Vision Research 45 (2005) 865–880 www.elsevier.com/locate/visres
The detection of motion in chromatic stimuli: first-order and second-order spatial structure q Simon J. Cropper
*
Department of Psychology, University of Melbourne, Victoria 3010, Australia Received 19 February 2004; received in revised form 31 August 2004
Abstract This study provides evidence for the existence of a low-level chromatic motion mechanism and further elucidates the conditions under which its operation becomes measurable in an experimental stimulus. Observers discriminated the direction of motion of amplitude modulated (AM) gratings that were defined by luminance or chromatic variation and masked with spatiotemporally broadband luminance or chromatic noise. The size and retinal location of the stimuli were varied and the effects of broadband noise and grating masks were both compared with the cohort of stimuli. Some significant disparities in the published literature were well explained by the results. In conclusion, evidence for a chromatically sensitive motion mechanism that evades the, detrimental effects of a luminance mask was found only at the fovea and only when the stimulus was small and centrally placed. 2004 Elsevier Ltd. All rights reserved. Keywords: Colour; Luminance; Noise; Gratings; Motion; Masking; First-order; Second-order; Low-level; High-level
1. Introduction For the last 30 or so years, peculiarities in the perceived motion of stimuli defined only by a change in colour have provided one of the main lines of evidence in support of the early segregation of signals (e.g. colour
q
Some of this work has been reported at the European Conference on Visual Perception in Oxford in 1998, The Association for Research in Vision and Ophthalmology in Ft Lauderdale in 2000 and Vision Sciences Society, Sarasota in 2004 and has therefore been published in abstract form (Cropper, S.J. (1998a). Combining the motion of firstorder and second-order, chromatic and luminance stimuli: A purely local process? Perception, 27, ECVP Abstracts; Cropper, S.J. (2000). Independent chromatic motion is limited to the fovea. Investigative Ophthalmology and Visual Science, 41 (4); Cropper, S.J. (2004a). Colour and motion––masking u¨ber alles. Journal of Vision (VSS abstracts)). * Tel.: +61 3 8344 6317; fax: +61 3 9347 6618. E-mail address:
[email protected] URL: http://deejayvision.org 0042-6989/$ - see front matter 2004 Elsevier Ltd. All rights reserved. doi:10.1016/j.visres.2004.09.043
and motion) within the visual system (Livingstone & Hubel, 1987; Marr, 1982; Zeki, 1978). However, as more data on chromatic motion has been collected, the findings have become increasingly discrepant, as have the conclusions drawn from them. Taken as a body of work, the published literature now supports every possible conclusion regarding the existence, or otherwise, of a purely chromatic motion mechanism. The current study examines and reconciles two disparate lines of data within the published work, ultimately falling in favour of the existence of a purely chromatic motion mechanism that is only revealed under specific conditions. Several studies of chromatic motion perception have used a stimulus composed of a rigidly displaced array of Gabor patches, termed a micropattern, to examine the effects of changes in the spatiotemporal structure of the pattern (Baker, Boulton, & Mullen, 1998; Boulton & Baker, 1993a, 1993b; Yoshizawa, Mullen, & Baker, 2000; Yoshizawa, Mullen, & Baker, 2003). The spatiotemporal structure of the micropatterns was manipulated so they preferentially stimulated either low-level
866
S.J. Cropper / Vision Research 45 (2005) 865–880
or high-level motion mechanisms. 1 The comparison of luminance and chromatic modulation of the first-order carrier of the Gabors, when presented in conjunction with dynamic noise masks showed only luminance modulated micropatterns to behave in a way consistent with being detected by low-level motion mechanisms. This led the authors reasonably to conclude that there is no low-level chromatic motion mechanism in the system. Rather, they suggested that the ability to discriminate motion in a chromatic carrier was mediated by the same mechanism that mediated the perception of motion in the patternÕs envelope. They defined their two motion systems empirically as quasi-linear and non-linear in their properties. To some degree, this classification is consistent with the hierarchy suggested by several other studies (Lu, Lesmes, & Sperling, 1999; Lu & Sperling, 2001; Cavanagh & Mather, 1989; Wilson, Ferrera, & Yo, 1992) although there is some inconsistency in the use of the terms, which originally referred to statistical order (Julesz, 1971), making such comparisons hard. Contrary data, however, was found using chromatic and luminance grating stimuli and masks (Cropper & Derrington, 1996). At a very short exposure duration (17 ms) it was shown that a purely chromatic stimulus was immune to the effects of a luminance (grating) mask, despite both luminance and colour stimuli being affected by masks of identical luminance or chromatic properties to the test. The short duration and strong masking effects when both test and mask are defined along a single axis of colour space (i.e. like test and mask properties), but not between axes (different properties), led these authors to conclude strongly in favour of the presence of a basic chromatic motion mechanism. The short duration is too brief to stimulate any non-linear mechanism sensitive to the motion of contrast modulation or to elicit any eye-movement, and so leaves the perception of the chromatic stimulus motion only possible through the use of a low-level mechanism (Cropper & Derrington, 1994; Derrington, Badcock, & Henning, 1993). Thus, the available data provide evidence for opposing conclusions and highlight the need to elucidate further the conditions where luminance and colour cross-masking does, and does not, occur. A study using jittered grating masks showed a strong contrast dependence for the cross-masking interaction (Stromeyer, 1
I will use the terminology ‘‘low-level’’ and ‘‘high-level’’ to distinguish between suggested motion mechanisms. These terms respectively refer either to a mechanism directly sensitive to the spatiotemporal orientation of the input (low-level), or to a mechanism which extrapolates that temporal orientation from the spatial properties of the input (high-level). Furthermore, I will use the terms ‘‘firstorder’’ and ‘‘second-order’’ to refer strictly to the statistical structure of the spatial modulation of the stimulus independent of the, as yet, undefined underlying mechanisms.
Chaparro, & Kronauer, 1996) and, more recently, an attempt was made to explain why different patterns of results were obtained (Mullen, Yoshizawa, & Baker, 2003; Yoshizawa et al., 2003) but a number of unresolved issues remain. The most significant of these issues is the question of why different results are obtained with noise masks and with grating masks. The purpose of this article is to try to reconcile the discrepant findings and to establish the conditions under which the action of the chromatic motion mechanism is evident, despite the presence of potentially deleterious luminance noise.
2. Methods 2.1. Apparatus and stimuli All patterns were digitally generated from sinusoidal modulations of colour or luminance and were displayed to a contrast-resolution of 14-bits per pixel by the TMS30c25 DSP chip on a VSG2/3 (Cambridge Research Systems) stimulus generator. The patterns were presented on a Sony GDM se2 17 inch colour monitor with a mean luminance of 90 cd/m2 and chromaticity having CIE co-ordinates of (0.333, 0.377). The monitor was driven at a frame rate of 75 Hz and a line rate of 52 kHz. All patterns were generated digitally during the line-flyback prior to presentation (i.e. there was no frame or line interleaving). The one-dimensional nature of the stimuli, oriented along the raster lines of the monitor, also minimised adjacent pixel non-linearities (Klein, Hu, & Carney, 1996). The voltage-to-luminance relationship of the display was regularly measured using a photometric head (Graseby S351G) and the non-linear relationship was corrected using internal lookup tables on the VSG. The curve fitting procedure gave an R value accounting for 0.998 of the variance. The total viewable display subtended a visual angle of 30 by 24 at the viewing distance of 0.5m with a pixel size of 0.036 by 0.036. The stimulus itself occupied a portion of the total display area as detailed below. A small dark fixation point was located at the centre of the display. Viewing was conducted in a semi-darkened room and was binocular with natural pupils. No head restraint was used. Observers were the author (SJC) and five paid observers naive to the aim of the experiment (TB, MW, LK, SL & SP). 2 2.2. Spatial structure The test stimuli were amplitude-modulated (AM) gratings with or without superimposed noise. Variants 2
The publisherÕs constraints upon the size of this article mean that much of the ‘‘non-essential’’ data are available in the supplementary material. I apologise to the critical reader for this unwelcome situation.
