Discrimination and identification of luminance contrast stimuli

Abstract

The goal of this study was to compare luminance contrast discrimination and polarity identification in the inferred Parvocellular (PC-) and Magnocellular (MC-) pathways. The position identification task tested ability to locate a contrast change within a stimulus. The polarity identification task tested ability to classify the contrast change as either brighter or darker. Three paradigms were employed to find these thresholds for the inferred MC- and PC-pathways: Pulsed-Pedestal, Steady-Pedestal and Pedestal-Δ-Pedestal. Position and polarity identification thresholds were the same for the Pulsed-Pedestal stimuli (inferred PC-pathway). The position identification thresholds were lower than the polarity identification thresholds for Steady-Pedestal stimuli (inferred MC-pathway). The position identification thresholds were lower than the polarity identification thresholds for Pedestal-Δ-Pedestal stimuli when the Δ-Pedestal contrast steps were sub-threshold or close to threshold. When the contrast steps were larger, position identification and polarity identification thresholds were similar for the Pedestal-Δ-Pedestal stimuli. Comparisons were also made between position identification and polarity identification thresholds for a short and a long stimulus presentation. There were no systematic differences dependent on presentation duration.

Introduction

Thomas (1985) provided a conceptual framework relating detection (the ability to distinguish a change within a stimulus) and identification (the ability to classify a change within a stimulus as one of a number of alternatives), postulating two sequential stages. The first stage is the encoding stage; the physical stimulus activates tuned pathways and is represented within the visual system according to the pathway responses. This stage should be the same for both detection and identification because the tasks use the same sensory information in the form of the encoded signal. The second stage is the decision stage, during which a judgment is made based on the combination of the pathway responses with respect to a decision rule. This stage should be different for the tasks because different decision rules are applied to the encoded signal (Thomas, 1985; Watson & Robson, 1981). Noise limits the accuracy of detection and identification due to random fluctuations inherent in the pathways. 

The identification task is often described in terms of “labeled lines”. These are tuned pathways that signal that a specific property of interest is present in a stimulus, e.g. spatial frequency or contrast. On- and Off-pathways are examples of labeled lines. Retinal ganglion cells in On-pathways increment their firing rate to increments in contrast; cells in Off-pathways increment their firing rate to decrements in contrast. For the detection task, we may apply an inclusive-or rule, which means that detection occurs if at least one pathway is activated (King-Smith & Carden, 1976; Sachs, Nachmias & Robson, 1971). For example, if the task is to detect a change in luminance from the steady state, activity in either the On- or the Off-pathway (or both) may signal detection. However, for the identification task, we may apply a simple maximum rule to the labeled line information. This information is used to classify the stimulus property of interest and to base the judgment according to which line or channel is most active (Thomas, 1985). In the example of a contrast change, if the On-pathway has the higher amount of activity, the stimulus will be identified as an increment in luminance. In this case, we are comparing the relative amounts of activity within different labeled lines and using that information to make a judgment about what happened. 

In the current experiment, we were interested in discrimination and identification of luminance contrast signals. Previous studies showed that chromatic stimuli were identified at the detection threshold, e.g. a “red-ward” pulse from white is seen as “reddish” and a “green-ward” pulse from white is seen as “greenish” (Gille, 1984; Mullen & Kulikowski, 1990; Smith, Pokorny & Sun, 2000). The result for chromatic stimuli is not surprising and can be explained by a model of chromatic processing in the PC pathway (Smith et al., 2000). Further, the information carried in the PC-pathway is graded. The observer can report the polarity of a chromatic pulse at the chromatic contrast discrimination threshold e.g. a red-ward increment on a “pinkish” pulse is seen as “more pink” and a decrement toward white is seen as “less pink”. The PC-pathways comprise four parallel sub-channels carrying information about color and luminance contrast (Derrington, Krauskopf & Lennie, 1984). A color contrast step from an equiluminant background causes firing in two of these sub-channels. It is an open question if labeled, graded responses are carried for both chromaticity and luminance in the PC-pathway. 

