Presentation
Search Abstracts | Symposia | Slide Sessions | Poster Sessions | Poster Slams
Meaning composition in two-word and naturalistic listening paradigms
Poster E38 in Poster Session E, Saturday, October 8, 3:15 - 5:00 pm EDT, Millennium Hall
Jixing Li1, Marco Lai2, Liina Pylkkänen1,2; 1New York University Abu Dhabi, 2New York University
Introduction Naturalistic paradigms have become popular in cognitive neuroscience, but connecting them to findings from controlled experiments remains rare. Here we address this gap for basic phrasal processing, commonly studied with a “minimal” two-word paradigm introduced in Bemis & Pylkkänen (2011). The original study compared phrases such as “red boat” to non-combinatory stimuli and found increased left anterior temporal lobe (LATL) activity in the presence of composition. How does this LATL effect manifest during naturalistic? The current MEG study tested how the neural reflections of composition as obtained from an auditory two-word paradigm generalize to natural story listening. Methods 20 right-handed native English speakers (15 females, M=27.8 years, SD=13.2) listened to both phrases (e.g., “green glass”) and single nouns that were preceded by a non-lexical mmm-sound, chosen for naturalness in a speech context (“mmmm glass”). After the auditory stimulus, subjects selected a matching picture from a set of 8. Participants also completed a naturalistic listening task where they listened passively to four audio excerpts from the YouTube channel “SciShow Kids” for about 12 minutes. MEG data were recorded at 1000 Hz (200 Hz low-pass filter), noise reduced and epoched from 100 ms before to 875 ms after the onset of the critical word in the two-word condition and of each word in the naturalistic condition. We conducted a whole-brain searchlight multivariate pattern classification analysis on the source-localized MEG data within the left and right temporal lobe for each subject. We trained a linear SVM classifier to a pairwise combination of the phrase and noun conditions in the two-word task with a leave-one-stimulus-pair-out cross-validation procedure. The trained classifier was then tested on data from the naturalistic stimuli to see if words at phrase boundaries can be successfully identified. Phrase boundaries in the naturalistic stimuli were manually annotated. The classifier was separately applied to all spatiotemporal timepoints, with a radius of 20 sources, and the performance was estimated by averaging across 20 permutations. At the group level, we performed a non-parametric cluster permutation test (Maris & Oostenveld, 2007) with 10,000 permutations to identify significant spatiotemporal clusters during which the classification accuracy was significantly above chance (p<0.05). Results For the two-word task, the classification analysis yielded a significant cluster for the phrase and noun distinction in the left anterior and middle temporal regions from about 125 to 265 ms after the onset of the critical word, consistent with prior LATL findings. When applied to data from the naturalistic listening condition, the classifier was significant in a similar left anterior and middle temporal region, yet with a much later time window from about 645 to 675 ms after the onset of the word. No significant clusters were found in the right temporal lobe. Conclusions The combinatory effect that has been previously observed in the LATL at around 150-250ms occurred much later during naturalistic story listening, suggesting different mechanisms of phrasal processing in these two contexts. The higher predictability and slower presentation rate of the two-word paradigm may underlie this contrast.
Topic Areas: Meaning: Combinatorial Semantics, Meaning: Lexical Semantics