Isolation of opponent-colour mechanisms at increment threshold

Vision Res.Vol.27,No.6,pp. 1017-1027. 1987 0042-6989/87 S3.00+0.00 Riotedin zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA G re a t Britain. All rights reserved CopyrightQ 1987 Pergamon Journals Ltd ISOLATION OF OPPONENT-COLOUR MECHANISMS AT INCREMENT THRESHOLD* ROSEMARY S. SNELGAR,DAVID H. FOSTERand MARK 0. SCASE Department of Communication and Neuroscience, University of Keele, Keele. Staffordshire ST5 SBG, England (Received 2 January1986; in revised fotm 18 August 1986) Ahatraet-An experimental examination was made of some paradigms designed to isolate the opponentcolour system at increment threshold. The effectiveness of a uniform white conditioning field spatially coincident with a l.OS-deg uniform test field was amessed by measuring intensity thresholds for simple detection and for colour discrimination. Values were obtained both by a method of adjustment and by a two-interval famed-choice procedure. For sutBciently high lmninances of the conditioning field (3000 td or greater) little or no difference was found between simple-detection and colour-discrimination thresholds over the critical test-flash spectral range 520-620nm. implying that the paradigm produced almost complete isolation of the opponent-colour system at increment threshold. A control experiment in which thresholds were. obtained for a conditioning field larger than the test field gave less satisfactory isolation; near 58Onm the luminance system was found to be at least 0.3 log unit more sensitive than the opponent-colour system. A comparison was also made of the spatially coincident field paradigm with a paradigm in which a modified test stimulus of low temporal and spatial frequency content was presented on a large conditioning field. Test spectral sensitivity curves for simple detection obtained by a method of adjustment showed little diflerence in etfectiveness in opponent-colour isolation. Gpponentcolour system Luminance system Spectral sensitivity Spectral sharpening tamed for the subjects who took part in the present study (methods as in Foster and SnelThere are three characteristic peaks at approxi- gar, 1983a, and as in Experiments 1 and 4 mately 440, 530, and 610 nm in the spectral below). Evidence suggesting that the peaks at sensitivity curve obtained by incrementabout 530 and 610 nm result from activity in the threshold measurements of a long-duration, cir- red-green opponent-colour channel of an cular, monochromatic test flash presented on a opponent-process system has been reviewed in large white conditioning field. The three peaks Foster and Snelgar (1983a). A specific assump have been demonstrated for the human eye in tion is that sensitivity of the non-opponent many studies (e.g. Stiles and Crawford, 1933; luminance system is depressed as a result of Sidley and Sperling, 1967; Sperling and Har- achromatic adaptation produced by the white werth, 1971; King-Smith and Carden, 1976; conditioning field (King-Smith and Kranda, Verriest and Uvijls, 1977; Harwerth and Levi, 1981). More detailed descriptions of post1977; Kuyk, 1982; Snelgar and Foster, 1982; receptoral colour processing have been offered Foster and Snelgar, 1983a; Zrenner, 1983). Ex- by Mollon (1982), Ingling and Martinezamples of such test spectral sensitivity curves are Uriegas (1983), and Hood and Finkelstein shown in Fig. 1 (open circles); they were ob- (1983). The dip at about 580 run in the curves shown in Fig. 1 (open circles) is also a feature of the test *Some of the data reported here were contained in a paper spectral sensitivity curve obtained under the presented to The Colour Group (Great Britain), January 1985, and in a Communication ptemnted to the conditions summarized above. The fact that the Physiological Society (Foster, Scase and Snelgar, 1986). dip is relatively shallow under those conditions Correspondence should be addressed to Dr D. H. is probably due to intrusion by the luminance Foster, Department of Communication and Neurosystem in that region of the spectrum (Kingscience, University of Keele, Keele, Staffordshire, ST5 Smith and Carden, 1976). It is, however, possi5BG, England. INTBODUCTION 1017 ROSEMARY S. SNELGARet al. zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQ 1018 Wa ve le ng th 400 I1 500 (nm l 600 700 1 white. Their assumption was that thresholds for colour detection are determined by and reflect the sensitivity of the opponent-colour system. This assumption is made here also. For a colour-detection criterion the test spectral sensitivity curve has a deep trough at about 580 nm (King-Smith and Carden, 1976), presumably representing a state of equilibrium in the redgreen opponent-colour channel. (The position of the equilibrium point need not be fixed at 580 nm; see Discussion.) Thresholds for colour detection may, however, be difficult to obtain, unless two-interval forced-choice discrimination methods are used. (2) Spatially coincident auxiliary field The test flash used in traditional measurements of test spectral sensitivity (some listed above) is usually bounded by sharp edges, both 15000 20000 25000 spatially and temporally. Such a stimulus, even Wa ve num b e r km - ‘) presented on a large white conditioning field, is unlikely to be optimal for measurements of Fig. 1. Test spectral sensitivity curves obtained for simple detection of a 1.05&g test flash presented on a large opponent-colour function for, in general, the (lO.Odeg) white auxiliary field (open symbols, broken line) luminance system is more sensitive than is the and on a small (l.OSdeg) white auxiliary field spatially opponent-colour system to the high spatial and coincident with the test field (solid symbols, continuous temporal frequencies present in the stimulus. line). Log reciprocal intensity of the test flash is plotted (Evidence in support of this assertion has been against wavenumber 1-r of the test flash. Thresholds were obtained by a method of adjustment. The white auxiliary reviewed in Foster, 1981; see also Stromeyer et 5eld gave retinal illuminance 3000 td and had colour temal., 1978, and Mullen, 1985.) This and the next perature 3400 K. The test flash had duration 200 msec. The class of paradigm exploit these differential sensipairs of curves, displaced downwards by 1.0 log unit tivities in different