Achromatic contrast signals also activate the MC pathways. Tolhurst & Dealy (1975) found a small, 0.05 log unit difference between detection and identification thresholds for achromatic contrast. The need for greater contrast for identification than detection in Tolhurst & Dealy’s experiment may reflect the fact that the pulse activates both On- and Off-pathways in multiple channels of matched spatial frequency. In other experiments using achromatic contrast, discrimination of different spatial frequencies was accurate at detection threshold (Furchner, Thomas & Campbell, 1977; Nachmias & Weber, 1975) when the ratio of the frequencies was greater than three. Presumably, the use of a spatial frequency variable selectively activated different spatial frequency channels. This result might suggest that activation of an appropriate labeled line is required for achromatic contrast identification in the MC-pathway. 

The rationale for the current study lies in our ability to segregate achromatic processing in MC- and PC-pathways psychophysically (Pokorny & Smith, 1997; Smith & Pokorny, 2003; Smith et al., 2000). In these experiments a four square array replaced a constant luminance background at a fixed increment or decrement in retinal illuminance. The contrast of one of the squares, chosen randomly, was varied to elicit an increment or decrement threshold. We used two paradigms that differed only in the stimulus pattern viewed during the interstimulus interval. In the Pulsed-Pedestal Paradigm, the observer adapted to the constant background and the entire four square array was pulsed on during the trial. In the Steady-Pedestal Paradigm, the observer adapted to the four square array and only one square varied during the trial. In the Pulsed-Pedestal Paradigm, for large contrast steps (Δ I/I > +/− 0.1), discrimination was ascribed to activity in PC-processing channels. In the Steady-Pedestal Paradigm, discrimination was ascribed to activity in MC-processing channels. In a third paradigm the Pedestal-Δ-Pedestal Paradigm, contrast discrimination at low contrast steps (Δ I/I < +/− 0.1) was ascribed to the MC-pathway. The goal of the present experiment was to assess contrast discrimination and polarity identification in these inferred PC- and MC-pathways. 

Methods

Equipment

The test stimulus consisted of an array of four 1° × 1° squares with a 0.07° separations between squares. The surround was 8° × 8°, filled the separations, and had a luminance that was held constant at 12 candelas per meter square (cd/m²;115 effective trolands (td)). The surround and test chromaticities were metameric to the equal energy spectrum. The monitor was viewed binocularly at a distance of 1 m. A chin rest was used for head stabilization. 

The stimuli were generated by a Macintosh Power PC G4/350 computer with a 10-bit Radius Thunder Power 30/1600 video card and were presented on a 17″ NEC Multisync FE750 color monitor. The resolution of the display was 832 × 624 pixels and the refresh rate was 75 Hz. An Optronics OL754 spectroradiometer was used to measure the spectral power distributions of the phosphors. The luminance of each phosphor was measured for 1024 levels of input integer values, by a Minolta LS-100 luminance meter. Look-up tables were constructed for each phosphor to convert the voltage integer value to luminance. 

Paradigms

Pulsed-Pedestal Paradigm

During the adaptation period, a spatially homogenous surround was presented continuously. A four square array (pedestal), briefly pulsed during the test period, replaced the surround at a fixed increment or decrement in retinal illuminance. One of the squares (the test square) either appeared at a higher or lower retinal illuminance than the other three. 

Steady-Pedestal Paradigm

During the adaptation period, the four square array appeared continuously in the surround, with the array at a fixed increment or decrement in retinal illuminance. Only the test square changed to a higher or lower retinal illuminance during the test period. 

Pedestal-Δ-Pedestal Paradigm

The four square array was presented at a steady pedestal retinal illuminance different from the surround during the adaptation period. During the test period, the entire four square array changed in retinal illuminance (the Δ-Pedestal) with the test square incremented or decremented from the other three by a different amount. 

Figure 1 is a movie illustrating increment and decrement stimuli for the three paradigms. 

 

Figure 1a

 

Figure 1b

 

Figure 1c

Figure 1

 

Illustrations of the stimuli of the three paradigms. Double-click on each image to activate the movie.

Procedure

The observer’s task was to determine the position of a luminance difference in the test array (position identification) and to report the polarity (increment or decrement) of the luminance difference (polarity identification). To eliminate external factors and ensure responses were given under similar conditions, both questions were answered on every trial (the 2×2 design). A position judgment is made concerning the location of the test in the four square array; then a polarity judgment requires the identification of the brightness of the test square relative to the remaining squares. The test square could be an increment or decrement relative to the array and four separate interleaved staircases were run, two based on correct position, two based on correct polarity identification. 

