View metadata, citation and similar papers at core.ac.uk brought to you by CORE

provided by Elsevier - Publisher Connector

Vision Research 39 (1999) 3329–3345 www.elsevier.com/locate/visres

Section 2 Minireview Stereoscopic (cyclopean) motion sensing

Robert Patterson *

Department of Psychology and Program in Neuroscience, Washington State Uni6ersity, Pullman, WA 99164-4820, USA

Received 22 July 1998; received in revised form 29 January 1999

Abstract

This paper reviews literature on the motion processing of dynamic change in binocular disparity, called stereoscopic (cyclopean) motion. Studies investigating the visual processing of stereoscopic motion in the Z-axis, stereoscopic motion in the X/Y plane, and cyclopean motion are discussed. It is concluded that stereoscopic motion is processed by a motion-sensing system composed of special-purpose mechanisms that function like low-level motion sensors. For animals with binocular vision, low-level motion processing may involve, at least in part, stereoscopic processing. © 1999 Elsevier Science Ltd. All rights reserved.

Keywords: Motion perception; Motion pathways; Stereoscopic; Cyclopean

1. Introduction whether dynamic change in disparity is processed by an actual motion-sensing system. The ability to detect the motion of an object moving This paper reviews the literature on the motion pro- through three-dimensional space has important survival cessing of dynamic change in binocular disparity, called value for an animal. Motion processing provides infor- stereoscopic motion. Stereoscopic motion processing is mation for proprioception, detection of pattern, esti- an interesting topic because it involves motion informa- mating time-to-collision, and segmentation of surfaces tion at cyclopean (i.e. binocular-integration) levels of (Nakayama, 1985). For an animal with binocular vision vision (see Sherrington, 1906; Julesz, 1960, 1971). such as a human observer (see Fox, 1978), motion Stereoscopic motion processing would demonstrate a processing may involve stereopsis. binocular substrate for a portion of the motion system Consider an object moving in front of a background because this kind of motion would be computed subse- and on a given trajectory in three-dimensional space. quent to the computation of binocular disparity To an observer with stereopsis, one binocular cue to (Sekuler, 1975; Patterson, Ricker, McGary & Rose, object movement would be information about dynamic 1992). change in the relative binocular disparity between ob- In this paper, the term stereoscopic motion refers to ject and background. For example, an object moving the movement of binocular disparity information, through the Z-axis and toward the observer’s head which should be distinguished from the movement of would produce an increase in the magnitude of relative luminance boundaries presented with binocular dispar- disparity. An object moving laterally across the observ- ity. With respect to the latter, a number of studies (e.g. er’s visual field would produce dynamic change (i.e. Mezrich & Rose, 1977; Erkelens & Collewijn, 1985; displacement) in the lateral direction of the relative Nawrot & Blake, 1989; Halpern, 1991; Lappin & Love, disparity without a change in mean disparity. One 1992; Johnston, Cumming & Landy, 1994; Verstraten, important issue for theories of motion processing is Verlinde, Fredericksen & van de Grind, 1994; Bradley, Qian & Andersen, 1995; Qian & Andersen, 1997; Lankheet & Palmen, 1998) examined the interaction * Fax: +1-509-3355043. between luminance motion processing and stereoscopic E-mail address: [email protected] (R. Patterson) processing. Consider the following examples. Nawrot

