Dec 7, 2004 - Early stages (P100) of face perception in humans as measured with event-related potentials (ERPs). M. J. Herrmann1, A.-C. Ehlis1, H. Ellgring2 ...
DOI 10.1007/s00702-004-0250-8 J Neural Transm (2005) 112: 1073–1081
Early stages (P100) of face perception in humans as measured with event-related potentials (ERPs) M. J. Herrmann1 , A.-C. Ehlis1 , H. Ellgring2 , and A. J. Fallgatter1 1
2
Department of Psychiatry and Psychotherapy, and Institute for Psychology, University of Wuerzburg, Germany
Received July 14, 2004; accepted October 16, 2004 Published online December 7, 2004; # Springer-Verlag 2004
Summary. According to current ERP literature, face specific activity is reflected by a negative component over the inferior occipito-temporal cortex between 140 and 180 ms after stimulus onset (N170). A recently published study (Liu et al., 2002) using magnetoencephalography (MEG) clearly indicated that a face-selective component can be observed at 100 ms (M100) which is about 70 ms earlier than reported in most previous studies. Here we report these early differences at 107 ms between the ERPs of faces and buildings over the occipito-temporal cortex using electroencephalography. To exclude contrast differences as the main factor for this P100 differences we replicated this study using pictures of faces and scrambled faces. Both studies indicated that face processing starts already at 100 ms with an initial stage which can be measured not only with MEG but also with ERPs. Keywords: Face processing, ERP, N170, P100.
Introduction The visual analysis of faces seems to be a highly specific skill in humans, which is probably based on the activation of a core system, consisting of occipitotemporal regions in extrastriate visual cortex (Haxby et al., 2002). This activity seems to be at least partly reflected by a bilateral negative occipital-temporal component of the event-related potential (ERP) peaking around 170 ms (N170) after stimulus onset (Bentin et al., 1996; B€ otzel et al., 1995; Eimer and McCarthy, 1999; Sagiv and Bentin, 2001). It has been shown that the N170 amplitude to faces depends on the presence of facial features, such as the mouth, the eyes or the nose (Eimer, 2000b). The N170 component probably reflects the processing of the configural information of facial features (Goffaux et al., 2003; Rossion et al., 2000; Taylor et al., 2001) and is independent of the emotional facial expression (Herrmann et al., 2002). In the same way Liu et al.
1074
M. J. Herrmann et al.
(2002) interpreted the M170 component of a magnetoencephalography (MEG) study as a later stage of face processing which is critical for identifying individual faces and which primarily relies on information about the configuration of facial components. Even more interestingly, Liu et al. (2002) proposed a M100 component deriving from the same cortical areas as the M170 component, with higher amplitudes to faces as compared to different control stimuli. In a series of experiments they showed that this M100 component is associated with face categorisation, which primarily relies on information about parts of the face. Furthermore, they showed that the amplitude of the M100 in each hemisphere does not differ for faces presented contralaterally versus ipsilaterally indicating that the source of this component must be located beyond the retinotopical cortex. Although some previous EEG studies reported face selective responses prior to 160 ms, (Halit et al., 2000; Herrmann et al., 2004; Itier and Taylor, 2002; Linkenkaer-Hansen et al., 1998; Taylor et al., 2001), these earlier effects are not investigated sufficiently and some inconsistencies remain. For example some studies reported shorter latencies for faces (Itier and Taylor, 2002, 2004; Linkenkaer-Hansen et al., 1998; Taylor et al., 2001) while other studies described no significant differences in latencies (Halit et al., 2000; Rossion et al., 1999). Probably these P100 latency effects were not caused by decreased latencies for faces, but increased latencies for specific control stimuli. In more detail the increased P100 latencies to the control condition were found for inverted faces (Itier and Taylor, 2002, 2004; LinkenkaerHansen et al., 1998) and were not found for the comparison between faces and scrambled faces (Linkenkaer-Hansen et al., 1998; Rossion et al., 1999), other objects (Itier and Taylor, 2004) or slightly modified faces (Halit et al., 2000). Most of the studies investigating the P100 effect for face processing reported amplitudes effects (Halit et al., 2000; Herrmann et al., 2004; Itier and Taylor, 2002, 2004; Taylor et al., 2001) but, again, not all of them (Linkenkaer-Hansen et al., 1998; Rossion et al., 1999; Rossion, 2003). It has to be mentioned that most studies of ERP correlates for face processing did not analyse, or at least did not report P100 effects. Therefore, a detailed analysis of the factors leading to increased P100 amplitudes could not be performed. Based on the literature analysing the P100 amplitude effect, an increased P100 was found for faces compared to contrast polarity faces (Itier and Taylor, 2002), faces with stretched features (Halit et al., 2000), buildings (Herrmann et al., 2004), phase scrambled faces (Taylor et al., 2001) and at last for different object categories (Itier and Taylor, 2004). As the other studies, which did not find increased P100 amplitudes, used comparable stimulus categories as control condition like cars (Rossion et al., 2003) or degraded face stimuli (LinkenkaerHansen et al., 1998), a clear conclusion regarding the amplitude effect could not be drawn. As in most of the cited studies with P100 amplitude effects differences in low level features between stimulus categories could not be ruled out (Herrmann et al., 2004; Itier and Taylor, 2002, 2004; Taylor et al., 2001), this could be an explanation for different findings across the studies. We believe that this simple explanation should be excluded before the functional meaning of the P100 effect can be further investigated.
