Search Abstracts | Symposia | Slide Sessions | Poster Sessions
Production and Perception Asymmetries in Phonological and Morphosyntactic Processing Decoded with Scalp EEG-MVPA
Poster Session A, Wednesday, September 30, 11:00 am - 1:00 pm, Wangari Maathai
Jacob Burger1, Neemias Souza Filho1, Azia Knox1, Abdulaziz Alromihi1, Arild Hestvik1; 1University of Delaware
We investigated whether there are discrete, temporally separated processing stages during speech production for abstract morphosyntactic affixation followed by phonological spell-out (e.g. a zero affix for singular and an overt -s for plural), and whether those stages are mirrored during perceptual representation building. Previous iEEG (Sahin et al., 2009) and MEG (Hauptmann et al., 2022) studies found evidence for these stages. However, these studies implicitly use morphosyntactic violations cueing participants to produce a corrected form of a word. In perception, these violations are indexed by reintegration processes such as the Left Anterior Negativity (LAN); it is unknown whether representing these violations is analogous between production and perception. Implementing a similar paradigm using scalp EEG, we asked: Will a LAN index morphosyntactic processing prior to an index of phonological spell-out in the same way during production as it does for perception? Participants (N=27; Female=25; Average Age=19.9) completed one block of a perception task, and one block of a production task, order counterbalanced within participants. In both blocks participants saw an inflectional cue followed by a target word. The inflectional cue had one of four different feature priming frames: “That is the __.” (singular), “Those are the __.” (plural), “Nowadays they __.” (present), or “Yesterday they __.” (past). Targets matched the cue in terms of word category, but were either congruently or incongruently inflected (balanced for length and Zipf frequency). Equal amounts of regular and irregular inflection were included to investigate different phonological spell-out processes (regular affixation versus suppletion). For perception, participants indicated via button press whether the target was good or bad in context of the cue. For production, participants overtly produced the target with correct inflection. EEG data was collected with 65 channels, bandpass filtered 0.1-40Hz, segmented to the target, artifact corrected, baseline corrected, and average rereferenced. Average RT was 655ms for production and 758ms for perception. ERP data was analyzed in two ways: a traditional ERP analysis investigated for a LAN, and a Multivariate Pattern Analysis (MVPA) investigated the timecourse of congruence and regularity. The incongruent-congruent difference wave reveals an anterior negativity peaking at 620ms for production, and 692ms for perception. An MVPA classifier trained to decode incongruent versus congruent trials revealed significant decoding from ~600ms-1100ms for production, whereas decoding for perception was sustained beginning at 550ms. While both blocks produced an anterior negativity, the timecourse is later than the typical timeframe for a LAN (300-500ms), suggesting late integration of morphosyntax. Training the classifier to distinguish regular versus irregular inflection shows decoding at ~480ms for production, whereas decoding for perception begins at 1000ms. For production, this shows phonological processing earlier than morphosyntactic, which was found by Hauptmann et al. (2022), but Sahin et al. (2009) found the reverse order. Late decoding for perception suggests this task does not require representing differences between inflectional types. Cross-decoding between the production and perception blocks showed no significance for either comparison. This indicates that the underlying neural representations failed to generalize between production and perception, supporting separate neural pathways for both.
Topic Areas: Language Production, Morphology