0042-6989/99/$ - see front matter © 1999 Elsevier Science Ltd. All rights reserved. PII: S0042-6989(99)00047-4 3330 R. Patterson / Vision Research 39 (1999) 3329–3345 and Blake (1989) found that adaptation to stereoscopic may help select features for subsequent processing (Lu depth influenced the perception of structure from (lumi- & Sperling, 1995a,b) or it may generate higher-order nance) motion. Johnston et al. (1994) suggested that motion signals itself (Cavanagh, 1992, 1995). Position luminance motion information may overcome the tracking mechanisms show lowpass temporal-frequency stereopsis distance-scaling problem. Erkelens and tuning (Nakayama & Tyler, 1981). Collewijn (1985) revealed that a visual frame of refer- Many studies investigating stereoscopic motion pro- ence was necessary for the perception of motion in cessing have used dynamic random-dot stereograms depth but not for lateral motion. Lankheet and Palmen (Julesz, 1971) to isolate mechanisms devoted to stereop- (1998) found that luminance motion contrast improved sis. In this type of display, each eye’s view typically sensitivity for stereoscopic depth segregation. Finally, consists of an array of many small randomly-positioned Qian and Andersen (1997) provided a physiologic luminance dots. Binocular disparity is created between model of luminance motion-stereopsis integration the two eyes’ views by shifting laterally a subset of dots within the context of the Pulfrich phenomenon. Al- in one eye’s view and leaving unshifted corresponding though interesting, these studies will not be discussed dots in the other eye’s view (the shift is camouflaged by further because they involved luminance motion (i.e. surrounding dots). The shape defined by the shifted non-cyclopean motion containing monocular cues) dots creates a stereoscopic (cyclopean) form that is which is different from stereoscopic motion (i.e. cy- defined by differences in binocular disparity that cannot clopean motion containing no monocular cues). be seen monocularly. To study stereoscopic motion Nonetheless, these studies are generally consistent with processing, the stereoscopic form is moved and the the main theme of the present paper by showing inter- observer makes a perceptual judgment about the move- action between motion processing and stereoscopic ment. To camouflage monocular cues associated with processing. the stereoscopic motion, the luminance dot arrays are This review covers research on the visual processing dynamic (i.e. dots replotted randomly across frames of the motion sequence). In this kind of study, the issue of of stereoscopic motion in the Z-axis (i.e. saggital direc- motion sensing versus position tracking applies to the tion normal to the frontal or X/Y plane), stereoscopic moving stereoscopic form and not to the dynamic motion in the X/Y or frontal plane, and cyclopean luminance dots. motion, in order to discover whether and how stereo- The generation of dynamic random-dot stereograms scopic motion may be processed by a motion-sensing is technically challenging because dot arrays containing system. This review also discusses the possible neuro- a large number of elements are generated, displayed physiological basis of stereoscopic motion processing. and updated continuously in both eyes of an observer Before turning to these topics, however, this review with the appropriate amount of disparity implemented. begins by considering whether stereoscopic motion is One method for generating dynamic random-dot processed by an actual motion-sensing system. stereograms is to employ a digital computer that gener- ates the dot arrays off-line, stores them in memory, and 6 1.1. Motion sensing ersus position tracking later presents them to an observer during an experi- ment. A second method is to develop a special-purpose A controversy exists as to whether stereoscopic mo- analog computer that generates and displays the dot tion is processed by a true motion-sensing system or by arrays in real time. A third method is to create a hybrid a position-tracking mechanism. Motion sensing in- system that employs an analog computer that generates volves computing the spatial displacement of an ob- and displays the dot arrays in real time and a digital ject’s boundaries per unit of time. One common model computer for controlling the disparity embedded in the for a motion sensor is a Reichardt detector (Reichardt, dot arrays. In any of these cases, it is important to 1961) which possesses two spatially-separated regions ensure that the stereograms are devoid of monocular of a receptive field that are activated in sequence by a cues which could arise from visible cross-talk between moving boundary. Signals from one region are delayed the eyes (i.e. left eye’s information leaking into the right and integrated with signals from the other region, creat- eye or vice versa) or from non-linearities in screen ing a local motion signal. A Reichardt detector is luminance. equivalent to a motion-energy sensor which is based Consider now the evidence for whether stereoscopic upon the processing of spatial and temporal frequency motion is processed by a motion-sensing system or by a (Watson & Ahumada, 1983; van Santen & Sperling, position-tracking mechanism. We begin with position 1984, 1985; Adelson & Bergen, 1985). Motion sensors tracking. show bandpass temporal-frequency tuning (Nakayama & Tyler, 1981). Position tracking involves computing or 1.2. E6idence for stereoscopic position tracking inferring motion by comparing the current position of the features of a stimulus with their previous position Evidence that stereoscopic motion is processed by a and noting the positional change. Position tracking position-tracking mechanism comes from studies that R. Patterson / Vision Research 39 (1999) 3329–3345 3331 have failed to find evidence for stereoscopic motion the stereoscopic stimulus was poor, presumably because sensing. For example, Papert (1964) investigated trackable features were absent. Lu and Sperling pro- whether adapting to stereoscopic motion would induce posed that stereoscopic motion was processed by a a classic motion aftereffect. Papert found that only a feature tracking system that involved a motion energy weak stereoscopic aftereffect was induced. Anstis and analysis operating on the outputs of feature detectors Rogers (1975) (cited in Anstis, 1980), Zeevi and Geri (see also Lu & Sperling, 1995a). (1985), and Cavanagh (1995) also found that stereo- To examine attention and stereoscopic motion pro- scopic motion induced only weak or non-existent mo- cessing, Cavanagh (1995) investigated stereoscopic mo- tion aftereffects. Of these four studies, two of them tion processing in the X/Y plane in a display that (Papert, Zeevi and Geri) reported the duration of adap- contained both stereoscopic and luminance motion sep- tation, which was relatively brief (30 s or less). These arated by an angular distance of about 6°. When the results suggested that the lack of a stereoscopic motion luminance motion was attentionally tracked, the per- aftereffect meant that there was no stereoscopic mo- ceived direction of the stereoscopic motion became tion-sensing system, which invited the possibility that ambiguous. Cavanagh posited that poor performance the stereoscopic motion was processed by a position- with stereoscopic motion under shifted-attention condi- tracking mechanism. tions resulted from a lack of a stereoscopic motion Chang (1990) examined the perceptual interaction system without attentional tracking of position (see also between stereoscopic and luminance motion in a dy- Cavanagh, 1992). namic random-dot display. The stereoscopic motion Harris, McKee and Watamaniuk (1998) investigated was in the same or different direction as the luminance detection of a single small luminance dot (target) mov- dot motion. Chang found that the perception of lumi- ing in the Z-axis versus detection of a single target dot nance motion dominated the perception of stereoscopic moving in the X/Y plane. In both cases, the target had motion, with the latter appearing weak and in the to be detected as it moved through a group of station- direction of the former. Chang conjectured that the ary noise dots, with target motion being very slow (i.e. weak perception of stereoscopic motion resulted from a 0.07 deg/s). These authors found that detection perfor- lack of a stereoscopic motion-sensing system and that mance was poor for motion in the Z-axis, whereas stereoscopic motion perception was based upon posi- performance was good for motion in the X/Y plane. tion tracking. They suggested that whereas the target moving in the To investigate speed discrimination of stereoscopic X/Y plane was detected by a true motion system, the and luminance motion in the Z-axis, Harris and Wata- target moving in the Z-axis was detected by a position- maniuk (1995) created stimuli that began with a crossed tracking mechanism (which presumably was why the disparity and moved through the horopter to an un- stationary noise dots degraded performance for the crossed disparity, with the stimulus momentarily disap- Z-axis motion). pearing as it went through the horopter. These authors Seiffert and Cavanagh (1998) investigated motion reported that speed discrimination was poor with detection thresholds for a slowly moving stereoscopic stereoscopic motion compared with luminance motion. grating and found that detection thresholds were deter- They concluded that there was no specialized mecha- mined by a minimum displacement. These authors sug- nism for processing the speed of stereoscopic motion in gested that position tracking was important for the the Z-axis. perception of stereoscopic motion. In a different study, Harris and Watamaniuk (1996) Studies that failed to find evidence for a true stereo- examined speed discrimination of stereoscopic and lu- scopic motion-sensing system seemed to provide com- minance motion in the X/Y plane employing relatively pelling evidence against its existence. Many of these small stereoscopic gratings as stimuli. These authors studies made the inference that failure to find evidence again found that speed discrimination was poor with for stereoscopic motion sensing was due to the lack of stereoscopic motion compared with luminance motion. a motion-sensing system for stereoscopic information. They concluded that there was no specialized mecha- This inference, however, was invalid. nism for processing the speed of stereoscopic motion in In deductive reasoning, one logical fallacy is called the X/Y plane. denial-of-the-antecedent, in which an individual infers Lu and Sperling (1995b) investigated direction dis- that the consequent of a conditional statement is false if crimination of a stereoscopic compound stimulus (i.e. the antecedent is false. If we are given the proposition, corrugated surface in depth) that contained stereo- ‘‘if P then Q’’, and then the proposition, ‘‘not P’’, we scopic motion in the X/Y plane but no trackable fea- should not infer, ‘‘not Q’’ or we will be committing the tures. This stimulus was presented for a duration of one denial-of-the-antecedent fallacy. In the present context, temporal cycle plus one frame (i.e. exposure duration of if given the proposition that high stereoscopic speed about1sorless, depending upon temporal frequency). sensitivity (or strong stereoscopic motion adaptation, These authors found that direction discrimination of etc.) would be evidence for the existence of a stereo- 3332 R. Patterson / Vision Research 39 (1999) 3329–3345 scopic motion-sensing system, but no evidence of high only weak motion aftereffects, leading these authors to speed sensitivity (or strong motion adaptation) is infer that there was no stereoscopic motion-sensing found, the inference that a stereoscopic motion-sensing system. However, Patterson, Bowd, Phinney, Pohndorf, system does not exist is invalid. The lack of evidence Barton-Howard and Angilletta (1994) investigated the for a stereoscopic motion-sensing system is not the effect of adaptation duration on the stereoscopic mo- same as evidence for the lack of such a system. But this tion aftereffect and found that an adaptation durations is exactly the kind of logical fallacy that the studies greater than 30 s was needed to produce reliable stereo- reporting negative evidence have made. scopic motion aftereffects. These authors concluded One implication of this denial-of-the-antecedent that the reason why previous studies failed to find fallacy is that one could assume that a stereoscopic evidence for a stereoscopic motion aftereffect was that motion-sensing system did not exist when, in fact, such adaptation duration was too brief. (A dynamic test a motion system did exist, and the lack of evidence display also may be important for inducing significant would be due to factors other than the lack of such a stereoscopic motion aftereffects; see Nishida & Sato, system. We now consider two such factors that may 1995.) account for the lack of evidence for stereoscopic mo- Chang (1990) examined the perceptual interaction tion sensing, namely intrinsically weak stereoscopic mo- between stereoscopic and luminance motion and found tion signals and the use of inappropriate stimulus that stereoscopic motion perception was weak, leading parameters. her to conjecture that there was no stereoscopic mo- With respect to the former idea, intrinsically weak tion-sensing system. However, Ito (1997) investigated internal stereoscopic motion signals may have con- the perceptual interaction between stereoscopic and tributed to the negative evidence reported in the above luminance motion using a display similar to Chang’s studies. Many motion phenomena appear perceptually display. Ito found that certain stimulus parameters weak when examined with stereoscopic stimuli. For controlled whether stereoscopic motion or luminance example, Donnelly, Bowd and Patterson (1997) created motion dominated perception. For example, long inter- random-walk cinematogram displays composed of ar- frame intervals and large spatial displacements favored rays of moving stereoscopic or luminance discs, and stereoscopic motion processing, while short interframe found that the threshold for detecting coherent global intervals and small spatial displacements favored lumi- stereoscopic motion was five times higher than the nance motion processing, the latter of which were simi- threshold for detecting coherent global luminance mo- lar to the parameters employed by Chang. Ito’s results tion. These authors concluded that moving stereoscopic suggested that the weak perception of stereoscopic mo- boundaries engendered weak responding by the motion tion reported by Chang was likely to be due to her system, which may have been because the cyclopean choice of stimulus parameters. information bypassed peripheral stages of visual pro- In the investigation of speed discrimination of stereo- cessing (Julesz, 1971). Intrinsically weak stereoscopic scopic motion in the Z-axis, Harris and Watamaniuk motion signals, coupled with inappropriate stimulus (1995) reported that such discrimination was poor, parameters (see below), may have conspired to produce leading these authors to suggest that there was no negative evidence in some or all of the studies discussed specialized mechanism for processing the speed of above.1 stereoscopic motion in the Z-axis. However, Portfors- With respect to the possibility that the lack of evi- Yeomans and Regan (1996) investigated speed discrimi- dence for stereoscopic motion sensing was due to the nation of stereoscopic and luminance motion in the use of inappropriate stimulus parameters, consider the Z-axis using a factorial design that permitted dissocia- following studies that reported positive evidence for tion of speed versus positional information. These au- stereoscopic motion sensing. thors found that speed discrimination of stereoscopic motion was good and equal to that of luminance 1.3. E6idence for stereoscopic motion sensing motion under conditions that controlled for position. Portfors-Yeomans and Regan concluded that the speed Evidence that stereoscopic motion is processed by a of stereoscopic motion in the Z-axis was computed by motion-sensing system comes from a number of studies. a speed-sensitive mechanism. Moreover, these authors Recall that Papert (1964) and Zeevi and Geri (1985) showed that the reason why Harris and Watamaniuk employed a relatively brief duration of adaptation (30 s (1995) reported poor discrimination for stereoscopic or less) and found that stereoscopic motion induced motion was likely to be because their stereoscopic stimulus momentarily disappeared as it went through the horopter, which may have degraded visual 1 It is unclear whether stereoscopic strength would be defined along a disparity contrast continuum or along an interocular correlation processing. continuum; see Cormack, Stevenson and Schor (1993) for a discus- Recall that Harris and Watamaniuk (1996) reported sion. that speed discrimination was poor for stereoscopic R. Patterson / Vision Research 39 (1999) 3329–3345 3333 motion in the X/Y plane, leading Harris and Watama- Finally, consider several other studies reporting posi- niuk to conclude that there was no specialized mecha- tive evidence for stereoscopic motion sensing that have nism for processing the speed of stereoscopic motion in addressed the issue of position tracking. Patterson et al. the X/Y plane. However, Portfors and Regan (1997) (1992) found that direction discrimination of stereo- investigated speed discrimination of stereoscopic and scopic motion in the X/Y plane was governed by speed luminance motion in the X/Y plane using a factorial and not by a constant spatial displacement. Because a design that permitted dissociation of speed and posi- constant spatial displacement would be expected if dis- tion. These authors found that speed discrimination of crimination was based on position information, these stereoscopic motion was good and equal to that of authors suggested that stereoscopic motion processing luminance motion under conditions that controlled for did not rely upon position tracking. position. They concluded that the speed of stereoscopic Employing a complex stereoscopic-motion display motion in the X/Y plane was computed by a speed-sen- that camouflaged position information, Johns, Rogers sitive mechanism. Moreover, Kohly and Regan (1999) and Eagle (1996) determined whether the processing of revealed that the reason why Harris and Watamaniuk stereoscopic motion in the X/Y plane was velocity (1996) found poor discrimination for stereoscopic mo- limited or displacement limited. These authors found tion was likely to be because the stimuli employed in that thresholds for oscillating stereoscopic motion were the latter study were small. limited by velocity rather than by a fixed spatial dis- Seiffert and Cavanagh (1998) found that stereoscopic placement. They suggested that such results provided motion-detection thresholds were determined by a min- evidence for the existence of a stereoscopic motion- imum displacement, leading these authors to propose sensing system separate from position tracking. that position tracking was the basis of stereoscopic Patterson, Donnelly, Phinney, Nawrot, Whiting and motion processing. However, these authors noted that Eyle (1997) investigated speed discrimination versus their stimuli moved very slowly and thus may not have spatial-displacement discrimination in the X/Y plane, in been optimal for engaging a stereoscopic motion-sens- a display that contained arrays of randomly-positioned ing system. Under such conditions, the existence of a stereoscopic discs that were moved or displaced bidirec- specialized stereoscopic motion detector that was less tionally. These authors found that speed could be dis- sensitive than a position tracking mechanism remained criminated under conditions in which spatial a possibility to these authors. displacement could not (the speed discrimination These studies reporting positive evidence for a stereo- thresholds were quite high in this study). They pro- scopic motion-sensing system demonstrated that the posed that stereoscopic motion was sensed in a way lack of evidence for such a system was likely due to the that could not be explained by position tracking. use of inappropriate stimulus parameters. These studies Smith and Scott-Samuel (1998) investigated percep- revealed that, with appropriate stimulus parameters, tion of stereoscopic motion using a stereoscopic miss- positive evidence for such a system may be obtained. ing-fundamental squarewave stimulus (defined in the While not every report of negative evidence has been disparity domain) that was laterally displaced. When a shown to be due to inappropriate stimulus parameters, missing-fundamental squarewave is displaced, the dom- those studies reporting negative evidence owing to inap- inant motion energy occurs in a direction opposite to propriate parameters invite the possibility that a similar the displaced features of the stimulus. These authors explanation would apply to other studies as well. found that stereoscopic motion was perceived in the For example, Lu and Sperling (1995b) found that direction of the cyclopean motion energy and not in the observers could not perceive the direction of a stereo- direction of trackable features. They concluded that scopic compound stimulus that contained no trackable stereoscopic motion was computed by a cyclopean mo- features, leading these authors to argue that there was tion-energy mechanism operating on binocular-dispar- no stereoscopic motion system without position track- ity modulations rather than by a position-tracking ing. However, the exposure duration employed by Lu mechanism. and Sperling (1995b) may have been too brief for The studies discussed above addressed the issue of stereoscopic motion sensing, an issue taken up later (see whether stereoscopic motion is processed by an authen- also Carney, 1997). tic motion-sensing system, with some studies reporting Recall that Harris et al. (1998) found that detection negative evidence and other studies reporting positive performance was poor for a single target dot moving in evidence, with the positive evidence being logically the Z-axis through a group of stationary noise dots, stronger. The remainder of this review covers studies leading them to propose that the target motion was reporting positive evidence for stereoscopic motion detected by a position-tracking mechanism. It was note- sensing and the characteristics of such sensing. To this worthy that the target dot was small and its speed was author, the positive evidence for the existence of a very slow, conditions that may have favored position- stereoscopic motion-sensing system is abundant and tracking over stereoscopic motion sensing. incontrovertible. 3334 R. Patterson / Vision Research 39 (1999) 3329–3345

