Do segment boundaries disrupt the processing of non-adjacent dependencies? Evidence from neural frequency tagging and ERP

Introduction: Language processing requires comprehenders to link the words and phrases of sentences. This can result in non-adjacent dependencies (NADs) that can span multiple words. In the current study, we ask whether the learning of NADs is constrained by segment boundaries in the stimulus. In particular, it has been proposed that dependencies do not cross the boundaries of cognitive segments, which is what we test here (Christiansen & Chater, 2005). Experiment: We will analyze participants' EEG responses during an artificial-grammar experiment. Six-syllable strings in German are designed and recorded as isochronous speech. During the learning phase, participants listen to the strings that contain NADs either within or across segments. An 80-ms silence is inserted as a segment boundary between each chunk to ensure that participants learn the segments. Then, in the test phase, participants listen to the strings that either contain NADs or not, in both within‰?? and across-segment conditions. The experiment is still in progress. Prediction: For the learning phase, if segments are learned, six-syllable frequency peaks can be observed in the power spectrum. For the testing phase, the ERP to syllables that constitutes the second element of an NAD will be contrasted with the ERP to syllables that are not part of an NAD. If NAD processing across segments is more difficult, the effect between NADs and non-NADs will be smaller in the across-segment condition than the within-segment condition. The results can help better understand the relationship between segmentation and dependency formation in human language acquisition and processing