First, the observer adapted for two minutes to the surround. There was an additional one-minute adaptation to the surround and pedestal in the Steady-Pedestal and the Pedestal-Δ-Pedestal Paradigms. A small cross that served as a fixation aid disappeared at the beginning of the trial period and reappeared following the trial period. The observer pressed one of four buttons on a Gravis USB Gamepad controller to indicate the spatial location of the test square and then pressed one of two buttons to indicate whether the luminance difference was an increment or a decrement. No feedback was given. The duration of the test was 26.7 ms given by two screen refreshes at the 75-Hz monitor refresh rate. Additional data were collected on two observers at 267 ms (twenty screen refreshes). 

Trials consisted of four random alternating staircases. There were two staircases for position identification, one for test square increment thresholds and the other for decrement thresholds. There were also two staircases for polarity identification, one for correct identification of increments and one for correct identification of decrements. All staircases began with an easily discriminable test contrast. Subsequently, the step size decreased by half following reversals, until reaching a criterion step size of 1% contrast after which there was no further change in the step size and staircases continued according to a reversal rule. For the position staircases, test contrast decreased after two correct responses and increased after one incorrect response. For the identification staircases, two correct position responses with correct identification led to a decrease in test contrast and one incorrect response for either position or identification or both led to an increase in test contrast. Thresholds for each staircase were estimated from the average of ten reversals per staircase. Each session lasted 20 to 30 minutes allowing the measurement of the four staircase thresholds for two or three of the starting pedestal retinal illuminances. For the Pulsed-Pedestal and Steady-Pedestal Paradigms and for each stimulus duration, two or three sessions were necessary to gather a complete set of thresholds. For the Pedestal-Δ-Pedestal Paradigm, six sessions were necessary to gather a complete set of thresholds. Thresholds for each condition were calculated from the average of the three or four repetitions. 

Observers

The five observers (EK, female age 25; AL, female age 30; ER, female age 21; LJ, female age 19; YS, male age 26) were all normal trichromats as assessed with the Ishihara pseudoisochromatic plates and the Neitz OT anomaloscope; all had Farnsworth 100-hue error scores of 32 or fewer. Observers EK (an author) and AL were familiar with the experimental design and psychophysical procedures. Observers ER, LJ and YS were untrained observers recruited for the experiment and were naïve to the purpose and design of the experiment. Observers EK, AL and YS participated in Experiment 1, EK and AL performed all conditions and observer YS provided data at 26.7 ms confirming the main results. Observers LJ and ER were recruited for Experiment 2. To provide training and for comparison purposes we first collected baseline data of Experiment 1. Observers EK, LJ and ER participated in Experiment 2. Written informed consent was obtained from all observers. 

Experiment 1

In Experiment 1, we replicated the previous study comparing Pulsed- and Steady-Pedestal Paradigms (Pokorny & Smith, 1997), but used the 2×2 design to include the polarity task. The pedestal had one of five retinal illuminances: two dimmer (75 and 91 td) than the surround, two brighter (145 and 182 td) and one equal (115 td) to the surround. No systematic differences were found between increment and decrement thresholds in pilot data and these thresholds were averaged. 

Results

Figure 2 shows results for four observers (clockwise panels AL, EK, YS, ER) for the 26.67 ms stimulus; data for LJ are presented in the Appendix. Data are shown for both the Pulsed- and Steady Pedestal Paradigms (open and closed symbols respectively). The delta log retinal illuminance (threshold from the pedestal) is plotted as a function of log retinal illuminance of the pedestal expressed in trolands. The panels show the position (circles) and polarity identification (squares) thresholds (±2 SE). 

Figure 2

 

Position (circles) and polarity identification (squares) for the Pulsed-Pedestal paradigm (open symbols) and the Steady-Pedestal paradigm (closed symbols). The log Δ retinal illuminance is plotted as a function of the log retinal illuminance of the pedestal. The arrow in the graph indicates surround retinal illuminance. V-shapes are fits to Equation 1 in the text; lines are fits to Equation 2 in the text. Error bars are ±2 SE. Panels show data for four observers, AL, EK, ER, and YS.

