low spatial-frequency channels in human vision: adaptation and ...
0042-6989~82.020225-09103.0010 Perpamon Press Ltd
c ,\,ont?<wwr~h Vol. 22.pp 225 10 233. 1982
Printed I” GreatBr~tiun
LOW SPATIAL-FREQUENCY CHANNELS IN HUMAN VISION: ADAPTATION AND MASKING C.
F.
STROMEYER
III.1,2
S.
KLEIN, 3.4 B. M. DAWSON~ and L. SPILLMANN~
‘Division of Applied Sciences, Pierce Hall. Harvard University, Cambridge, MA 02138, U.S.A. ‘Department of Neurology and Neurophysiology, University of Freiburg, 78 Freiburg im Br., West Germany. JDivision of Biology, California Institute of Technology, Pasadena, CA 91125, and 4Joint Science Department, Claremont College, Claremont, CA 91711. ‘Department of Psychology, Massachusetts Institute of Technology. Cambridge, MA 02139. U.S.A. (Recrit~d
18 Nocrmber
1980; in rrcisrdform
27 July 1981)
Abstract-Previous work showed that adapting to low spatial frequency gratings (below 1.5 cycles/ degree) may cause maximal spatial adaptation at a significantly higher spatial frequency, It has been suggested that there are no adaptable spatial-frequency channels tuned to below 1.5 c/deg. Contrary to this view, we found that adaptation and masking with low spatial frequencies (0.12-1.0 c/deg) produced maximal threshold elevations when the test patterns were the same spatial frequency as the adapting or masking pattern, These results were obtained using test patterns that turned on and off gradually or sharply. The results suggest that there are form mechanisms optimally sensitive to very low spatial frequencies. Adaptation was selective to position (phase) and orientation at low spatial frequencies; masking was observed to be selective to orientation at a spatial frequency as low as 0.2 c/deg. A clear dichotomy between transient, motion channels and sustained. form channels at low spatial and temporal frequencies may represent an unrealistic simplification. There may exist directionally-selective motion mechanisms sensitive to very slow motion, and these may play a role in the discrimination of form. The discussion considers the bandwidths of the low spatial frequency mechanisms.
INTRODUCTION Visual form information may be processed by many independent channels each responding maximally to a limited band of spatial frequencies (Campbell and Robson, 1968). Adaptation studies provide evidence for spatial-frequency channels. Adapting to a highcontrast sinusoidal grating selectively raises the threshold for detecting gratings of similar spatial frequencies (Pantle and Sekuler, 1968; Blakemore and Campbell, 1969) and orientations (Gilinsky. 1968). The full bandwidth of adaptation at different spatial frequencies is about 1.5 octaves (Blakemore and Campbell, 1969). Comparable bandwidths have been measuring for masking, using mask patterns that were dynamic noise gratings (Stromeyer and Julesz, 1972) or single sinusoidal gratings (Legge, 1978). At low spatial frequencies the maximum threshold elevation produced by an adapting grating may be displaced upward from the adapting spatial frequency (Blakemore and Campbell, 1969). Tolhurst (1973) measured such a displacement with stationary rest gratings which were presumably detected by shape or form analyzers. With mocing test patterns, the threshold elevation was always maximal when the adapting and test spatial frequencies matched. The displacement of the peak observed with stationary test patterns does not imply that there are no shape analyzing mechanisms optimally sensitive to low spatial frequencies, but rather that these mechanisms may nor be reudily adqred. If there were no such mechanisms. then the threshold elevation at the bH
222
H
225
adapting frequency should remain high, for adapted mechanisms tuned to a higher spatial frequency would be used to detect the low spatial frequency patterns, And, thus, the adaptation function would have a low-pass spatial frequency characteristic. Only if there were low-frequency mechanisms that do not adapt would there be little threshold elevation at the adapting frequency. Our experiments further examined whether there is an adaptable lowest spatialfrequency channel. Legge (1978), using a masking paradigm, obtained evidence for a lowest spatial frequency, sustained mechanism. A test grating of 0.38 c/deg and 100 msec (10deg field) was masked with a sinusoidal grating that was presented briefly at the onset and offset of the test grating so as to selectively mask transient mechanisms. with the consequence that the test grating would be detected by sustained mechanisms. The test grating was not maximally masked by a grating of the same spatial frequency (presented spatially inphase, personal communication). but rather by mask gratings of higher spatial frequencies, 0.75 and 1.5 c/deg. Legge (1978) concluded that the lowest spatial frequency sustained mechanism peaks at 1.5 cjdeg. Legge (1979) also measured masking with sinusoidal mask and test gratings presented simultaneously for 200 msec (13 by 20 deg field). With monocular presentation, a 0.12 c/deg test grating was maximally masked by a 0.25 c/deg mask grating, and a 0.25 c/deg test grating was maximally masked by a 1 c/deg grating. Our experiments also examined the spatial-frequency selectivity of masking at low spatial fre-
qucncies. using a dynamic noise masking grating and test patterns that turned nn and off gradually. In contrast to these previous studies. \\t‘ observed that adaptatmn and masking were maximal when the spatial frcqucnc) of the test pattern matched that of the adapting or maskmg pattern The results also show ~)rlcntation-selectt~ity and position or phasc\clcctivity at Ioh spatial frequencies.