ways. aucocssively, arc for di5erent subjects as indicated. Each If the large white conditioning field is made point is the mean of six readings and the vertical bars show f 1 SEM where su5%zimtlylarge. spatially coincident with the test field then the high spatial-frequency adaptation or masking that occurs at the boundary of the stimulus is ble to depress further the sensitivity of the thought to depress further the sensitivity of the luminance system relative to that of the luminance system (Snelgar and Foster, 1982; Foster and Snelgar, 1983a; see Foster, 1981, for opponent-colour system by manipulating the experimental paradigm according to the discussion of this paradigm in measures of field spectral sensitivity). The conditioning field spad&rential sensitivities of the two systems. The tially coincident with the test field is referred to activity of the opponent-colour system may then be revealed more completely: the dip at as the small auxiliary field, and the large condiabout 580 nm becomes a deeper trough and the tioning field as the large auxiliary field, following the notation introduced by W. S. Stiles. Test peaks at about 530 and 610 nm become sharper and more clearly defined. Three classes of ex- spectral sensitivity curves for simple detection obtained with a small auxiliary field are shown perimental paradigms have been described that in Fig. 1 (solid circles) for the subjects who took secure or facilitate detection by the opponentcolour system. The critical features of these part in the present study (methods as in Foster paradigms were as follows. and Snelgar, 1983a, and as in Experiments 1 and 4 below). In comparison with the test spectral ( 1) zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA Colour - discrimination criterion sensitivity curve obtained with the large auxilAs well as determining threshold by the usual iary field (open circles), there was a marked criterion of simple detection of the test flash, increase in the depth of the trough at about King-Smith and Carden (1976) used a criterion 580 nm and an improvement in the delineation of colour detection or discrimination from of the peaks at about 530 and 610 nm with the Opponent-colour mechanisms at threshold small auxiliary field (solid circles). The latter curves are similar to those obtained by KingSmith and Carden (1976) using a colourdetection criterion and a large white auxiliary field. (3) Low -frequency test stimuli Some approaches to the isolation of the opponent-colour system have employed test stimuli with modified spatial and temporal profiles that eliminate the spatial and temporal transients at the stimulus edges. The “lowfrequency” test stimulus of Thornton and Pugh (1983a,b) had bell-shaped spatial and temporal distributions which were designed so that the Fourier spectra had most energy concentrated at low frequencies. The test spectral sensitivity curves obtained with this low-frequency test stimulus presented on a large white auxiliary field (Thornton and Pugh, 1983a) were similar to those in Fig. 1 for the small auxiliary field (solid circles). If the extent and periodicity of the stimulus are increased the relative power at low frequencies may be still further enhanced. Thus Mullen (1987) used monochromatic sinusoidal gratings superimposed on a large white auxiliary field; for spatial frequencies less than 1 deg-’ and a temporal frequency of 0.8 Hz the spectral sensitivity curves obtained for contrast sensitivity were also similar to the curves in Fig. 1 for the small auxiliary field (solid circles). Evidence was presented, however, which led to the conclusion that the opponent-colour system did not contribute to detection at 577 nm, even though sensitivity of the luminance system was depressed in that region. There were two main objectives in undertaking the present experiments. The first was to assess whether use of a small white auxiliary field did fully isolate the opponent-colour system at increment threshold. This was decided operationally, by reference to the corresponding test spectral sensitivity curve for detection of colour. As noted below, this approach, although safe, may have been rather conservative. Nevertheless, of all the above-cited studies describing opponent-colour spectral sensitivities, only King-Smith and Carden (1976) reported direct comparisons with colour detection data. Because wavelength-independent variations in sensitivity were not of interest here, adjustment of the relative vertical positions of simpledetection and colour-detection curves were allowed to achieve maximum overlap on a logaVII 27,6-l. 1019 rithmic scale, under the constraint that at no wavelength should simple-detection threshold exceed colour-detection threshold. Coincidence of the curves implied isolation of the opponentcolour system over the spectral range studied; departures from coincidence, most evident in the region about 580 run, gave a direct measure of the amount by which the luminance system was more sensitive at threshold than the opponent-colour system over that portion of the spectrum. This estimate was conservative for although perception of colour necessarily entailed activity in the opponent-colour system, the converse need not have applied. The second objective was to compare directly the small-auxiliary-field paradigm with the lowfrequency-test-stimulus paradigm. The two are in some respects complementary, but comparisons of test spectral sensitivities obtained by the two methods have not previously been made in the same subjects. In fact, under the same conditions, little difference in effectiveness was found between the two paradigms. In initial measurements, described in Experiment 1 below, thresholds were obtained by a method of adjustment. This traditional method provides an accurate and efficient procedure for obtaining simple-detection thresholds, but, for colour-detection judgements, threshold settings are more difficult to make, and a two-interval forced-choice (ZIFC) method was therefore introduced. A range of luminance levels of the auxiliary field was tested in these measurements to determine whether there was a limiting effect in the depression of the sensitivity of the luminance system relative to the opponent-colour system. Sperling et al. (1967) and Harwerth and Levi (1977) showed that the dip at about 