Abstract
Human visual object recognition is thought to begin within 150-170 ms of stimulus onset. Here we used magnetoencephalography (MEG) to show that recognition begins substantially earlier: by 100 ms after stimulus onset MEG responses over occipitotemporal sensors are significantly higher to faces than to nonface control stimuli. The amplitude of this “M100” response is higher when a degraded face stimulus is correctly categorized as a face than when the same stimulus is incorrectly categorized, showing that the M100 reflects not simply the low-level visual features present in the stimulus, but the perceptual experience of the subject. Furthermore, the magnitude of the M100 is not correlated with correct identification of a specific face, although the magnitude of the later face-selective M170 is. Thus, categorizing a stimulus as a face apparently begins substantially earlier than identifying a specific face. More generally, by sharply reducing current estimates of the earliest latency of stimulus categorization, our findings place powerful new empirical constraints on theories of human visual recognition.
Supported by: Reed Fund and NEI grant EY13455 to N. Kanwisher