Fig. 1. Hypothetical motion-sensing system for stereoscopic (cyclopean) motion. Left-eye and right-eye signals converge at a binocular-integration stage at the level of V1. At the level of V1, stereoscopic boundaries are detected and binocular disparity is scaled or calibrated by viewing distance information. Stereoscopic motion signals are computed by a cyclopean motion-energy mechanism at a relatively early level of the motion stream, possibly at the level of V2. The stereoscopic motion signals are pooled with luminance and texture motion signals for the purpose of computing a two-dimensional pattern-motion code at the level of MT (note that the luminance and texture motion signals are initially computed at the levels of V1 and V2, respectively, the pathways of which are not shown in the diagram).

That stereoscopic motion can be perceived under Note that, as shown in Fig. 1, the possibility exists conditions that control for positional information indi- that stereoscopic motion sensing may involve scaled or cates that position tracking may be ruled out as a calibrated disparity. The difference between calibrated necessary mechanism for stereoscopic motion process- disparity versus raw disparity is related to the distinc- ing. That is, although position tracking by attention tion between disparity and depth. Disparity is defined may influence stereoscopic motion processing, as it may as an interocular difference in the position of corre- influence the motion processing of other kinds of stimu- sponding retinal images. Depth is defined as the Z-axis lus attributes, it does not appear to be necessary for interval between a given stimulus and fixation processing stereoscopic motion. Such results challenge (horopter). To derive a metric of stereoscopic depth, the validity of theories of stereoscopic motion process- disparity must be calibrated (scaled) by viewing dis- ing that invoke only position tracking as an explanation tance information because the same disparity value will (e.g. Chang, 1990; Cavanagh, 1995; Harris & Watama- correspond to different magnitudes of depth, depending niuk, 1995, 1996; Lu & Sperling, 1995b). Rather, on viewing distance (Wallach & Zuckerman, 1963; Ono stereoscopic motion processing is velocity-limited (e.g. & Comerford, 1977; Patterson & Martin, 1992). Inter- Patterson et al., 1992; Johns et al., 1996), direction-se- estingly, calibrated disparity seems to be represented lective (Patterson & Becker, 1996; Phinney, Bowd & early in visual cortex: Trotter, Celebrini, Stricanne, Patterson, 1997), and based on cyclopean motion-en- Thorpe and Imbert (1992) showed that the activity of ergy (Smith & Scott-Samuel, 1998). Therefore, stereo- many disparity-tuned cells in V1 was affected by view- scopic motion appears to be processed by an authentic ing distance, suggesting that these cells were represent- motion-sensing system. For an overview of this pro- ing calibrated disparity information. At present, it is posed system, see Fig. 1. not clear whether stereoscopic motion sensing involves R. Patterson / Vision Research 39 (1999) 3329–3345 3335 raw disparity or calibrated disparity, although the latter sidered stereoscopic motion in this review. Thus, mo- remains a distinct possibility. tion in three-dimensional space is related to The question arises as to the locus of stereoscopic stereoscopic motion insofar as the latter may be consid- motion sensing. One possibility is that the computation ered to be a binocular cue for the former. Regan (1993), of stereoscopic motion occurs relatively late in the Cumming and Parker (1994), and Brenner, van den motion stream, which in this paper will be taken to be Berg and van Damme (1996) all found that dynamic at the level of, or levels subsequent to, an area in change in binocular disparity was an important cue for humans homologous to monkey area MT. Another perceiving the motion of objects in three-dimensional possibility is that the computation of stereoscopic mo- space (for review of the earlier literature, see Regan, tion occurs relatively early in the motion stream, which Kaufman & Lincoln, 1988). will be taken to be at a level prior to area MT. As the For example, Regan (1993) employed a dynamic literature on stereoscopic motion sensing is reviewed in random-dot stereogram display devoid of monocular this paper, it should become clear that the literature is cues and found that the apparent direction of stereo- consistent with the idea that stereoscopic motion sens- scopic motion in three-dimensional space was given by ing occurs relatively early in the motion-processing the ratio of translational velocity to the rate of change stream (see Fig. 1). in disparity. He showed that the ratio of translational velocity to disparity change was a sufficient cue for motion perception in three-dimensional space. 2. Stereoscopic motion sensing Cumming and Parker (1994) measured detection thresholds for stereoscopic motion in the Z-axis using Motion sensing of stereoscopic information has been temporally-uncorrelated random-dot stereograms (de- investigated within a number of different paradigms. void of interocular retinal-image velocity differences) These paradigms may be classified into two categories: and temporally-correlated random-dot stereograms Stereoscopic motion in the Z-axis and stereoscopic (containing interocular retinal-image velocity differ- motion in the X/Y plane. These categories are discussed ences). These authors found that thresholds were lower below. In addition, a paradigm involving cyclopean for the uncorrelated stereogram that lacked interocular motion without disparity will also be covered. This velocity differences than for the correlated stereogram section ends with a discussion of the possible neuro- that contained such differences. They also found that physiology of stereoscopic motion sensing. the perception of motion in three-dimensional space was not evoked when observers viewed a stimulus that 2.1. Stereoscopic motion in the Z-axis contained interocular retinal-image velocity differences coupled with dynamic changes in disparity beyond the To an observer with binocular vision, an object mov- spatiotemporal resolution limit of stereopsis. Cumming ing in front of a background and on a given trajectory and Parker argued that the temporal derivative of in three-dimensional space would provide two binocu- disparity was adequate to explain the perception of lar cues about such movement. The first cue involves motion in three-dimensional space. the relationship between the velocities (i.e. direction Brenner et al. (1996) examined the relative influence and speed) of the retinal-images in the two eyes. If the of changes in target vergence (i.e. target’s position object moves in the Z-axis, the interocular retinal-im- relative to the two eyes), target retinal image size, and age velocities would be different. If the object moves disparity on perceived motion in three-dimensional laterally, the two retinal-image velocities would be the space. These authors found that all three factors af- same or very similar. For this cue, the visual system fected such motion perception. They suggested that would be computing the retinal-image velocities in the changing disparity may be one of several cues for two eyes first and then comparing them. motion perception in three-dimensional space. The second cue involves dynamic change in the rela- Thus, the perception of motion in three-dimensional tive binocular disparity between object and back- space is related to the perception of stereoscopic motion ground. If the object moves in the Z-axis, the in the Z-axis because the former seems to be based, at magnitude of the relative disparity would change. If the least in part, upon computing the latter (Regan, 1993; object moves laterally, the relative disparity would Cumming & Parker, 1994; Brenner et al., 1996). Indeed, change dynamically (i.e. it would be displaced) in the evidence for the existence of a separate stereoscopic lateral direction without a change in mean disparity. motion system is its sensitivity for Z-axis motion For this cue, the visual system would be computing the (Tyler, 1971) and different temporal frequency tuning binocular disparity first and then computing the motion relative to monocular motion (Tyler, 1975; also see of the disparity. Tyler, 1990). The remaining portion of this section The second cue for motion in three-dimensional reviews studies on the visual processing of stereoscopic space—dynamic change in binocular disparity—is con- motion in the Z-axis. 3336 R. Patterson / Vision Research 39 (1999) 3329–3345