METHODS
Sinusoidal gratings w’ere presented on a Tektronix 602 oscilloscope with a P-4. white phosphor. The circular stimulus field had a mean luminance of 8 cd,m’ (unless otherwise stated) and dark surround. A dark fixation point or cross-hair was placed in the center of the field. The adaptmg patterns were drifting gratings or a randomly positioned counterphase grating that was designed to homogeneously adapt the retina and avoid afterimages which might occur. especially w,ith low spatial frequencies. The latter adapting pattern was a series of frames of a sinusoidal grating that was constant in contrast. orientation. and spatial frcquencq. To approximate random positioning. the grating was presented in tivr possible phase positions (separated by 36 deg of phase) and their respective antiphase positions. Each position was chosen randomly. and the grating was presented in this position for a random number of frames. II = 1. 2 or 3. Immedtately folloumg this. the grating \ras presented in the antiphase positlon for an equl\alent II frame. Each frame lasted 19.5 msec. with II I.0 msec inter-frame interbat. Each adapting period lasted 5 see and consisted of the same sequence of frames. For the last several frames. II decreased 3. 7. I so as to dampen out an) afterimage. .A computer triggered the \--axis sweep of the oscilloscope and then. after a precise delay. gated on a Wa\etek 112 oscillator that sinusoidally modulated the :-axis. The delay determined the phase of the adapting grating. The adapting pattern appeared to dance about with no regular motion, The appropriate positioning of the adapting grating was confirmed by observing one small spot on the oscilloscope with a photomultiplicr tube. Dynamic masking gratmgs \\cre produced by passmg white nolsc through liltcrs (with cutoffs of approximately 30dB octa\e) and feeding the signal to the :-axis (Legge VI i/l.. 1978). The noise bandwidth was specified by the 3-dB attenuation points of the tiltcrs. The noise contrast is cil, Lo. where L,, is the mean field lummancc and (T,,is the standard deviation of the GaussIan noise (Stromeycr and Julesr. 1972). The \--axis sweep was NO H7
A smgle adapting or masking pattern and test pattern was used for each run of 100 trials. The observer monocularly fixated the central mark and adapted
5 min to the stimulus sequence before data was collected. The patterns were vertical unless stated otherwise. For the adapting experimcnts~ a 5-set exposure to the adapting pattern (IO-XC for Fig. I) was alternated with a 3-set test period. ‘The test pattern had two or four contrast values. including zero contrast. which occurred about equally. according to ;I random sequence. The observer rated the visibility of each test pattel-n on a whole-number scale of 1 5. and was giben feedback about teht cuntrast. The detectablhcs. tl. of the patterns were calculated from the ratings (Stromeyer (‘r trl.. 1977). Runs w’er’ctypically combined when the same conditions were used (Stromeyer 6’1trl.. 1977). The figures display log d’ LS log test contrast. Straight lines were fit to the data, using a least squares procedure in which each datum is weighted by the invcrsc of its variance. The lines arc constrained to be parallel within each panel of the figures (Klein and Stromeyer. 1980). The lines are represented by the function tl’ = /CC”. where c‘ is the contrast of the test grating. The axes in the figures are scaled so that lines of 45 deg slope correspond to n = 2. The data are approximately fitted by lines of 45 deg slope for both the adaptation and masking experiments. The effect of adaptation or masking is shown by the Iatcral displacement (along the contrast axis) of the functions obtained with and without the masking or adaptation pattern. The relative threshold elevation (RTE) is defined as the ratio of test contrasts for II given tl’ value with the adaptation or mask gratmg present versus absent. minus one (Blakemore and Campbell. 1969). A RTE of 1.0. for example. means that the mask or adapting pattern raises the threshold b\ IOO”,,.