S.J. Cropper / Vision Research 45 (2005) 865–880
on this stimulus structure were the use of gratings as either a test or mask and are individually detailed in the relevant sections. All patterns were one-dimensional in their spatial variation and horizontally oriented. The chromatic properties of the stimuli are described by a vector in a cardinal colour space using a three-dimensional coordinate system (Derrington, Krauskopf, & Lennie, 1984). The spatial modulation function, L(y), of the AM gratings was of the form: LðyÞ ¼ Lm f1 þ C c sin 2pðfc y þ /c Þ0:5 ð1 þ M cos 2pðfenv y þ /env ÞÞg
867
conditions, first and second-order modulations (carrier and envelope) of the stimuli were displaced in opposite directions by a given spatial phase angle. In the Ôrigid motionÕ condition (Experiment 2), stimuli were displaced as coherent, rigid patterns, with the relative spatial phase of carrier to envelope (/c to /env) fixed throughout the presentation interval. In all conditions, the starting phase of the pattern relative to the display was randomised between each interval. 3 2.4. Noise masks
ð1Þ
where Lm is the mean luminance of the display, M is the depth of modulation (0 6 M 6 1) of carrier contrast (Cc), fc is the spatial frequency of the carrier expressed in cycles per degree (cpd) and fenv is the spatial frequency of the contrast envelope. The spatial phase angles of the carrier and envelope are denoted by /c and /env, respectively. All the experiments reported here were carried out with a carrier of 1 cpd modulated by a 0.2 cpd envelope. All conditions were replicated with a 0.1 cpd envelope without any systematic differences in the results. The spatial configuration of the stimulus was either a 20 disk, a 4 disk or a 20/8 annulus to control for foveal versus parafoveal presentation. All stimuli were placed centrally and restricted by a rectangular spatial envelope. Examples of the test stimuli and comparative illustrations of micropattern stimuli are shown in Fig. 1. 2.3. Temporal (motion) structure The stimuli were displaced between the two central frames of a 76 frame (1 s) presentation interval, which was temporally modulated by a single cycle of a raised cosine (half-width 500 ms). The spatial envelope and the carrier of the stimuli were independently manipulable which allowed the response to the motion of the first-order and second-order variations in the stimuli to be independently examined. In the non-rigid motion
In the majority of stimulus conditions, a one-dimensional random 1/f (pink) noise field was superimposed on the AM gratings. The noise was digitally generated in the same way as were the gratings and the composite stimulus (noise and grating) calculated on each line prior to presentation. The masks were presented at a peak contrast that was scaled to the detectability of the mask alone and were modulated in either luminance or colour along one axis of the cardinal space. The detectability of the masks was independently measured for each observer using the staircase described below. The noise threshold value used was derived using static rather than dynamic masks to remove the addition flicker cue present in the latter (Derrington & Henning, 1981). This approach tends to overestimate detection threshold and therefore to bias the mask contrasts toward a higher value increasing their potential as noise. The fundamental spatial frequency of the noise in all experiments was 0.044 cpd (i.e. approximately one cycle per display-screen). Whilst the test and mask were both present in every frame, the mask structure was only updated every third frame. When the test pattern was displaced, the noise profile was also updated to ensure optimum mask-test interactions in the temporal domain.
3. Psychophysical methods 3.1. Detection thresholds Detection thresholds for the grating carrier of the AM waveform and for the noise and grating masks were independently measured for all stimulus configurations in a standard two-alternative forced-choice (2AFC) detection task. A stimulus was presented in one of two intervals and the observer indicated which interval the stimulus had appeared in by means of a mouse button-press. A feedback tone was provided. The contrast
Fig. 1. Density plots of example stimuli. The top two panels plot sparse and dense micropattern stimuli, as used in previous work. The bottom panels plot 100% and 50% modulated AM patterns used in the current work. Carrier contrast is normalised to 1 in these illustrations but was appropriately scaled in the experiments.
3 The temporal structure of the motion sequence is an important parameter of the stimulus but for the current purposes a single displacement is most appropriate and shows data consistent with either 5-frame or continuous motion Cropper (1998a, 2004a).
868
S.J. Cropper / Vision Research 45 (2005) 865–880
of the stimulus was adjusted according to the observerÕs response by a modified PEST staircase procedure (Findlay, 1978) which gave a final threshold estimate of performance at the level of 75% correct. The contrasts of the test and mask stimuli used subsequently were scaled to the individualÕs detection threshold for each stimulus type alone.
sures, in line with previous work (Anstis & Cavanagh, 1983; Cropper, Mullen, & Badcock, 1996; Wagner & Boynton, 1972).
4. Results 4.1. Experiment 1
3.2. Direction discrimination thresholds A 2AFC direction-discrimination procedure combined with the method of constant stimuli was used to measure observersÕ ability to discriminate the direction of motion of the stimuli. Observers had to indicate the interval in which the stimulus moved upwards. In the conditions in which carrier and envelope moved in opposite directions, observers were instructed to give their overall impression of the direction of motion. No feedback was given. The data points represent the mean of between 40 and 80 trials. 3.3. Temporal frequency discrimination The final experiment of the article briefly examines the general nature of the noise mask in terms of a different temporal task: namely temporal frequency discrimination. This examined the possible general nature of the noise mask compared to the grating mask in order to try and explain the differences in the data obtained for each when the task was direction discrimination. Observers were presented with two intervals and required to indicate via the mouse which interval contained the stimulus that was flickering more slowly. One stimulus was always 5 Hz, the partner stimulus was flickering either faster or slower. 3.4. Subjective equiluminance A measurement of minimal perceptual flicker of a 5 Hz counterphased grating (heterochromatic flicker photometry) at a contrast of 40 times detection threshold was measured and taken as the subjective equiluminant point for each observer and each stimulus configuration. This setting was also checked regularly. The observers had to adjust the luminance angle (as specified in the colour space) until the flicker appeared minimal. The stimulus automatically refreshed to a new random luminance angle after 15 s to avoid prolonged exposure and subsequent habituation being a confounding factor in the measurement. The final step size in the adjustment was 0.32 within the colour space and the mean of 10 estimates was taken for each stimulus. It was verified that both minimum motion and quadrature phase measures gave the same result; data are available in the supplementary material (Figure s1). There was no significant difference between mea-