2.1.1. Temporal resolution quence. Each frame contained four stereoscopic discs Tyler (1971), Richards (1972), Regan and Beverley positioned at locations equidistant from one another on (1973c), Beverley and Regan (1974a), Norcia and Tyler the circumference of an imaginary circle. On one frame, (1984), and White and Odom (1985) investigated the the discs occupied positions at 12, 3, 6 and 9 o’clock. upper limit of temporal resolution of stereopsis in the The disparity/depth of the discs alternated across posi- Z-axis by measuring depth perception while disparity tions, such that discs at the 12 and 6 o’clock positions was temporally varied, from a crossed to an uncrossed had one disparity value while discs at the 3 and 9 value, across a range of temporal frequencies. Observ- o’clock positions had a lesser disparity. On subsequent ers perceived a depth plane oscillating in depth toward frames, the positions of the discs were rotated clockwise and away from them up to a frequency of about 1–5 by some angular amount. The observer’s task was to Hz, above which temporal summation of disparity in- indicate perceived direction of rotation (clockwise or formation occurred and motion perception failed. counterclockwise). These results showed that temporal resolution in the In this kind of display, motion is typically perceived disparity domain was about a factor of 10 worse than in the direction corresponding to the shortest distance. temporal resolution in the luminance domain. Green and Odom investigated whether this spatial proximity rule applied to the Z-axis. They found that a 6 2.1.2. Direction selecti ity trade-off existed between distance in the X/Y-plane and Regan and Beverley (1973a) investigated the effects the Z-axis. When subsequent frames were rotated of adaptation on the perception of rotation in depth clockwise relative to previous frames by an amount created by varying interocular retinal-image motion sufficiently greater than 45°, counterclockwise motion which also varied binocular disparity. These authors was seen, even though that perception entailed seeing found that adaptation to motion along a given path of motion in the Z-axis. Thus, the spatial proximity rule rotation in three-dimensional space decreased sensitiv- applied to stereoscopic apparent motion in three dimen- ity for that direction of rotation, but such adaptation sions. Norcia and Tyler (1984), cited earlier under the increased or left unaffected sensitivity to the opposite temporal resolution section, were the first to study direction of rotation. They suggested the existence of cyclopean apparent motion in the Z-axis. different classes of disparity detectors tuned to different directions of motion (see also Beverley & Regan, 1974b; 2.1.5. Changing size Regan & Beverley, 1973d). In an examination of the relationship between chang- Recording electrical brain responses to stimuli that ing size (looming) and the perception of motion in the moved in the Z-axis toward or away from the plane of Z-axis, Regan and Beverley (1978) and Beverley and fixation, Regan and Beverley (1973b) found that the Regan (1979) found that adaptation to a changing-size electrical brain responses were different for stimuli that carried a mean crossed disparity versus a mean un- stimulus whose dimensions increased over time made a crossed disparity. The electrical brain responses were test stimulus appear to move continuously away in also different for stimuli that moved toward the plane depth, while adaptation to the changing-size stimulus of fixation relative to stimuli that moved away from the whose dimensions decreased over time made the test plane of fixation. According to these authors, these stimulus appear to move closer in depth. These authors results provided evidence for different classes of mecha- proposed that the mechanisms sensitive to changing size nisms encoding motion in three dimensions. fed into the mechanisms that mediated the perception of motion in the Z-axis. 2.1.3. Speed discrimination Gray and Regan (1996) examined the discrimination Recall from an earlier section that Portfors-Yeomans of stereoscopic and luminance motion in the Z-axis and Regan (1996) and Portfors and Regan (1997) inves- under conditions of disparity oscillation, size oscilla- tigated speed discrimination of stereoscopic and lumi- tion, and oscillatory motion within the frontoparallel nance motion in the Z-axis. Both studies used a plane. These authors found that thresholds for stereo- factorial design that permitted the dissociation of speed scopic motion produced by disparity oscillation were and position. Both studies found that speed discrimina- similar to thresholds for luminance motion. Moreover, tion of stereoscopic motion was equal to that of lumi- the perception of stereoscopic or luminance motion in nance motion under conditions that controlled for the Z-axis could be cancelled by pitting disparity oscil- position. Both studies concluded that a speed-sensitive lation against size oscillation. Gray and Regan con- mechanism existed for stereoscopic motion. cluded that the stimulus for perceiving motion in the Z-axis was the rate of change in disparity, and that 2.1.4. Apparent motion signals produced by changing size and changing dispar- Green and Odom (1986) created a display that con- ity converged onto common mechanisms that signaled sisted of several frames of an apparent motion se- motion in three dimensions. R. Patterson / Vision Research 39 (1999) 3329–3345 3337

Brenner et al. (1996), cited earlier, also investigated the direction of stereoscopic motion was encoded by a the effects of changes in disparity and image size on the distribution of adaptable, direction-selective mecha- perception of motion in three dimensions. These au- nisms as proposed for luminance motion (e.g. Levinson thors reported that changes in both disparity and image & Sekuler, 1976; Marshak & Sekuler, 1979; Moulden, size were important cues for three-dimensional motion 1980). Patterson and Becker also found that the repul- perception. sion aftereffect transferred between the stereoscopic and Thus, research on the processing of stereoscopic mo- luminance domains (i.e. aftereffect with stereoscopic tion in the Z-axis suggests that there are special-pur- and luminance patterns employed as adapting and test pose mechanisms for computing such motion. The next stimuli, respectively, and vice versa), thus stereoscopic section covers a different kind of paradigm for investi- and luminance motion were processed by a common gating stereoscopic motion sensing. direction-selective mechanism and substrate. In the investigation of direction discrimination of 2.2. Stereoscopic motion in the X/Y plane stereoscopic and luminance global motion cited earlier, Donnelly et al. (1997) used stereoscopic and luminance To an observer with binocular vision, an object mov- random-walk cinematograms equated for effective ing laterally in front of a background would provide strength by presenting them with signal strength set at two binocular cues about such movement: (1) The equal multiples of global-motion detection threshold. interocular retinal-image velocities would be the same Under equal strength conditions, direction discrimina- or very similar; and (2) the relative disparity between tion thresholds were equal for stereoscopic and lumi- object and background would change dynamically in nance global motion. These authors suggested that the the lateral direction without a change in mean dispar- directional precision of global motion pooling was the ity. Thus, an alternative paradigm to studying stereo- same for stereoscopic and luminance motion signals. scopic motion in the Z-axis is to investigate stereoscopic motion in the X/Y plane. This section 2.2.3. Speed discrimination reviews studies on the visual processing of stereoscopic Recall that Portfors and Regan (1997) investigated motion in the X/Y plane. speed discrimination of stereoscopic and luminance mo- tion in the X/Y plane using a factorial design that 2.2.1. Temporal resolution permitted the dissociation of speed and position. Speed Patterson et al. (1992) investigated the upper limit of discrimination of stereoscopic motion was equal to that temporal resolution of stereopsis in the X/Y plane by of luminance motion under conditions that controlled measuring direction discrimination for a disparity- for position. These authors concluded that the speed of defined grating that moved laterally at a given temporal stereoscopic motion in the X/Y plane was computed by rate. They found that stereoscopic motion in the X/Y a speed-sensitive mechanism. plane was perceived up to a frequency of 8 Hz, above Patterson et al. (1997) investigated speed discrimina- which temporal summation of disparity information tion versus displacement discrimination in the X/Y occurred and motion perception failed. This value of 8 plane using a bidirectional stereoscopic motion display. Hz was similar to, albeit slightly higher than, the 1–5 These authors found that speed could be discriminated Hz limit of temporal resolution for stereoscopic motion under conditions in which displacement could not. The in the Z-axis. speed discrimination thresholds were higher in this study than in the Portfors and Regan (1997) study, a 2.2.2. Direction selecti6ity difference that may have been due to different experi- Patterson and Becker (1996) investigated the effects mental paradigms. Patterson et al. suggested that of adapting to stereoscopic motion in the X/Y plane on stereoscopic motion was sensed in a way that could not the perceived direction of subsequently-viewed stereo- be explained by position tracking. scopic test motion. These authors found that adapting to stereoscopic motion in a given direction made the 2.2.4. Apparent motion direction of the test motion appear repulsed away from Julesz and Payne (1968) briefly presented two station- its true direction, a direction-selective repulsion afteref- ary stereoscopic stimuli in temporal succession. These fect. Phinney et al. (1997) investigated the effects of authors found that such stimulation induced the per- adapting to stereoscopic motion in the X/Y plane on ception of stereoscopic apparent motion in the X/Y direction discrimination of stereoscopic test motion. plane. This result indicated that the stereoscopic stimuli Phinney et al. found that direction-discrimination engaged a motion system. thresholds were elevated 20–30° away from the direc- Cavanagh, Arguin and von Grunau (1989) briefly tion of adaptation, a direction-selective threshold-eleva- presented stationary stimuli in temporal succession. tion aftereffect. As pointed out by both studies, Within a given temporal sequence, the stimuli were direction-selective adaptation provided evidence that defined by differences in luminance, texture, color, or 3338 R. Patterson / Vision Research 39 (1999) 3329–3345 binocular disparity. These authors found that apparent above fixation combined with leftward stereoscopic mo- motion was perceived when the temporal sequence con- tion below fixation) induced a bidirectional motion tained stimuli defined by different attributes, including aftereffect. Adaptation duration ranged from 30 s to 64 disparity (i.e. inter-attribute apparent motion percep- min per trial. These authors found that robust stereo- tion). This result suggested that binocular disparity fed scopic motion aftereffects of many seconds were in- into the same motion process and substrate as other duced with aftereffect duration proportional to the boundary cues. square root of adaptation duration, similar to the lumi- In an examination of the spatial-displacement limit nance motion aftereffect. Moreover, two opposite af- for stereoscopic apparent motion, Phinney, Wilson, tereffects, with a distinct border between them where Hays, Peters and Patterson (1994) found that, for the the oppositely-moving stimuli met, were induced simul- same exposure duration and stimulus onset asynchrony, taneously. Bowd et al. suggested that the stereoscopic stereoscopic apparent motion was perceived with spa- motion aftereffect was retinotopic. tial displacements three times greater than the displace- Nishida and Sato (1995) investigated whether adapt- ment limit for luminance apparent motion perception. ing to stereoscopic motion induced a motion aftereffect These authors interpreted this result as suggesting that when a flickering versus stationary non-stereoscopic the receptive fields of the stereoscopic motion mecha- test pattern was viewed. Adaptation duration was 30 s. nisms were larger than the receptive fields of luminance Although adaptation duration was relatively brief, motion mechanisms (see also Patterson et al., 1992). these authors revealed that strong stereoscopic afteref- In an investigation of bistable apparent motion in a fects of many seconds were induced with the flickering, stereoscopic Ternus display, Patterson, Hart and but not the stationary, test pattern. Thus, a dynamic Nowak (1991) and Petersik (1995) revealed that element test display may also be important for inducing signifi- motion was perceived at short interstimulus intervals, cant stereoscopic motion aftereffects (most studies that whereas group motion was seen at long interstimulus have reported the existence of stereoscopic motion af- intervals, similar to the luminance Ternus display. tereffects employed dynamic test displays). Nishida and These results suggested that the mechanism that pro- Sato proposed that the static motion aftereffect was duces the bistability in the Ternus display may be the induced by a first-order motion mechanism, while the same for stereoscopic and luminance motion. flicker motion aftereffect was induced by a second-or- 2.2.5. Classic motion aftereffect der motion mechanism. According to these authors, the stereoscopic motion aftereffect would be considered a In several brief reports, Lehmkuhle and Fox (1977), 2 Fox, Patterson and Lehmkuhle (1982), and Stork, second-order motion phenomenon . Crowell and Levinson (1985) examined whether adapt- Several studies have investigated the disparity contin- ing to stereoscopic motion induced a motion aftereffect. gency of the stereoscopic motion aftereffect. A dispar- Adaptation duration was 45 s in the first study, 90 s in ity-contingent motion aftereffect is an aftereffect that is the second study, and unreported in the third study. contingent upon the binocular disparity of the adapt (Such studies typically controlled fixation by establish- and test stimuli. Disparity contingency suggests that the ing a fixation point in the middle of the motion dis- visual system contains mechanisms that code for both play.) These studies found that stereoscopic motion direction of motion and binocular disparity (e.g. Maun- induced strong motion aftereffects lasting many sell & Van Essen, 1983a,b), and that common mecha- seconds. nisms are engaged only when adapt and test stimuli Recall that Patterson et al. (1994) investigated the have the same or similar disparity. effects of varying the duration of adaptation on the In the brief reports mentioned earlier, Lehmkuhle stereoscopic motion aftereffect. These authors found and Fox (1977) and Fox et al. (1982) adapted observers that an adaptation duration longer than 30 s was to stereoscopic motion in one depth plane and tested required for robust stereoscopic aftereffects. They also for the aftereffect in the same or different depth plane. found that the motion aftereffect transferred between Both studies found that differences in disparity between the stereoscopic and luminance domains, which indi- adapt and test stimuli made the stereoscopic motion cated that stereoscopic and luminance motion process- aftereffect decline in strength. ing were mediated, at least in part, by a common mechanism and substrate. These results were replicated 2 According to Julesz (1971) and Cavanagh and Mather (1989), by Webster, Panthradil and Conway (1998). Patterson first-order motion processing refers to the processing of stimulus and Becker (1996) replicated the results of Patterson et boundaries defined by differences in first-order statistics (e.g. lumi- al. (1994) within a repulsion aftereffect paradigm. nance-defined boundaries) whereas second-order motion processing refers to the processing of stimulus boundaries defined by differences Bowd, Rose, Phinney and Patterson (1996) investi- in second-order statistics (e.g. disparity-defined boundaries); the latter gated whether prolonged adaptation to bidirectional process may involve more complex pre-processing before the motion stereoscopic motion (i.e. rightward stereoscopic motion extraction stage than the former process. R. Patterson / Vision Research 39 (1999) 3329–3345 3339

