LETTER COLOR
TO THE
AND EDGE SENSITIVE STEREO-DEPTH
Held and Shattuck (1971) demonstrated that a tilt aftereffect (AE) can be made contingent upon the color of a test grating. The adapting procedure involved alternating 5-set presentations of a red grating tilted clockwise off vertical by 15’ with a green grating tilted counterclockwise off vertical by 15‘. After the observers had scanned these gratings for IO min. a vertical green test grating appeared tilted slightly clockwise. and a vertical red grating appeared tilted slightly counterclockwise. These aftereffects. of opposite direction contingent upon color. demonstrate adaptation specific to each edge-color pair and imply a close linkage between channels selective for edge orientation and those selectiv-e for color in the visual system. Similar implications have been drawn from studies of the McCollough aftereffect (McCollough, 1969; Stromeyer. 1972). from work on rivalry of colored gratings (Campbell, Gilinsky, Howell, Riggs and Atkinson. 1973; Rauschecker. Campbell and Atkinson. 1973). and from the evoked potential studies of Regan (I 973). One would like to know at what level in the visual system this linkage ofcolor and edge-orientation information occurs. The type and degree of interaction between the two eyes can provide a clue. For example. substantial interocular transfer of an AE is usually interpreted as evidence for a locus of processing at or beyond the level of combination of information from the two eyes. The orientation-contingent color AE. demonstrated by McCollough. is reported to transfer either not at all or to a very slight degree (McCollough. 1965; Coltheart, 1973). That is. if one eye alone is exposed to the adapting gratings, the other eye shows either no demonstrable contingent AE. or a very weak one. The color fringe AE produced at luminance gradients. such as are induced by a prism, are reported not to transfer at all (Hajos and Ritter, 1965; Hay, Pick and Rosser, 1963). Dichoptic presentation. with color presented to one eye and edge orientation to the other. is reported to yield an AE less than halfas strong as that ’ This ivork was first presentsd at the Association for Research in Vision and Ophthalmology meeting in 1973 (program abstract p. 35. No. 2). The research was supported National of Health. Institutes Grant NO. by I ROI EYOI 191-01, and National Aeronautics and Space Administration. Grant No. NGL X-009-308.
EDITORS
CHANNELS CONVERGE ANALYZERS’
ON
for monoptic presentation. with both in the same eve (MacKay and XlacKay. 1973). Moreover. this result-is interpreted as -. taking place before the stage at which form and color are associated”. which is compatible with the notion that color-edge channels originating in separate eyes are independent. Taken together, these results indicate at most a vveak central interaction between coloredge information from the separate eyes at the level where this color-contingent tilt AE is generated. Monocular color-specific orientation channels may. however, converge at SOW level on binocular channels. Julesz (1971) suggested such a possibility. but was unable to demonstrate it with his stereograms. Using the color-contingent tilt aftereffect we have investigated. in the present experiments. the interaction between information reaching the separate eyes by testing for: (1) interocular transfer; (2) simultaneous induction of opposing AEs in the two eves: and (3) change in stereoscopic depth accompancing (2) and consequent on convergence of monocular channels on binocular analyzers of stereoscopic depth.