4.1.1. Masking of motion by dynamic noise The main source of discrepancy in the published data is that under some conditions, chromatic motion perception is degraded by the addition of luminance noise (Baker et al., 1998; Mullen et al., 2003; Yoshizawa et al., 2000; Yoshizawa et al., 2003) whereas under different conditions, there appears to be no interaction between a moving chromatic grating stimulus and a luminance grating mask (Cropper & Derrington, 1996; Zaidi & DeBonet, 2000). Experiment 1 examines the interaction between an AM grating and a noise mask to see whether it is the test structure which dictates whether there is an interaction between chromatic motion and luminance noise. Figs. 2a and b plot data for one observer, TB, across a range of mask contrasts, envelope displacements, and for two envelope modulation depths. Fig. 2a plots the results when a luminance defined AM grating is masked with a dynamic luminance noise mask; Fig. 2b plots equivalent data for a L–M defined chromatic AM grating masked with a dynamic L–M noise mask. The boxed sections of the figure provide a focus for comparing the data across envelope displacements at a near-optimal carrier displacement of 0.25 cycles (Watson, 1990), which is also ambiguous in its direction of motion as far as a mechanism sensitive to unsigned chromatic borders is concerned (Dobkins & Albright, 1993). These values are then replotted in Fig. 4. Initially, it was important to ascertain that AM gratings modulated by 100% and 50% reflected the properties of sparse and dense micropatterns respectively. The left hand panels in Figs. 2a and b plot data for the unmasked luminance carrier modulated by a 100% (upper) and 50% (lower) envelope for observer TB. The envelope motion only has an effect upon the perceived motion of the stimulus in the 100% condition, and reverses the perceived direction completely when displaced by 0.25 cycles. The carrier completely dominates perceived direction in the 50% condition. This reflects the micropattern data but without the need for an additional temporal interval to be introduced, relying only on the spatial modulation for the properties of the perceived motion. The principal characteristics of the masked conditions, shown in the remaining panels of Figs. 2a and b, are as follows. In the luminance masking luminance 100% condition (Fig. 2a, top row), increasing contrast
S.J. Cropper / Vision Research 45 (2005) 865–880
869
Fig. 2. (a) Performance in a direction discrimination task (two interval 2AFC) for an amplitude-modulated (AM) grating; 0.2cpd contrast envelope modulating a 1cpd carrier modulated in luminance. The mask contrast (one-dimensional 1/f luminance noise) is indicated for each panel, which plot the number judged to move in the carrier direction by the observer (TB) against the displacement in cycles of the carrier. Each curve on the graph is a different envelope displacement (in cycles of the envelope modulation) in the opposite direction denoted by the different symbols. The upper panels present data for a AM luminance grating with a carrier 1.0 log units above detection threshold and an envelope modulated by 100%, the lower panels panel plot data for a 50% modulated grating of similar carrier contrast. Each data point is the result of between 40 and 80 observations and error bars plot ± 1SEM. The boxed section draws attention to the 0.25 cycle carrier displacement, which is then replotted in Fig. 4. (b) The effect of masking a chromatic (L–M) AM grating with a dynamic chromatic (L–M) noise mask; data for observer TB. Other details are as Fig. 2a.
870
S.J. Cropper / Vision Research 45 (2005) 865–880
of the noise mask (0.5 and 1.0 log units above threshold) decreases the percept of motion in the carrier direction to chance. Perception of envelope motion is either improved with the reduced carrier percept or left unaffected. In the 50% condition (lower row), carrier motion is again degraded to chance; there is little or no envelope motion in any condition. In the L–M chromatic test and mask condition (Fig. 2b) the only effect of the chromatic mask under any condition is to reduce the envelope percept, leaving the carrier unaffected. This reduced effect of chromatic noise compared to luminance noise on like tests is consistent with published data (Mullen et al., 2003; Yoshizawa et al., 2000). Data for two further observers are available in their full form in the supplementary material (Figures s2a and s2b) and are summarised in Fig. 4. To assess whether chromatic carrier motion is masked by luminance noise, the critical comparison is the one between luminance test and mask conditions and chromatic test and luminance mask combinations. Fig. 3 plots data for observer TB for a chromatic L– M test grating and superimposed luminance noise mask. In the 100% modulated condition, the percept of motion in the carrier direction is sharply and selectively reduced by the addition of a luminance mask. This result is consistent with what is found with the sparse micropattern stimuli and, if considered in isolation, may be taken as evidence against an independent chromatic motion mechanism mediating motion perception in the carrier
direction in the 100% modulated AM grating. The results for the 50% modulated grating (lower row), however, are quite different because there is significantly less masking effect in this stimulus, the structure of which favours perception of motion in the carrier (i.e. first-order spatial modulation) direction. This data is contrary to that found with the dense micropattern. If one adopts exactly the same theoretical grounds as those in which the micropattern work is couched, and uses the distinction between the 100% and 50% modulation depths drawn above, in conjunction with relative carrier and envelope motion strengths, this result is evidence for an independent chromatic motion mechanism whose functioning is relatively unaffected by luminance noise. It is important to remember that this same luminance mask, all but removed carrier motion with a luminance AM test grating (Fig. 2a). Data for the other two observers are presented in the supplemental data (Figure s3) and show the same effect more clearly. The data at a carrier displacement of 0.25 cycles are summarised in Fig. 4 for all three observers for a static envelope (open symbols), and an envelope displacement of 0.25 cycles in the opposite direction to the carrier (closed symbols). The bolder lines indicate the chromatic L–M test combined with a luminance noise mask. The slope of the lines indicates the effect of as the mask contrast increases along the x-axis. The most important comparison for present purposes is between the effect of the luminance mask upon the luminance test (upward
Fig. 3. The effect of combing a chromatic (L–M) test and luminance mask; data for observer TB. Other details are as Fig. 2a and b.
S.J. Cropper / Vision Research 45 (2005) 865–880
871
Fig. 4. Summary data for all three observers, taken from Figs. s2a & b, s2a & b, 3 and s3. The data points are the performance achieved in the direction discrimination task at each noted condition at a carrier displacement of 0.25 cycles. i.e. the highlighted data points in the figures.