Early face perception
1075
Therefore, in a first ERP-study we tried to replicate the P100 effects reported previously (Halit et al., 2000; Herrmann et al., 2004; Itier and Taylor, 2002, 2004; Taylor et al., 2001). In a second study we used pictures of faces and scrambled faces to rule out the possibility that these P100 effects were caused by contrast differences or luminance between stimulus categories. Material and methods In a first study we investigated 39 right-handed subjects (20 male, mean age 32.6 12.1 years) after written informed consent was obtained. The task (Herrmann et al., 2002) consisted of 12 different faces (Ekman and Friesen, 1976) and 12 different buildings, which were presented repeatedly three times each in a pseudo-randomised order (visual horizontal angle of 4.6 , vertical angle of 6.3 ). The presentation time of each stimulus was 500 ms with a constant interstimulus interval (ISI) of 1500 ms. Mean grey levels of the pictures were similar between faces and buildings. Subjects were instructed to decide whether a face or a building was presented. During this task the EEG was recorded from 21 scalp electrodes positioned according to the international 10–20 system with linked mastoids as reference. Three additional electrodes were placed at the outer canthi of both eyes and below the right eye to monitor eye blinks and movements. The EEG was sampled continuously at a rate of 256 Hz with a bandpass from 0.1 to 70 Hz. Impedances were kept at 5 k or below. Epochs (100 ms before to 500 ms after stimulus onset) with amplitudes or with a voltage step exceeding 50 mV in any of the EEG or EOG channels were excluded from further analyses. After calculation of the average reference, the artifact-free trials (at least 20 epochs) were averaged separately for each subject and condition. The ERPs were filtered with a bandpass from 1 Hz to 30 Hz (Liu et al., 2002). No baseline correction was calculated. To determine the time windows for the peak detection we chose the time point of the preceding and following maximum for the N170 according to the grand mean curves (minimum for the P100 component). Based on this criterion, the N170 was automatically detected as the maximum negative peak in the time window from 109 ms–219 ms for the electrodes T5 and T6. For the P100 component the most positive peak between 70 ms–160 ms at the electrodes O1 and O2 was determined. Peak amplitudes and latencies were statistically analysed using a 2 (side: T5, T6)2 (condition: face, buildings) ANOVA with repeated measures for the factors ‘side’ and ‘condition’.
Fig. 1. Example of the stimulus material of the second study. On the left side an intact and on the right side a scrambled face is presented
1076
M. J. Herrmann et al.
In the second study, 21 right-handed subjects (11 male, mean age 24.4 2.5 years) were investigated while showing 20 pictures of faces and scrambled faces (3 times each for 200 ms, ISI 800 ms, visual angel: horizontally 4.8 , vertically 7.0 ). We scrambled each picture by cutting 1515 pixel quadrates out of the picture and resampling it (see Fig. 1 for an example of the stimulus material). EEG was registered with 1000 Hz against one reference located between Cz and Fz and recalculated offline against average reference. According to the grand mean curves, the face specific potential N170 was detected as the maximum negative peak in the time window from 105 ms–194 ms for the electrodes T5 and T6. For the P100 component the most positive peak between 75 ms–145 ms at the electrodes O1 and O2 was determined. All other parameters were the same as in study 1.
Results Sample 1 For the P100 amplitude a significant main effect ‘‘condition’’ (F[1, 38] ¼ 38.2, p < 0.0001) without a significant main effect ‘‘side’’ (F[1, 38] ¼ 1.6, n.s.) or interaction effect ‘‘sidecondition’’ (F[1, 38] ¼ 1.4, n.s.) was found (see Fig. 2).
Fig. 2. Grand mean event related potentials over the left (T5) and right (T6) temporal (N170) and occipital (O1, O2; P100) cortex after the presentation of faces and buildings for sample 1
Early face perception
1077
The P100 peak was significantly higher for faces as compared to buildings for the left side (O1: mean amplitude faces: 4.48 3.03 mV; mean amplitude buildings: 3.51 2.65 mV; t [38] ¼ 5.9, p
