On the evidence for a 'pure' binocular process in human vision

BRIAN TIMNEY, LAURIE M. WILCOX and ROBERT ST. JOHN* Department of Psychology, University of Western Ontario, London, Ontario, Canada N6A 5C2

Received13 October 1987;accepted 28 February 1989

Abstract-Wolfe (1986,Psychol. Rev. 93, 269-282)proposed a model of human based on the assumption of two functionallydistinct classesof binocular . These may be regarded as logicalAND and OR gates.In the present paper we assessthe evidencerelevant to this assumption.We find that whileboth types of binocular neuron have been describedin the cortex of cat and monkey,there is no indicationthat they form functionallyseparate populations.Critical analysis of the psychophysicalevidence for AND and OR channelsin human vision suggeststhat much of the data presentedin favor of an AND channelis subject to alternativeinterpretations. We concludethat the availabledata are not consistentwith the existenceof separate channels as proposed by Wolfe.

INTRODUCTION In a recent paper, Wolfe (1986) proposed a model of human binocular vision that attempted to integrate the mechanisms of and . His model depends critically upon two functionally distinct classes of binocular neuron. The neurons of one class may be described as logical AND gates. They can be activated only through simultaneous stimulation of both eyes and are assumed to play a major role in stereopsis. Neurons in the second class may be considered as OR gates. They can be stimulated through either eye and are assumed to play a major role in binocular rivalry. The OR neurons are considered to be secondary to the AND cells in the mediation of stereopsis. Wolfe cites several lines of evidence, both physiological and psychophysical, to support his arguments. In the present paper we shall argue that although there is physiological evidence for both AND and OR neurons, there is little support for the view that these should be considered as making up two separate channels, each serving a different purpose. We shall argue also that, while there is some psychophysical evidence for an AND process, there are major inconsistencies in the available data, and that it is premature to attempt to build a model of the kind proposed by Wolfe (1986).

PHYSIOLOGICALEVIDENCE FOR 'AND' NEURONS Hubel and Wiesel's (1962) original classification of visual neurons into seven categories was based on an assumption that binocular cells functioned as logical OR gates; that is, they would respond to stimulation through either eye.

* Presentaddress: Royal Roads Military College, FMO Victoria, British Columbia VOSIBO, Canada 2

They suggested that neurons falling into different ocular dominance classes served to mediate the perception of different depths, a notion which has received some support from the experiments of Poggio and Fischer (1977) and Ferster (1981). In making their classification, Hubel and Wiesel explicitly excluded a small proportion of binocular neurons that responded weakly or not at all to stimulation of either eye alone, but gave a strong response to simultaneous stimulation of both eyes. However, they did report that many of the cells falling into ocular dominance categories 2 through 6 showed binocular facilitation, giving an enhanced response when both eyes were stimulated together. In order for a neuron to subserve stereopsis, it must be sensitive to small differences in horizontal retinal disparity. Such disparity sensitive cells were first described by Barlow et al. (1967). These authors also reported that the majority of neurons they tested gave a strongly enhanced binocular response when presented with stimuli having the appropriate retinal disparity. They did not report that any of their units were exclusively binocular. Similar results have been obtained in a number of different studies of disparity selectivity. Few, if any, of the binocular neurons described in these studies could be considered strictly as AND neurons (Nikara et al., 1968; Pettigrew et al., 1968; Joshua and Bishop, 1970; Bishop et al., 1971; Von der Heydt et al., 1978; Fischer and Kruger, 1979; Ferster, 1981). Both Ferster (1981) and Gardner and Raiten (1986) have reported excellent disparity sensitivity among neurons classified as belonging to groups 1 and 7, the nominally monocular groups of Hubel and Wiesel (1962). Other studies have reported in more detail on exclusively binocular cells, although the proportions encountered are typically small (Poggio and Fischer, 1977; Poggio and Talbot, 1981). Poggio and Talbot studied stereoscopic responsiveness in the foveal cortex of rhesus monkey. Of 202 binocular cells studied, 75% were binocular; of these, 45% fell into a group that Poggio and Talbot classified as 'tuned excitatory'. Among these tuned excitatory cells approximately 40% were exclusively binocular. Thus, about 13% of binocular cells could be considered as AND units. All of the AND cells found by Poggio and Talbot were classified as 'tuned excitatory'. The distinguishing feature of these (and a second group of 'tuned inhibitory') cells was that they were selectively sensitive to a narrow range of disparities around 0 deg. Gardner and Raiten (1986), who found 4% of their units in cat to be AND types, also reported that these neurons were highly disparity-selective. Presumably, such neurons would be important in maintain- ing high levels of stereoacuity. Poggio and Talbot (1981) classified a second group of cells that responded over a much broader range of disparities as 'near' and 'far' cells. They suggested that these neurons could provide information for qualitative estimates of depth. Two conclusions may be drawn from these data. The first is that there do not appear to be any distinguishing characteristics of binocular AND cells, other than their binocular exclusivity. The response properties of tuned excitatory AND cells and tuned excitatory OR cells do not appear to differ. The second conclusion is that the majority of depth sensitive cells fall into Wolfe's class of binocular OR cells. This is the opposite of what is required by Wolfe's model. While it seems likely that AND cells do play an important role in stereopsis, there appears to be no reason to consider them as func- tionally different from binocular OR cells. We have been unable to find any evidence supporting the view that cells which respond only to binocular stimulation have also some special stereoscopic response properties that make them distinct from OR cells.