- Email: [email protected]
Multiple gain control processes in contrast–contrast phenomena
Vision Research 39 (1999) 3983 – 3987 www.elsevier.com/locate/visres
Rapid communication
Multiple gain control processes in contrast–contrast phenomena Lynn A. Olzak a,*, Pentti I. Laurinen b a
Department of Psychology, Miami Uni6ersity of Ohio, Oxford, OH 45056, USA b Department of Psychology, Uni6ersity of Helsinki, Helsinki, Finland Received 31 December 1998; received in revised form 24 May 1999
Abstract Spatial interactions among orientation-tuned gain control processes are presumed to mediate center-surround contrast – contrast phenomena. In this paper, we assess contributions of gain control processes that pool over orientation. We measured the apparent contrast of a luminance-modulated center disk embedded in various modulated surrounds. In all conditions, observers compared the apparent contrast of the test center to an identically modulated disk with no surround. When center and surround are simple, vertical sinusoids and presented in phase, suppression depends upon surround contrast and is marked at high contrasts. When components are presented 180° out of phase, no suppression occurs at any contrast. When a horizontal component is added to the surround, much less suppression occurs. However, strong suppression is reinstated when both center and surround are plaids. Neither of the latter two effects are phase dependent. We suggest that two different sources of gain control are revealed by the simple sinusoidal and the plaid stimuli. One is orientation tuned and phase-dependent. The other pools over all orientations and includes neurons tuned to multiple phases. © 1999 Elsevier Science Ltd. All rights reserved. Keywords: Control processes; Spatial interactions; Contrast–contrast phenomena
1. Introduction Ohzawa, Sclar and Freeman (1982) were the first to report that the majority of neurons in cat visual cortex adjust their response levels to prevailing contrast, similar to adjustments made in retinal cells to prevailing illumination. They pointed out that one obvious benefit of contrast gain control is to help the system maintain a high degree of sensitivity for contrast differences over wide ranges of contrasts, analogous to retinal control of brightness sensitivity. They also suggested that early normalization processes could provide contrast-independent input to a number of higher-level mechanisms. Gain control processes operating laterally over space have been widely suggested as the mechanism underlying contrast–contrast phenomena, a term coined by
* Corresponding author. Tel.: +1-513-529-1754; fax: + 1-513-5292420. E-mail address: [email protected] (L.A. Olzak)
Chubb, Sperling and Solomon (1989). In this phenomenon, the apparent contrast of a luminance or chromatically modulated pattern is reduced by the addition of an adjacent or surrounding pattern. A number of authors have proposed models of luminance or chromatic contrast induction that modify activity by lateral inhibition within a pool of neurons (e.g. Cannon & Fullenkamp, 1991a,b; Zaidi, Yoshimi, Flanigan & Canova, 1992; Solomon, Sperling & Chubb, 1993; Singer & D’Zmura, 1994; D’Zmura & Singer, 1996; De Bonet & Zaidi, 1997). However, it is not clear from existing studies whether the pool includes only neurons with similar tuning functions, or whether the pool is broadband with respect to spatial frequency, orientation, and phase (see Chubb et al., 1989; Cannon & Fullenkamp, 1991a; Solomon et al., 1993). Furthermore, it is not clear whether the same lateral processes underlie masking effects observed in detection tasks (Foley, 1994), contrast discrimination (Olzak & Thomas, 1992) and contrast appearance, as has been suggested by some (e.g. Snowden & Hammett, 1998).
0042-6989/99/$ - see front matter © 1999 Elsevier Science Ltd. All rights reserved. PII: S 0 0 4 2 - 6 9 8 9 ( 9 9 ) 0 0 1 3 1 - 5
3984
L.A. Olzak, P.I. Laurinen / Vision Research 39 (1999) 3983–3987
Of particular relevance to the present paper are conflicting results with respect to phase or polarity. Ejima and Takahashi (1985) reported that with vertically abutting inducing and test gratings (vertical sinusoids), no suppressive effects were found when the two patches were out of phase and therefore differed in polarity along the line of collinearity. However, using center-surround textures composed of micro-elements, Solomon et al. (1993) reported that the relative polarity of the elements had no effect on the magnitude of the suppressive effect. No phase dependencies were reported by Zenger and Sagi (1996), who used a contrast detection task with spatially displaced but vertically aligned Gabor patches. These apparently conflicting results might be reconciled if more than a single gain control mechanism contributes to contrast – contrast. Cannon and Fullenkamp (1991a,b), for example, found hints of two processes underlying contrast-contrast effects in two different experiments. Evidence for multiple sources of gain control exists in other literatures, as well. Pertinent to the current paper is recent work in complexpattern discrimination (Olzak & Thomas, 1992, 1997; Thomas & Olzak, 1997; Olzak & Thomas, 1999). One source isolated in their work is highly tuned with respect to spatial frequency, orientation, and phase, and is presumed to occur within a low-level, tuned pathway. In a second type of process, at least two different types of higher-level mechanisms normalize over a broad but selective pool of neurons and then sum over more specific (and different) sets of neurons to signal particular types of information. One of these mechanisms pools over all orientations, but only within a limited frequency band. This mechanism provides information about the contrast and/or spatial grain of a textured surface (Olzak & Thomas, 1992, 1999) and therefore might play a role in contrast– contrast phenomena. A similar broadly-tuned pooled gain-control process has been proposed by Foley (1994) in his model of contrast detection. However, all tests of these higher-level pools have to date been made with spatially overlapping components. Thus, nothing is yet known about their spatial pooling abilities. In this paper, we test the hypothesis that multiple contrast gain control mechanisms exist and operate laterally over space. Specifically, we ask whether properties of gain control mechanisms revealed with simple sinusoidal grating patterns differ from those tapped with more complex stimuli. We further ask how higher-level processes revealed with plaid stimuli in contrast discrimination experiments might contribute to contrast–contrast effects.