For the Pulsed-Pedestal Paradigm, the thresholds for position and polarity identification increased similarly as the pedestal contrast increased with respect to the surround, forming a V-shape. The minimum threshold values occurred when the pedestal was at the same retinal illuminance as the surround (indicated by arrow on graphs). When the pedestal has equal retinal illuminance to the surround, the two paradigms are replications. This data point is not included in the analysis of the Pulsed-Pedestal Paradigm. Pokorny and Smith (1997) described the data as reflecting the contrast response of the PC-pathway to achromatic contrast. Their equation, revised in Smith, Sun & Pokorny, (2001) is expressed as logΔI:  

 

(1)

where K_c represents a positive criterion (Pokorny & Smith,1997) and K_P accounts for threshold sensitivity and gain. For the Pulsed Pedestal Paradigm, the data did not approach saturation and K_c can be set at 0.001. The dashed lines show fits obtained by varying C_sat and K_P. The fits are good, with similar parameters for the three observers. Values of C_sat were 0.76 – 1.5, consistent with previous work (Pokorny & Smith, 1997), showing that position judgments are always contrast discriminations. 

For the Steady-Pedestal Paradigm, the thresholds increased monotonically as pedestal retinal illuminance increased for increment pedestals. Pokorny & Smith (1997) showed that the data could be fit by a unit slope, implying Weber’s Law. For pedestal decrements, a few thresholds departed from monotonicity. This phenomenon was noted in earlier studies (Pokorny & Smith, 1997) and was ascribed to stray light since the effect was increased with reduction in size of the four-square array (Smith, Sun & Pokorny, 2001) and decreased with reduction in surround width. The solid lines are linear fits to the data; obtained by varying K_M:  

 

(2)

The observers showed little variation in the one free parameter. This result confirms previous data and suggests that for the Steady Pedestal Paradigm, the position thresholds were always detections from the pedestal luminance. 

Thresholds for polarity identification were consistently higher (∼0.14 log unit) than position thresholds for all observers. The figures additionally showed that the polarity identification thresholds for the Steady-Pedestal Paradigm were below the contrast discrimination threshold for the Pulsed-Pedestal Paradigm. This result suggests that polarity identification for the Steady-Pedestal Paradigm was mediated in the presumed MC-pathway. 

We calculated the difference thresholds between position and polarity identification for the two paradigms, for each observer. For the Pulsed-Pedestal Paradigm, the difference threshold at the surround luminance was combined with the Steady-Pedestal Paradigm data. The difference thresholds were evaluated by ANOVA. For the Pulsed-Pedestal Paradigm, the difference thresholds ranged from −0.011 to 0.051 among observers and were not significant. For the Steady-Pedestal Paradigm, the difference thresholds ranged from 0.085 to 0.189 among observers and were significant. For two observers we collected data using a 267 ms duration. Thresholds showed an increase in sensitivity and the data were parallel to the 26.7 ms data. Polarity identification for long duration stimuli in the Steady-Pedestal Paradigm did not improve due to temporal separation of the pulse onset and offset. 

Our average separation between position and polarity discrimination on the Steady-Pedestal Paradigm was larger than noted by Tolhurst & Dealy (1975). For the zero pedestal, the four observers showed greater sensitivity than predicted for detection and the separation between detection and polarity identification was less, more similar to Tolhurst & Dealy (1975). 

Experiment 2

Experiment 1 showed that polarity identification thresholds were slightly higher than position thresholds in the Steady-Pedestal Paradigm and we have argued that both thresholds were mediated in the MC-pathways. Two possibilities suggest themselves. First, we have argued that the MC-pathway adapts to the Steady-Pedestal. The pulse thus excites both On- and Off-pathways as suggested in the introduction for the zero pedestal condition. Polarity identification might depend on pathway isolation. Previous work by Tolhurst (1975) revealed that at low spatial frequencies responses to the onset and offset of a pulse were about equal when they were independently detectable events spaced in time. We used the long duration pulse in an attempt to separate the responses of the On- or the Off-pathway by separating their onset and offset in time. We thought that the observer might be able to attend separately to the onset and offset. However, use of the longer duration did not improve polarity identification. A second explanation might be that the contrast pedestal does favor the correct pathway but the signal is noisy. Polarity identification could require a larger signal in order to maintain accuracy. We can isolate the MC-pathway On-and Off-pathways by using pulsed pedestals of very low contrast (Pokorny & Smith, 1997). However there is only a small range to observe the MC-pathway contrast response since there is a transition to presumed PC-pathway function for contrast steps > 0.15 (Pokorny & Smith, 1997). By placing the entire array on a pedestal, we can extend the available range between the presumed PC- and MC-pathways; this is the Pedestal-Δ-Pedestal Paradigm. The Pedestal-Δ-Pedestal Paradigm revealed the contrast discrimination behavior of the presumed MC-pathway with a steep V-shape (Pokorny & Smith, 1997). Further, a study of the time course of recovery from the Δ-Pedestal showed that increment and decrement thresholds showed different time courses (Pokorny, Sun & Smith, 2003). The data suggested that increment discriminations at Δ-Pedestal onset were mediated in On-pathways and decrement discriminations were mediated in Off-pathways. The increment and decrement staircases should thus be averaged separately although this separation was not performed in the initial study. Further, the smallest contrast steps may fall in the range where threshold summation has been reported (Foley & Legge, 1981), although this also was not observed in previous work (Pokorny & Smith, 1997; Pokorny et al., 2003). 