oriented triangles) and the effect of the luminance mask on the chromatic L–M test (circles and bold lines). In all but one condition for one observer the effect of the luminance mask on the luminance test is greater than the luminance mask on the chromatic test. Three factors combine to make this result strong evidence for a purely chromatic motion response. Firstly, any luminance signal present in the chromatic grating, will be at a lower effective contrast than the signal produced by a (real) luminance-modulated stimulus. If motion in a chromatic grating were mediated by a weak luminance signal, one would expect it to be more susceptible to masking by a strong luminance mask than would a pure luminance stimulus. However, the data show the converse. Secondly, when one considers that the LM cone modulation required to detect a chromatic L–M grating is far lower than required to detect a luminance grating (Chaparro, Stromeyer, Huang, Kronauer, & Eskew, 1993; Cropper, 1998b; Cropper & Derrington,
1994; Metha, Vingrys, & Badcock, 1994; Switkes & Crognale, 1999) and that the stimuli are scaled to detection threshold, the only reasonable conclusion is that the strength of any luminance component in the chromatic stimulus is very low indeed. Third, luminance masks have a strong contrast dependent effect on direction discrimination of a luminance test (Mullen et al., 2003). Given this dependency, any result but a total destruction of all chromatic motion by a luminance mask is good evidence for the existence of chromatic motion mechanism sensitive to the carrier motion and a suggests that the mask may be having a different, more general, effect than originally proposed. In summary, the results of Experiment 2 support the existence of a motion mechanism selectively sensitive to the chromatically modulated carrier. The moderate differences between observers, which tend to be when the carrier is 100% modulated, in conjunction with the discrepancies already present in the literature, raise
872
S.J. Cropper / Vision Research 45 (2005) 865–880
questions not so much about the underlying mechanisms involved in the motion task but about the nature and specificity of noise and grating masks in the context of the stimulus configurations used to date. This issue is examined in the remaining three experiments, each of which, in addition, consistently provide further support for a purely chromatic motion mechanism. 4.2. Experiment 2 4.2.1. Rigidity of motion and stimulus size both influence the data There are two aspects of the two stimuli, AM gratings and micropatterns, that suggest themselves as reasons for the discrepant findings. One property is the stimulus size and retinal location; the other is the rigidity of the spatiotemporal structure. The micropatterns used in previous work varied in their placement from, on the one hand two 20 · 4 strips above and below fixation (Baker et al., 1998) to, on the other, a single 20 · 4 strip passing through the fovea (Yoshizawa et al., 2000). The grating mask configuration used by Cropper and Derrington (1996) was a centrally placed 8 disc. This raises the possibility that a stimulus that extends significantly outside the fovea may be more susceptible to luminance masking than one which is restricted to the central visual field. Both the anatomy of the retina and the behavioural data give some expectation of this. In addition, the micropatterns were moved rigidly and relied upon the different spatial periodicity of carrier and envelope to provide the different directions of motion of each component. This rigidity is a potent cue to any analysis system that
needs to identify that the moving stimulus is a single rigid body, while signalling different directions of motion at different scales of analysis. To examine the effect of rigidity and size, the spatial configuration of the stimulus was changed from a 20 disk (as in Experiments 1 and 2) to an annulus with 2, 4 and 8 central apertures. This manipulation results in a stimulus which is comparable to the micropatterns used previously which were also presented extrafoveally (Baker et al., 1998; Yoshizawa et al., 2000). As in the previous experiments, the stimulus contrast was rescaled to the individualÕs threshold for each configuration and the subjective equiluminant point was remeasured. All properties of the test patterns were the same as in the previous experiments except that the whole pattern moved rigidly, thereby maintaining the relative spatial phase of carrier and envelope constant throughout the interval. To present the results from this experiment, carrier displacement is plotted against the proportion judged to be moving in the carrier direction. With a 1 cpd carrier and a 0.2 cpd envelope, a carrier displacement range of 0–1 cycle corresponds to an envelope displacement range of 0–0.2 cycles. As a result, the envelope was always moving in one direction but the carrier reversed its motion once it had moved more than half its own period. Thus a cyclic curve of performance against displacement on the figures indicates carrier motion was perceived, whereas a non-cyclic function indicates envelope motion or ambiguous motion was perceived. Fig. 5 shows data for observer TB; data for observers SJC and LK are available in the supplementary information (Figures s4a). As one of the factors of interest
Fig. 5. The effect of a luminance mask of low contrast (0.5 log units above threshold) on the rigid motion of a chromatic L–M AM grating of different stimulus configurations (a disk and three annuli). 100% and 50% modulated gratings shown in left and right hand panels respectively and the observer is TB.
S.J. Cropper / Vision Research 45 (2005) 865–880
in this experiment was the difference between central and peripheral presentation, a low contrast luminance mask (0.5 log units above threshold) was used to accentuate this difference. That is, if an effect of stimulus placement is seen with a low contrast mask, rather than a high contrast mask, then it is even stronger evidence for the importance of its influence in the overall data set. The independence of the chromatic motion percept when presented centrally in the presence of a high contrast mask has already been shown in Experiment 1 and is also further supported by a high contrast mask condition with the different stimulus configurations, data for which is available in the supplementary material (Figure s4b). Each panel of Fig. 5 plots data collected from four stimulus conditions: a 20 disc and three annuli with progressively increasing central apertures with either 100% or 50% envelope modulation depths. The y-axis shows the proportion of stimuli judged to be moving in the direction in which the whole rigid pattern was displaced. The different effects of carrier and envelope motion are clear for the 20 disk (square symbols) for 100% and 50% modulated stimuli. Performance in the 100% modulation condition is above 75% correct and in the ÔpatternÕ direction (performance above 75% correct) except at the smallest carrier displacement. In the 50% condition, however, there is a shift in perception from one direction of motion to the other, which is revealed by the cyclic form of the curve. This indicates that in the 100% condition, the envelope primarily determined the motion percept, whereas in the 50% condition, the carrier determined the overall perceived direction. This result replicates the findings from the previous experiments for non-rigidly moving patterns and is consistent with what has been found with the rigidly moving micropatterns (Baker et al., 1998; Yoshizawa et al., 2000; Yoshizawa et al., 2003). The remaining curves show the effect of progressively enlarging the central aperture on the motion percept, while all other aspects of the stimuli were kept suitably scaled for both sensitivity and subjective equiluminance. The results show that enlarging the aperture has a much greater effect in the 50% conditions than on the 100% conditions. For all observers, the cyclic nature of the motion (carrier dominating) was either completely or mostly eliminated at the largest aperture (20/8 annulus) indicating a much greater effect of the luminance mask when there was no foveal stimulus. These data clearly show that stimulus placement is critical to whether there is an effect of a luminance mask upon chromatic carrier motion perception. Furthermore, the interaction between chromatic test and luminance mask is clearly quite strong in the near periphery, given the low mask contrast used in the experiment. Given this result, one would expect to find
873
some interaction between colour and luminance with micropattern stimuli, all of which extended into the near periphery. The rigid motion structure of the stimulus appears to amplify this interaction. The use of different types of mask for essentially the same purpose, to examine the luminance sensitivity of the mechanism mediating the percept of chromatic motion, is at the root of the discrepant experimental findings investigated in this article. The results of Experiment 1 suggest that noise masks may have quite a general effect upon chromatic stimuli, and may possibly account for both the interobserver differences and for the different conclusions drawn from these experiment and the micropattern experiments. Two obvious questions are: First, how do the effects of grating masks compare to those already measured for noise masks with the same test stimulus? Second, do noise masks act as general temporal masks rather than as masks specifically affecting the underlying motion analysis? Experiments 3 and 4 address these questions.
5. Experiment 3: Masking of motion by gratings The effect of replacing the noise mask with a static grating mask is shown for two of the observers (TB and SJC) in Fig. 6. The test stimuli were otherwise the same as in Experiment 1 except that both a 20 disk and a 4 disk were used. In the plots, the perceived direction of motion of a non-rigidly moving AM grating stimulus is shown as a function of carrier displacement. The mask was a static grating, presented in either 0 or 180 starting phase. The upper panels of Fig. 6 show data for TB; the lower panels show data for SJC. All stimuli were 50% modulated to bias them toward carrier motion and the envelope was static. The left hand panels show data for a 20 configuration (as in Experiment 2) and the right hand panels show data for a small central 4 stimulus configuration. As in previous experiments, all contrasts were scaled to detection threshold for each configuration with the carrier presented 1.0 log unit above threshold and the mask at 1.2 log units above threshold. The plots show the results for and L–M test stimuli and a luminance mask, which is the pertinent condition for the purposes of isolating a purely chromatic motion mechanism. The effect of the grating mask on the perceived motion of the test stimulus is ascertained from the difference in the performance in the two relative phase conditions. Although the theoretical background to this remains controversial (Cropper, 2000; Cropper, 2004a; Cropper & Derrington, 1996; Lu et al., 1999; Lu & Sperling, 1995; Lu & Sperling, 2001; Zaidi & DeBonet, 2000; Zemany, Stromeyer, Chaparro, & Kronauer, 1998) and forms the basis for a complete investigation reported
874
S.J. Cropper / Vision Research 45 (2005) 865–880
Fig. 6. The effect of replacing the noise mask with a static grating mask, either in 0 or 180 starting phase with the carrier. Both mask and carrier have a spatial frequency of 1cpd; the carrier contrast is 1.0 log unit above threshold, the mask is 1.2 log units above threshold. See text for further explanation.
elsewhere (Cropper, 2004b), data pertinent to the current study are presented in Fig. 6. The smaller inset panels at the bottom of the figure show data for SJC under conditions where both test and mask grating are alike (L–M/L–M or Lum/Lum). Two stimulus configurations, 20 (left) and 4 (right) discs are shown. According to one theoretical position (Cropper & Derrington, 1996) (Zemany et al., 1998), under these conditions an interaction between the test and mask would be expected. The data strongly support this position, as the two relative phase conditions cause equal and opposite effects on
the perceived motion; that is, the mask reverses the perceived motion of the test stimulus. If the results in these inset panels are compared with those in the four main panels of Fig. 6, a clear difference is apparent. In the 20 conditions in Fig. 6, the luminance mask disrupted (but did not reverse) motion of the L–M test for Observer TB but not SJC. This is consistent with the pattern of interobserver difference found previously. However, when the size of the stimulus was reduced to 4, neither observer showed an effect of the mask.
