Distributed Memory Representations Generate The Perceptual-Magnet Effect

Kuhl’s native language magnet (NLM) theory of phonetic learning is examined in the light of an exemplar-based model. According to the NLM theory the perceptual space is partitioned into phonetic categories as a consequence of the magnet-like properties of the category’s best exemplar  the category’s prototype. The NLM theory predicts that the perceptual space is warped in the neighborhood of these perceptual magnets because a prototype attracts exemplars that fall within its zone of influence. The present paper introduces a very simple theoretical model that accounts for the perceptual-magnet effect as an emergent consequence of an exemplar-based phonetic memory. The model is inspired by the neuronal group selection theory (Edelman, 1987) and shows that the notions of prototype or warping of the perceptual space are not necessary to account for the perceptual-magnet effect.