2. Methods
2.1. Obser6ers Three observers participated in both experiments. Observer PL is an experienced observer with normal uncorrected vision. Observers TK and SK are relatively inexperienced observers, emmetropes, and naı¨ve to the purpose of the study.
2.2. Stimuli and apparatus Stimuli were created using NIH Image and displayed on a linearized Nokia Multigraph 445× color monitor using a Cambridge Research systems V2.2 graphics board in a PC 486 system. Viewing distance was 114 cm. Test and comparison disks were displayed simultaneously in a left–right configuration. The comparison disk consisted of a 1°, luminance-modulated disk of variable contrast viewed against a uniform background subtending 5.5° of visual angle. The test stimulus consisted of a 1° modulated disk embedded in the center of a 5.5° modulated surround, located adjacent to the comparison on the display. Comparison and test were thus each located at eccentricities of 2.25°. The spatial frequency of all modulations was 2.75 cpd. Contrast of the test center was held constant in all experiments and conditions at 0.18. Contrast of the surround modulation of the test stimulus was varied in different conditions around the mean luminance of 21 cd/m2, but was held constant in any one condition.
2.3. General procedures In all experiments and conditions, the apparent contrast of the test disk was measured in a 2-alternative forced choice method of constant stimuli procedure. Observers initiated each trial following presentation of a centered fixation point. Test and comparison stimuli were presented simultaneously for 1 s, with abrupt onset and offset. Left–right presentation of test and comparison stimuli was counterbalanced and randomly intermingled within a session. The contrast of the comparison disk was varied in all experiments in ten steps over a range that depended upon condition. The observer indicated by pressing one of two buttons whether the left or right disk appeared to have the greater contrast. Each contrast level was presented ten times for a given condition, for a total of 100 trials per condition. Two conditions were always randomly intermingled in a block of trials, for a total of 200 trials per session. Each observer participated in three replications of each condition. The frequency of left–right responses were converted to probabilities of responding that the test disk ap-
L.A. Olzak, P.I. Laurinen / Vision Research 39 (1999) 3983–3987
Fig. 1. Examples of stimuli. Top row: center and surround in phase. Bottom row: center and surround 180° out of phase. Left-hand column: simple combinations of sinusoids used in both experiments 1 and 2. Center: sinusoidal center with plaid surround used in experiment 2. Right-hand column: Plaid center and surrounds used in experiment 2.
peared to have more contrast than the comparison disk, as a function of contrast level. A cumulative normal was fit to these data using least squares criterion. The mean of this distribution was taken as the point of subjective equality (PSE), and its standard deviation served as a measure of error. Contrast suppression was calculated as (Ccenter −CPSE)/Ccenter. This measure ranges from 0 (no suppression) to 1.0.
3. Experiment 1 A preliminary experiment was performed to test whether we could replicate the Ejima and Takahashi (1985) results with sinusoidal, center-surround stimuli. We also wanted to be assured that that contrast suppression results reported by others in previous work could be demonstrated in our observers, as differences
3985
among individuals have been reported (Cannon & Fullenkamp, 1993). We also wanted to ensure that the phenomenon we were measuring was not a simple by-product of contrast reduction by the grating induction effect first reported by McCourt (1982). Induced gratings appear counterphase to the inducing grating, and would affect perceived contrast by reducing apparent contrast when center and surround are in phase, and enhancing it when they are out of phase. Accordingly, we performed a partial replication of Ejima and Takahashi (1985), using high contrast values and our own particular stimulus configuration and observers. The contrast of the modulated surround took one of eight values, ranging from + 0.39 to − 0.39, resulting in surrounds that were both in phase and 180° out of phase with the center. When stimuli were out of phase, an additional measurement was made with surround contrast equal to the test contrast of 0.18. Example of the in-phase and out-of-phase sinusoidal stimuli used in this experiment are shown, respectively in the upper and lower lefthand panels of Fig. 1.