In Experiment 2, the steady pedestal retinal illuminance was 182 td and there were 12 values of the Δ-Pedestal varying between 91 and 363 td. These included eight small luminance steps (155,162,170,174, 191, 195, 205, and 214 td) adjacent to the steady pedestal and four large luminance steps (91, 115, 289, and 364 td). The stimulus duration was 26.7 ms. Increment and decrement thresholds for both position and polarity identification were averaged separately. 

Results

The data of Experiment 2 are displayed in Figures 3–5 for observers EK, LJ and ER, respectively. Each figure compares results for position (circles) and polarity identification (squares) in separate panels for the increment (upper panel) and decrement (lower panel) staircases. The dashed lines are predictions based the steady Pedestal Paradigm of Experiment 1. The dashed V-shape is from the PC-pathway prediction used for the Pulsed Pedestal paradigm of Experiment 1, adjusted to predict the conditions of the Pedestal-Δ-Pedestal Paradigm. There are two separate regions of the graphs to consider: the outlying regions where the Δ-Pedestal contrasts are high and the region near the steady pedestal where the Δ-Pedestal contrasts are low. When the Δ-Pedestal contrasts are > +/−0.35, the thresholds (the four outlying points) correspond to the dashed line prediction of the PC-pathway. We interpret these thresholds at high Δ-Pedestal contrasts to reflect the PC-pathway contrast response function. Position and polarity identification are equivalent for both increment and decrement thresholds, consistent with the Pulsed Pedestal data of Figure 2. Where the Δ-Pedestal contrasts are < +/−0.15, thresholds rise sharply from the steady pedestal threshold. This sharp rise shows the rapidly saturating contrast response of the MC-pathway. The V-shape indicates that these are contrast discriminations mediated in isolated On- and Off-pathways. Contrast discrimination and polarity identification are equivalent. 

Figure 3

 

Comparison of position (circles) and polarity identification (squares) for the Pedestal-Δ-Pedestal Paradigm. The upper panel shows increment data and the lower panel shows decrement data. The dashed lines are the fits to the Steady-Pedestal data of Experiment 1 and the V-shape is the predicted PC-pathway function from the Pulsed-Pedestal Paradigm adjusted to the Pedestal-Δ-Pedestal Paradigm. The solid V-shapes are fits to the combined data for both panels based on Equation 3 in the text. These fits are described further in the text. The data are for observer EK.

Figure 4

 

Comparison of position and polarity identification for the Pedestal-Δ-Pedestal Paradigm. The plot format is as for Figure 3; the observer is LJ.

Figure 5

 

Comparison of position and polarity identification for the Pedestal-Δ-Pedestal Paradigm. The plot format is as for Figure 3; the observer is ER.

Within the region of rapid threshold rise, the increment and decrement thresholds show an asymmetry about the steady pedestal luminance. Increment thresholds are lower than decrements on increment Δ-Pedestals and decrement thresholds are lower than increment thresholds on decrement Δ-Pedestals. The results are consistent across observers and for both discrimination and polarity identification tasks. A similar result was previously reported by Pokorny, (Pokorny, Sun & Smith, 2003). For increment staircases on increment Δ-Pedestals and for decrement staircases on decrement Δ-Pedestals, the effective contrast is reduced by a factor (0<κ<1). Equation (1) was thus rewritten to reflect the MC-pathway response:  

 

(3)

 

The solid lines are fits to the combined position and polarity identification data using Equations (1) and (3) with individual adjustment of C_sat, κ, and K_M. The value of K_c was set at 0.01. Compared with the fits of equation (1) to Pulsed Pedestal data, the value C_sat was at a lower value (usually near 0.1–0.15) consistent with the steep MC-pathway saturation (Pokorny & Smith, 1997). For observer LJ, we allowed an additional constant added to the polarity identification for decrements (see Appendix). The fits were good with little variation in the parameters among the observers. There were over-estimations near the steady pedestal luminance where the Δ-Pedestal contrast (∼0.05) is sub-threshold for our observers. At pedestal contrast near 0.05, measured thresholds were more sensitive than for the zero pedestal for increments measured on the increment pedestal (upper panels) and for decrements measured on the decrement pedestals (lower panels). These data may reflect the effect of sub-threshold summation, when the staircase and the Δ-Pedestal are in the same direction. 

