Exp Brain Res (1998) 118:501±510  Springer-Verlag 1998 RESEARCH ARTICLE S.J. Swithenby ´ A.J. Bailey ´ S. Bräutigam O.E. Josephs ´ V. Jousmäki ´ C.D. Tesche Neural processing of human faces: a magnetoencephalographic study Received: 28 May 1996 / Accepted: 22 July 1997 S.J. Swithenby ( ) ) ´ S. Bräutigam ´ O.E. Josephs The Open University, Department of Physics, Milton Keynes MK7 6AA, UK A.J. Bailey MRC Child Psychiatry Unit, Institute of Psychiatry, De Crespigny Park, London SE5 8AF, UK V. Jousmäki ´ C.D. Tesche Low Temperature Laboratory, Helsinki University of Technology, 02150 Espoo, Finland Abstract This is a whole head magnetoencephalographic (MEG) study of the neural processing of briefly presented images of human faces in 14 normal subjects. The exper- iments involved three tasks of increasing complexity, in- volving image categorisation, image comparison and the identification of emotion. The analyses were based on av- erage responses to repeated stimuli in the different image categories. These averages were processed to give numer- ical measures of the power within defined regions and la- tency spans. The only statistically significant difference in these data between the response to faces and other images is in the right occipito-temporal channels at a latency of 140 ms. The face-specific response is largely independent of the task. Source modelling suggests an extended source in the ventral occipito-temporal region. The analysis sup- ports the notions of both face-specificity and right hemi- sphere dominance for all image types at early latencies. Key words Face processing ´ Magnetoencephalography ´ Evoked responses ´ Human subjects Introduction In this study, magnetoencephalographic (MEG) measure- ments are used to explore the neural processing that fol- lows brief presentation of static images of human faces. Although the protocol is designed to allow exploration of the various stages of face processing, the focus in this paper is on activity at short latencies and in posterior re- gions. There have been extensive microelectrode studies of face processing in animals, principally in monkeys (Gross et al. 1972; Bruce et al. 1981; Perrett et al. 1982, 1992; Yamane et al. 1988; Rolls 1992). These have shown that the temporal cortex contains neurones that respond specif- ically to faces (as well as to images of other parts of the body). The face-specific neurones are found in the con- vexity of the inferior temporal cortex and in the superior temporal sulcus. The selectivity of the neurones varies: factors such as perspective and familiarity influence the response of subsets of the cells. Face specificity is appar- ent in infant monkeys in the second month of life (Rod- man et al. 1993). Evidence for localised processing of faces in humans comes from several sources. A face processing deficit, prosopagnosia, is associated with lesions in posterior re- gions of the brain (Meadows 1974; Michel et al. 1989; Damasio et al. 1990; Sergent and Signoret 1992). Such studies suggest that the right hemisphere may play the more crucial role (De Renzi et al. 1994; Calesimo and Caltegirone 1995). Scalp electrodes studies have demonstrated activity that is specific to face stimuli (Bötzel and Grüsser 1989, Jeffreys et al. 1992; Seeck and Grüsser 1992; Bötzel et al. 1995; Bentin et al. 1996). The earliest latency at which face specificity is observable ranges from 154 ms (repeat- ed black and white drawings) to 180 ms (non-repeated photographs), the latency differences possibly being the result of priming. These studies used relatively sparse ar- rays of electrodes and have provided limited information about the generator sites. Halgren et al. (1994a, b) recorded responses to faces using multi-contact depth electrodes implanted in patients during pre-surgical evaluation of drug-resistant epilepsy. Patients responses to the face stimuli were considered to be the same as in normal subjects. Widespread activity was found, with the most prominent post-primary signals in the basal occipito-temporal cortex at latencies of 130, 180 and 240 ms. Although this study lacked the control stimuli needed to provide unambiguous evidence of face-specific processing, the authors argued for such spec-