Patterson, Bowd, Phinney, Fox and Lehmkuhle In an investigation employing stereoscopic plaid pat- (1996) investigated whether the stereoscopic motion terns (i.e. plaids created by crossing two moving stereo- aftereffect was affected by differences in disparity be- scopic gratings each defined by disparity modulation), tween adapt and test stimuli, or by differences in dis- Bowd, Donnelly and Patterson (1997) examined the parity between stimuli and fixation (horopter). The effects of adaptation to a moving stereoscopic plaid, or relative disparity among adapt, test, and fixation stim- its components, on the perceived coherence of a lumi- uli was varied. These authors revealed that the stereo- nance test plaid, and vice versa. These authors found scopic aftereffect was greatest when adapt and test were that adapting to a moving plaid or its components of presented with zero disparity and in the plane of fixa- one stimulus type (stereoscopic or luminance) affected tion, and the aftereffect declined as the disparity of the coherence of a moving test plaid of the other adapt and/or test increased away from fixation. Patter- stimulus type. They suggested that stereoscopic and son et al. suggested that robust stereoscopic motion luminance motion signals fed into the same two-dimen- aftereffects occurred when adapt and test stimuli en- sional pattern-motion process. gaged common mechanisms that encoded positions Alais, van der Smagt, Verstraten and van de Grind near fixation (horopter). (1996) found that adaptation to dichoptically-presented Shorter, Bowd, Donnelly and Patterson (1998) inves- luminance gratings did not significantly affect perceived tigated whether the stereoscopic motion aftereffect was coherence of a luminance test plaid, which led these selective for either the spatial frequency or orientation authors to conclude that plaid coherence was mediated of disparity modulation. These authors found that the by a purely monocular mechanism. However, these strongest stereoscopic motion aftereffect was induced authors employed only a 30-s adaptation duration, when adapt and test had the same spatial frequency and which may have been too brief to induce an adaptation orientation (i.e. the stereoscopic motion aftereffect was effect (see Patterson et al., 1994). The results of Bowd et al. (1997) indicated that plaid coherence was medi- selective for spatial frequency and orientation). They ated, at least in part, by a cyclopean mechanism. suggested that the stereoscopic motion aftereffect was Donnelly, Bowd and Patterson (1996) employed ran- mediated by oriented stereoscopic spatial-frequency dom-walk cinematograms composed of arrays of mov- mechanisms. ing stereoscopic or luminance discs, and examined the For a recent review of the stereoscopic motion af- effects of adaptation to global stereoscopic motion on tereffect, see Moulden, Patterson and Swanston (1998). the perceptual coherence of global luminance motion, and vice versa. These authors found that adapting to 2.2.6. Local motion-signal pooling and surface the global motion of one stimulus type (stereoscopic or representation luminance) affected the perceived coherence of global In order to represent the motion of coherent two-di- motion of the other stimulus type. They suggested that mensional patterns, it is thought that localized motion stereoscopic and luminance motion signals fed into the signals from a lower processing level are pooled at same global-motion process. higher levels. To study this motion pooling process, Given that plaid patterns and global motion displays Adelson and Movshon (1982) (see also Movshon, Adel- are thought to activate motion-integration mechanisms son, Gizzi & Newsome, 1985; Movshon & Newsome, in area MT (Movshon et al., 1985; Newsome & Pare, 1996) created moving two-dimensional plaid patterns 1988; Movshon & Newsome, 1996), results of Donnelly by crossing and superimposing two moving luminance et al. (1996) are consistent with results of Bowd et al. gratings (called components). Under certain conditions, (1997). Both studies showed that stereoscopic and lumi- the two gratings phenomenally cohered into a single nance motion signals feed into a common motion-inte- plaid pattern, while under other conditions, coherence gration mechanism (likely in a human area homologous was absent and two gratings were perceived as sliding to monkey area MT; see Fig. 1). across one another and moving in their individual Patterson, Bowd and Donnelly (1998) showed that directions. Movshon and colleagues proposed a two- the barber pole illusion (i.e. grating pattern appearing stage process wherein the first stage detected the motion to move in the direction of the long axis of a rectangu- of the individual components while the second stage lar aperture) was perceived with a moving stereoscopic integrated the first-stage signals (producing perceptual grating and a stereoscopic aperture. The barber pole coherence) and computed the motion of the entire illusion is usually interpreted as being a product of two-dimensional plaid. Movshon et al. (1985) and mechanisms that generate and propagate local motion Movshon and Newsome (1996) presented neurophysio- signals in order to represent coherently-moving rigid logical evidence that the (luminance) component-mo- surfaces. These authors suggested that the generation tion signals were computed in primate area V1, whereas and propagation of stereoscopic motion signals at cy- the two-dimensional plaid-motion signals were com- clopean levels of vision played a part in the representa- puted in area MT. tion of coherently-moving surfaces. 3340 R. Patterson / Vision Research 39 (1999) 3329–3345