Each eye of the observer was tested separately, before and after a 20-min exposure of only one eye to two square wave gratings of 3.0 cycles per degree, alternated once every 5 sec. The red exposure was tilted off vertical by 15’ in one direction and the green tilted off vertical in the other. The observer was instructed to continuallv shift his fixation within the central region of the grating. The test gratings were circular, subtended 4>, and were divided mto upper and lower halves composed of gratings of differing color and equal but opposite tilt with respect to vertical. The gratings were either red above. green below or green above. red Ixlow. The tilts ofeach half varied in steps of IO’of arc from 60’ clockwise to 60’ counterclockwise off vertical. The combined tilts of the top and bottom halves of the test fields formed a series of angles slightly deviating from ISO’. These were presented to the observer in a double staircase procedure (Cornsweet. 1961) in which “red above. green below” test gratings were always alternated with ‘-green above. red below”. in order to equalize color adaptation during testing. Two projectors were used to
309
310
Letter 10 the Edttors
present the test gratings. The dark bars were maintained at a luminance of 2.5 ft-L. the light bars at 63 h-L for both test and exposure gratings. The size of the color-contingent tilt AE was measured as the change. from before to after exposure. in the angle (between bars of upper and lower test half fields) which was perceived by the observer as 1YO’ or straight. On each trial the observer either increased or decreased by one step the angle formed by the gratings in accord with his judgment of their deviation from apparent straightness. The mean shift for I2 observers was X4 in the eye exposed to the adapting stimuli. and less than 1’ in the unexposed eye. indicating no appreciable interocular transfer of the AE. One of the observers was run five times and one four times. In both cases. the mean shift in the exposed eye was significantly different {at the O-05level by t-test) from that in the unexposed eye. and the latter did not differ signifrcantiy from zero. This result stands in sharp contrast to the high degree of interocular transfer (approaching 100 per cent) for tilt induced by adaptation to edges defined by luminance gradients alone (Campbell and Maffei, 197I).
Ifcolor-contingent tilt AEs of opposite direction can be established simultaneously in the two eyes. we can ask two further questions: (a) what are the magnitudes ofthese AEs relative to those generated by exposure of only one eye to adapting stimuli” A sign&cant difference would suggest interocuiar interaction: (b) different frontal tilts of edges in corresponding retinal loci in the separate eyes are equivalent to a binocular disparity. Do the tilts produced as AEs (see Fig. I) act as if they altered binocular disparity; i.e. are they convertcd into a change in stereoscopic depth’! If so. we may conclude that information from monocuIar coioredge orientation channels converges onto binocular channels at the level of stereoscopic depth processing. A dove prism was set in front of one eye to reverse (left to right) the tilts of both the adapting and test gratings presented to that eye. Each eye was tested separately, before and after Xmin of adaptation. as described above. The mean shifts in the angle of subjective straightness were 31’ in the right eye and 37’ in the opposite direction in the left eye, for eight subjects: an effect of larger magnitude. but not significantly so. than that obtained with adaptation ofa single eye. This result supports the hypothesis that color-edge orientation information from the two eyes is processed independently.
A slightly modified experimental procedure allowed us to decide whether the induction of color-contingent tilt AEs. opposite in the two eyes. leads to a measureable change in judgements of stereoscopic depth. When both eyes simultaneously view a tilted test stimulus with reversing prism in place. the two eyes view opposite tilts. As a result. the upper and lower
Fig. I. Stereo depth
induction
(bars on left deleted).
planes of the test gratings appear to be tilted in depth, forming a dihedral angle pointing awa>: from or toward the observer (Fig. I ). Under these Circumstances, the sequential presentation of test stimuti produces variation in tilt disparity in steps of 20’ (IO’ per eye). We used the double-statrcase procedure to measure the change in frontal target angles required to make the target appear subjectively flat (dihedral angle of ISO’) after opposite adaptation of the two eyes. That change measured the amount of conversion of the frontal tilt AE to stereoscopic depth. Since shifts in frontal tilt of settings to subjective straightness were also measured for each eye separately. the shift in stereoscopic depth found in the binocular test can be expressed as a percentage of the average shift in frontal tilt for the two eyes. This is seen in Fig. 2. where the line at 45’ represents 100 per cent conversion from the two opposite monocular frontal tilt shifts to the predicted shift in stereoscopic depth. Six observers were run once monocularly and once binocularly in separate sessions (open circles). One observer (RH) was tested in eight sessions, four on frontal tilt and four on depth ‘%lt constituting four combined but uncorre~ated runs (square shows mean. bars show range). and one observer (SS) was tested in four sessions. each time both monocularly and binocularly during the same session (filled circles).
