Visual Cortex: Seeing Motion

What happens in primary visual cortex (V1) when we look at a complex moving image? While motion responses have been investigated thoroughly at the level of single neurons [1–4], studies of motion representation at the level of maps have been restricted to stimuli composed of long bars. These stimuli are inherently ambiguous as to their direction of motion [1]. A recent study by Fitzpatrick and collaborators [5] overcomes this limitation. Using optical imaging of intrinsic signals [6], the authors measured population responses to a variety of moving stimuli. Their results contradict the literature on optical imaging, but are in line with previous results obtained by recording from single neurons. As is common in optical imaging, Fitzpatrick and collaborators [5] recorded responses to gratings — arrays of long bars such as the one in Figure 1A — to measure a map of preferred orientation, that is, to assign a preferred orientation to each pixel in the imaged portion of V1. They then recorded responses to short tilted bars drifting diagonally — such as that in Figure 1B — and found that these responses were inconsistent with the map. Nonetheless, for each direction of the short bar, an ‘equivalent long bar’ could be found that elicited a similar pattern of response. For example, even though their orientations differ by 25°, the long bar in Figure 1A is ‘equivalent’ to the short bar in Figure 1B. Indeed, the two give rise to similar patterns of responses (Figure 1C). Manipulations of bar length, direction and speed brought the orientation of the ‘equivalent long bar’ closer to or further from the orientation of the short bar. These results flatly contradict the bulk the literature on optical imaging, which postulates or reports explicit maps of preferred orientation, spatial frequency and direction [7–11]. According to this literature, a stimulus with a given orientation, spatial frequency and direction should have elicited responses in those pixels that lie at the intersection of the corresponding maps. The results come as less of a surprise, however, if one considers the literature on responses of single neurons. After a debate that raged in the 1970s, it was largely agreed that V1 neurons do not isolate this or that feature of the stimulus. For example, an elegant study by the De Valois group [12] demonstrated that V1 neurons do not encode orientation independently of other features: preferred orientation depends on stimulus spatial frequency as predicted by a simple model based on the receptive field [13]. Similarly, it has also been known for a long time that V1 neurons do not encode direction of motion independently of other features. This issue was settled by the studies of responses of single V1 neurons to plaids — sums of two gratings — by Movshon and colleagues [1,2]. Dispatch Current Biology, Vol. 13, 000–000, December, 2 2003, ©2003 Elsevier Science Ltd. All rights reserved. DOI 10.1016/j.cub.2003.XX.XXX