580 nm obtained for a simple-detection criterion with a large white auxiliary field deepened with increasing luminance of that field, reaching a maximum in depth for ltinances of 10,000 or 3000Td respectively. It seemed plausible that in&easing the luminance of the small auxiliary field would also decrease sensitivity of the luminance system, offering the possiblity that the trough at about 580 nm obtained for simple detection would eventually converge on the trough obtained for colour detection. EXPERIMENT 1 Simple-detection and Colour-detection Thresholrls by Method of Adjustment As explained in the Introduction, one aim of 1020 ROSEMARY S. SNELGARet al. the present study was to assess the effectiveness of the small auxiliary field in isolating the opponent-colour system at threshold for simple detection. In this exploratory experiment, simple-detection and colour-detection thresholds were measured on the small auxiliary field over a range of wavelengths spanning the critical region of the trough at about 580 nm and over a range of luminances of the small auxiliary field corresponding to 300, 1000, 3000, 10,000, and 30,000 td. Instead of absolute judgements of stimulus colour in the colour-detection measurements, a reference white was introduced and the test stimulus judged against that. Formally the task was therefore one of colour discrimination. Methoa3 Stimuli and apparatus. The monochromatic test flash was circular, of dia. 1.05 deg, duration 2OOmsec, and wavelength ranging from 520 to 623 nm. In the colourdiscrimination experiments a white comparison flash identical in size, shape, and duration to the monochromatic test flash was also used; its colour temperature was adjusted with colourcorrecting filters to be close to that of the auxiliary field. The steady small white auxiliary field was circular, of dia. 1.05 deg, and of varying luminance, with colour temperature 3400 K. The monochromatic test flash and the white comparison flash were each presented concentrically on the small auxiliary field either separately or as a test sequence with an interstimulus interval of 1 sec. The stimuli were produced by a three-channel Maxwellian-view optical system (described fully in Foster, 1981) with as light source a single tungsten-halogen lamp run from a regulated d.c. power supply. Transmittances of the channels were controlled by compensated neutral density wedges and Wratten neutral density filters (NDFs). The spectral compositions of the monochromatic test fields, produced by channel 2, were controlled with Balzers I340 interference filters with half-height full-bandwidths each less than 9 nm. The comparison white flash was produced by channel 1. The time-courses of the stimulus flashes (monochromatic and white) were controlled by silent electromagnetic shutters that interrupted the relevant channels at intermediate foci. Rise and fall times of each flash were each less than 2 msec. The white auxiliary field, produced by channel 3, could be set to give a retinal illuminance of 300, 1000, 3000, 10,000 or 30,000 td. The settings were obtained by means of minimally distinct border matches (Wagner and Boynton, 1972) against a 562 nm reference field. The spatial geometry of the stimuli was determined by photographic masks inserted in the channels. Precautions were taken to minimize instrumental stray light. The stimuli were viewed through a 2-mm artificial pupil with an achromatizing lens and correcting lenses for non-emmetropic subjects. Stability of head position was maintained with the aid of a dental bite-bar. Further details of the apparatus and its calibration are given in Foster (1981). Subject. One subject, R.S.S. (co-author) participated in this experiment. She had normal colour vision and corrected Snellen acuity of 6/5, and was aged 31 yr. Procedure. At the beginning of each experimental session, the positions of the masks were adjusted so that the stimulus fields appeared concentric and in focus. Where necessary, small corrections to the alignments of the masks were made as the wavelength of the test flash was varied. The subject dark-adapted for 10 min before commencing the observation session. Viewing was monocular and the subject fixated the centre of the small auxiliary field with the right eye. The intensity of the test flash and its onset were controlled by the subject using pushbutton switch-boxes. Stimuli were not presented more rapidly than once every 2 sec. Measurements of simple-detection and colourdiscrimination thresholds for the monochromatic test flash were obtained by a method of adjustment, for 10 wavelengths ranging from 520 to 615 nm. Simple-detection thresholds were obtained as described in Foster and Snelgar (1983a). In each experimental session, measurements were made with test-flash wavelengths in ascending and then descending order for one auxiliary-field luminance. In the same session, detection threshold for the white comparison flash was determined for use in subsequent measurements of colour-discrimination threshold. Colour-discrimination thresholds were set as follows. In each experimental session, auxiliary-field luminance was fixed. The neutral density wedges in channels 1 and 2 were first set by the experimenter so that the white comparison flash and the monochromatic test flash were each at detection threshold as determined previously. Next, the wedge in channel 1 was electronically yoked to that in channel 2 so that the push-button control box varied the intensities of the white comparison flash and the Opponcntcolour mechanisms at threshold 1021 zyxwvutsrqpon for simple detection and open symbols for colour discrimination. The vertical bars show f 1 SEM where sufhciently large. The pairs of curves (each pair displaced successively down300 Td wards by 0.5 log unit) are for increasing luminance of the small auxiliary field as indicated. 1000 Td For simple detection the trough at about 580 nm became progressively deeper with increasing auxiliary-field luminance. The trough 3000 Td for colour discrimination, however, remained substantially the same over auxiliary-field lumi10000 Td nances of 300-10,000 td. (At 30,000 td the relative sensitivity for colour discrimination at short 30000 Td and long wavelengths was different from that at lower luminances, and also the SEMs were larger than at any other luminance.) M Simple detection To quantify these changes in trough depth, o-0 Colour detection zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA the difference between mean sensitivity over 577 zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHG and 582nm and over the regions 520-566 and -13." 