Receptive fields for flexible face categorizations
Smith, M.L. and Gosselin, F. and Schyns, P.G. (2004) Receptive fields for flexible face categorizations. Psychological Science 15 (11), pp. 753-761. ISSN 0956-7976.
Abstract
Examining the receptive fields of brain signals can elucidate how information impinging on the former modulates the latter. We applied this time-honored approach in early vision to the higher-level brain processes underlying face categorizations. Electroencephalograms in response to face-information samples were recorded while observers resolved two different categorizations (gender, expressive or not). Using a method with low bias and low variance, we compared, in a common space of information states, the information determining behavior (accuracy and reaction time) with the information that modulates emergent brain signals associated with early face encoding and later category decision. Our results provide a time line for face processing in which selective attention to diagnostic information for categorizing stimuli (the eyes and their second-order relationships in gender categorization; the mouth in expressive-or-not categorization) correlates with late electrophysiological (P300) activity, whereas early face-sensitive occipitotemporal (N170) activity is mainly driven by the contralateral eye, irrespective of the categorization task. One of the most pressing issues in relating brain function to perception and cognition concerns the functional interpretation of brain responses to complex visual stimuli. Although it is obvious that visual stimuli elicit brain signals, it is a challenge to narrow down the relevant visual features that modulate the signal, and to relate these to the behavior of the observer. For example, single-cell studies concerned with the mechanisms of visual categorization have established that infero-temporal neurons respond specifically to the sight of complex objects such as faces and hands (Desimone, Albright, Gross, & Bruce, 1984; Gross, Rocha-Miranda, & Bender, 1972; Perrett, Rolls, & Cann, 1982). However, further investigations have revealed that the effective stimulus is represented in an abstract-feature space (Desimone et al., 1984; K. Tanaka, Saito, Fukada, & Moriya, 1991) or parameter space (Op de Beeck, Wagemans, & Vogels, 2001; Sigala & Logothetis, 2002; see also Pasupathy & Connor, 2002, for V4 neurons) of much lower dimensionality. In a related vein, electroencephalogram (EEG), magnetoencephalogram (MEG), and neuroimaging studies have established sensitivity of brain signals to faces (e.g., the N170 in EEG—Bentin, Allison, Puce, Perez, & McCarthy, 1996; Carmel & Bentin, 2002), and neuroimaging has shown that faces elicit activity in the middle fusiform gyrus (Gauthier, Tarr, et al., 2000; Kanwisher, McDermott, & Chun, 1997), but further studies have revealed that similar brain responses are elicited when people categorize objects in other areas of expertise (Gauthier, Skudlarski, Gore, & Anderson, 2000) and when they categorize novel objects (Gauthier, Tarr, Anderson, Skudlarski, & Gore, 1999), leaving unresolved the question of the specific features that determine the identified brain activity. The functional interpretation of a brain signal represents the solution of a multidimensional credit-assignment problem (Barlow, 1959): What stimulus features determine the amplitude of the brain signal? To resolve this question, a suitable method must allow for the flexible testing of many sorts of features (have low bias) while minimizing spurious features (keep low variance), in a reasonable amount of trials (Geman, Bienenstock, & Doursat, 1992; Stone, 1982). Because of this dilemma between bias and variance (Geman et al., 1992), it is questionable whether a faithful mapping between stimulus features and brain signals can be computed in the limited number of trials that is typical of categorization experiments (Edelman & Intrator, 1997). In the study we report here, we established this mapping between scalp EEG activity and stimulus features in the context of the categorization of realistic faces (gender, expressive or not). Faces are a natural stimulus category with several advantages over others: They have a privileged biological significance to the normal functioning of the human species (Farah, 1996; Nachson, 1995), levels of expertise in processing faces are similar across observers (Gauthier, Skudlarski, et al., 2000), and they afford multiple categorizations and are fairly homogeneous as a class. We searched the EEG for signals modulated by facial information, using a 4-ms recording window from stimulus onset and focusing on electrode sites that were known from prior studies to be sensitive to the encoding of facial information relatively early (low occipito-temporal P9 and P10; Bentin et al., 1996; Carmel & Bentin, 2002; Rossion & Gauthier, 2002) and to decision making later on (centro-parietal Pz; Donchin & Coles, 1998; Verleger, 1997, 1998), and that displayed the largest amplitude of the components of interest.
Metadata
| Item Type: | Article |
|---|---|
| School: | Birkbeck Faculties and Schools > Faculty of Science > School of Psychological Sciences |
| Depositing User: | Sarah Hall |
| Date Deposited: | 16 Mar 2020 15:28 |
| Last Modified: | 02 Aug 2023 17:58 |
| URI: | https://eprints.bbk.ac.uk/id/eprint/31352 |
Statistics
Additional statistics are available via IRStats2.
Archive Staff Only (login required)
 |
Edit/View Item |