3.1. Results of experiment 1 The results are plotted in Fig. 2 for two observers. The third observer’s data was highly similar to the results of PL. The results of all three observers replicate Ejima and Takahashi (1985) and others in that we find suppression only when the vertically modulated center and surround are in phase, and suppression increases with increasing surround contrast. However, we observed no effects of any sort when center and surround were out of phase, regardless of surround contrast. We therefore can rule out grating induction as the causal factor in our observed contrast–contrast effect. To-
Fig. 2. Reduction of perceived contrast of center grating as a function of surround grating contrast for observers TK (left panel) and PL (right panel). Normalized suppression is plotted on the ordinate (1.0 = maximum suppression). Surround contrast is plotted on the abscissa. Positive values to the right of the ordinate indicate center and surround were in phase. Negative values to the left of the abscissa indicate center and surround were 180° out of phase. Error bars indicate one standard deviation of the fitted cumulative normal distributions (see text). Error bars are often smaller than the size of the plotted points.
3986
L.A. Olzak, P.I. Laurinen / Vision Research 39 (1999) 3983–3987
Fig. 3. Reduction of perceived contrast of center grating for the three different center-surround pattern combinations used in experiment 2. Normalized suppression is again plotted on the ordinate. Error is indicated as in Fig. 2.
gether, these results suggest a phase-specific gain control pool is operating under our stimulus conditions.
4. Experiment 2 In this experiment, contrast of the surround was held constant at 0.39. Six test configurations were used that mixed different center and surround modulation types (sinusoidal or plaid) and phase relationships between center and surround components. All six are shown in Fig. 1. Upper and lower rows show stimuli presented with center and surround in phase and 180° out of phase, respectively. The leftmost two configurations, referred to as sin/sin, were identical to the vertical sinusoidal center and surrounds of experiment I. The center pair added a horizontal grating to the surround to create a plaid surround with a vertical sinusoidal center. The third pair contained plaids in both center and surround. The variable comparison disk always matched the modulation pattern of the center test stimulus. For each of the three stimulus types, in phase and out-of phase configurations were run in a single intermingled session.
4.1. Results of experiment 2 The results for all six conditions are shown in Fig. 3 for two observers. Again, the results of the third observer closely matched those of PL. Consider first the sin/sin results. When both center and surround are in phase, we again replicate the results that suppression occurs only when components are in phase. Little, if any suppression occurs when components are 180° out of phase. This result agrees with results of experiment 1 and with those of Ejima and Takahashi (1985).
Now consider the effect of adding a second horizontal component to the surround. The phase-specificity of the effect is still evident in the data of both. Adding the second, orthogonal component did not change the results of TK; results were still highly phase-specific. Observer PL (and our third observer) shows some reduction of phase specificity, perhaps reflecting individual differences found by others, but the effect is still marked. Finally, consider results when both center and surround components were plaids. In this condition, the phase specificity of the effect disappears entirely in all three of our observers’ data. Regardless of the phase relationships between the components, large suppression effects reliably occur. The effects are of the magnitude reported previously by ourselves and others for simple gratings when center and surround are in phase.
5. Discussion Our results suggest that there are at least two sources of gain control that operate in contrast–contrast phenomena. Both reflect processes that operate laterally over space. One process or pool, which is revealed with simple sinusoidal grating stimuli, appears to operate only over spatially aligned pathways with similar phase or polarity tuning. The second process is revealed when center and surround are complex and are modulated along more than a single orientation. This process is not specific to phase, and appears not only to operate over all orientations, but also over neurons tuned to different phases. The reduced or eliminated interaction when the center is a simple sinusoid and the surround a plaid suggests that the two processes operate more or less independently, although some cross-talk may exist. The current results may help reconcile the apparently contradictory results regarding phase or polarity spe-
L.A. Olzak, P.I. Laurinen / Vision Research 39 (1999) 3983–3987
cificity in lateral gain control. Polarity-specific results have been obtained with simple grating stimuli (Ejima & Takahashi, 1985). In the Olzak and Thomas (1999) model, pooled gain control processes are transparent to such stimuli and reveal only within-pathway processes. The polarity-nonspecific results reported by Solomon et al. (1993), on the other hand, were obtained not with simple gratings, but with dot patterns that are complex in the Fourier domain. Such stimuli would reflect operations of higher-level gain control pools. The existence of two different sources of gain control may be relevant in understanding phase effects found in other studies of lateral interactions, such as those reported by Levi and Waugh (1996), Moulden (1994) and Zenger and Sagi (1996). Ozawa et al. suggested that normalization might precede input to higher-level mechanisms. What might those higher-level mechanisms be? It has been suggested that mechanisms involved in the perception of transparency (D’Zmura, Colantoni, Knoblauch & Laget, 1997) could view the stimulus as a background seen through a veiling substance. Our phase-specific results are consistent with such a view because the phase shift would disrupt the perception of transparency. However, the disappearance of specificity with plaid patterns argues against transparency as a general explanation of contrast – contrast phenomena. Another possibility includes the higher level summing circuits involved in contrast discrimination tasks. We note that the processes we have isolated here share many striking similarities with the two types of gain control processes described by Thomas and Olzak (1997) and Olzak and Thomas (1999) to account for masking in fine discrimination experiments. As noted in Section 1, in their model normalized responses feed into specialized summing circuits, one of which sums over orientations and signals information used to make fine discriminations about the contrast of surface textures. These mechanisms may contribute to apparent contrast as well. Georgeson and Meese (1997) recently described a higher-level summing mechanism that also pools over orientation and is involved in the appearance of plaids. Whether these two mechanisms are the same, and are actually involved in contrast–contrast phenomena remains to be seen.