1 ,
0
v
0
10
20
30
40
.W (min./eye)
Monocular tilt aftereffect Fig. 2. Shift in perceived stereo depth as a function in perceived frontal tilt after adaptation to opposite tilts in the two eyes.
of shift frontai
311
Letter to the Editors
Clearly. there is substantial conversion to stereo depth: all observers show a positive change although its magnitude varies considerably. It is unlikely that this is the result of direct a~p~tion of binocular stereoscopic depth channels, since the disparities of the adapting tilts are 30” and the stimuli are rivalrous during the adapting exposure. Moreover, we have found substantial stereo depth shifts even after the separate eyes were adapted sequentially, rather than simuitaneously, to opposite tilts; in this case. no direct adaptation of binocular channels could occur. SU>l;\tARY The results confirm the conclusion, drawn from studies of the orientation-contingent color AE (McCollough effect), that information about combined color and edge orientation is initially processed separately in monocular channels. It does. however, converge on binocular channels, at least at the level of processing for stereoscopic depth. Other investigators have been unable to demonstrate a depth-specific color AE after binocular adaptation to red and green gratings producing different stereoscopic depths which presumably should adapt coior selective binocular channels if they exist (Brown, Stromeyer and Dawson. 1974).These findings suggest that the binocular channels themselvesare not selective for color, even though they receive information from monocular channels which are color selective. Taken together. present knowledge drawn from psychophysical research implies that the linkage between cotor and edge orientation processors occurs prior to and not in binocular channels. Massachusetts Institute of Technology, Department of Psychology, 79 Amherst Stree, Caprice, ~as~&hu~tts 02139, U.S.A.
REFERESCES Brown J. L.. Stromeyer C. F. and Dawson B. M. (1971) Personal communication. Campbell F. W.. Gilinsky A. S.. Howell E. R_ Riggs L. A. and Atkinson J. (39733The dependence of monocular rivalry on orientation. Prrcrptiorl. 2, 173-125. Camnhell F. W. and Kulikowski J. J. (1966) Orientation
selectivity of the human visual system. J.
PhysioL
Lend.
187,437. Campbell F. W. and Maffei L. (1971) The tilt aftereffect: a fresh Iook. C’fsiottRes. 11, 833540. Cornsweet T. N. (1962) The staircase method in psychophysics. Ant. .f. Psyclrol. 75,185.
Coltheart M. (1973)Colour-specificity and monocularity in the visual cortex. CisiortRex 13, 259.5. Hajos A. and Ritter >l. (1965) Experiments to the problem of interocular transfer. Acra psyhol. 24,81-90. Hay J. C.. Pick H. L. and Rosser E. (1963) Adaptation to chromatic aberration by the human visua1 system. Scierzce. ‘V.Y.14t, 167. Held R. and Shattuck S. R. (197 I) Color and edge sensitive channels in the human visual system: tuning for orientation. Scirrtcr. N.Y. 174, 314. Julesz B. (1971) Fourrdatiou of C_wlopum Prrception. University of Chicago Press, Chicago. MacKay D. M. and liacKay V. (1973) Orientation sensitive aftereffects of dichoptically presented colour and form. Naturr.
Lorrd. 242, 477.
McCollough C. (1965) Color adaptation ofedge detectors in the human visual system. Scirrm, N.Y. 149, I I IS. Rauschecker J. P.. Campbell F. W. and Atkinson J. (1973) Color-opponent neurons in the human visual system. Natnrr. Land. 245, 42. Regan D. (L973) Evoked potentials specific to spatial patterns of luminance and color. iisiotr Rrs. 13.2381. Stromeyer C. F. (197’) Edge-contingent color aftereffects: spatial frequency specificity. Msiorl Rcs. 12, 7 17.
