Single-trial EEG dynamics of object and face visual processing
Introduction
A negative bilateral occipito-temporal ERP component peaking at about 140–200 ms is systematically recorded in response to faces. This component, the N170, is generally larger for faces than for many control objects (e.g., Bötzel et al., 1995, Bentin et al., 1996, Rossion et al., 2000, Itier and Taylor, 2004, Rousselet et al., 2004, Rousselet et al., 2005). This amplitude difference (the ‘face effect’) often is taken as the hallmark of the activity of face-specific generators. The vast majority of ERP studies describing the N170 have relied on a very successful approach consisting, through an averaging process, in analyzing only signals that are time-locked to stimulus onset.
However, by focusing on the evoked activity, this approach might miss useful information available in single EEG trials (Makeig et al., 2004). First, averaging single-trial data cancels signals that are not time-locked to stimulus onset, thus potentially masking differences in the so-called induced activity (Klopp et al., 1999, Pfurtscheller and Lopes da Silva, 1999, Tallon-Baudry and Bertrand, 1999, Makeig et al., 2004). Second, the classic averaging technique is unable to distinguish between the effects of amplitude modulation and phase resetting, whereby modulations of mean amplitude can be caused by the partial temporal alignment of ongoing oscillations that were previously out of phase. Such phase resetting of ongoing EEG activity has been proposed to explain the generation of early visual-evoked potentials like the P1 and the N1 (Makeig et al., 2002, Klimesch et al., 2004, Gruber et al., 2005; this idea was first introduced for auditory ERPs by Sayers et al., 1974). Furthermore, induced amplitude and phase modulations appear to be frequency specific, occurring in relatively narrow EEG frequency bands. Although a peak in the amplitude spectrum does not necessarily imply the existence of neuronal oscillations at the corresponding frequencies (Bullock et al., 2003), the presence of ongoing neuronal oscillations is well documented at different levels of organization, from ion channels and single neurons to large neuronal networks (Destexhe and Sejnowski, 2001, Steriade, 2003). The integration of these various mechanisms at different scales leads to oscillations of large populations of neurons in cortico-cortical and thalamo-cortical networks that can be recorded on the scalp as EEG waves with different amplitude spectra (Destexhe and Sejnowski, 2001, Wright et al., 2001, Steriade, 2003, David et al., 2005, Robinson et al., 2005).
Thus, although amplitude differences at the level of the N170 might initially appear to result from pure phase-locked amplitude differences, other possibilities remain. In particular, this difference in ERP amplitude may alternatively be explained by a mixture of evoked amplitude increase and evoked phase resetting of ongoing oscillations in specific frequency bands. Although these hypotheses have been put forward, they have never been tested in a rigorous manner in the context of scalp ERPs recorded in response to complex objects (Fig. 1). Thus, we used wavelet time–frequency analyses to determine the extent to which the face effect could be explained by a resetting of ongoing activity. Providing a time–frequency, single-trial, description of ERPs to complex objects might allow us to gain insights into the underlying neuronal mechanisms.
When analyzing inter-trial phase coherence, it is important to keep in mind that an increase in phase coherence is not synonymous with phase resetting. Indeed, a phasic-evoked response, with no phase resetting, will appear as a sharp increase in phase coherence (Yeung et al., 2004, Mazaheri and Jensen, 2006). This is because time–frequency analyses can turn, artificially, a sudden burst of activity into what looks like an oscillation. In other words, phase coherence could result simply from evoked phase locking due to the presence of an ERP of fixed amplitude and phase in single-trial data, instead of really being induced by the stimulus presentation. To control for this potential shortcoming, the mean ERP was subtracted from each individual trial before performing the wavelet transform. In addition, a simulation using a purely evoked ERP model was used for comparison with the experimental results. The subtraction technique and the model follow a very simple idea according to which the stimulus-evoked activity results from the linear summation of fluctuating ongoing activity and a reproducible evoked response (Arieli et al., 1996). Although our analyses cannot rule out totally the involvement of phase resetting in generating, even partially, ERPs to faces and objects, they clearly demonstrate that an evoked response, without any phase resetting, is sufficient to account for the experimental data.
Section snippets
Participants
Sixteen subjects participated in this experiment. All subjects gave written informed consent and had normal or corrected-to-normal visual acuity. Here we report only data from 11 subjects, because 5 subjects presented too many artifact-contaminated trials (e.g., due to eye blinks) when using an extended time window of ± 1000 ms around stimulus onset necessary to perform the time–frequency analyses. Those 11 subjects ranged in age from 20 to 39 (mean age 26). All observers other than ABS, GAR,
Results
ERP results were consistent with the pattern reported in the literature, with the main effect being a stronger response to faces compared to houses and textures between 140 and 200 ms after stimulus onset. Most subjects had also a stronger response to houses compared to textures in the same time window. ERPs from 4 subjects are shown in Fig. 2, Fig. 3 along with the main results regarding amplitude and phase modulations. Although there was a relatively consistent pattern across subjects, the
Discussion
In this paper we provided a systematic time–frequency analysis of the single-trial ERPs recorded in response to faces, houses and textures. Because these stimuli had the same spatial frequency amplitude spectra, differences in the EEG signal cannot be attributed to differences in the global amplitude spectra. Instead, we suggest that such differences reflect the visual system's interpretation of shape and object categories, based on differences in the stimuli's spatial phase spectra (Bennett
Acknowledgments
We thank two anonymous reviewers for their insightful comments on an earlier version of this manuscript. In particular, Reviewer 1 suggested critical alternative analyses to determine whether an evoked model could account for the data.
This work was supported by NSERC Discovery Grants 42133 and 105494, Canada Research Chairs, and infrastructure support from the CFI and OIT to P.J.B. and A.B.S.; an NSERC PGS-D award to J.S.H.; and a CIHR fellowship grant to G.A.R.
References (45)
- et al.
Are the electroencephalograms mainly rhythmic? Assessment of periodicity in wide-band time series
Neuroscience
(2003) - et al.
Modelling event-related responses in the brain
NeuroImage
(2005) - et al.
Eeglab: an open source toolbox for analysis of single-trial EEG dynamics including independent component analysis
J. Neurosci. Methods
(2004) - et al.
Identification of band-pass filtered letters and faces by human and ideal observers
Vision Res.
(1999) - et al.
Oscillatory characteristics of face-evoked neuromagnetic responses
Int. J. Psychophysiol.
(2006) - et al.
Phase-locked alpha and theta oscillations generate the P1–N1 complex and are related to memory performance
Brain Res. Cogn. Brain Res.
(2004) - et al.
Distinguishing the evoked response from phase reset: A comment to Makinen et al.
NeuroImage
(2006) - et al.
Face-selective spectral changes in the human fusiform gyrus
Clin. Neurophysiol.
(1999) - et al.
Mining event-related brain dynamics
Trends Cogn. Sci.
(2004) - et al.
Auditory event-related responses are generated independently of ongoing brain activity
NeuroImage
(2005)