Thus, research on the processing of stereoscopic mo- In an investigation of the cyclopean motion afteref- tion in the X/Y plane suggests that there is a special- fect, Carney and Shadlen (1993) revealed that adapting purpose mechanism for computing such motion (Fig. to cyclopean motion created from dichoptically-viewed 1). The next section covers research on the processing flickering gratings induced a strong motion aftereffect. of cyclopean motion without disparity. These authors also reported that motion discrimination was possible when dichoptic versions of random-texture 2.3. Cyclopean motion motion displays were viewed (the latter of which pre- sumably revealed early motion processing). Carney and This section covers studies on the visual processing of Shadlen suggested that a binocular substrate existed for non-stereoscopic cyclopean motion. This form of cy- luminance motion processing. clopean motion is discussed in this review because it Lu and Sperling (1995b) examined direction discrimi- involves motion information that depends upon binoc- nation of cyclopean motion created from dichoptically- ular integration and therefore is related to the topics viewed flickering compound stimuli that contained no discussed in this paper. trackable features. The stimuli were exposed for a Anstis and Moulden (1970) created an apparent mo- duration of one full temporal cycle of 3.0 Hz plus one tion display in which lights comprising frames of a frame (about 375 ms). These authors found that direc- motion sequence stimulated the two eyes which pro- tion discrimination of cyclopean motion was poor pre- duced either monocular or interocular apparent mo- sumably because trackable features were absent. They tion. Adaptation to monocular apparent motion concluded that poor performance with cyclopean mo- induced a monocular motion aftereffect while adapta- tion in the absence of trackable features meant that tion to interocular apparent motion induced a cy- there was no cyclopean-motion processing without fea- clopean motion aftereffect. These authors suggested ture tracking. that the motion aftereffect had both monocular and However, Carney (1997) examined direction discrimi- cyclopean components. nation of cyclopean motion using dichoptically-viewed flickering compound stimuli that contained no track- Shadlen and Carney (1986) showed that cyclopean able features, similar to Lu and Sperling (1995b), but motion perception can be induced from dichoptically- the stimuli were exposed for a longer duration of 2 s. viewed flickering luminance gratings presented in spa- Carney found that direction discrimination of cy- tio-temporal quadrature. These motion stimuli were clopean motion was good even though trackable fea- based upon the principle of decomposing traveling tures were absent. He concluded that cyclopean motion sinewave gratings into the sum of two waves in spatial was processed by a binocular motion-energy system and temporal quadrature. The interocular quadrature without feature-tracking. phase relationship engaged motion sensors at binocu- The studies that failed to find evidence for the exis- lar-integration levels of vision without binocular tence of a cyclopean motion-sensing system suggested disparity. that cyclopean motion was processed solely by a fea- In a study examining cyclopean motion created from ture-tracking system. But, as discussed earlier, we dichoptically-viewed flickering gratings, Georgeson and should not commit the fallacy in conditional reasoning Shackleton (1989) found that, relative to motion per- called denial-of-the-antecedent. We should not infer ception under monocular conditions, motion perception that a cyclopean motion system does not exist on the was poor under dichoptic conditions and displayed basis of a lack of evidence for its existence. characteristics inconsistent with a motion sensor (e.g. The studies that reported positive evidence for the p low incidence of reverse motion when the contrast of existence of a cyclopean motion-sensing system sug- one frame of a dichoptically-viewed motion sequence gested that cyclopean motion processing may possess was reversed). Georgeson and Shackleton posited that characteristics similar to early motion sensing, therefore early motion sensors were not activated dichoptically early motion sensing may have, at least in part, a and that cyclopean motion was processed by a position- binocular substrate. The existence of cyclopean motion tracking mechanism. sensing would be consistent with the existence of stereo- However, Carney and Shadlen (1992) reviewed the scopic motion sensing insofar as both would involve the Georgeson and Shackleton (1989) study and offered an computation of motion subsequent to binocular alternative interpretation. In doing so, Carney and integration. Shadlen pointed out that there was positive evidence To summarize this review up to this point, a number for cyclopean motion sensing without feature tracking of studies failed to find evidence for stereoscopic mo- under many of Georgeson and Shackleton’s conditions, tion sensing. Other studies, however, showed that and that poor performance with cyclopean motion un- stereoscopic information is processed by a motion sys- der other conditions was likely due to the use of tem composed of special-purpose mechanisms that inappropriate stimulus conditions. function like early low-level motion sensors. Mecha- R. Patterson / Vision Research 39 (1999) 3329–3345 3341 nisms for stereoscopic motion in the X/Y plane appear ing to interocular differences in retinal-image velocity to be retinotopic (Bowd et al., 1996; Shorter et al., or to dynamic change in disparity. 1998), selective for spatial-frequency and orientation Poggio, Motter, Squatrito and Trotter (1985) (see (Shorter et al., 1998), and based upon cyclopean mo- also Poggio & Poggio, 1984) found that many cells in tion-energy (Smith & Scott-Samuel, 1998). Mechanisms primate area V1 responded to stereoscopic boundaries for stereoscopic motion both in the X/Y plane and in embedded in a dynamic random-dot stereogram. These the Z-axis appear to be selective for direction (Regan authors suggested that stereoscopic boundary and form & Beverley, 1973a,b,c; Beverley & Regan, 1974b; Pat- information was extracted at the level of area V1. terson & Becker, 1996; Phinney et al., 1997). These Finally, DeAngelis, Cumming and Newsome (1998) results are consistent with some of the research on electrically stimulated clusters of disparity-tuned neu- cyclopean motion processing that suggests a binocular rons in area MT of rhesus monkey and examined the substrate for early motion sensing. effect of such stimulation on perceptual judgments of The next section discusses the possible neurophysio- depth in a random-dot display. These authors found logical basis of stereoscopic motion sensing. that electrical stimulation of disparity-tuned cells bi- ased depth judgments in a way predictable from the 2.4. Neurophysiology of stereoscopic motion sensing disparity preference of the cells at the stimulation site. They posited that behaviorally relevant signals for At present, there appears to be no direct evidence stereoscopic depth were present in area MT. for the existence of cells that respond selectively to Taken together, this research is consistent with the dynamic change in binocular disparity. To show selec- speculation that stereoscopic motion signals are com- tive responding to dynamic change in disparity, such puted relatively early in the motion stream, namely in cells would need to have a negligible monocular re- areas of the motion stream prior to an area in humans sponse and a directional response to disparity change, homologous to monkey area MT, such as area V2 (e.g. an issue that should be resolvable with future research. Smith, Greenlee, Singh, Kraemer & Hennig, 1998; see However, there is some research that provides insight Fig. 1). Stereoscopic motion signals may feed onto as to the possible neurophysiology of stereoscopic mo- common pattern-motion mechanisms in area MT tion sensing. Speculation from neurophysiological re- where the stereoscopic motion signals are integrated search is tentative because visual cortical areas are with luminance and texture motion signals in repre- multifunctional and more complex than has been pre- senting the motion of coherent two-dimensional pat- viously recognized (Schiller, 1996). The circuitry of terns (Bowd et al., 1997). Such speculation is consistent striate and extrastriate cortex is a complex interacting with the results of psychophysical studies showing that system involving feedback (Zipser, Lamme & Schiller, stereoscopic motion is likely to be computed by spe- 1996). Despite this complexity, one can speculate about cial-purpose mechanisms that function like low-level neurophysiology from considering the following stud- motion sensors. ies. A number of studies (e.g. Hubel & Wiesel, 1962; Schiller, Finlay & Volman, 1976; Zeki, 1978; Maunsell 3. Concluding remarks & Van Essen, 1983a,b; Hammond & Pomfreff, 1989; Roy & Wurtz, 1990; Roy, Komatsu & Wurtz, 1992) 1. Previous failures to find evidence for stereoscopic revealed the existence of cells with both direction selec- motion sensing were likely due to the use of stimu- tivity and binocularity in early and late cortical areas lus parameters ill suited for revealing such sensing, of the motion stream of cats and monkeys. These rather than to the lack of a stereoscopic motion- studies, however, did not explicitly test whether any of sensing system. Such stimulus parameters included these cells were selectively sensitive to dynamic change brief duration of adaptation, brief duration of expo- in binocular disparity. Nonetheless, these studies sure, small stimulus size, and slow speed. showed that motion and binocularity were represented 2. It is doubtful that position-tracking of stereoscopic early in the motion-processing stream. motion would show characteristics usually at- Cells that responded selectively to motion toward or tributable to motion sensing, such as being away from an animal were found in area MT of rhesus retinotopic; selective for spatial-frequency, orienta- monkey by Zeki (1974a,b), in area 18 of cat by Cy- tion and direction; compute cyclopean motion-en- nader and Regan (1978), and near the area 17/18 ergy; and yield effects such as apparent motion, border of cat by Regan and Cynader (1982). In these motion adaptation, barber pole illusion, plaid-mo- studies, the animals were presented with monocularly- tion perception, and global-motion perception (ran- visible moving edges and bars for which retinal image dom-walk cinematogram). Position tracking should motion would have been visible. Therefore, it was involve general-purpose mechanisms possessing a unclear in these studies whether the cells were respond- low-pass tuning function for temporal variation. 3342 R. Patterson / Vision Research 39 (1999) 3329–3345