SITFANIE SHAITJCK
RICHARDHELD~ ‘Author addressed.
to whom requests for reprints should be
TO THE
AND EDGE SENSITIVE STEREO-DEPTH
Held and Shattuck (1971) demonstrated that a tilt aftereffect (AE) can be made contingent upon the color of a test grating. The adapting procedure involved alternating 5-set presentations of a red grating tilted clockwise off vertical by 15’ with a green grating tilted counterclockwise off vertical by 15‘. After the observers had scanned these gratings for IO min. a vertical green test grating appeared tilted slightly clockwise. and a vertical red grating appeared tilted slightly counterclockwise. These aftereffects. of opposite direction contingent upon color. demonstrate adaptation specific to each edge-color pair and imply a close linkage between channels selective for edge orientation and those selectiv-e for color in the visual system. Similar implications have been drawn from studies of the McCollough aftereffect (McCollough, 1969; Stromeyer. 1972). from work on rivalry of colored gratings (Campbell, Gilinsky, Howell, Riggs and Atkinson. 1973; Rauschecker. Campbell and Atkinson. 1973). and from the evoked potential studies of Regan (I 973). One would like to know at what level in the visual system this linkage ofcolor and edge-orientation information occurs. The type and degree of interaction between the two eyes can provide a clue. For example. substantial interocular transfer of an AE is usually interpreted as evidence for a locus of processing at or beyond the level of combination of information from the two eyes. The orientation-contingent color AE. demonstrated by McCollough. is reported to transfer either not at all or to a very slight degree (McCollough. 1965; Coltheart, 1973). That is. if one eye alone is exposed to the adapting gratings, the other eye shows either no demonstrable contingent AE. or a very weak one. The color fringe AE produced at luminance gradients. such as are induced by a prism, are reported not to transfer at all (Hajos and Ritter, 1965; Hay, Pick and Rosser, 1963). Dichoptic presentation. with color presented to one eye and edge orientation to the other. is reported to yield an AE less than halfas strong as that ’ This ivork was first presentsd at the Association for Research in Vision and Ophthalmology meeting in 1973 (program abstract p. 35. No. 2). The research was supported National of Health. Institutes Grant NO. by I ROI EYOI 191-01, and National Aeronautics and Space Administration. Grant No. NGL X-009-308.
EDITORS
CHANNELS CONVERGE ANALYZERS’
ON
for monoptic presentation. with both in the same eve (MacKay and XlacKay. 1973). Moreover. this result-is interpreted as -. taking place before the stage at which form and color are associated”. which is compatible with the notion that color-edge channels originating in separate eyes are independent. Taken together, these results indicate at most a vveak central interaction between coloredge information from the separate eyes at the level where this color-contingent tilt AE is generated. Monocular color-specific orientation channels may. however, converge at SOW level on binocular channels. Julesz (1971) suggested such a possibility. but was unable to demonstrate it with his stereograms. Using the color-contingent tilt aftereffect we have investigated. in the present experiments. the interaction between information reaching the separate eyes by testing for: (1) interocular transfer; (2) simultaneous induction of opposing AEs in the two eves: and (3) change in stereoscopic depth accompancing (2) and consequent on convergence of monocular channels on binocular analyzers of stereoscopic depth.