S.J. Cropper / Vision Research 45 (2005) 865–880
6. Experiment 4: The noise mask is a general temporal mask The second question posed by the results of Experiments 1 and 2 regarded the generality of dynamic noise masks and their utility in isolating a motion mechanism. This final experiment addressed this issue in the context of the major questions raised in this article. No attempt has been made here to carry out an exhaustive examination of the effect, which will be reported elsewhere. However, evidence showing the general nature of dynamic noise masks is presented in Fig. 7. The figure shows data on the ability of two observers (SL and SP) to discriminate temporal flicker frequency in a chromatic (L–M) stimulus, scaled to detection threshold (1.0 log units above) and subjective equiluminance (measured independently at each temporal frequency). The stimulus was an L–M grating with a superimposed dynamic luminance noise mask. Data for both masked and unmasked stimuli are shown, for two different mask profiles (flat and 1/f) and two different noise contrasts (+0.5 and +1.0 above threshold). Both forms of mask had a strongly detrimental effect upon performance, even at the lower contrast. This result suggests that the noise mask is an effective general temporal mask and is not
875
specific to motion extraction––even if flicker and motion do have common coding components (Levinson & Sekuler, 1975; Ramachandran & Gregory, 1978; Watson, Thompson, Murphy, & Nachmias, 1980). Furthermore, the result could well explain the lack of any masking effect for detection of a chromatic stimulus compared to the strong effect obtained for discriminating the motion of the same stimulus (Baker et al., 1998; Yoshizawa et al., 2000). The difference between detection and discrimination has often been cited as evidence for the specificity, and consequent utility, of the noise mask (Mullen et al., 2003), but, this may be more accurately cast in terms of a spatial, as compared to a temporal, judgement. This, then, is further support for the view that data obtained with grating masks provide a clearer picture of the underlying mechanisms than data obtained with noise masks and strengthens the case for an independent chromatic motion mechanism.
7. General discussion The experiments in this article address the issue of whether we possess a motion mechanism sensitive to the displacement of purely chromatic stimuli.
Fig. 7. The effect of dynamic luminance noise on temporal (flicker) frequency discrimination. The proportion judged to be flickering slower than the 5Hz reference is plotted against the temporal frequency of the test. The luminance noise mask was updated at 25Hz.
876
S.J. Cropper / Vision Research 45 (2005) 865–880
Specifically, the effects of luminance and chromatic masks on the perception of motion for luminance and chromatic stimuli were measured in an attempt to resolve a disagreement in the published literature. AM grating stimuli were used because their critical properties lie in between those of the micropattern and grating stimuli that have been previously used in the literature. With these intermediate stimuli it was found that both the presence and absence of a masking effect could be obtained from a single test/mask stimulus combination. Overall, the results show that a motion response to a chromatic stimulus can be obtained that is independent of any detrimental effect of a luminance mask. However, this independence was only evident within a restricted range of spatial extent and location of the stimulus. Clear and categorical independence was only shown when a specific grating mask was used. The size dependence of the effect, in otherwise scaled and standardised stimuli, is very strong support for a result uncontaminated by luminance artefacts. Previous studies using gratings with superimposed grating masks found evidence for a motion mechanism that is sensitive to chromatic input independent of any luminance signal (Cropper & Derrington, 1996). However, when micropatterns were masked with noise masks in a combined motion paradigm, no such evidence was found (Baker et al., 1998; Mullen et al., 2003; Yoshizawa et al., 2000; Yoshizawa et al., 2003). The AM stimuli used in these experiments have properties similar to both these stimulus types. The carrier grating acts like an unmodulated grating when the envelope is static, and particularly so when it is only 50% modulated; such stimuli act like dense micropattern stimuli. The displaced envelope of the AM grating acts very like the envelope of the individual micropattern elements when modulating the carrier by 100%; overall this stimulus was comparable to the sparse micropattern stimuli. Furthermore, the results show that while envelope and carrier motions can be selectively disrupted by the addition of a dynamic noise mask of the same colour and luminance properties, this effect does not necessarily transfer when the test and mask are composed of colour and luminance respectively; nor do colour and luminance noise have comparable effects within their own domain. When the stimulus is located centrally and constructed to optimise the conditions necessary to reveal a chromatic motion mechanism, there is virtually no detrimental effect of a luminance noise mask upon the detection of motion in a purely chromatic carrier. Two potentially important differences between gratings and micropatterns that may account for the difference in results are the rigid motion structure and the stimulus placement. Although the results of Experiment 2 showed that both factors influenced the effect of a luminance mask on a chromatic test, the critical factor is the size and location of the stimulus. The data suggest
that, in order to isolate a mechanism sensitive to the motion of purely chromatic motion, independent of any luminance signal, the stimulus must be located foveally and be small in size. When one considers the structure of the retina in relation to the distribution of different photoreceptor subtypes, their potential function, and the relation of this function to their dendritic fields and wiring characteristics, this is not a surprising result (Boycott & Dowling, 1969; Boycott & Wa¨ssle, 1991; Dacey & Petersen, 1992; Leventhal, Rodiek, & Dreher, 1981; Perry & Cowey, 1981, 1984; Perry & Cowey, 1985; Perry, Oehler, & Cowey, 1984; Rodieck, 1965, 1973; Rodieck & Rushton, 1976; Wa¨ssle, Boycott, & Illing, 1981; Wa¨ssle, Gru¨nert, Ro¨hrenbeck, & Boycott, 1990; Wa¨ssle, Peichl, & Boycott, 1981). Furthermore, the foveal specialisation of colour vision lends strong psychophysical support to this position (Anderson, Mullen, & Hess, 1991; Metha et al., 1994; Mullen, 1985; Mullen, 1987; Mullen, 1991). When the stimulus motion has a rigid spatiotemporal structure, the effect of stimulus size is accentuated and the differences already seen between the micropattern and the grating data are even more apparent. The rigid stimulus structure imposes a phase-coherence on the envelope and carrier of the stimulus that is consistent with the spatiotemporal properties of a moving rigid object. This coherence has also been shown to be important in the identification and localisation of borders across spatial scale in an image (Hayes, 1989, 1990, 1994, 1998) and appears to be having a similar influence upon the system when conflicting motion signals are processed in order to provide some overall percept (Morrone & Burr, 1997). This observation makes some sense of the effect of stimulus rigidity shown in Experiment 2 where, like the micropatterns, there appeared to be a much stronger disparity between the percept of either carrier motion or envelope motion. This apparent dominance of one or other signal may assist the system in the grouping of several different motion signals to identify a single moving object. This observation raises the point that a careful examination of the spatiotemporal structure of a test stimulus is critical in determining the utility of that stimulus as a tool for the dissociation of interacting, underlying processes. There are two final methodological points worthy of mention; the mean luminance of the display and the format of the noise mask. With regard to the mean luminance of the stimulus display, the two main bodies of research examined and reconciled here were carried out using very different mean luminance levels. The grating studies used relatively high mean luminance levels, similar to that used in the present study (45–90 cd/m2) whereas the micropattern, and related studies, used a lower mean luminance level (6.4 cd/m2). Given the high potential for residual rod activity at this low mean luminance, a point raised in explanation of related work
S.J. Cropper / Vision Research 45 (2005) 865–880
