Behavioural Brain Research 112 (2000) 1 – 11 Research report Electrophysiological correlates of category goodness Maritza Rivera-Gaxiola a, *, Mark H. Johnson b , Gergely Csibra b , Annette Karmiloff-Smith c a Laboratory B06 Psicofisiologı ´a, Centro de Neurobiologı ´a, Campus UNAM-UAQ Juriquilla, Proincia Juriquilla, C.P. 76230 Quere ´taro, Mexico b Centre for Brain and Cognitie Deelopment, Birkbeck College, Uniersity of London, Mallet St, London WC1E 7HX, UK c Neurocognitie Deelopment Unit, Institute of Child Health, 30 Guildford St, London WC1N IEH, UK Received 1 March 1999; received in revised form 22 October 1999; accepted 6 December 1999 Abstract We report the results obtained from a behavioural and electrophysiological study. A synthesised continuum going from labial /ba/ to retroflex /d  a/ through dental /da/ was tested for category goodness. Native English speakers rated different tokens from each category as good, bad or ambiguous. The results showed that not all of the representatives of each category were ideal and that the categories tested have an internal structure. The electrophysiological study evaluated whether event related potentials (ERPs) mirrored the goodness judgements. During a passive oddball task, the same participants were exposed to native /ba/ – /da/, Hindi dental /da/-retroflex /d  a/ and within-category /ba/ – /ba/ contrasts. Results showed that participants pre-attentively perceive the differences in all cases, as shown by mis-match negativities (MMN), late positive deflections (LPD) or greater N1 and/or P2 components for deviant stimuli. Acoustic sensitivities, categorical perception and category goodness all contributed to the waveforms obtained. We attribute the ERP effects to a combination of (1) prototypes built from initial sensitivities, (2) reinforcement with exposure to one’s native language and (3) no permanent loss of the initial boundaries explains the effects observed. © 2000 Elsevier Science B.V. All rights reserved. Keywords: Category-goodness; Cognitive neuroscience; Event-related potentials (ERP); Late positive deflections (LPD); Mis-match negatives (MMN); Perceptual magnet effect; Synthesised speech www.elsevier.com/locate/bbr 1. Introduction Categorisation is a fundamental aspect of human perceptual processing. There is evidence showing that speech categories, for example, have a rich internal structure. This means that members of the same cate- gory are not necessarily equivalent and that categories might be organised around- or with reference to a ‘prototype’ [6]. Evidence indicates that there are mem- bers of categories that are better adaptors in selective adaptation studies [17,24] and are preferred over other exemplars in dichotic listening tasks [15,21]. Tokens of the same category also vary in the ‘representativeness’ or ‘goodness’ with which (adult) participants rate them [3,6,14,27]. In other words, categories are graded both quantitatively and qualitatively. There are prototypes, which are encoded more rapidly, last longer in memory and are often preferred over other members of the category [6,16]. For example, Kuhl and collaborators [6,7,9] have shown that a good exemplar of a category perceptually attracts other members of the category, whilst bad exemplars tend to be differentiated more easily. They have called this the ‘perceptual magnet effect’. Although the existence of an internal structure within linguistic categories is an established fact, the underlying nature of such structures is not known. In a previous study carried out by our group [22], we re- ported differential processing for different types of con- trasts (native (ba/da), non-native (dental da/ retroflex d  a), and within-category (two /ba/ tokens) as reflected by different event related potential (ERP) responses. * Corresponding author. Tel.: +52-42-381050. E-mail address: rivera@calli.cnb.unam.mx (M. Rivera-Gaxiola) 0166-4328/00/$ - see front matter © 2000 Elsevier Science B.V. All rights reserved. PII:S0166-4328(00)00218-7