Adapting and masking experiments were performed to determine if the peak threshold elevation occurs at the adapting or masking spatial frequency. The test pattern had a symmetrical triangular temporal waveform and turned on linearly for 1.5 set and turned off linearly for the next I.5 sec. The gradual onset and offset was used to attempt to selectively stimulate sustained mechanisms. T/Iiototrr. Vol. Vfil. Pe~~[,~ti~~~ (Edited bv Held R.. Leibowirz H. “W. and Teuber’ H.-L.). Springer-Verlag. Berlin. Sekuler R. W.. Rubin E. L. and Cushman W. H. (1968) Seiectivities of human visual mechanisms for direction of movement and contour orientation. J. opt. Sot,. Am. 58, 1146-1150. Sharpe C. R. and Tolhurst D. J. (1973) The effects of temporal modulation on the orientation channels of the human visual system. Pe~~epfi~~ 2, 23-29. Stromeyer C. F. 111 and Dawson B. M. (1978) Form-colour aftereffects: selectivity to local luminance contrast. Pcrception 7, 407-415. Stromeyer C. F. III and Julesz B. (1972) Spatial-frequency masking in vision: critical bands and spread of masking. 1. opt. Sot. Am. 62, 1221&1232. Stromeyer C. F. III and Klein S. (1974) Spatial frequency channeis in human vision as asymmetric (edge) mechanisms Vision Res. 14, 1409-1420. Stromeyer C. F. III. Klein S. and Sternheim C. E. (1977) Is spatial adaptation caused by prolonged inhibttion ‘I Vision Res. 17, 603-606. Stromeyer C. F. III. Madsen J. C. and Klein S. (1979) Direction-selective adaptation with very slow motion. J. opt. Sot. Am. 69, 1039-1041. Tolhurst D. J. (1973) Separate channels for the analysts of the shape and the movement of a moving visual stimulus. J. Pk~iol.. Land. 231, 385402. Tolhurst D. J. (1975) Sustained and transient channels in human vision. Vision Res. 15. 1151-l 155. Watson A. B. and Nachmias J. (1977) Patterns of temporal interaction in the detection of gratings. Vkion Res. 17, 893-902. Watson .4. B. and Robson J. G. (1981) Discrimination at threshold: Labelled defectors in human vision. Visioit Rex 21, 1115-1122.
c ,\,ont?<wwr~h Vol. 22.pp 225 10 233. 1982
Printed I” GreatBr~tiun
LOW SPATIAL-FREQUENCY CHANNELS IN HUMAN VISION: ADAPTATION AND MASKING C.
F.
STROMEYER
III.1,2
S.
KLEIN, 3.4 B. M. DAWSON~ and L. SPILLMANN~
‘Division of Applied Sciences, Pierce Hall. Harvard University, Cambridge, MA 02138, U.S.A. ‘Department of Neurology and Neurophysiology, University of Freiburg, 78 Freiburg im Br., West Germany. JDivision of Biology, California Institute of Technology, Pasadena, CA 91125, and 4Joint Science Department, Claremont College, Claremont, CA 91711. ‘Department of Psychology, Massachusetts Institute of Technology. Cambridge, MA 02139. U.S.A. (Recrit~d
18 Nocrmber
1980; in rrcisrdform
27 July 1981)
Abstract-Previous work showed that adapting to low spatial frequency gratings (below 1.5 cycles/ degree) may cause maximal spatial adaptation at a significantly higher spatial frequency, It has been suggested that there are no adaptable spatial-frequency channels tuned to below 1.5 c/deg. Contrary to this view, we found that adaptation and masking with low spatial frequencies (0.12-1.0 c/deg) produced maximal threshold elevations when the test patterns were the same spatial frequency as the adapting or masking pattern, These results were obtained using test patterns that turned on and off gradually or sharply. The results suggest that there are form mechanisms optimally sensitive to very low spatial frequencies. Adaptation was selective to position (phase) and orientation at low spatial frequencies; masking was observed to be selective to orientation at a spatial frequency as low as 0.2 c/deg. A clear dichotomy between transient, motion channels and sustained. form channels at low spatial and temporal frequencies may represent an unrealistic simplification. There may exist directionally-selective motion mechanisms sensitive to very slow motion, and these may play a role in the discrimination of form. The discussion considers the bandwidths of the low spatial frequency mechanisms.