(Lindsey & Teller, 1990), and the lower sensitivity of specialised chromatic mechanisms at very low light levels (Walkey, Barbour, Harlow, & Makous, 2001), it is not surprising that there are greater effects of luminance masks at lower mean luminance. The second issue concerns the spatial frequency content of the noise mask. The micropattern work used a flat-spectrum noise mask (binary) and a Butterworth filter to reduce high spatial frequency components in the case of the chromatic noise; this compares to the 1/f noise mask used here. While there are a host of theoretical and ecological reasons to use a 1/f noise mask (Brady & Field, 1995; Field, 1987, 1994), a particularly important reason for the current work is that the two mask types predict different effects on a model L–M receptive field. Typical chromatically-opponent L–M receptive fields predict a response to low spatial frequency chromatic modulation and high spatial frequency luminance modulation (Schiller & Colby, 1983). This suggests that the spatial frequency profile of the masking noise may be critical in determining the nature of the masking effect; an observation supported by recent data (Cropper, 2004b; Cropper & Johnston, 2001). Since the interaction of interest here was whether there was a specific common sensitivity to chromatic and luminance motion signals (at the same spatial frequency, or at least within an octave-band), the use of a flat spatial frequency profile noise mask may compromise the clarity of the collected data simply due to the additional high frequencies present in the noise. This is particularly problematic when attempting to compare noise data with grating data. While this issue is partly addressed by a comparison of detection and motion discrimination thresholds (Baker et al., 1998; Mullen et al., 2003; Yoshizawa et al., 2000; Yoshizawa et al., 2003) this approach does not really dissect the tasks in terms of the spatiotemporal structure of the stimulus, nor does it take into account the different demands placed upon the system by detection judgements and motion discrimination judgements. A point explicitly brought out by the data of Experiment 4 and further bolstered by the intermediate effect of low contrast independence seen when the mask is a jittering grating (Stromeyer et al., 1996). 7.1. Parity with previous data—the stimuli It was argued in relation to Experiment 1 that AM gratings provide a suitable intermediate stimulus to examine why micropatterns and gratings yield different effects of masking of a displaced chromatic test with a luminance mask. Although the principal similarity between dense and sparse micropatterns on the one hand, and 50% and 100% modulated AM gratings on the other, is in their spatial structure, the reversal of the perceived motion of the micropattern stimulus was de-
877
pended partly on the temporal interval between successive frames. This manipulation creates a temporal structure in which motion is more likely to be detected by a comparatively high level process. If the temporal continuity of the stimulus is interrupted with an interval of zero contrast energy, any mechanism sampling the spatiotemporal signal relatively frequently will have a noisier output. So whilst a given mechanism may actually be sensitive to the stimulus motion, its performance will be compromised by the change in temporal profile. As a limiting case, is has been shown that if the interstimulus interval is greater than the temporal response function for luminance signals, motion perception cannot be mediated by a first-order sensitive motion mechanism (Stromeyer, 2003; Stromeyer & Martini, 2003). Furthermore, the response function is not of a set critical period but changes with stimulus contrast. A Ôfeature-sensitiveÕ mechanism, however, should not be affected by this property of the system because, arguably, the critical moving features of the stimulus remain unperturbed by the inter-stimulus interval. So, as long as the feature sensitive mechanism does not rely on the coherence of the signal over time to elicit a response but rather operates on some level of explicit successive spatial analysis of the input, there should be no affect of an interstimulus interval upon its output. The micropattern stimuli specifically exploit these properties of the system to bias the stimuli toward detection by low-level motion mechanisms in the dense condition and highlevel feature mechanisms in the sparse condition. Given the nature of the low-level chromatic motion mechanism and its relative obscurity compared to the luminance sensitive subsystem, it is not surprising that targeting different motion mechanisms in this way misses the somewhat more subtle influence of low-level chromatic motion signals. If the temporal structure of the stimulus is maintained, and the spatial structure alone is used to elicit differences in motion properties, as is the case with the stimuli used in this article, the result is a gradual change from one perceived direction of motion to the other, that is, from the carrier to the envelope percept. This observation indicates that the conflicting motion signals are both accessible in the stimulus under both dense and sparse conditions, a claim supported by results obtained here using a range of envelope displacements. The advantage of the less extreme approach to the stimulus manipulation used here is that it allows the effect of the spatial modulation of the stimulus on the underlying motion mechanisms to be examined. This result is particularly useful when put in the context of recent studies on first-order and second-order stimuli. Strictly defined, the first-order versus second-order (or higher order) distinction explicitly limits the stimulus description to the spatial domain and relates that description to the temporal structure and its motion. In the current context,
878
S.J. Cropper / Vision Research 45 (2005) 865–880
this may be interpreted as maximising the likelihood that some low level mechanism sensitive to chromatic motion may be revealed, even though it is generally accepted that such a mechanism, if it exists, is less efficient than an equivalent luminance mechanism, if considered in terms of the strength of motion percept rather than its contrast sensitivity (Chaparro et al., 1993; Cropper & Derrington, 1994). In conjunction with the time-critical nature of the first-order response, particularly where differential L and M cone responses are concerned (Cropper & Derrington, 1994; Derrington et al., 1993; Stromeyer, Kronauer, Ryu, Chaparro, & Eskew, 1995; Stromeyer, Chaparro, Tolias, & Kronauer, 1997), AM stimuli such as those used here are more likely to reveal the properties of any first-order colour-sensitive motion mechanisms, which, presumably, co-exist in the same spatiotemporal region as any luminance sensitive mechanism. The reduced effect of dynamic chromatic noise seen in the results presented here, and in the micropattern and associated studies (Mullen et al., 2003; Yoshizawa et al., 2000; Yoshizawa et al., 2003) is consistent with the powerful general temporal masking effect of the dynamic luminance noise found in Experiment 5. The temporal contrast sensitivity function of chromatic stimuli compared to luminance stimuli (Kelly, 1974; Kelly, 1975, 1977, 1983; Mullen, 1985), upon which the theory of flicker photometry is based, would predict that chromatic noise would have a reduced effect, as compared to luminance noise, only if the principal effect of the dynamic noise mask was a general temporal disturbance rather than a specific degradation of the motion signal. 7.2. The persistent artefact The common criticism levelled at data that shows chromatic stimuli as having properties consistent with their motion being detected by low-level mechanisms is that a luminance artefact, whether internal or external in its origin, is mediating the performance. Even ignoring the clear stimulus size dependent nature of any masking effect, this explanation for the current data is implausible for several reasons. The maximum size of the artefact cannot exceed the contrast of the chromatic stimulus itself, even with the most optimistic of estimates (Mullen et al., 2003). The luminance and chromatic stimuli, both tests and masks used here were scaled to detection threshold, so any effect of the luminance mask upon the test must have been greater for the luminance test that the chromatic test. When considered in conjunction with the known contrast dependence of the masking effect (Mullen et al., 2003), and the significantly lower cone modulation at threshold produced by an L–M stimulus (Chaparro et al., 1993; Cropper & Derrington, 1994; Switkes & Crognale, 1999), it is unlikely that any luminance artefact would even approach
detection threshold. As a gross estimate, the highest possible luminance artefact (17 deviation from photometric equiluminance for recorded for observer SP in Figure s1) would elicit an L + M cone modulation 0.23 times its own detection threshold in a chromatic L–M stimulus ten times its own detection threshold (Chaparro et al., 1993; Cropper & Derrington, 1994; Cropper et al., 1996). This makes it a most unlikely mediator of performance even if the data did support its existence, which they do not. It is also the case that the majority of the data arguing against the existence of a chromatic motion mechanism have relied on the different effects of the noise mask on the tasks of detection and discrimination. The results of Experiment 4 show that the noise mask is a powerful general temporal mask and so weakens the argument that it has a specific effect on motion discrimination mechanisms. It is possible that the temporal properties of the suggested artefact (Mullen et al., 2003) are more closely related to the psychophysical judgement than the stimulus itself. Most importantly, however, any argument for a luminance artefact, wherever it may be generated, must explain why there are multiple examples of performance, as clearly shown here, where perception of motion in a chromatic stimulus is immune to a luminance mask which, in turn, will destroy luminance coded motion.
