INTRODUCTION

The capacity of the human visual system to process, very quickly, complex natural scenes and to extract relevant information extremely fast may be crucial in human social interactions. Recently, a number of electrophysiological studies in non-human primates and humans have reported evidence for such rapid visual analysis in face or object recognition, categorization of complex stimuli, and visuo-spatial perception.

In monkeys, it has been shown that neurons in high-level visual structures such as the superior temporal sulcus, the inferior temporal (IT) and the medial superior temporal (MST) areas could respond very quickly (40± 140 ms from stimulus onset) during discrimination tasks performed on complex stimuli, such as human or monkey faces, or food [1±5]. For example, the response latency of face-selective neurons in the macaque IT was as short as 20±40 ms, indicating that sufficient visual information was processed to just allow identification in a 20±40 ms period of the spike train [3,4]. Similarly, visual tracking tasks in monkeys induced neural responses in MST prior to 60 ms from stimulus onset [2].

In humans, ERP measurements during a go/no-go visual categorization task have shown that subjects can detect the presence of an animal (or food) in previously unseen photographs flashed for a very short period, within 150 ms [6]. In subsequent experiments, the same authors showed that monkeys could perform the task even faster [7]. Very early ERP effects also distinguished between linguistic and non-linguistic (face) stimuli [8], inverted and non-inverted faces [9], and between repeated and non-repeated faces [10] at latencies around 90, 120 and 50±90 ms, respectively. These findings suggest that higher aspects of visual processing, particularly of biological stimuli, may be achieved very rapidly.

In a previous experiment primarily aimed at investigat-ing the neural correlates of gender processing in human faces, we had found, unexpectedly, similar early ERP effects. Although gender categorization was not required, ERPs with a latency between 45 and 85 ms distinguished between a condition in which male and female faces were mixed (thus possibly inducing incidental categorization of gender) and a condition in which the gender was separated in different blocks (preventing any gender categorization); the results were replicated for hand stimuli [11]. These early ERP differences, found for both faces and hands, were interpreted as possibly reflecting a low-level, incidental categorization process between two wide classes of stimuli (male, female). These results however challenge the current view on the transmission of the information fow through the visual neural pathways. The aim of the present study was therefore to test the robustness of these very early ERP effects in face processing, and to examine whether similar effects can also be elicited by non-biological geometrical shapes. To achieve our goals, we used a paradigm similar to that of our previous experiment [11], comparing the early ERP responses to task-irrelevant stimuli according to whether they include one class or two stimulus classes.