References Cannon, M. W., & Fullenkamp, S. C. (1991a). Spatial interactions in apparent contrast: inhibitory effects among grating patterns of different spatial frequencies, spatial positions and orientations. Vision Research, 31, 1985–1998.
.
3987
Cannon, M. W., & Fullenkamp, S. C. (1991b). A transducer model for contrast perception. Vision Research, 31, 983 – 998. Cannon, M. W., & Fullenkamp, S. C. (1993). Spatial interactions in apparent contrast: individual differences in enhancement and suppression effects. Vision Research, 33, 1685 – 1695. Chubb, C., Sperling, G., & Solomon, J. A.. (1989). Texture interactions determine perceived contrast. Proceedings of the National Academy of Sciences of the United States of America, 86, 9631– 9635. De Bonet, J. S., & Zaidi, Q. (1997). Comparison between spatial interactions in perceived contrast and perceived brightness. Vision Research, 37, 1141 – 1155. D’Zmura, M., Colantoni, P., Knoblauch, K., & Laget, B. (1997). Color transparency. Perception, 26, 471 – 492. D’Zmura, M., & Singer, B. (1996). Spatial pooling of contrast in contrast gain control. Journal of the Optical Society of America A, 13, 2135 – 2140. Ejima, Y., & Takahashi, S. (1985). Apparent contrast of a sinusoidal grating in the simultaneous presence of peripheral gratings. Vision Research, 25, 1223 – 1232. Foley, J. (1994). Human luminance pattern-vision mechanisms: masking experiments require a new model. Journal of the Optical Society of America A, 11, 1710 – 1719. Georgeson, M. A., & Meese, T. S. (1997). Perception of stationary plaids: the role of spatial filters in edge analysis. Vision Research, 37, 3255 – 3271. Levi, D. M., & Waugh, S. J. (1996). Position acuity with oppositecontrast polarity features: evidence for a nonlinear collector mechanism for position acuity. Vision Research, 36, 573–588. McCourt, M. E. (1982). A spatial frequency dependent grating-induction effect. Vision Research, 22, 119 – 134. Moulden, B. (1994). Collator units: second-stage orientational filters. In Higher order processing in the 6isual system, Ciba Foundation Symposium 184. Chichester: Wiley. Ohzawa, I., Sclar, G., & Freeman, R. D. (1982). Contrast gain control in the cat visual cortex. Nature, 298, 266 – 268. Olzak, L. A., & Thomas, J. P. (1992). Configural effects constrain Fourier models of pattern discrimination. Vision Research, 32, 1885 – 1898. Olzak, L. A., & Thomas, J. P. (1997). Models of cortical contrast gain control: A test. Paper presented at the Optical Society of America Meeting, Rochester, NY Supplement to Optics & Photonics News, 8, 97 (abs). Olzak, L. A., & Thomas, J. P. (1999). Neural recoding in human pattern vision: model and mechanisms. Vision Research, 39, 231– 256. Singer, B., & D’Zmura, M. (1994). Color contrast induction. Vision Research, 34, 3111 – 3126. Solomon, J. A., Sperling, G., & Chubb, C. (1993). The lateral inhibition of perceived contrast is indifferent to on-center/off-center segregation, but specific to orientation. Vision Research, 33, 2671 – 2683. Snowden, R. J., & Hammett, S. T. (1998). The effects of surround contrast on contrast thresholds, perceived contrast and contrast discrimination. Vision Research, 38, 1935 – 1945. Thomas, J. P., & Olzak, L. A. (1997). Contrast gain control and fine spatial discriminations. Journal of the Optical Society of America A, 14, 2392 – 2405. Zaidi, Q., Yoshimi, B., Flanigan, N., & Canova, A. (1992). Lateral interactions within color mechanisms in simultaneous induced contrast. Vision Research, 32, 1695 – 1707. Zenger, B., & Sagi, D. (1996). Isolating excitatory and inhibitory nonlinear spatial interactions involved in contrast detection. Vision Research, 36, 2497 – 2513.