8. Conclusions The experiments reported here isolated a motion mechanism that is sensitive to purely chromatic stimuli and which is unaffected by luminance masking. This isolation, however, was possible only under optimal stimulus conditions. Specifically, the stimulus needed to be located in the central 4 of the visual field. Outside this area the influence of the purely chromatically sensitive motion mechanism is quickly diluted by signals from other, less specific, mechanisms that, while sensitive to chromatic motion, seem also to be sensitive to luminance and luminance-contrast modulation. The properties of the purely chromatically sensitive mechanism indicate that it operates very similarly to a luminance sensitive mechanism and independently of any higher-level feature-based extraction of the motion signal.
Acknowledgments I wish to thank Patrick Cavanagh and Charles Stromeyer for their significant input to this work. I also thank Bradley Wolfgang and Phillip Smith for their comments and Brad for Fig. 1. This work was supported by grants from the Australian Research Council and the Welcome Trust over the past 5 years.
S.J. Cropper / Vision Research 45 (2005) 865–880
Supplementary material Supplementary data associated with this article can be found, in the online version, at doi:10.1016/ j.visres.2004.09.043.
References Anderson, S. J., Mullen, K. T., & Hess, R. F. (1991). Human peripheral spatial resolution for achromatic and chromatic stimuli: Limits imposed by optical and retinal factors. Journal of Physiology (London), 442, 47–64. Anstis, S. M., & Cavanagh, P. (1983). A minimum motion technique for judging equiluminance. In J. D. Mollon & L. T. Sharpe (Eds.), Colour vision: physiology and psychophysics (pp. 154–166). London: Academic Press. Baker, C. L., Jr., Boulton, J. C., & Mullen, K. T. (1998). A nonlinear chromatic motion mechanism. Vision Research, 38(2), 291–302. Boulton, J. C., & Baker, C. L. (1993a). Different parameters control motion perception above and below a critical density. Vision Research, 33, 1803–1811. Boulton, J. C., & Baker, C. L. Jr., (1993b). Dependence on stimulus onset asynchrony in apparent motion: Evidence for two mechanisms. Vision Research, 33, 2013–2019. Boycott, B. B., & Dowling, J. E. (1969). Organization of primate retina. Philosophical Transactions of the Royal Society series B, 255, 109–184. Boycott, B. B., & Wa¨ssle, H. (1991). Morphological classification of bipolar cells of the primate retina. European Journal of Neuroscience, 3, 1069–1088. Brady, N., & Field, D. J. (1995). WhatÕs constant in contrast constancy. The effects of scaling on the percieved contrast of bandpass patterns. Vision Research, 35, 739–756. Cavanagh, P., & Mather, G. (1989). Motion: The long and short of it. Spatial Vision, 4, 103–129. Chaparro, A., Stromeyer, C. F., III, Huang, E. P., Kronauer, R. E., & Eskew, R. T. Jr., (1993). Colour is what the eye sees best. Nature, 361, 348–350. Cropper, S. J. (1998a). Combining the motion of first-order and second-order, chromatic and luminance stimuli: A purely local process? Perception, 27, ECVP Abstracts. Cropper, S. J. (1998b). The detection of luminance and chromatic contrast modulation by the visual system. Journal of the Optical Society of America A, 14(8), 1969–1986. Cropper, S. J. (2000). Independent chromatic motion is limited to the fovea. Investigative Ophthalmology and Visual Science, 41(4). Cropper, S. J. (2004a). Colour and motion––masking u¨ber alles. Journal of Vision, (VSS abstracts). Cropper, S. J. (2004b). The detection of motion in chromatic stimuli: Pedestals and masks. Vision Research, submitted. Cropper, S. J., & Derrington, A. M. (1994). Motion of chromatic stimuli: First-order or second-order? Vision Research, 34(1), 49– 58. Cropper, S. J., & Derrington, A. M. (1996). Rapid colour-specific detection of motion in human vision. Nature, 379(Jan 4), 72–74. Cropper, S. J., & Johnston, A. (2001). The motion of contrast envelopes: Peace and noise. Journal of the Optical Society of America A, 18(9), 2237–2254. Cropper, S. J., Mullen, K. T., & Badcock, D. R. (1996). Motion coherence across cardinal axes. Vision Research, 36(16), 2475–2488. Dacey, D. M., & Petersen, M. R., (1992). Dendritic field size and morphology of midget and parasol ganglion cells of the human retina. In: Proceedings of the National Academy of Sciences of the USA, 89 (pp. 96660–99670).