593-623 nm was calculated for simple detection Wavenumber km-‘) and for colour discrimination at each auxiliaryFig. 2. Test spectral sensitivity curves obtained for simple field luminance from 300 to 30,000 td. For detection (solid symbols) and for colour discrimination simple detection, there was a monotonic in(open symbols). The l.OS-deg test tlash had duration crease in trough depth relative to that at 300 td 200 mscc and was presented on a small white auxiliary field at each successive luminance level: values were spatially coincident with the test field. Thresholds were +0.06, +0.12, +0.19, and +0.26log unit as obtained by a method of adjustment. The pairs of curves, dispJaad downwards successively by 0.5 log unit, are for luminance increased from 1000 td. For colour different luminances of the small auxiliary field as indicated. discrimination, the differences in trough depth Each point is the mean of six readings and the vertical bars relative to that at 300 td were essentially conshow f 1 SEM where sufficiently large. Subject: R.S.S. stant: values were + 0.07, - 0.10, - 0.02, and Other details as for Fig. 1. - 0.17 log unit (where there was a change in curve shape) as luminance increased from monochromatic test flash in unison (the gra1000 td. The difference between trough depths dients of the two wedges being almost identical). for simple detection and colour discrimination The two flashes were presented as a test se- decreased with increase in auxiliary-field lumiquence: the monochromatic test flash followed nance: values were + 0.42, + 0.43, + 0.19, by the white comparison flash with an inter+ 0.20, and - 0.01 log unit as luminance instimulus interval of 1 sec. The subject, using a creased from 300 to 30,000 td (a positive value method of adjustment, then set the yoked indicating a less deep trough for simple dewedges so that the monochromatic test flash tection). could just be distinguished from the white comparison flash on the basis of colour. This procedure was repeated at each wavelength of the Comment For luminances of the small auxiliary field of monochromatic test flash. Wavelength was varied in ascending and then descending order (or 30,900 td isolation of the opponent-colour sysvice-versa). For both simple-detection and tem was effectively complete. For auxiliary-field colour-discrimination measurements, three conluminances of 3000 and 10,000 td a small resisecutive threshold settings were recorded at each dual difference (of the order of 0.2 log unit) wavelength on each traverse. Mean threshold between simple-detection and colourdiscriminvalues were thus based on six measurements. ation thresholds was evident at wavelengths close to 580 nm. Thresholds for colour discrimResults ination are, however, notoriously difficult to set Results are shown in Fig. 2. Sensitivities in by a method of adjustment, even for an experilog quanta. set-’ deg-* are plotted against testenced observer. For subsequent experiments flash wavenumber. Solid symbols indicate data involving colourdiscrimination thresholds a Wavelength 400 500 (nm) 600 700 1 1022 ROSEM ARYS. SNELGARer al. array of four tiny white lights. Otherwise, the apparatus was as in Experiment 1. For the measurement of luminance dependence, the wavelength of the monochromatic test flash was EXPERIMENT2 zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA fixed at 582 nm, and the luminance of the small Simple- detection and Colour- discrimination auxiliary field was varied over the range by 21FC M ethod 300-30400 td. Other conditions were as in In contrast to the procedure followed in Experiment 1. Subjects. R.S.S. and M.O.S. (co-author) took Experiment 1, simple-detection and colourdiscrimination thresholds were measured here in part in this experiment. M.O.S. was male, had normal colour vision and corrected Snellen acuthe same session to offset day-to-day variations in sensitivity; blocks of the two types of mea- ity of 615, and was aged 22 yr. Procedure. Data for only one test-flash wavesurement were interleaved to offset ordering and carry-over effects within sessions. Differences in length or for only one auxiliary-field luminance were collected in a session. Subjects fixated with thresholds were then computed. A 2IFC method for determining threshold was used both as a the right eye the centre of the small auxiliary control on Experiment 1 and to obtain data field or, for the large auxiliary field, the centre from a less practised observer. There were two of the fixation display. At the beginning of the parts to this experiment. The first part experimental session, six preliminary meaquantified the difference between simplesurements of simple-detection threshold were detection and colour-discrimination thresholds made by a method of adjustment first for the over the same spectral region as explored in monochromatic test ffash and then for the white Experiment 1. It was expected that any residual comparison flash. The wedges in channels 1 and differences would be greatest in the region about 2 were then set to the mean positions thus 580 nm. A fixed auxiliary-field luminance of obtained and remained fixed for the remainder 3000 td was used, for this gave reasonably good of the experiment. The intensity of the stimulus isolation of the opponentcolour system, as flashes was now controlled by the experimenter Experiment 1 showed, without being uncomusing NDFs inserted at a point common to fortably bright. Thresholds were determined on channels 1 and 2. Intensities were altered the small auxiliary field by two subjects, and, to pseudorandomly over the range approximately provide a quantitative control on the effects of - 0.5 to + 0.5 log unit in steps of approxiauxiliary-field size (see Fig. 1), thresholds were mately 0.1 log unit. In the 2IFC determinations also determined by one of the subjects on the of colour-discrimination threshold, one interval large auxiliary field. The second part of the contained the monochromatic test flash and the experiment quantified the effect on