EEG dynamics underlying visual ERP to faces and objects
In the study of visual object processing, the EEG literature has been relatively agnostic until recently to the existence of ongoing brain activity. Indeed, the main approach uses event-related potentials (ERP), a technique that makes the implicit assumption that the neuronal events of interest are purely evoked by stimulus presentation out of a flat baseline, with the main ERP components systematically phase-locked to stimulus onset. According to this model, ERP components can be explained by the fixed latency and fixed polarity, bottom-up phasic modulation of cortical generators. I will present experiments using the classic ERP technique that have focused on the N170, a posterior lateral ERP component found consistently in the time window of 130-200 ms. The larger amplitude of the N170 to faces than to other objects has often been taken as a hallmark of face-specific processing. I will conclude that inherently, this approach cannot answer questions about the timing and, in the end, the mechanisms involved in face and object processing. However, time-frequency analyses can open the door to the underlying on-going physiological activity producing the N170 and other visual evoked potentials. Specifically, wavelet analyses reveal a consistent pattern of increase in EEG power triggered by the stimulus presentation and an event-related phase resetting of ongoing EEG activity in specific frequency bands. In this framework, I will present evidence showing that ‘early’ visual ERP components P1, N170 and P2 are associated with a strong increase in phase coherence across the 5-20 Hz frequency band, with maximum coherence occurring near 10 Hz. In general, the increase in phase coherence precedes an increase in EEG power at the same frequencies. Phase coherence is very strong at the latency of the N170 for faces and other objects. The larger N170 for faces find its origin in larger and earlier peaks of power and phase coherence in the 5-20 Hz frequency band. These analyses help disentangle the EEG dynamics that drive the N170 and other visual ERP components. Finally, I will argue that single-trial analyses in the frequency domain might be linked to specific neuronal mechanisms, while averaged ERPs break this link.