INTRODUCTION Visual form information may be processed by many independent channels each responding maximally to a limited band of spatial frequencies (Campbell and Robson, 1968). Adaptation studies provide evidence for spatial-frequency channels. Adapting to a highcontrast sinusoidal grating selectively raises the threshold for detecting gratings of similar spatial frequencies (Pantle and Sekuler, 1968; Blakemore and Campbell, 1969) and orientations (Gilinsky. 1968). The full bandwidth of adaptation at different spatial frequencies is about 1.5 octaves (Blakemore and Campbell, 1969). Comparable bandwidths have been measuring for masking, using mask patterns that were dynamic noise gratings (Stromeyer and Julesz, 1972) or single sinusoidal gratings (Legge, 1978). At low spatial frequencies the maximum threshold elevation produced by an adapting grating may be displaced upward from the adapting spatial frequency (Blakemore and Campbell, 1969). Tolhurst (1973) measured such a displacement with stationary rest gratings which were presumably detected by shape or form analyzers. With mocing test patterns, the threshold elevation was always maximal when the adapting and test spatial frequencies matched. The displacement of the peak observed with stationary test patterns does not imply that there are no shape analyzing mechanisms optimally sensitive to low spatial frequencies, but rather that these mechanisms may nor be reudily adqred. If there were no such mechanisms. then the threshold elevation at the bH
222
H
225
adapting frequency should remain high, for adapted mechanisms tuned to a higher spatial frequency would be used to detect the low spatial frequency patterns, And, thus, the adaptation function would have a low-pass spatial frequency characteristic. Only if there were low-frequency mechanisms that do not adapt would there be little threshold elevation at the adapting frequency. Our experiments further examined whether there is an adaptable lowest spatialfrequency channel. Legge (1978), using a masking paradigm, obtained evidence for a lowest spatial frequency, sustained mechanism. A test grating of 0.38 c/deg and 100 msec (10deg field) was masked with a sinusoidal grating that was presented briefly at the onset and offset of the test grating so as to selectively mask transient mechanisms. with the consequence that the test grating would be detected by sustained mechanisms. The test grating was not maximally masked by a grating of the same spatial frequency (presented spatially inphase, personal communication). but rather by mask gratings of higher spatial frequencies, 0.75 and 1.5 c/deg. Legge (1978) concluded that the lowest spatial frequency sustained mechanism peaks at 1.5 cjdeg. Legge (1979) also measured masking with sinusoidal mask and test gratings presented simultaneously for 200 msec (13 by 20 deg field). With monocular presentation, a 0.12 c/deg test grating was maximally masked by a 0.25 c/deg mask grating, and a 0.25 c/deg test grating was maximally masked by a 1 c/deg grating. Our experiments also examined the spatial-frequency selectivity of masking at low spatial fre-
qucncies. using a dynamic noise masking grating and test patterns that turned nn and off gradually. In contrast to these previous studies. \\t‘ observed that adaptatmn and masking were maximal when the spatial frcqucnc) of the test pattern matched that of the adapting or maskmg pattern The results also show ~)rlcntation-selectt~ity and position or phasc\clcctivity at Ioh spatial frequencies.
METHODS
Sinusoidal gratings w’ere presented on a Tektronix 602 oscilloscope with a P-4. white phosphor. The circular stimulus field had a mean luminance of 8 cd,m’ (unless otherwise stated) and dark surround. A dark fixation point or cross-hair was placed in the center of the field. The adaptmg patterns were drifting gratings or a randomly positioned counterphase grating that was designed to homogeneously adapt the retina and avoid afterimages which might occur. especially w,ith low spatial frequencies. The latter adapting pattern was a series of frames of a sinusoidal grating that was constant in contrast. orientation. and spatial frcquencq. To approximate random positioning. the grating was presented in tivr possible phase positions (separated by 36 deg of phase) and their respective antiphase positions. Each position was chosen randomly. and the grating was presented in this position for a random number of frames. II = 1. 2 or 3. Immedtately folloumg this. the grating \ras presented in the antiphase positlon for an equl\alent II frame. Each frame lasted 19.5 msec. with II I.0 msec inter-frame interbat. Each adapting period lasted 5 see and consisted of the same sequence of frames. For the last several frames. II decreased 3. 