Each eye of the observer was tested separately, before and after a 20-min exposure of only one eye to two square wave gratings of 3.0 cycles per degree, alternated once every 5 sec. The red exposure was tilted off vertical by 15’ in one direction and the green tilted off vertical in the other. The observer was instructed to continuallv shift his fixation within the central region of the grating. The test gratings were circular, subtended 4>, and were divided mto upper and lower halves composed of gratings of differing color and equal but opposite tilt with respect to vertical. The gratings were either red above. green below or green above. red Ixlow. The tilts ofeach half varied in steps of IO’of arc from 60’ clockwise to 60’ counterclockwise off vertical. The combined tilts of the top and bottom halves of the test fields formed a series of angles slightly deviating from ISO’. These were presented to the observer in a double staircase procedure (Cornsweet. 1961) in which “red above. green below” test gratings were always alternated with ‘-green above. red below”. in order to equalize color adaptation during testing. Two projectors were used to
309
310
Letter 10 the Edttors
present the test gratings. The dark bars were maintained at a luminance of 2.5 ft-L. the light bars at 63 h-L for both test and exposure gratings. The size of the color-contingent tilt AE was measured as the change. from before to after exposure. in the angle (between bars of upper and lower test half fields) which was perceived by the observer as 1YO’ or straight. On each trial the observer either increased or decreased by one step the angle formed by the gratings in accord with his judgment of their deviation from apparent straightness. The mean shift for I2 observers was X4 in the eye exposed to the adapting stimuli. and less than 1’ in the unexposed eye. indicating no appreciable interocular transfer of the AE. One of the observers was run five times and one four times. In both cases. the mean shift in the exposed eye was significantly different {at the O-05level by t-test) from that in the unexposed eye. and the latter did not differ signifrcantiy from zero. This result stands in sharp contrast to the high degree of interocular transfer (approaching 100 per cent) for tilt induced by adaptation to edges defined by luminance gradients alone (Campbell and Maffei, 197I).
Ifcolor-contingent tilt AEs of opposite direction can be established simultaneously in the two eyes. we can ask two further questions: (a) what are the magnitudes ofthese AEs relative to those generated by exposure of only one eye to adapting stimuli” A sign&cant difference would suggest interocuiar interaction: (b) different frontal tilts of edges in corresponding retinal loci in the separate eyes are equivalent to a binocular disparity. Do the tilts produced as AEs (see Fig. I) act as if they altered binocular disparity; i.e. are they convertcd into a change in stereoscopic depth’! If so. we may conclude that information from monocuIar coioredge orientation channels converges onto binocular channels at the level of stereoscopic depth processing. A dove prism was set in front of one eye to reverse (left to right) the tilts of both the adapting and test gratings presented to that eye. Each eye was tested separately, before and after Xmin of adaptation. as described above. The mean shifts in the angle of subjective straightness were 31’ in the right eye and 37’ in the opposite direction in the left eye, for eight subjects: an effect of larger magnitude. but not significantly so. than that obtained with adaptation ofa single eye. This result supports the hypothesis that color-edge orientation information from the two eyes is processed independently.
A slightly modified experimental procedure allowed us to decide whether the induction of color-contingent tilt AEs. opposite in the two eyes. leads to a measureable change in judgements of stereoscopic depth. When both eyes simultaneously view a tilted test stimulus with reversing prism in place. the two eyes view opposite tilts. As a result. the upper and lower
Fig. I. Stereo depth
induction
(bars on left deleted).
planes of the test gratings appear to be tilted in depth, forming a dihedral angle pointing awa>: from or toward the observer (Fig. I ). Under these Circumstances, the sequential presentation of test stimuti produces variation in tilt disparity in steps of 20’ (IO’ per eye). We used the double-statrcase procedure to measure the change in frontal target angles required to make the target appear subjectively flat (dihedral angle of ISO’) after opposite adaptation of the two eyes. That change measured the amount of conversion of the frontal tilt AE to stereoscopic depth. Since shifts in frontal tilt of settings to subjective straightness were also measured for each eye separately. the shift in stereoscopic depth found in the binocular test can be expressed as a percentage of the average shift in frontal tilt for the two eyes. This is seen in Fig. 2. where the line at 45’ represents 100 per cent conversion from the two opposite monocular frontal tilt shifts to the predicted shift in stereoscopic depth. Six observers were run once monocularly and once binocularly in separate sessions (open circles). One observer (RH) was tested in eight sessions, four on frontal tilt and four on depth ‘%lt constituting four combined but uncorre~ated runs (square shows mean. bars show range). and one observer (SS) was tested in four sessions. each time both monocularly and binocularly during the same session (filled circles).
