarly trending pattern was observed in the LIFG at 375-410ms. Dependency: Instead of eliciting increased activation, sentences with dependencies trended towards lower activation both in the LIFG and BA22 at ~400-450ms. Most likely, this pattern reflects not dependency formation, but rather more straightforward composition operations when no dependency needs to be formed. Interaction: In multiple temporal and prefrontal regions, dependency and morphology interacted significantly around 400ms in the following, somewhat complex, way: For monomorphemic and transparent compound targets, activity was higher in non-dependency contexts (following the trending main effect described above), but for opaque compounds, the opposite was true: there was greater activity for the dependency conditions. Thus, semantic opacity robustly impacted the effect of our dependency manipulation in the 400-450ms time-window, suggesting that retrieval mechanisms of the opaque meaning may occur at this time. CONCLUSION In this study, the effects of two very different composition operations converged on a single word, revealing a morphological effect in the LATL and, in an overlapping time-window, a broadly distributed interaction between morphological complexity and dependency resolution in left temporal and prefrontal cortices. We have provided evidence that language processing mechanisms do not simply co-occur, but rather interact in complex ways.
Topic Area: Grammar: Syntax