Rapid communication
Multiple gain control processes in contrast–contrast phenomena Lynn A. Olzak a,*, Pentti I. Laurinen b a
Department of Psychology, Miami Uni6ersity of Ohio, Oxford, OH 45056, USA b Department of Psychology, Uni6ersity of Helsinki, Helsinki, Finland Received 31 December 1998; received in revised form 24 May 1999
Abstract Spatial interactions among orientation-tuned gain control processes are presumed to mediate center-surround contrast – contrast phenomena. In this paper, we assess contributions of gain control processes that pool over orientation. We measured the apparent contrast of a luminance-modulated center disk embedded in various modulated surrounds. In all conditions, observers compared the apparent contrast of the test center to an identically modulated disk with no surround. When center and surround are simple, vertical sinusoids and presented in phase, suppression depends upon surround contrast and is marked at high contrasts. When components are presented 180° out of phase, no suppression occurs at any contrast. When a horizontal component is added to the surround, much less suppression occurs. However, strong suppression is reinstated when both center and surround are plaids. Neither of the latter two effects are phase dependent. We suggest that two different sources of gain control are revealed by the simple sinusoidal and the plaid stimuli. One is orientation tuned and phase-dependent. The other pools over all orientations and includes neurons tuned to multiple phases. © 1999 Elsevier Science Ltd. All rights reserved. Keywords: Control processes; Spatial interactions; Contrast–contrast phenomena
1. Introduction Ohzawa, Sclar and Freeman (1982) were the first to report that the majority of neurons in cat visual cortex adjust their response levels to prevailing contrast, similar to adjustments made in retinal cells to prevailing illumination. They pointed out that one obvious benefit of contrast gain control is to help the system maintain a high degree of sensitivity for contrast differences over wide ranges of contrasts, analogous to retinal control of brightness sensitivity. They also suggested that early normalization processes could provide contrast-independent input to a number of higher-level mechanisms. Gain control processes operating laterally over space have been widely suggested as the mechanism underlying contrast–contrast phenomena, a term coined by
* Corresponding author. Tel.: +1-513-529-1754; fax: + 1-513-5292420. E-mail address: [email protected] (L.A. Olzak)
Chubb, Sperling and Solomon (1989). In this phenomenon, the apparent contrast of a luminance or chromatically modulated pattern is reduced by the addition of an adjacent or surrounding pattern. A number of authors have proposed models of luminance or chromatic contrast induction that modify activity by lateral inhibition within a pool of neurons (e.g. Cannon & Fullenkamp, 1991a,b; Zaidi, Yoshimi, Flanigan & Canova, 1992; Solomon, Sperling & Chubb, 1993; Singer & D’Zmura, 1994; D’Zmura & Singer, 1996; De Bonet & Zaidi, 1997). However, it is not clear from existing studies whether the pool includes only neurons with similar tuning functions, or whether the pool is broadband with respect to spatial frequency, orientation, and phase (see Chubb et al., 1989; Cannon & Fullenkamp, 1991a; Solomon et al., 1993). Furthermore, it is not clear whether the same lateral processes underlie masking effects observed in detection tasks (Foley, 1994), contrast discrimination (Olzak & Thomas, 1992) and contrast appearance, as has been suggested by some (e.g. Snowden & Hammett, 1998).
0042-6989/99/$ - see front matter © 1999 Elsevier Science Ltd. All rights reserved. PII: S 0 0 4 2 - 6 9 8 9 ( 9 9 ) 0 0 1 3 1 - 5
3984
L.A. Olzak, P.I. Laurinen / Vision Research 39 (1999) 3983–3987
Of particular relevance to the present paper are conflicting results with respect to phase or polarity. Ejima and Takahashi (1985) reported that with vertically abutting inducing and test gratings (vertical sinusoids), no suppressive effects were found when the two patches were out of phase and therefore differed in polarity along the line of collinearity. However, using center-surround textures composed of micro-elements, Solomon et al. (1993) reported that the relative polarity of the elements had no effect on the magnitude of the suppressive effect. No phase dependencies were reported by Zenger and Sagi (1996), who used a contrast detection task with spatially displaced but vertically aligned Gabor patches. These apparently conflicting results might be reconciled if more than a single gain control mechanism contributes to contrast – contrast. Cannon and Fullenkamp (1991a,b), for example, found hints of two processes underlying contrast-contrast effects in two different experiments. Evidence for multiple sources of gain control exists in other literatures, as well. Pertinent to the current paper is recent work in complexpattern discrimination (Olzak & Thomas, 1992, 1997; Thomas & Olzak, 1997; Olzak & Thomas, 1999). One source isolated in their work is highly tuned with respect to spatial frequency, orientation, and phase, and is presumed to occur within a low-level, tuned pathway. In a second type of process, at least two different types of higher-level mechanisms normalize over a broad but selective pool of neurons and then sum over more specific (and different) sets of neurons to signal particular types of information. One of these mechanisms pools over all orientations, but only within a limited frequency band. This mechanism provides information about the contrast and/or spatial grain of a textured surface (Olzak & Thomas, 1992, 1999) and therefore might play a role in contrast– contrast phenomena. A similar broadly-tuned pooled gain-control process has been proposed by Foley (1994) in his model of contrast detection. However, all tests of these higher-level pools have to date been made with spatially overlapping components. Thus, nothing is yet known about their spatial pooling abilities. In this paper, we test the hypothesis that multiple contrast gain control mechanisms exist and operate laterally over space. Specifically, we ask whether properties of gain control mechanisms revealed with simple sinusoidal grating patterns differ from those tapped with more complex stimuli. We further ask how higher-level processes revealed with plaid stimuli in contrast discrimination experiments might contribute to contrast–contrast effects.