879
Derrington, A. M., Badcock, D. R., & Henning, G. B. (1993). Discriminating the direction of second-order motion at short stimulus durations. Vision Research, 33(13), 1785–1794. Derrington, A. M., & Henning, G. B. (1981). Pattern discrimination with flickering stimuli. Vision Research, 21, 597–602. Derrington, A. M., Krauskopf, J., & Lennie, P. (1984). Chromatic mechanisms in lateral geniculate nucleus of macaque. Journal of Physiology (London), 357, 241–265. Dobkins, K. R., & Albright, T. (1993). What happens if it changes color when it moves? Psychophysical experiments on the nature of chromatic input to motion detectors. Vision Research, 33, 1019–1036. Field, D. J. (1987). Relations between the statistics of natural images and the response properties of cortical cells. Journal of the Optical Society of America A, 4, 2379–2394. Field, D. J. (1994). What is the goal of sensory coding? Neural Computation, 6, 559–601. Findlay, J. M. (1978). Estimates on probability functions: A more virulent pest. Perception and Psychophysics, 23, 181–185. Julesz, B. (1971). Foundations of cyclopean perception. Chicago, IL: University of Chicago Press. Kelly, D. H. (1974). Spatio-temporal frequency characteristics of color-vision mechanisms. Journal of the Optical Society of America A, 64, 983–990. Kelly, D. H. (1975). Luminous and chromatic flickering patterns have opposite effects. Science, 188, 371–372. Kelly, D. H. (1977). Visual contrast sensitivity. Optica Acta, 24(2), 107–129. Kelly, D. H. (1983). Spatio-temporal variation of chromatic and achromatic contrast thresholds. Journal of the Optical Society of America, 73, 742–750. Klein, S. A., Hu, Q. J., & Carney, T. (1996). The adjacent pixel nonlinearity: Problems and solutions. Vision Research, 36, 3167–3181. Leventhal, A. G., Rodiek, R. W., & Dreher, B. (1981). Retinal ganglion cell classes in the old world monkey: Morphology and central projections. Science, 213, 1139–1142. Levinson, E., & Sekuler, R. (1975). The independence of channels in human vision selective for direction of movement. Journal of Physiology (London), 250, 347–366. Lindsey, D. T., & Teller, D. Y. (1990). Motion at isoluminance: Discrimination/detection ratios for moving isoluminant gratings. Vision Research, 30(11), 1751–1761. Livingstone, M. S., & Hubel, D. H. (1987). Psychophysical evidence for separate channels for the perception of form, color, movement, and depth. Journal of Neuroscience, 7, 3416–3468. Lu, Z.-L., Lesmes, L. A., & Sperling, G. (1999). The mechanism of isoluminant chromatic motion. In: Proceedings of the National Academy of Sciences of the USA, 96 (pp. 8289–8294). Lu, Z.-L., & Sperling, G. (1995). The functional architecture of human visual motion perception. Vision Research, 35(19), 2697–2722. Lu, Z.-L., & Sperling, G. (2001). Three systems theory of human visual motion perception: Review and update. Journal of the Optical Society of America A, 18(9), 2331–2370. Marr, D. (1982). Vision: A computational investigation into the human representation and processing of visual information. San Francisco: W.H. Freeman and Company. Metha, A. B., Vingrys, A. J., & Badcock, D. R. (1994). Detection and discrimination of moving stimuli: The effects of color, luminance and eccentricity. Journal of the Optical Society of America A, 11(4), 724–734. Morrone, M. C., & Burr, D. C. (1997). Capture and transparency in coarse quantized images. Vision Research, 37(18), 2609–2629. Mullen, K. T. (1985). The contrast sensitivity of human color vision to red/green and blue/yellow chromatic gratings. Journal of Physiology (London), 359. Mullen, K. T. (1987). Spatial influences on colour opponent contributions to pattern detection. Vision Research, 27(5), 829–840.
880
S.J. Cropper / Vision Research 45 (2005) 865–880
Mullen, K. T. (1991). Colour vision as a post-receptoral specialization of the central visual field. Vision Research, 31, 119–130. Mullen, K. T., Yoshizawa, T., & Baker, C. L. Jr., (2003). Luminance based mechanisms mediate the motion perception of ‘‘isoluminant’’ chromatic gratings; the role of ‘‘temporal chromatic aberration’’. Vision Research, 43. Perry, V. H., & Cowey, A. (1981). The morphological correlates of xand y- like retinal ganglion cells in the retina of monkeys. Experimental Brain Research, 43, 226–228. Perry, V. H., & Cowey, A. (1984). Retinal ganglion cells that project to the superior colliculus and pretectum in the macaque monkey. Neuroscience, 12(4), 1125–1137. Perry, V. H., & Cowey, A. (1985). The ganglion cell and cone distributions in the monkeyÕs retina: Implications for central magnification factors. Vision Research, 25, 1795–1810. Perry, V. H., Oehler, R., & Cowey, A. (1984). Retinal ganglion cells that project to the dorsal lateral geniculate nucleus in the macaque monkey. Neuroscience, 12, 1101–1123. Ramachandran, V. S., & Gregory, R. L. (1978). Does colour provide an input to human motion perception? Nature, 275, 55–56. Rodieck, R. W. (1965). Quantitative analysis of cat retinal ganglion cell responses to visual stimuli. Vision Research, 5, 583–601. Rodieck, R. W. (1973). The vertebrate retina. San Fransisco: W.H. Freeman. Rodieck, R. W., & Rushton, W. A. H. (1976). Isolation of rod and cone contributions to cat ganglion cells by a method of light exchange. Journal of Physiology (London), 254, 759–773. Schiller, P. H., & Colby, C. L. (1983). The responses of single cells in the lateral genicualte nucleus to color and luminance contrast. Vision Research, 23, 1631–1641. Stromeyer, C. F. (2003). Temporal assymetry in motion masking: A shortening of the temporal impulse response. Vision Research, 43(3), 261–268. Stromeyer, C. F., Chaparro, A., & Kronauer, R. E. (1996). The colour and motion of moving patterns are processed independently. Investigative Ophthalmology and Visual Science, 37(Suppl), s916. Stromeyer, C. F., III, Kronauer, R. E., Ryu, A., Chaparro, A., & Eskew, R. T. Jr., (1995). Contributions of human long-wave and middle-wave cones to motion detection. Journal of Physiology (London), 485, 221–243. Stromeyer, C. F., & Martini, P. (2003). Human temporal impulse response speeds up with increased stimulus contrast. Vision Research, 43(3), 285–298. Stromeyer, C. F. I., Chaparro, A., Tolias, A., & Kronauer, R. E. (1997). Colour adaptation modifies the long-wave versus middlewave cone weights and temporal phases in human luminance (but
not red-green) mechanism. Journal of Physiology (London), 499(1), 227–254. Switkes, E., & Crognale, M. A. (1999). Comparison of color and luminance contrast: Apples vs. Oranges? Vision Research, 39, 1893–1906. Wagner, G., & Boynton, R. M. (1972). A comparison of four methods of heterochromatic photometry. Journal of the Optical Society of America, 62(12), 1508–11515. Walkey, H. C., Barbour, J. L., Harlow, J. A., & Makous, W. (2001). Measurements of chromatic sensitivity in the mesopic range. Color research and application, 26(supplement), 36–42. Wa¨ssle, H., Boycott, B. B., & Illing, R. B., (1981). Morphology and mosaic of on- and off- beta cells in the cat retina and some functional considerations. In: Proceedings of the Royal Society of London B, B 212 (pp. 177–195). Wa¨ssle, H., Gru¨nert, U., Ro¨hrenbeck, J., & Boycott, B. B. (1990). Retinal ganglion cell density and cortical magnification factor in the primate. Vision Research, 30(11), 1897–1911. Wa¨ssle, H., Peichl, L., & Boycott, B. B. (1981). Morphology and topography of on- and off-alpha cells in the cat retina. In: Proceedings of the Royal Society of London B, B 212 (pp. 157– 175). Watson, A. B. (1990). Optimal displacement in apparent motion and quadrature models of motion sensing. Vision Research, 30(9), 1389–1393. Watson, A. B., Thompson, P. G., Murphy, B. J., & Nachmias, J. (1980). Summation and discrimination of gratings moving in opposite directions. Vision Research, 20, 341–347. Wilson, H. R., Ferrera, V. P., & Yo, C. (1992). A psychophysically motivated model for two-dimensional motion perception. Visual Neuroscience, 9, 79–97. Yoshizawa, T., Mullen, K. T., & Baker, C. L. Jr., (2000). Absence of a chromatic linear motion mechanism in human vision. Vision Research, 40, 1993–2010. Yoshizawa, T., Mullen, K. T., & Baker, C. L. Jr., (2003). Failure of signed chromatic apparent motion with luminance masking. Vision Research, 43, 751–759. Zaidi, Q., & DeBonet, J. S. (2000). Motion energy versus position tracking: Spatial, temporal and chromatic parameters. Vision Research, 40, 3613–3635. Zeki, S. M. (1978). Functional specialization in the visual cortex of the rhesus monkey. Nature, 274, 423–428. Zemany, L., Stromeyer, C. F. S., III, Chaparro, A., & Kronauer, R. E. (1998). Motion detection on flashed, stationary pedestal gratings: Evidence for an opponent-motion mechanism. Vision Research, 38(6), 795–812.