7. I so as to dampen out an) afterimage. .A computer triggered the \--axis sweep of the oscilloscope and then. after a precise delay. gated on a Wa\etek 112 oscillator that sinusoidally modulated the :-axis. The delay determined the phase of the adapting grating. The adapting pattern appeared to dance about with no regular motion, The appropriate positioning of the adapting grating was confirmed by observing one small spot on the oscilloscope with a photomultiplicr tube. Dynamic masking gratmgs \\cre produced by passmg white nolsc through liltcrs (with cutoffs of approximately 30dB octa\e) and feeding the signal to the :-axis (Legge VI i/l.. 1978). The noise bandwidth was specified by the 3-dB attenuation points of the tiltcrs. The noise contrast is cil, Lo. where L,, is the mean field lummancc and (T,,is the standard deviation of the GaussIan noise (Stromeycr and Julesr. 1972). The \--axis sweep was NO H7
A smgle adapting or masking pattern and test pattern was used for each run of 100 trials. The observer monocularly fixated the central mark and adapted
5 min to the stimulus sequence before data was collected. The patterns were vertical unless stated otherwise. For the adapting experimcnts~ a 5-set exposure to the adapting pattern (IO-XC for Fig. I) was alternated with a 3-set test period. ‘The test pattern had two or four contrast values. including zero contrast. which occurred about equally. according to ;I random sequence. The observer rated the visibility of each test pattel-n on a whole-number scale of 1 5. and was giben feedback about teht cuntrast. The detectablhcs. tl. of the patterns were calculated from the ratings (Stromeyer (‘r trl.. 1977). Runs w’er’ctypically combined when the same conditions were used (Stromeyer 6’1trl.. 1977). The figures display log d’ LS log test contrast. Straight lines were fit to the data, using a least squares procedure in which each datum is weighted by the invcrsc of its variance. The lines arc constrained to be parallel within each panel of the figures (Klein and Stromeyer. 1980). The lines are represented by the function tl’ = /CC”. where c‘ is the contrast of the test grating. The axes in the figures are scaled so that lines of 45 deg slope correspond to n = 2. The data are approximately fitted by lines of 45 deg slope for both the adaptation and masking experiments. The effect of adaptation or masking is shown by the Iatcral displacement (along the contrast axis) of the functions obtained with and without the masking or adaptation pattern. The relative threshold elevation (RTE) is defined as the ratio of test contrasts for II given tl’ value with the adaptation or mask gratmg present versus absent. minus one (Blakemore and Campbell. 1969). A RTE of 1.0. for example. means that the mask or adapting pattern raises the threshold b\ IOO”,,.
Adapting and masking experiments were performed to determine if the peak threshold elevation occurs at the adapting or masking spatial frequency. The test pattern had a symmetrical triangular temporal waveform and turned on linearly for 1.5 set and turned off linearly for the next I.5 sec. The gradual onset and offset was used to attempt to selectively stimulate sustained mechanisms. T/Iiototrr. Vol. Vfil. Pe~~[,~ti~~~ (Edited bv Held R.. Leibowirz H. “W. and Teuber’ H.-L.). Springer-Verlag. Berlin. Sekuler R. W.. Rubin E. L. and Cushman W. H. (1968) Seiectivities of human visual mechanisms for direction of movement and contour orientation. J. opt. Sot,. Am. 58, 1146-1150. Sharpe C. R. and Tolhurst D. J. (1973) The effects of temporal modulation on the orientation channels of the human visual system. Pe~~epfi~~ 2, 23-29. Stromeyer C. F. 111 and Dawson B. M. (1978) Form-colour aftereffects: selectivity to local luminance contrast. Pcrception 7, 407-415. Stromeyer C. F. III and Julesz B. (1972) Spatial-frequency masking in vision: critical bands and spread of masking. 1. opt. Sot. Am. 62, 1221&1232. Stromeyer C. F. III and Klein S. (1974) Spatial frequency channeis in human vision as asymmetric (edge) mechanisms Vision Res. 14, 1409-1420. Stromeyer C. F. III. Klein S. and Sternheim C. E. (1977) Is spatial adaptation caused by prolonged inhibttion ‘I Vision Res. 17, 603-606. Stromeyer C. F. III. Madsen J. C. and Klein S. (1979) Direction-selective adaptation with very slow motion. J. opt. Sot. Am. 69, 1039-1041. Tolhurst D. J. (1973) Separate channels for the analysts of the shape and the movement of a moving visual stimulus. J. Pk~iol.. Land. 231, 385402. Tolhurst D. J. (1975) Sustained and transient channels in human vision. Vision Res. 15. 1151-l 155. Watson A. B. and Nachmias J. (1977) Patterns of temporal interaction in the detection of gratings. Vkion Res. 17, 893-902. Watson .4. B. and Robson J. G. (1981) Discrimination at threshold: Labelled defectors in human vision. Visioit Rex 21, 1115-1122.