opponentother interval contained the white comparison colour isolation of varying the luminance of flash; the interstimulus interval was 1 sec. Orthe small auxiliary field. Simple-detection and dering of the intervals was pseudorandom. The colour-discrimination thresholds were detersubject initiated the test sequence on an audible mined at 582 nm over a range of auxiliary-field cue, and indicated (forced-choice) whether the luminances. monochromatic test flash occurred in the first or second interval. In the 21FC determinations of M ethods simple-detection threshold, the procedure was Stimuli and apparatus. For the measurement similar except that one interval contained the of wavelength dependence, the monochromatic monochromatic test flash and the other interval test flash and the white comparison flash were was empty. The subject indicated the interval in as in Experiment 1 except that the wavelength which the flash occurred. Each threshold was of the monochromatic test flash was varied over based on 500 trials comprising five blocks of 100 the range 520-615 nm in rather larger steps. The trials, each block consisting of ten trials at each small auxiliary field was also as in Experiment of ten stimulus intensities. Blocks of colour1, except that its luminance was fixed at 3000 td. discrimination trials were alternated with blocks The large auxiliary field was circular of dia. of simple-detection trials. An experimental session lasted approximately 3.5 hr. lOdeg, and had the same colour temperature and luminance as the small auxiliary field. When the large auxiliary field was in use stimuli were Results Results for wavelength dependence are shown centred at the middle of a 3-deg, square, fixation two-interval forced-choice therefore used. (ZIFC) method was Opponent-colour mechanisms at threshold Wavelength 400 :: f = ii 0.0 I 500 (ntnl 600 700 1023 tection at 582 nm is plotted as a function of auxiliary-field luminance. Differences are again shown positive if more light was required for colour discrimination. There is an evident monotonic decrease in threshold difference with increasing auxiliary-field luminance. For R.S.S. [Fig. 4(a)] the difference was reduced from 0.38 log unit at 300 td to 0.14 log unit at 3000 td and at 10,000 td, and to - 0.01 log unit at 30,000 td. For M.O.S. the difference was reduced from 0.40 log unit at 300 td, to 0.01 log unit at 10,000 td and to 0.09 log unit at 30,000 td. _____________________~.__.... t- Comment q -‘.” Wavenumber [cm-‘) Fig. 3. (a) Difference in sensitivity for simple detection and for colour discrimination of a 1.05&g test gash presented on a large white auxiliary field (open symbols) or on a small white auxiliary field spatially coincident with the test field (solid symbols). Difference in log intensities of the test gash is plotted against wavenumber 1 -I of the test gash. Thresholds were obtained by a two-interval forced-choice procedure. Each point shows the difference between the thresholds each of which was based on 500 trials and the vertical bars shows f 1 SEM where sufhciently large. Subject: R.S.S. Other details as for Fig. 1. (b) Subject: M.O.S. Measurements were made for the small auxiliary geld only. Other details as for (a). in Fig. 3. The difference between threshold for colour discrimination and that for simple detection is plotted as a function of test-flash wavenumber; if more light was required for colour discrimination then the result is shown positive. Solid symbols show differences in threshold on the small auxiliary field and open symbols on the large auxiliary field [Fig. 3(a) only]. Vertical bars show f 1 SEM where sufficiently large. For subject R.S.S. [Fig. 3(a)] the differences between simple-detection and colour-discrimination thresholds on either auxiliary field did not exceed + 0.04 or - 0.06 log units at all wavelengths except 582 nm where the difference was 0.13 and 0.37 log unit on the small and large auxiliary fields respectively. For. subject M.O.S. [Fig. 3(b)], who made measurements on the small auxiliary field only, the difference between the two thresholds at 582 nm was 0.08 log unit, and not more than 0.06 log unit elsewhere. Results for luminance dependence are shown in Fig. 4. The difference between threshold for colour discrimination and that for simple de- These forced-choice measurements confirmed the preliminary conclusions of Experiment 1. First, a large white conditioning field was not sufficient to isolate the opponent-colour system over the whole trough region. Some indication of the greater sensitivity of the luminance system at 582 nm is given by the difference, 0.37 log unit, between colourdiscrimination and simpledetection thresholds obtained by subject R.S.S. on the large field. The small auxiliary field was clearly more effective in isolating the opponentcolour system: at 582mn the difference between colour-discrimination and simple-detection thresholds was 0.13 log unit for R.S.S., and 0.08 log unit for M.O.S. Second, almost complete isolation of the opponentcolour system can evidently be zyxwvutsrq l.Or ?- $ 9 ‘; 0.0 ~, -----------+----.__ _---- ___-_____________ Y m E z i 3 s (a) , RSS, -1.0 l.o;O , , , , 1000 10000 100000 1000 10000 100000 0.0 k z ii -1.0 100 Auxiliary-field luminance zyxwvutsrqponmlkjihg (Td) Fig. 4. (a) Difference in sensitivity for simple detection and colour discrimination for a 1.05deg test gash of wavelength 582nm presented on a small auxiliary geld of various luminances. Each point shows the dikence in log units between threaholds value each of which was based on 500 trials; the vertical bars show f 1 SEM where sufhciently large. Subject: R.S.S. Other details as for Fig. 3. (b) Subjecn M.O.S. Other details as (a). ROSEMARY S. SNEffiAR ef al. 1024 achieved at simple-detection threshold by use of a sufhciently intense auxiliary field spatially coincident with the test field. If there is a limiting effect in the action of this auxiliary field it would seem to occur at 10,000 or 30,000 td. For a large auxiliary field, a limiting effect may be at 3000 td (Harwerth and Levi, 1977) or 10,000 td (Sperling et al., 1967). EXPERIMENT 3 Comparison with Low-frequency-test-stimulus Thresholds In this experiment, the small auxiliary-field paradigm was compared with the lowfrequency-test-stimulus paradigm of Thornton and Pugh (1983a). Simple-detection thresholds for a low-frequency test stimulus were determined by a method of adjustment, with wavelengths spanning most of the visible spectrum. Experiments 1 and 2 showed that the small auxiliary field could produce almost compkte isolation of the opponent-colour system; the results obtained there were accordingly used to assess the effectiveness of the low-frequency test stimulus. Methods Stimuli and apparatus. The low-frequency *The spatial profile of the low-frequency test stimulus was obtained by placing a photographic mask at an appropriate position in the channel to give a defocussed image. This position was determined and the image quality assessed by means of a linear photodiode system mounted on a travelling microscope placed after the exit pupil. The profile of the sharp l.OS-deg stimulus was also measured by the same method and, with this as a reference, the mask giving rise to the low-frequency stimulus was adjusted so that the stimulus had a half-height full-width of approximately 1 deg. Its profile was sampled at 14 points along a diameter of the field. The data were analysed by means of a generalized linear interactive modelling technique GLIM (Baker and Nelder, 1978) and the profile was found to be well fitted by a Gaussian function (proportion of variance accounted for was 95.2%). The temporal profile of the low-frequency test stimulus was controlled by two polaroid filters placed in the channel, the one static and the other rotated through 180” by a stepping motor controlled by a microcomputer. The speed of rotation of the stepping motor was set so that the half-height full-width of the time-varying intensity was 200ms. Light output was sampled at 24 points over its time-course. The data were analysed by means of GLIM and the temporal profile was also found to be well fitted by a Gaussian function (proportion of variance accounted for was 96.8%). monochromatic test flash had the spatial profile of a radially symmetric Gaussian with halfheight full-width of 1.Odeg, and the temporal profile of a Gaussian with half-height full-width of 200 ms*. The white comparison flash was not used. Other stimuli were as described above. The additional interference filters used had characteristics as described in Experiment 1, except for a 461~nm filter which had half-height full-bandwidth of 14 nm. Subjects. R.S.S., M.O.S., D.H.F. (co-author), and W.F.B. took part in this experiment. D.H.F. and W.F.B. were male with normal colour vision. D.H.F. had corrected Snellen acuity of 614, and was aged 40yr. W.F.B. had corrected Snellen acuity of 6/6 and was aged 34yr. A.R.R. who made measurements for the sharp test flash only on the small and large auxiliary field (Fig. 1) was female, had normal colour vision and corrected Snellen acuity of 615, and was aged 22 yr. Procedure. Threshold was measured for the monochromatic low-frequency test stimulus, with the spatial and temporal profiles described above, presented on a large white auxiliary field. Threshold was also measured for the monochromatic test flash with sharp-edged spatial and temporal profiles presented on a spatially coincident white auxiliary field. Both auxiliary fields were of luminance 3000 td. Subjects made these measurements by a method of adjustment for simple detection only. For each stimulus paradigm, measurements were made in two sessions, the one with ascending and the other with descending order of wavelengths. Other details of the measurement procedure were as in Experiment 1. Results and Comment Results are shown in Fig. 5. Sensitivity in log quanta. set-’ sdeg-’ is plotted against test-flash wavenumber. Solid symbols show thresholds obtained with the sharp stimulus on the small auxiliary field, replotted from Fig. 1, and open symbols show thresholds obtained with the lowfrequency stimulus on the large auxiliary field. It can be seen that, although relative sensitivity to the two types of test stimulus varied from subject to subject (possibly because of different relative threshold criteria), the shapes of the two test spectral sensitivity curves particularly in the region about 580 nm were broadly similar. For each subject, the difference in sensitivity between 582 nm and the mean over the regions 520-562 and 601615 nm was calculated. The Opponentcolour mechanisms at threshold 1025 at auxiliajr-field luminances of values 3000-10,000 td, where the curves were very 400 500 600 700 I 1 IIll I close to those obtained for colour discrimi1 zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDC nation in the same experiment. In contrast, for colour discrimination, the trough in spectral sensitivity remained fairly stable over all luminances of the auxiliary field. The concurrence of detection and discrimination troughs at auxiliary-field huninances of 3000 td or higher provided strong support for the hypothesis that the small auxiliary field secured isolation of the opponentcolour system at simple-detection threshold. h -12.0This hypothesis was further supported by C zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA the two-interval forced-choice measurements of simple-detection and colour-discrimination threshold in Experiment 2. The difference beO-0 +, tween the two thresholds was greatest at -14.0 , I 582 nm, and this difference was reduced almost I I I 15000 20000 25000 to zero for auxiliary-field luminances of Wavenumber (cm-‘) lO,OOO-30,000td. It was also evident that use of a large auxiliary field with a spatially and temFig. 