3. Stereoscopic motion appears to be processed by a Anstis, S. M. (1980). The perception of apparent movement. Philo- motion-sensing system composed of special-purpose sophical Transactions of the Royal Society, London, B, 290, 153– 168. mechanisms that function like early motion sensors, Anstis, S. M., & Moulden, B. P. (1970). Aftereffect of seen move- with characteristics such as being retinotopic; selec- ment: evidence for peripheral and central components. Quarterly tive for spatial-frequency, orientation and direction; Journal of Experimental Psychology, 22, 222–229. compute cyclopean motion-energy; and yield effects Beverley, K. I., & Regan, D. (1974a). Temporal integration of such as apparent motion, motion adaptation, barber disparity information in stereoscopic perception. Experimental pole illusion, plaid-motion perception, and global- Brain Research, 19, 228–232. Beverley, K. I., & Regan, D. (1974b). Visual sensitivity to disparity motion perception. Motion sensing should involve pulses: evidence for directional selectivity. Vision Research, 14, special-purpose mechanisms possessing a band-pass 357–361. tuning function for temporal variation. Beverley, K. I., & Regan, D. (1979). Separable aftereffects of chang- 4. If stereoscopic motion is computed by a cyclopean ing-size and motion-in-depth: different neural mechanisms. Vision motion-energy mechanism, early motion processing Research, 19, 727–732. of different stimulus attributes would be based upon Bowd, C., Donnelly, M., & Patterson, R. (1997). Cross-domain adaptation to component gratings affects perceived coherence of one kind of motion computation because luminance cyclopean and luminance plaids. In6estigati6e Ophthalmology and motion (Adelson & Bergen, 1985) and texture mo- Visual Science (Supplement), 38, S903. tion (Chubb & Sperling, 1989; Wilson, Ferrera & Bowd, C., Rose, D., Phinney, R. E., & Patterson, R. (1996). Endur- Yo, 1992) are computed as motion-energy in the ing stereoscopic motion aftereffects induced by prolonged adapta- luminance domain (for texture motion, the energy tion. Vision Research, 36, 3655–3660. Bradley, D. C., Qian, N., & Andersen, R. A. (1995). Integration of computation would follow a rectification operation). motion and stereopsis in the middle temporal cortical area of The computation of cyclopean motion energy would macaques. Nature, 373, 609–611. involve integrating the spatio-temporal envelope of Brenner, E., van den Berg, A. V., & van Damme, W. J. (1996). the disparity profile with some integration window Perceived motion in depth. Vision Research, 36, 699–706. (although the spatial extent of integration has yet to Carney, T. (1997). Evidence for an early motion system which be determined, the temporal extent of integration is integrates information from the two eyes. Vision Research, 37, 2361–2368. about 125 ms; see e.g. Patterson et al., 1992). This Carney, T., & Shadlen, M. N. (1992). Binocularity of early motion would mean that models of motion processing that mechanisms: comments on Georgeson and Shackleton. Vision neglect stereoscopic motion and binocularity are Research, 32, 187–191. incomplete. Carney, T., & Shadlen, M. N. (1993). Dichoptic activation of the 5. Stereoscopic motion may be computed early in the early motion system. Vision Research, 33, 1977–1995. Cavanagh, P. (1992). Attention-based motion perception. Science, motion stream (i.e. prior to an area in humans 257, 1563–1565. homologous to monkey area MT). Cavanagh, P. (1995). Is there low-level motion processing for non-lu- 6. From an ecological perspective, the visual analysis minance-based stimuli? In T. V. Papathomas, C. Chubb, A. of dynamic change in binocular disparity may be an Gorea, & E. Kowler, Early 6ision and beyond. Cambridge, MA: integral part of early motion processing for primates MIT. and other animals with binocular vision. To under- Cavanagh, P., Arguin, M., & von Grunau, M. (1989). Interattribute apparent motion. Vision Research, 29, 1197–1204. stand motion processing, it may be necessary to Cavanagh, P., & Mather, G. (1989). Motion: the long and short of it. understand stereopsis. Spatial Vision, 4, 103–129. Chang, J. J. (1990). New phenomena linking depth and luminance in stereoscopic motion. Vision Research, 30, 137–147. Acknowledgements Chubb, C., & Sperling, G. (1989). Two motion perception mecha- nisms revealed through distance-driven reversal of apparent mo- tion. Proceedings of the National Academy of Science, 86, The author would like to thank Stephanie Shorter, 2985–2989. Robert Fox and Christopher Bowd for comments on an Cormack, L. K., Stevenson, S. B., & Schor, C. M. (1993). Disparity- earlier version of this manuscript. tuned channels of the human visual system. Visual Neuroscience, 10, 585–596. Cumming, B. G., & Parker, A. J. (1994). Binocular mechanisms for detecting motion-in-depth. Vision Research, 34, 483–495. References Cynader, M., & Regan, D. (1978). Neurons in cat parastriate cortex sensitive to the direction of motion in three-dimensional space. Adelson, E. H., & Bergen, J. R. (1985). Spatiotemporal energy Journal of Physiology, 274, 549–569. models for the perception of motion. Journal of the Optical DeAngelis, G. C., Cumming, B. G., & Newsome, W. T. (1998). Society of America, A2, 284–299. Cortical area MT and the perception of stereoscopic depth. Adelson, E. H., & Movshon, J. A. (1982). Phenomenal coherence of Nature, 394, 677–680. moving visual patterns. Nature, 300, 523–525. Donnelly, M., Bowd, C., & Patterson, R. (1996). Bandwidth of Alais, D., van der Smagt, M. J., Verstraten, F. A. J., & van de Grind, directionally-selective mechanisms for cyclopean (stereoscopic) W. A. (1996). Monocular mechanisms determine plaid motion motion. In6estigati6e Ophthalmology and Visual Science (Supple- coherence. Visual Neuroscience, 13, 615–626. ment), 37, S285. R. Patterson / Vision Research 39 (1999) 3329–3345 3343

Donnelly, M., Bowd, C., & Patterson, R. (1997). Direction discrimi- Levinson, E., & Sekuler, R. (1976). Adaptation alters perceived nation of cyclopean (stereoscopic) and luminance motion. Vision direction of motion. Vision Research, 16, 779–781. Research, 37, 2041–2046. Lu, Z.-L., & Sperling, G. (1995a). Attention-generated apparent Erkelens, C. J., & Collewijn, H. (1985). Motion perception during motion. Nature, 377, 237–239. dichoptic viewing of moving random-dot stereograms. Vision Lu, Z.-L., & Sperling, G. (1995b). The functional architecture of Research, 25, 583–588. human visual motion perception. Vision Research, 35, 2697–2722. Fox, R. (1978). Binocularity and stereopsis in the evolution of Marshak, W., & Sekuler, R. (1979). Mutual repulsion between mov- vertebrate vision. In S. J. Cool, & E. L. Smith, Frontiers in 6isual ing visual targets. Science, 205, 1399–1401. science. New York: Springer-Verlag. Maunsell, J. H. R., & Van Essen, D. C. (1983a). Functional proper- Fox, R., Patterson, R., & Lehmkuhle, S. W. (1982). Effect of depth ties of neurons in middle temporal area of the macaque monkey. position on the motion aftereffect. In6estigati6e Ophthalmology I. Selectivity for stimulus direction, speed and orientation. Journal and Visual Science (Supplement), 22, 144. of Neurophysiology, 49, 1127–1147. Georgeson, M. A., & Shackleton, T. M. (1989). Monocular motion Maunsell, J. H. R., & Van Essen, D. C. (1983b). Functional proper- sensing, binocular motion perception. Vision Research, 29, 1511– ties of neurons in middle temporal area of the macaque monkey. 1523. II. Binocular interactions and sensitivity to binocular disparity. Gray, R., & Regan, D. (1996). Cyclopean motion perception pro- Journal of Neurophysiology, 49, 1148–1167. duced by oscillations of size, disparity and location. Vision Re- Mezrich, J. J., & Rose, A. (1977). Coherent motion and stereopsis in search, 36, 655–665. dynamic visual noise. Vision Research, 17, 903–910. Green, M., & Odom, J. V. (1986). Correspondence matching in Moulden, B. (1980). After effects and the integration of patterns of apparent motion: evidence for three-dimensional spatial represen- neural activity within a channel. Philosophical Transactions of the tation. Science, 233, 1423–1429. Royal Society of London B, 290, 39–55. Halpern, D. L. (1991). Stereopsis from motion-defined contours. Moulden, B., Patterson, R., & Swanston, M. (1998). The retinal Vision Research, 31, 1611–1617. image, ocularity, and cyclopean vision. In G. Mather, F. Hammond, P., & Pomfreff, C. J. D. (1989). Directional and orienta- Verstraten, & S. Anstis, The motion aftereffect: a modern perspec- 6 tional tuning of feline striate cortical neurons: correlation with ti e. Cambridge, MA: MIT. neuronal class. Vision Research, 29, 653–662. Movshon, J. A., Adelson, E. H., Gizzi, M. S., & Newsome, W. T. (1985). The analysis of moving patterns. In C. Chagas, R. Gat- Harris, J. M., & Watamaniuk, S. N. J. (1995). Speed discrimination tass, & C. Gross, Pattern recognition mechanisms. New York: of motion-in-depth using binocular cues. Vision Research, 35, Springer. 885–896. Movshon, J. A., & Newsome, W. T. (1996). Visual response proper- Harris, J. M., & Watamaniuk, S. N. J. (1996). Poor speed discrimina- ties of striate cortical neurons projecting to area MT in macaque tion suggests that there is no specialized speed mechanism for monkeys. Journal of Neuroscience, 16, 7733–7741. cyclopean motion. Vision Research, 36, 2149–2157. Nakayama, K. (1985). Biological image motion processing: a review. Harris, J. M., McKee, S. P., & Watamaniuk, S. N. J. (1998). Visual Vision Research, 25, 625–660. search for motion-in-depth: stereomotion does not pop out from Nakayama, K., & Tyler, C. W. (1981). Psychophysical isolation of disparity noise. Nature Neuroscience, 1, 165–1168. movement sensitivity by removal of familiar position cues. Vision Hubel, D. H., & Wiesel, T. N. (1962). Receptive fields, binocular Research, 21, 427–433. interaction and functional architecture in cats visual cortex. Jour- Nawrot, M., & Blake, R. (1989). Neural integration of information nal of Physiology, 160, 106–154. specifying structure from stereopsis and motion. Science, 244, Ito, H. (1997). The interaction between stereoscopic and luminance 716–718. motion. Vision Research, 37, 2553–2559. Newsome, W. T., & Pare, E. B. (1988). A selective impairment of the Johns, A. M., Rogers, B. J., & Eagle, R. A. (1996). Thresholds for motion perception following lesions of the middle temporal visual detecting moving cyclopean forms are limited by velocity not area MT. Journal of Neuroscience, 5, 2201–2211. 6 6 displacement. In estigati e Ophthalmology and Visual Science Nishida, S., & Sato, T. (1995). Motion aftereffect with flickering test (Supplement), 37, S285. patterns reveals higher stages of motion processing. Vision Re- Johnston, E. B., Cumming, B. G., & Landy, M. (1994). Integration of search, 35, 477–490. stereopsis and motion shape cues. Vision Research, 34, 2259– Norcia, A. M., & Tyler, C. W. (1984). Temporal frequency limits for 2275. stereoscopic apparent motion processes. Vision Research, 24, 395– Julesz, B. (1960). Binocular depth perception of computer-generated 401. patterns. Bell System Technical Journal, 30, 1125–1162. Ono, H., & Comerford, T. (1977). Stereoscopic depth constancy. In Julesz, B. (1971). Foundations of cyclopean perception. Chicago: Uni- W. Epstein, Stability and constancy in 6isual perception: mecha- versity of Chicago. nisms and processes (pp. 91–128). New York: Wiley. Julesz, B., & Payne, R. A. (1968). Differences between monocular Papert, S. (1964). Stereoscopic synthesis as a technique for localizing and binocular stroboscopic movement perception. Vision Re- visual mechanisms. M.I.T. Quarterly Progress Report No., 73, search, 8, 433–444. 239. Kohly, R. P., & Regan, D. (1999). Evidence for a mechanism Patterson, R., & Becker, S. (1996). Direction-selective adaptation and sensitive to the speed of cyclopean motion. Vision Research, 39, simultaneous contrast induced by stereoscopic (cyclopean) mo- 1011–1024. tion. Vision Research, 36, 1773–1781. Lankheet, M. J. M., & Palmen, M. (1998). Stereoscopic segregation Patterson, R., Bowd, C., & Donnelly, M. (1998). The cyclopean of transparent surfaces and the effect of motion contrast. Vision (stereoscopic) barber pole illusion. Vision Research, 38, 2119– Research, 38, 659–668. 2125. Lappin, J. S., & Love, S. R. (1992). Planar motion permits the Patterson, R., Bowd, C., Phinney, R., Fox, R., & Lehmkuhle, S. W. perception of metric structure in stereopsis. Perception and Psy- (1996). Disparity tuning of the stereoscopic (cyclopean) motion chophysics, 51, 86–102. aftereffect. Vision Research, 36, 975–983. Lehmkuhle, S., & Fox, R. (1977). Stereoscopic motion aftereffects. Patterson, R., Bowd, C., Phinney, R., Pohndorf, R., Barton-Howard, Paper presented to Midwestern Psychological Association, W., & Angilletta, M. (1994). Properties of stereoscopic motion Chicago, IL. aftereffects. Vision Research, 34, 1139–1147. 3344 R. Patterson / Vision Research 39 (1999) 3329–3345