Subthreshold effects have been noted for simple spatial displays in the achromatic (Bowen, 1995) and in the chromatic domains (Cole Stromeyer & Kronauer, 1990). Although sub-threshold summation was originally attributed simply to physical addition of light, this explanation was discredited with the arrival of signal detection accounts of discrimination (Nachmias & Kocher, 1970). The current explanations for dipper effects include a non-linear transducer (Chen, Foley & Brainard, 2000; Graham, 1989; Nachmias & Kocher, 1970) and uncertainty reduction (Pelli, 1985). Neither approach explains the bumper effect (Bowen, 1995). For our simple displays, the physical addition of light offers the simplest account of both phenomena. We assume that with sub-threshold pedestals the observers revert to detecting the odd square against the background rather than discriminating among four squares. This conclusion for simple color stimuli was also reached by Eskew (1999). These threshold interactions do not occur uniformly among observers. We reviewed data collected in our laboratory for the Pedestal-Δ-Pedestal paradigm (Pokorny & Smith, 1997), but did not find consistent evidence of sub-threshold effects. In this previous work, sub-threshold pedestals were not included since assessment of threshold interactions was not the goal of the study. Additionally, some of the variation may reflect observer variability. Our three observers in Experiment 2 were not as highly trained as in the previous work, and showed higher thresholds overall. 

The predictions describe both contrast discrimination and polarity identification above detection threshold. In this region, the appropriate pathway is isolated and contrast discrimination and polarity identification are the same. 

Discussion

In the Pulsed-Pedestal Paradigm, thresholds for contrast detection and polarity identification were the same for all non-zero pedestal contrasts. Previous studies found no differences between chromatic contrast detection (Gille, 1984; Mullen & Kulikowski, 1990) or chromatic contrast discrimination (Smith, Pokorny & Sun, 2000) and identification. These results suggest that the PC-pathway does indeed have labeled lines, which signal polarity identification for both chromaticity and luminance at detection threshold. The results suggest that the four retinal pathways, +L+M, +L−M, −L+M and −L−M, combine at higher cortical areas in different combinations to separate luminance and chromatic information. This interpretation has been called demultiplexing (Ingling & Martinez-Uriegas, 1983; Lennie & D’Zmura, 1988). However, physiological evidence of demultiplexing has not yet been found (Lennie, Krauskopf & Sclar, 1990). Of note is that both for color and for luminance the cortex extracts a continuous graded signal. 

The situation in the MC-pathway is different. At steady state, both On- and Off-pathways are active and contribute to detection. Polarity identification for our display required about 0.14 log unit more contrast. We thought that observers might have better identification performance with long pulses for which they could separate attention to the onset and offset of the Pulsed Pedestal. However, they apparently did not use this strategy. With the Pedestal-Δ-Pedestal paradigm, increment contrast discrimination was determined in On-pathways and decrement contrast discrimination was determined in Off-pathways. When the pathway was isolated, polarity identification was at contrast discrimination threshold except in the region where the Δ-Pedestals were sub-threshold. This result supports the hypothesis that polarity identification requires isolation of an appropriate pathway. 

