A dynamic neural field model of memory, attention and cross-situational word learning

Recent empirical studies have affirmed the fundamental role of attention and memory processes in statistical word learning tasks. These processes interact in complex ways to guide spontaneous looking behaviors of learners as well as determine their overall learning performance. On the modelling side, studies have made it clear that computational models must provide process-based rather than only computational accounts of word learning, because these can connect to the empirically observed behaviors at a moment-to-moment timescale. Thus, here we present a neurally-grounded process model of word learning called WOLVES (Word-Object Learning Via Visual Exploration in Space) that integrates visual dynamics and word-object binding across multiple timescales. WOLVES integrates multiple established dynamic neural field models to allow fine-grained indexing of component processes driving the looking-learning loop. We report simulation results for three empirical cross-situational word learning experiments to validate the model.