2. Methods
2.1. Obser6ers Three observers participated in both experiments. Observer PL is an experienced observer with normal uncorrected vision. Observers TK and SK are relatively inexperienced observers, emmetropes, and naı¨ve to the purpose of the study.
2.2. Stimuli and apparatus Stimuli were created using NIH Image and displayed on a linearized Nokia Multigraph 445× color monitor using a Cambridge Research systems V2.2 graphics board in a PC 486 system. Viewing distance was 114 cm. Test and comparison disks were displayed simultaneously in a left–right configuration. The comparison disk consisted of a 1°, luminance-modulated disk of variable contrast viewed against a uniform background subtending 5.5° of visual angle. The test stimulus consisted of a 1° modulated disk embedded in the center of a 5.5° modulated surround, located adjacent to the comparison on the display. Comparison and test were thus each located at eccentricities of 2.25°. The spatial frequency of all modulations was 2.75 cpd. Contrast of the test center was held constant in all experiments and conditions at 0.18. Contrast of the surround modulation of the test stimulus was varied in different conditions around the mean luminance of 21 cd/m2, but was held constant in any one condition.
2.3. General procedures In all experiments and conditions, the apparent contrast of the test disk was measured in a 2-alternative forced choice method of constant stimuli procedure. Observers initiated each trial following presentation of a centered fixation point. Test and comparison stimuli were presented simultaneously for 1 s, with abrupt onset and offset. Left–right presentation of test and comparison stimuli was counterbalanced and randomly intermingled within a session. The contrast of the comparison disk was varied in all experiments in ten steps over a range that depended upon condition. The observer indicated by pressing one of two buttons whether the left or right disk appeared to have the greater contrast. Each contrast level was presented ten times for a given condition, for a total of 100 trials per condition. Two conditions were always randomly intermingled in a block of trials, for a total of 200 trials per session. Each observer participated in three replications of each condition. The frequency of left–right responses were converted to probabilities of responding that the test disk ap-
L.A. Olzak, P.I. Laurinen / Vision Research 39 (1999) 3983–3987
Fig. 1. Examples of stimuli. Top row: center and surround in phase. Bottom row: center and surround 180° out of phase. Left-hand column: simple combinations of sinusoids used in both experiments 1 and 2. Center: sinusoidal center with plaid surround used in experiment 2. Right-hand column: Plaid center and surrounds used in experiment 2.
peared to have more contrast than the comparison disk, as a function of contrast level. A cumulative normal was fit to these data using least squares criterion. The mean of this distribution was taken as the point of subjective equality (PSE), and its standard deviation served as a measure of error. Contrast suppression was calculated as (Ccenter −CPSE)/Ccenter. This measure ranges from 0 (no suppression) to 1.0.
3. Experiment 1 A preliminary experiment was performed to test whether we could replicate the Ejima and Takahashi (1985) results with sinusoidal, center-surround stimuli. We also wanted to be assured that that contrast suppression results reported by others in previous work could be demonstrated in our observers, as differences
3985
among individuals have been reported (Cannon & Fullenkamp, 1993). We also wanted to ensure that the phenomenon we were measuring was not a simple by-product of contrast reduction by the grating induction effect first reported by McCourt (1982). Induced gratings appear counterphase to the inducing grating, and would affect perceived contrast by reducing apparent contrast when center and surround are in phase, and enhancing it when they are out of phase. Accordingly, we performed a partial replication of Ejima and Takahashi (1985), using high contrast values and our own particular stimulus configuration and observers. The contrast of the modulated surround took one of eight values, ranging from + 0.39 to − 0.39, resulting in surrounds that were both in phase and 180° out of phase with the center. When stimuli were out of phase, an additional measurement was made with surround contrast equal to the test contrast of 0.18. Example of the in-phase and out-of-phase sinusoidal stimuli used in this experiment are shown, respectively in the upper and lower lefthand panels of Fig. 1.