Patterson, R., Donnelly, M., Phinney, R. E., Nawrot, M., Whiting, Roy, J. P., & Wurtz, R. H. (1990). The role of disparity-sensitive A., & Eyle, T. (1997). Speed discrimination of stereoscopic (cy- cortical neurons in signalling the direction of self-motion. Nature, clopean) motion. Vision Research, 37, 871–878. 348, 160–162. Patterson, R., Hart, P., & Nowak, D. (1991). The cyclopean Ternus van Santen, J. P. H., & Sperling, G. (1984). Temporal covariance display and the perception of element versus group movement. model of human motion perception. Journal of the Optical Society Vision Research, 31, 2085–2092. of America, 1A, 451–473. Patterson, R., & Martin, W. L. (1992). Human stereopsis. Human van Santen, J. P. H., & Sperling, G. (1985). Elaborated Reichardt Factors, 34, 669–692. detectors. Journal of the Optical Society of America, A2, 300–321. Patterson, R., Ricker, C., McGary, J., & Rose, D. (1992). Properties Schiller, P. H. (1996). On the specificity of neurons and visual areas. of cyclopean motion perception. Vision Research, 32, 149–156. Beha6ioral Brain Research, 76, 21–35. Petersik, J. T. (1995). A comparison of varieties of second-order Schiller, P. H., Finlay, S. F., & Volman, J. (1976). Quantitative motion. Vision Research, 35, 507–517. studies of single-cell properties in monkey striate cortex. I. Spatio- Phinney, R. E., Bowd, C., & Patterson, R. (1997). Direction-selective temporal organization of receptive fields. Journal of Neurophysiol- coding of stereoscopic (cyclopean) motion. Vision Research, 37, ogy, 39, 1288–1319. 865–869. Seiffert, A. F., & Cavanagh, P. (1998). Position displacement, not Phinney, R., Wilson, R., Hays, B., Peters, K., & Patterson, R. (1994). velocity, is the cue to motion detection of second-order stimuli. Spatial displacement limits for cyclopean apparent motion percep- Vision Research, 38, 3569–3582. tion. Perception, 23, 1287–1300. Sekuler, R. (1975). Visual motion perception. In E. C. Carterette, & Poggio, G. F., Motter, B. C., Squatrito, S., & Trotter, Y. (1985). M. P. Friedman, Handbook of perception. In: Seeing, vol. 5 (pp. Responses of neurons in visual cortex (V1 and V2) of the alert 387–430). New York: Academic. macaque to dynamic random-dot stereograms. Vision Research, Shadlen, M., & Carney, T. (1986). Mechanisms of human motion 25, 397–406. perception revealed by a new cyclopean illusion. Science, 232, Poggio, G. F., & Poggio, T. (1984). The analysis of stereopsis. Annual 95–97. 6 6 Re6iew of Neuroscience, 7, 379–412. Sherrington, C. S. (1906). The integrati e action of the ner ous system. Portfors, C. V., & Regan, D. M. (1997). Just-noticeable difference in New Haven, CT: Yale University. the speed of cyclopean motion in depth and the speed of cy- Shorter, S., Bowd, C., Donnelly, M., & Patterson, R. (1998). Spatial- 6 clopean motion within a frontoparallel plane. Journal of Experi- frequency tuning of the stereoscopic motion aftereffect. In estiga- ti6e Ophthalmology and Visual Science (Supplement), 39, S1081. mental Psychology: Human Perception and Performance, 23, Smith, A. T., Greenlee, M. W., Singh, K. D., Kraemer, F. M., & 1074–1086. Hennig, J. (1998). The processing of first- and second-order Portfors-Yeomans, C. V., & Regan, D. (1996). Cyclopean discrimina- motion in human visual cortex assessed by functional magnetic tion thresholds for the direction and speed of motion in depth. resonance imaging (fMRI). Journal of Neuroscience, 18, 3816– Vision Research, 36, 3265–3279. 3830. Qian, N., & Andersen, R. A. (1997). A physiological model for Smith, A. T., & Scott-Samuel, N. E. (1998). Stereoscopic and con- motion-stereo integration and a unified explanation of Pulfrich- trast-defined motion in human vision. Proceedings of the Royal like phenomena. Vision Research, 37, 1683–1698. Society of London B, 265, 1573–1581. Regan, D. M. (1993). Binocular correlates of the direction of motion Stork, D. G., Crowell, J. A., & Levinson, J. Z. (1985). Cyclopean in depth. Vision Research, 33, 2359–2360. motion aftereffect in the presence of monocular motion. In6estiga- Regan, D., & Beverley, K. I. (1973a). Disparity detectors in human ti6e Ophthalmology and Visual Science (Supplement), 26, 55. depth perception: evidence for directional selectivity. Science, 181, Trotter, Y., Celebrini, S., Stricanne, B., Thorpe, S., & Imbert, M. 877–879. (1992). Modulation of neural stereoscopic processing in primate Regan, D., & Beverley, K. I. (1973b). Electrophysiological evidence area V1 by the viewing distance. Science, 257, 1279–1281. for existence of neurons sensitive to direction of depth movement. Tyler, C. W. (1990). A stereoscopic view of visual processing streams. Nature, 246, 504–506. Vision Research, 30, 1877–1895. Regan, D., & Beverley, K. I. (1973c). Some dynamic features of depth Tyler, C. W. (1975). Characteristics of stereomovement suppression. perception. Vision Research, 13, 2369–2379. Perception and Psychophysics, 17, 225–230. Regan, D., & Beverley, K. I. (1973d). The dissociation of sideways Tyler, C. W. (1971). Stereoscopic depth movement: two eyes less movements from movements in depth: psychophysics. Vision Re- sensitive than one. Science, 174, 958–961. search, 13, 2403–2415. Verstraten, F. A. J., Verlinde, R., Fredericksen, R. E., & van de Regan, D., & Beverley, K. I. (1978). Illusory motion in depth: Grind, W. A. (1994). A transparent motion aftereffect contingent aftereffect of adaptation to changing size. Vision Research, 18, on binocular disparity. Perception, 23, 1181–1188. 209–212. Wallach, H., & Zuckerman, C. (1963). The constancy of stereoscopic Regan, D., & Cynader, M. (1982). Neurons in cat visual cortex tuned depth. American Journal of Psychology, 76, 404–412. to the direction of motion in depth: effect of stimulus speed. Watson, A. B., & Ahumada, A. J. (1983). A look at motion in the In6estigati6e Ophthalmology and Visual Science, 22, 535–550. frequency domain. In J. K. Tosotsos, Motion: perception and Regan, D. M., Kaufman, L., & Lincoln, J. (1988). Motion in depth representation (pp. 1–10). New York: Association for Computing and visual acceleration. In K. Boff, L. Kaufman, & J. T. Thomas, Machinery. Handbook of perception and human performance. In: Sensory pro- Webster, W. R., Panthradil, J. T., & Conway, D. M. (1998). A cesses and perception, vol. 1. New York: Wiley (Chapter 19). rotational stereoscopic 3-dimensional movement aftereffect. Vi- Reichardt, W. (1961). Autocorrelation, a principle for the evaluation sion Research, 38, 1745–1752. of sensory information by the central nervous system. In W. A. White, K. D., & Odom, J. V. (1985). Temporal integration in global Rosenblith, Sensory communication. New York: Wiley. stereopsis. Perception and Psychophysics, 37, 139–144. Richards, W. (1972). Response functions for sine- and square-wave Wilson, H. R., Ferrera, V. P., & Yo, C. (1992). A psychophysically modulations of disparity. Journal of the Optical Society of Amer- motivated model for two-dimensional motion perception. Visual ica, 62, 907–911. Neuroscience, 9, 79–97. Roy, J. P., Komatsu, H., & Wurtz, R. H. (1992). Disparity sensitivity Zeevi, Y. Y., & Geri, G. A. (1985). A purely central movement of neurons in monkey extrastriate area MST. The Journal of aftereffect induced by binocular viewing of dynamic visual noise. Neuroscience, 12, 2478–2492. Perception and Psychophysics, 38, 433. R. Patterson / Vision Research 39 (1999) 3329–3345 3345

Zeki, S. M. (1974a). Cells responding to changing image size and Zeki, S. M. (1978). Uniformity and diversity of structure and function disparity in the cortex of the rhesus monkey. Journal of Physiol- in rhesus monkey prestriate visual cortex. Journal of Physiology, ogy, 242, 827–841. 277, 273–290. Zeki, S. M. (1974b). Functional organization of a visual area in the Zipser, K., Lamme, V. A. F., & Schiller, P. H. (1996). Contextural posterior bank of the superior temporal sulcus of the rhesus modulation in primary visual cortex. Journal of Neuroscience, 16, monkey. Journal of Physiology, 236, 549–573. 7376–7389.