5. Test spectral sensitivity curves obtained for simple porally sharp-edged test flash did not secure detection of the previously used sharp-edged test Bash presented on a small white auxiliary field spatiey coincomplete isolation of the opponent-colour syscident with the test field (solid symbols, continuous line, tem, failing by 0.37 log unit at 582nm for a replotted from Fig. 1) and for simple detection of a lowluminance of 3000 td. frequency test stimulus presented on a large white auxiliary These results confirmed previous suggestions field (open symbols, broken tine). The low-frequency test of Foster and Snelgar (1983a) concerning the stimulus had Gaussian spatial and temporal pro&s with half-height full-widths of 1 deg and 200 msec rqectively. effectiveness of a small auxiliary field in reThresholds were set by a method of adjustment. The pairs vealing opponent-colour function at simpleof curves, displaced downwards by 1.Olog tit successively, detection threshold. There, as in Experiment 1 are for different subjects as indicated. Each point is the here, a method of adjustment was used. It may mean of six readings and the vertical bars show f 1 SEM be noted that although colourdiscrimination where sticiently large. Other detailsas for Fig. 1. thresholds provide, in principle, a direct measure of opponent-colour sensitivity, the preferred procedure entails some 21FC method, differences in trough depth for W.F.B., D.H.F., which, in turn, requires many more preM.O.S., and R.S.S. were respectively -0.02, sentations of the stimulus. The 21FC thresholds + 0.09, -0.11, and - 0.04 log unit, where a obtained in Experiment 2 were based on 500 positive result corresponds to a less deep trough trials each, which may be compared with about for the low-frequency test stimulus. This 60 trials with the method of adjustment. difference was significant for both D.H.F. and Experiment 3 compared the effectiveness of M.O.S. (respectively t = 2.27 and - 2.25, the small-auxiliary-field paradigm with the lowd.f. = 30, each P < 0.05, two-tailed), but the frequency-test-stimulus paradigm. For all four differences were in opposite directions and nusubjects, there was essentially no difference bemerically small. Overall, it was concluded that tween the two sets of results in the region of the the two paradigms were equally effective in troughs near 580 nm: differences in depths were revealing activity of the opponent-colour system. 0.11 log unit or less. The supposed mechanisms of isolation were, however, different in the two DI!XU!Z3ION cases: the small auxiliary field actively caused high spatial-frequency adaptation or masking In Experiment 1 simple-detection threshold of the luminance system, whereas the lowwas measured by a method of adjustment. The frequency test stimulus provided little or no trough in spectral sensitivity at about 580 nm stimulus transients to which the luminance sysprogressively deepened with increasing tem might respond. auxiliary-field luminance and reached its lowest Wavelength (nml 1 I I 1026 ROSEMARY s. SNELO.UterzyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONML al. tern. Thus, a test flash which is small, or brief, The paradigm of an auxiliary field spatially coincident with the test field may be applied in or both (properties which should favour detection by the luminance system) may under a variety of investigations of opponent-colour processes where complete or near-complete iso- suitable conditions give rise to field spectral sensitivity data that show shifted and sharpened lation is desirable, both in the fovea and at other funclocalized retinal sites. Most of the studies in peaks characteristic of opponent-colour tion (Foster, 198 1; Finkelstein and Hood, 198 1, which a large auxiliary field was employed, 1982). Moreover, the shape of the spectral sensome cited in the Introduction, used central system, presentation of the test flash. Three peaks in the sitivity curve of the opponent-colour determined by either test or field variation, need test spectral sensitivity curve have been shown not be fixed, and will alter if the equilibrium at stimulus eccentricities of 1.5” (Harwerth and points of the constituent opponent-colour chanLevi, 1977), 5” (Stiles and Crawford, 1933), and nels are shifted by chromatic adaptation. Thus 6” (Verriest and Uvijls, 1977). For the more for the red-green channel, the deep trough in the peripheral retina, however, it seemed that the peaks at 530 and 610 nm were much less promtest spectral sensitivity curve and the cross-over inent (Wooten et zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA al., 1975; Verriest and Uvijls, point of the sharpened long- and medium1977). Instead, a broad maximum was usually wavelength field spectral sensitivity curves obfound with peak sensitivity at about 550 nm and tained by Foster and Snelgar (1983a) both it was generally assumed that opponent-colour occurred at about 580 nm because the small processes do not operate or operate weakly on auxiliary field used in those experiments was signals from receptors situated in the peripheral white. If the chromaticity of the auxiliary field, retina. Interestingly, Kuyk (1982) found that the either large or small, is changed, however, then three peaks in the test spectral sensitivity curve the position of the trough in the test spectral could be obtained in the peripheral retina if the sensitivity curve may also change (Sperling and test field was made sufficiently large (e.g. 5.5 deg Harwerth, 1971; Thornton and Pugh, 1983b; of visual angle at 45” eccentricity). The paraFoster and Snelgar, 1986). A parallel effect digm of the small auxiliary field, developed by occurs in the cross-over point of the sharpened Foster (1979) for measurements of field spectral long- and medium-wavelength field spectral sensensitivity (see below), was also used by Krastel sitivity curves, a result which suggests a quasiet al. (1983, 1984) who showed that a threeinvariance of opponent-colour processes (Foster peaked test spectral sensitivity curve, with a and Snelgar, 1986). deep trough at about 580 nm, could be obtained In conclusion, the present results show that a in the peripheral retina if the test and auxiliary small auxiliary field spatially coincident with the fields were large (e.g. 16 deg of visual angle at test field provides an effective device for depress33” eccentricity). ing the sensitivity of the luminance system and A small auxiliary field was first used to reveal revealing opponent-colour activity. It has, activity of the opponent-colour system in meamoreover, the advantage that opponent-colour sures of field spectral sensitivity with either function may be explored over small regions of a monochromatic (Foster, 1979, 1980, 1981; the visual field, a property which may have Foster and Snelgar, 1983b) or white auxiliary particular significance in the investigation of field (Foster and Snelgar, 1983a). An almost retinal and optic-nerve pathologies (Foster ef exactly analogous paradigm, exploiting the sen- al., 1985). sitivity of the luminance system to zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFED temporal Acknowledgemenrr-We