3.1. Results of experiment 1 The results are plotted in Fig. 2 for two observers. The third observer’s data was highly similar to the results of PL. The results of all three observers replicate Ejima and Takahashi (1985) and others in that we find suppression only when the vertically modulated center and surround are in phase, and suppression increases with increasing surround contrast. However, we observed no effects of any sort when center and surround were out of phase, regardless of surround contrast. We therefore can rule out grating induction as the causal factor in our observed contrast–contrast effect. To-
Fig. 2. Reduction of perceived contrast of center grating as a function of surround grating contrast for observers TK (left panel) and PL (right panel). Normalized suppression is plotted on the ordinate (1.0 = maximum suppression). Surround contrast is plotted on the abscissa. Positive values to the right of the ordinate indicate center and surround were in phase. Negative values to the left of the abscissa indicate center and surround were 180° out of phase. Error bars indicate one standard deviation of the fitted cumulative normal distributions (see text). Error bars are often smaller than the size of the plotted points.
3986
L.A. Olzak, P.I. Laurinen / Vision Research 39 (1999) 3983–3987
Fig. 3. Reduction of perceived contrast of center grating for the three different center-surround pattern combinations used in experiment 2. Normalized suppression is again plotted on the ordinate. Error is indicated as in Fig. 2.
gether, these results suggest a phase-specific gain control pool is operating under our stimulus conditions.
4. Experiment 2 In this experiment, contrast of the surround was held constant at 0.39. Six test configurations were used that mixed different center and surround modulation types (sinusoidal or plaid) and phase relationships between center and surround components. All six are shown in Fig. 1. Upper and lower rows show stimuli presented with center and surround in phase and 180° out of phase, respectively. The leftmost two configurations, referred to as sin/sin, were identical to the vertical sinusoidal center and surrounds of experiment I. The center pair added a horizontal grating to the surround to create a plaid surround with a vertical sinusoidal center. The third pair contained plaids in both center and surround. The variable comparison disk always matched the modulation pattern of the center test stimulus. For each of the three stimulus types, in phase and out-of phase configurations were run in a single intermingled session.
4.1. Results of experiment 2 The results for all six conditions are shown in Fig. 3 for two observers. Again, the results of the third observer closely matched those of PL. Consider first the sin/sin results. When both center and surround are in phase, we again replicate the results that suppression occurs only when components are in phase. Little, if any suppression occurs when components are 180° out of phase. This result agrees with results of experiment 1 and with those of Ejima and Takahashi (1985).
Now consider the effect of adding a second horizontal component to the surround. The phase-specificity of the effect is still evident in the data of both. Adding the second, orthogonal component did not change the results of TK; results were still highly phase-specific. Observer PL (and our third observer) shows some reduction of phase specificity, perhaps reflecting individual differences found by others, but the effect is still marked. Finally, consider results when both center and surround components were plaids. In this condition, the phase specificity of the effect disappears entirely in all three of our observers’ data. Regardless of the phase relationships between the components, large suppression effects reliably occur. The effects are of the magnitude reported previously by ourselves and others for simple gratings when center and surround are in phase.
5. Discussion Our results suggest that there are at least two sources of gain control that operate in contrast–contrast phenomena. Both reflect processes that operate laterally over space. One process or pool, which is revealed with simple sinusoidal grating stimuli, appears to operate only over spatially aligned pathways with similar phase or polarity tuning. The second process is revealed when center and surround are complex and are modulated along more than a single orientation. This process is not specific to phase, and appears not only to operate over all orientations, but also over neurons tuned to different phases. The reduced or eliminated interaction when the center is a simple sinusoid and the surround a plaid suggests that the two processes operate more or less independently, although some cross-talk may exist. The current results may help reconcile the apparently contradictory results regarding phase or polarity spe-
L.A. Olzak, P.I. Laurinen / Vision Research 39 (1999) 3983–3987
cificity in lateral gain control. Polarity-specific results have been obtained with simple grating stimuli (Ejima & Takahashi, 1985). In the Olzak and Thomas (1999) model, pooled gain control processes are transparent to such stimuli and reveal only within-pathway processes. The polarity-nonspecific results reported by Solomon et al. (1993), on the other hand, were obtained not with simple gratings, but with dot patterns that are complex in the Fourier domain. Such stimuli would reflect operations of higher-level gain control pools. The existence of two different sources of gain control may be relevant in understanding phase effects found in other studies of lateral interactions, such as those reported by Levi and Waugh (1996), Moulden (1994) and Zenger and Sagi (1996). Ozawa et al. suggested that normalization might precede input to higher-level mechanisms. What might those higher-level mechanisms be? It has been suggested that mechanisms involved in the perception of transparency (D’Zmura, Colantoni, Knoblauch & Laget, 1997) could view the stimulus as a background seen through a veiling substance. Our phase-specific results are consistent with such a view because the phase shift would disrupt the perception of transparency. However, the disappearance of specificity with plaid patterns argues against transparency as a general explanation of contrast – contrast phenomena. Another possibility includes the higher level summing circuits involved in contrast discrimination tasks. We note that the processes we have isolated here share many striking similarities with the two types of gain control processes described by Thomas and Olzak (1997) and Olzak and Thomas (1999) to account for masking in fine discrimination experiments. As noted in Section 1, in their model normalized responses feed into specialized summing circuits, one of which sums over orientations and signals information used to make fine discriminations about the contrast of surface textures. These mechanisms may contribute to apparent contrast as well. Georgeson and Meese (1997) recently described a higher-level summing mechanism that also pools over orientation and is involved in the appearance of plaids. Whether these two mechanisms are the same, and are actually involved in contrast–contrast phenomena remains to be seen.
