Search Abstracts | Symposia | Slide Sessions | Poster Sessions
The neural basis of derivational morpheme acquisition under semantic uncertainty
Poster Session B, Friday, September 12, 4:30 - 6:00 pm, Field House
This poster is part of the Sandbox Series.
Marianne Azar1, Alec Marantz1; 1New York University
Our knowledge of derivational affixes ([-er], [-ify]) helps us understand new words based on their familiar parts, whose meaning we generalized across words that contain them. This generalization is essential in first and second language acquisition, during which new derived words are frequently encountered. It also occurs in code-switching when affixes are borrowed across languages. However prevalent, the brain basis of meaning generalization is largely unknown. Further complicating matters, affixes often mean disparate things, like -ery in brewery and savagery. How do we generalize the meaning of a morpheme with disparate meanings, and how is the ambiguity represented? Building on our previous experiments in Arabic that have found a neural substrate for the meaning typicality of affixes, with methods refined by studies on artificial morphological acquisition, we attempt to elucidate this question by teaching English speakers new derivational suffixes. Within an English-base artificial grammar paradigm, we introduce two novel but plausible nominalizing suffixes (e.g., ‘[paint][le]’, ‘[runn][ock]’) with contrasting typical meanings: places and objects. Following rule generalization findings1, 2, 3, we employ a skewed frequency distribution (many low-frequency types per class) and a skewed distribution of typical/atypical meanings (40 typical, 10 atypical words per suffix) with one suffix typically creating place names and the other object names. Suffix meanings are counterbalanced across participants, and all stimuli are normed for well-formedness prior to the study. Using magnetoencephalography, participants are first presented with the target stimuli (‘paintle’) in isolation, matched with existing complex words (‘violinist’) and new-suffixed words that will stay untrained (‘lighteen’). Next, they are trained on all target stimuli in sentences that disambiguate the suffix’s general meaning for the stimulus stem (‘The drinkle is closed on Mondays’). The subsequent test phase includes trained and novel words with the target suffixes to test for meaning generalization. Training sets continue for each participant until they reach a satisfactory learning cutoff, with a minimum/maximum amount of training blocks. Finally, after participants learned the suffix meanings, they are presented with the stimuli in isolation again, mixed with novel words with the learned suffixes, to allow us to measure the effect of training on the processing of our novel suffixes. We predict participants to generalize suffix meaning onto untrained words. We expect learning to be influenced by suffix-meaning type frequency and by the meaning disjunction, with a skew in interpreting novel words towards the typical meaning of their suffix. After training, we predict brain activity for trained suffix words to indicate multimorphemic processing unlike untrained new suffix words, as evidenced by stem-to-suffix surprisal effects on the M170 response from the Visual Word Form Area4, and a difference in processing meaning-typical vs. -atypical words like in our Arabic study. This ongoing study importantly connects across neuroscience, rule generalization studies, artificial grammar learning paradigms, and linguistic theory. It addresses an important cross-linguistic phenomenon – affix ambiguity -- which is currently overlooked. We are currently finalizing our study design and expect to have analyzed data collected from a substantial set of our participants by the time of the conference.
Topic Areas: Morphology, Meaning: Lexical Semantics