are grateful to R. Knapper for transients, has also been used (Finkelstein and technical assistance and to A. R. Roberts for making Hood 198 1, 1982). In this method a test tlash preliminary observations. We thank S. R. Pratt for critical was presented on a steady, large white auxiliary reading of the manuscript. Support was provided by field and a flashed large main field, the onset of project grants from the Medical Research Council and which was temporally coincident with that of from the Multiple Sclerosis Society of Great Britain and Northern Ireland. the test flash. A single field spectral sensitivity curve was obtained with peaks at 600nm and REFERENCES 540 nm, and a dip at 580 nm. Raker R. J. and Nelder J. A. (1978) The GLIM system. It should be noted that, in general, the impoNumerical Algorithms Group, Oxford. sition of a particular stimulus property or set of Finkelstein M. A. and Hood D. C. (1981) Cone system properties need not always ensure isolation of saturation: more than one stage of sensitivity loss. Vision the opponent-colour system or luminance sysRes. 21, 319-328. Gpponent-colour mechanisms at threshold Finkelstein M. A. and Hood D. C. (1982) Opponent-color cells can influence detection of small, brief lights. Vision Res. 22, 89-95. Foster D. H. (1979) El&t of a small blue adapting field on the spectral sensitivity of the red-sensitive colour mechanism of the human eye. Oprica Acra 26, 293-296. Foster D. H. (1980) Gpponency of red- and green-sensitive cone mechanisms in field spectral sensitivity measurements. J. Physiol., L.ond. 298, 2 1P. Foster D. H. (1981) Changes in field spectral sensitivities of red-, green- and blue-sensitive colour mechanisms obtained on small background fields. Vision Res. 21, 1433-1455. Foster D. H., Scase M. 0. and Snelgar R. S. (1986) Functional isolation of normal human opponent-colour processes at increment threshold. J. Physiol., Lord 377, 44P. Foster D. H. and Snelgar R. S. (1983a) Test and field spectral sensitivities of colour mechanisms obtained on small white backgrounds: action of unitary opponentcolour processes ? Vision Res. 23,181-791. Foster D. H. and Snelgar R. S. (1983b) Initial analysis of opponent-colour interactions revealed in sharpened field spectral sensitivities. In Colow Vision (edited by Mellon J. D. and Sharpe L. T.), pp. 303-311. Academic Press, New York. Foster D. H. and Snelgar R. S. (1986) Quasi-independence of opponent-colour processes under chromatic adaptation. 1. Physiol., L.ond. 377, 42P. Foster D. H., Snelgar R. S. and Heron J. R. (1985) Nonselective losses in fovea1 chromatic and luminance sensitivity in multiple sclerosis. Inaesr. Ophthd. visd Sci. 26, 1431-1441. Hanverth R. S. and Levi D. M. (1977) Increment threshold spectral sensitivity in anisometric amblyopia. Vision Res. 17, 585-590. Hood D. C. and Finkelstein M. A. (1983) A case for the revision of textbook models of color vision: the detection and appearance of small brief lights. In Colour Vision (edited by Mollon J. D. and Sharpe L. T.), pp. 385-398. Academic Press, New York. lngling C. R. Jr and Martinex-Uriegas E. (1983) The relationship between spectral sensitivity and spatial sensitivity for the primate r-g X-channel. Vision Res. 23, 1495-1500. King-Smith P. E. and Carden D. (1976) Luminance and opponent-color contributions to visual detection and adaptation and to temporal and spatial integration. J. opt. Sot. Am. 66, 109-111. King-Smith P. E. and Kranda K. (1981) Photopic adaptation in the red-green spectral range. Vision Res. 21, 565-572. Krastel H., Jaeger W. and Braun S. (1983) An incmnentthreshold evaluation of mechanisms underlying colour constancy. In Colour Vision (editedby Mollon J. D. and 1027 Sharpe L. T.), pp. 545-551. Academic Press, New York. Krastel H., Jaeger W. and Braun S. (1984) The contribution of spectral increment thresholds to the interpretation of color perimetry. Devl Ophthal. 9, 171-181. Kuyk T. K. (1982) Spectral sensitivity of the peripheral retina to large and small stimtdi. Vtiion Res. 2% 1293-1297. Mellon J. D. (1982) Color vision. Ann. Reu. Psycho/. 33, 41-85. Mullen zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONML K . T. (1985) The contrast sensitivity of human colour vision to red-green and blue-yellow chromatic gratings. J. Physiol., Land. %9, 381+Kb Mullen K. T. (1987) Spatial influences on colour opponent contributions to pattern detection. Vision Res. 27, 829-839. Sidley N. A. and Sperling H. G. (1967) Photopic spectral sensitivity in the Rhesus monkey. J. opt. Sot. Am. 57, 816-818. Snelgar R. S. and Foster D. H. (1982) Isolation of opponent-colour channels. Presented at: London Meeting of the Experimental Psychology Society, January, 1982. Sperling H. G. and Hanverth R. S. (1971) Red-green cone interactions in the increment-threshold spectral sensitivity of primates. Science, N.Y. 172, 180-184. Sperling H. G., King V. and Schoessler J. (1967) Incrementthreshold spectral sensitivity as a function of the intensity of a white background field. J. opt. Sot. Am. sl, 1424A. Stiles W. S. and Crawford B. H. (1933) The liminal brightness increment as a function of wave-length for different conditions of the fovea1 and parafoveal retina. Proc. R. Sot. B113,496-530. Stromeyer C. F. III, Kranda K. and Sternheim C. E. (1978) Selective chromatic adaptation at different spatial frequencies. Vision Res. 1% 421-431. Thornton J. E. and Pugh E. N. Jr (1983a) Relationship of opponent colours cancellation measures to coneantagonistic signals deduced from increment threshold data. In Colour Virion (edited by Mollon J. D. and Sharpe L. T.), pp. 361-373. Academic Press, New York. Thornton J. E. and Pugh E. N. Jr (1983b) Red/green color opponency at detection threshold. Science, N.Y. 219, 191-193. Verriest G. and Uvijls A. (1977) Central and peripheral increment thresholds for white and spectral lights on a white background in different kinds of congenitally defective colour vision. Atti. Fond. Giorgio Ronchi 32, 213-254. Wagner G. and Boynton R. M. (1972) A comparison of four methods of heterochromatic photometry. J. opt. Sot. Am. 62, 1508-1515. Wooten B. R., Ftdd K. and Spillman L. (1975) Photopic spectral sensitivity of the peripheral retina. J. opr. Sot. Am. 65, 334-342. Zrenner E. (1983) Neurophysiological Aspects of Color Vision in Primates Springer, Berlin.