References Cannon, M. W., & Fullenkamp, S. C. (1991a). Spatial interactions in apparent contrast: inhibitory effects among grating patterns of different spatial frequencies, spatial positions and orientations. Vision Research, 31, 1985–1998.
.
3987
Cannon, M. W., & Fullenkamp, S. C. (1991b). A transducer model for contrast perception. Vision Research, 31, 983 – 998. Cannon, M. W., & Fullenkamp, S. C. (1993). Spatial interactions in apparent contrast: individual differences in enhancement and suppression effects. Vision Research, 33, 1685 – 1695. Chubb, C., Sperling, G., & Solomon, J. A.. (1989). Texture interactions determine perceived contrast. Proceedings of the National Academy of Sciences of the United States of America, 86, 9631– 9635. De Bonet, J. S., & Zaidi, Q. (1997). Comparison between spatial interactions in perceived contrast and perceived brightness. Vision Research, 37, 1141 – 1155. D’Zmura, M., Colantoni, P., Knoblauch, K., & Laget, B. (1997). Color transparency. Perception, 26, 471 – 492. D’Zmura, M., & Singer, B. (1996). Spatial pooling of contrast in contrast gain control. Journal of the Optical Society of America A, 13, 2135 – 2140. Ejima, Y., & Takahashi, S. (1985). Apparent contrast of a sinusoidal grating in the simultaneous presence of peripheral gratings. Vision Research, 25, 1223 – 1232. Foley, J. (1994). Human luminance pattern-vision mechanisms: masking experiments require a new model. Journal of the Optical Society of America A, 11, 1710 – 1719. Georgeson, M. A., & Meese, T. S. (1997). Perception of stationary plaids: the role of spatial filters in edge analysis. Vision Research, 37, 3255 – 3271. Levi, D. M., & Waugh, S. J. (1996). Position acuity with oppositecontrast polarity features: evidence for a nonlinear collector mechanism for position acuity. Vision Research, 36, 573–588. McCourt, M. E. (1982). A spatial frequency dependent grating-induction effect. Vision Research, 22, 119 – 134. Moulden, B. (1994). Collator units: second-stage orientational filters. In Higher order processing in the 6isual system, Ciba Foundation Symposium 184. Chichester: Wiley. Ohzawa, I., Sclar, G., & Freeman, R. D. (1982). Contrast gain control in the cat visual cortex. Nature, 298, 266 – 268. Olzak, L. A., & Thomas, J. P. (1992). Configural effects constrain Fourier models of pattern discrimination. Vision Research, 32, 1885 – 1898. Olzak, L. A., & Thomas, J. P. (1997). Models of cortical contrast gain control: A test. Paper presented at the Optical Society of America Meeting, Rochester, NY Supplement to Optics & Photonics News, 8, 97 (abs). Olzak, L. A., & Thomas, J. P. (1999). Neural recoding in human pattern vision: model and mechanisms. Vision Research, 39, 231– 256. Singer, B., & D’Zmura, M. (1994). Color contrast induction. Vision Research, 34, 3111 – 3126. Solomon, J. A., Sperling, G., & Chubb, C. (1993). The lateral inhibition of perceived contrast is indifferent to on-center/off-center segregation, but specific to orientation. Vision Research, 33, 2671 – 2683. Snowden, R. J., & Hammett, S. T. (1998). The effects of surround contrast on contrast thresholds, perceived contrast and contrast discrimination. Vision Research, 38, 1935 – 1945. Thomas, J. P., & Olzak, L. A. (1997). Contrast gain control and fine spatial discriminations. Journal of the Optical Society of America A, 14, 2392 – 2405. Zaidi, Q., Yoshimi, B., Flanigan, N., & Canova, A. (1992). Lateral interactions within color mechanisms in simultaneous induced contrast. Vision Research, 32, 1695 – 1707. Zenger, B., & Sagi, D. (1996). Isolating excitatory and inhibitory nonlinear spatial interactions involved in contrast detection. Vision Research, 36, 2497 – 2513.