Tony Smith (Yale) @ Department of Economics - Wyman W603
Sep 27 @ 3:30 pm – 5:00 pm
Arie Beresteanu (University of Pittsburgh) @ Department of Economics - Wyman W537
Sep 28 @ 11:00 am – 11:50 am
Sohini Mahapatra (JHU) @ Department of Economics - Wyman W603
Sep 28 @ 3:30 pm – 5:00 pm
Bonnie Breining Dissertation Presentation @ Krieger 111
Sep 29 @ 10:30 am – 11:30 am

“Effects of Semantic and Segmental Similarity on the Production and Learning of Spoken and Written Words”

This dissertation investigates mechanisms of word production and word learning, focusing specifically on how semantic and segmental similarity affect the production and learning of words.  Incremental learning models of spoken word production propose that learning occurs each time a familiar word is produced.  In recent work, I argued that this type of model applies across stages and modalities (spoken and written) of word production.  Here, I extend the model from word production to the learning of new words, incorporating insights from the literature regarding the role of learning difficulty to make predictions about both training effects and long-term learning outcomes.  The extended incremental learning model (e-ILM) framework I propose makes testable predictions about the effects of training new words in semantic or segmentally related blocks of words compared to training in unrelated contexts.  According to these predictions, semantic blocking is a desirable difficulty that initially causes interference during training but improves long-term learning by enhancing the distinctiveness of the learned representations.  In contrast, segmental blocking is predicted to negatively affect both training and long-term learning because it reduces distinctiveness.

Two studies tested these predictions.  In the first, neurotypical adults learned names for novel items in semantic, segmental, and unrelated blocks in separate written and spoken experiments.  Findings were consistent with the e-ILM predictions about training: although training in both types of related vs. unrelated blocks produced interference, semantic blocking increased distinctiveness while segmental blocking reduced distinctiveness.  The predictions regarding long-term learning were less clearly supported because blocking did not consistently affect retention.

In the second study, individuals with acquired dysgraphia relearned previously known word spellings in semantic, segmental, and unrelated blocks.  Group results did not consistently support the e-ILM predictions about training.  However, there was evidence that both semantic and segmental blocking served as desirable difficulties.  Specifically, both improved long-term retention to the degree to which they increased training difficulty, suggesting that the general learning principle of desirable difficulty applies in this situation.

Overall, the proposed e-ILM account accurately describes the underlying mechanisms that produce effects of blocking by semantic or segmental similarity during the (re)learning of words.  The work raises potentially important questions regarding differences between new learning and relearning in neurotypical and brain damaged individuals that motivate further research.

Department of Biology Seminar Series – D. Allan Drummond
Sep 29 @ 4:00 pm – 5:00 pm

‘Adaptive, stress-responsive self-assembly of poly (A)-binding protein’

D. Allan Drummond
Department of Biochemistry & Molecular Biology
University of Chicago
Host:  Christian Kaiser