1 ,
0
v
0
10
20
30
40
.W (min./eye)
Monocular tilt aftereffect Fig. 2. Shift in perceived stereo depth as a function in perceived frontal tilt after adaptation to opposite tilts in the two eyes.
of shift frontai
311
Letter to the Editors
Clearly. there is substantial conversion to stereo depth: all observers show a positive change although its magnitude varies considerably. It is unlikely that this is the result of direct a~p~tion of binocular stereoscopic depth channels, since the disparities of the adapting tilts are 30” and the stimuli are rivalrous during the adapting exposure. Moreover, we have found substantial stereo depth shifts even after the separate eyes were adapted sequentially, rather than simuitaneously, to opposite tilts; in this case. no direct adaptation of binocular channels could occur. SU>l;\tARY The results confirm the conclusion, drawn from studies of the orientation-contingent color AE (McCollough effect), that information about combined color and edge orientation is initially processed separately in monocular channels. It does. however, converge on binocular channels, at least at the level of processing for stereoscopic depth. Other investigators have been unable to demonstrate a depth-specific color AE after binocular adaptation to red and green gratings producing different stereoscopic depths which presumably should adapt coior selective binocular channels if they exist (Brown, Stromeyer and Dawson. 1974).These findings suggest that the binocular channels themselvesare not selective for color, even though they receive information from monocular channels which are color selective. Taken together. present knowledge drawn from psychophysical research implies that the linkage between cotor and edge orientation processors occurs prior to and not in binocular channels. Massachusetts Institute of Technology, Department of Psychology, 79 Amherst Stree, Caprice, ~as~&hu~tts 02139, U.S.A.
REFERESCES Brown J. L.. Stromeyer C. F. and Dawson B. M. (1971) Personal communication. Campbell F. W.. Gilinsky A. S.. Howell E. R_ Riggs L. A. and Atkinson J. (39733The dependence of monocular rivalry on orientation. Prrcrptiorl. 2, 173-125. Camnhell F. W. and Kulikowski J. J. (1966) Orientation
selectivity of the human visual system. J.
PhysioL
Lend.
187,437. Campbell F. W. and Maffei L. (1971) The tilt aftereffect: a fresh Iook. C’fsiottRes. 11, 833540. Cornsweet T. N. (1962) The staircase method in psychophysics. Ant. .f. Psyclrol. 75,185.
Coltheart M. (1973)Colour-specificity and monocularity in the visual cortex. CisiortRex 13, 259.5. Hajos A. and Ritter >l. (1965) Experiments to the problem of interocular transfer. Acra psyhol. 24,81-90. Hay J. C.. Pick H. L. and Rosser E. (1963) Adaptation to chromatic aberration by the human visua1 system. Scierzce. ‘V.Y.14t, 167. Held R. and Shattuck S. R. (197 I) Color and edge sensitive channels in the human visual system: tuning for orientation. Scirrtcr. N.Y. 174, 314. Julesz B. (1971) Fourrdatiou of C_wlopum Prrception. University of Chicago Press, Chicago. MacKay D. M. and liacKay V. (1973) Orientation sensitive aftereffects of dichoptically presented colour and form. Naturr.
Lorrd. 242, 477.
McCollough C. (1965) Color adaptation ofedge detectors in the human visual system. Scirrm, N.Y. 149, I I IS. Rauschecker J. P.. Campbell F. W. and Atkinson J. (1973) Color-opponent neurons in the human visual system. Natnrr. Land. 245, 42. Regan D. (L973) Evoked potentials specific to spatial patterns of luminance and color. iisiotr Rrs. 13.2381. Stromeyer C. F. (197’) Edge-contingent color aftereffects: spatial frequency specificity. Msiorl Rcs. 12, 7 17.
SITFANIE SHAITJCK
RICHARDHELD~ ‘Author addressed.
to whom requests for reprints should be