The data of Figures 3 – 5 showed very high thresholds for decrement staircases measured with increment pedestals and increment staircases measured with decrement pedestals. The thresholds were in fact higher than the predicted PC-pathway thresholds. This phenomenon might reflect a cognitive component, e.g. the observer is monitoring only the MC-pathway until it saturates and the Δ-Pedestals are above PC-pathway threshold. Alternatively, perhaps the close alignment of the squares interferes with the observers’ judgment. We thought it would be instructive to look at the actual illuminances at the threshold ΔI. Figures 6 – 8 show this calculation for the three observers. This plot is in linear units of retinal illuminance. The abscissa shows the Δ-Pedestal illuminance; the ordinate shows the threshold illuminance at the discrimination step for position (open symbols) and for polarity (closed symbols). The vertical and horizontal lines dividing the plot frame into quadrants are placed at the steady pedestal illuminance. Increment Δ-Pedestals occur in the right hemi field; decrement Δ-Pedestals occur in the left hemi field. Increment thresholds with increment pedestals fall in the upper right quadrant above the diagonal (shown as a dashed line) and decrement thresholds with decrement pedestals fall in the lower left quadrant below the diagonal. These results were as expected. The unexpected results shown by the plot concern the decrement staircases with increment Δ-Pedestals. These thresholds fell in the lower right quadrant indicating that discrimination and polarity identification of a decrement required a decrement from the steady pedestal luminance. A parallel result occurred for the increment staircases with decrement Δ-Pedestals. These thresholds fell in the upper left quadrant indicating that both discrimination and polarity identification required an increment from the steady pedestal luminance. It can be seen that this untoward event occurred only for the Δ-Pedestals that were ascribed to the MC-pathway. At higher contrasts (stippled area), where we have suggested that the MC-pathway saturates and the PC-pathway mediates achromatic discrimination, the thresholds are distributed in the appropriate quadrants: increments above the diagonal, decrements below the diagonal. These results are consistent with the continuous graded achromatic response signaled by the PC-pathway. The results suggest that the MC-pathways are labeled by pathway only and do not signal a graded response. Under our stimulus conditions, the response “brighter” is signaled only by On-pathways while the response “dimmer” is signaled only by Off-pathways. 

Figure 6

 

The test square illuminance at the position and the polarity identification thresholds plotted as a function of the Δ-Pedestal illuminance in trolands. Open symbols are for position identification; closed symbols are for polarity identification. Increment thresholds are shown by circles and decrement thresholds are shown by squares. The solid horizontal and vertical lines indicate the retinal illuminance (183 td) of the steady pedestal. The dashed line indicates the illuminance of the Δ-Pedestal. Icons across the top correspond to (from left to right): a decremental pedestal (decrement test), a pedestal equal to the surround (decrement test) and an incremental pedestal (increment test). The observer is EK.

Figure 7

 

The test square illuminance at the position and the polarity identification thresholds plotted as a function of the Δ-Pedestal luminance in trolands. The plot format is as for Figure 6; the observer is LJ.

Figure 8

 

The test square illuminance at the position and the polarity identification thresholds plotted as a function of the Δ-Pedestal luminance in trolands. The plot format is as for Figure 6; the observer is ER.

In summary, we suggest that graded polarity signals are generated in the PC-pathway for both chromatic and achromatic contrast. For the MC-pathway, isolation of On- and Off-pathways can lead to correct polarity identification but this information is not graded. 

Acknowledgments

NEI Research Grant EY00901 and NIMH Training Grant T32 MH20029 supported this research. We thank our observers for their time and attention, and Linda Glennie for assistance with programming. We thank Patrick Monnier and Barry Lee for useful discussions of the results. Publication supported in part by an unrestricted grant to the Department of Ophthalmology and Visual Science from Research to Prevent Blindness. 

Commercial relationships: none. 

Appendix

Response bias: In previous work with contrast discrimination paradigms, we have noted no differences in increment and decrement staircases using the Pulsed- and Steady-Pedestal Paradigm. It was on this basis that we averaged data for increment and decrement thresholds. In fact, no differences occurred in the present work for the discrimination task. Further, no differences in the increment and decrement staircases were found for the position task for the Pulsed-Pedestal Paradigm. 

Our method does allow for the possibility of response bias in polarity identification in the Steady-Pedestal paradigm where many trials are detected but not identified. Since there are only two choices for identification, “brighter” or “dimmer”, response bias could occur if the observer tended to favor one response on trials where the position was correctly identified but the polarity identification was below threshold. Since staircases are driven by the correct responses, response bias could preferentially drive the polarity staircase. Suppose the observer tends to respond “dimmer”. The number of correct identifications on the decrement staircase would be above chance. Decrement staircases would approach the detection threshold. In examining the data of Experiment 1, two observers showed no evidence of response bias for polarity identification. One observer (AL) showed a bias to “dimmer”. For the 26.7ms pulses her average (Polarity — Detection) difference of 0.107 reflected a difference of 0.1625 on the increment staircase and 0.05 on the decrement staircase. Two new observers were recruited for Experiment 2. The data of ER (shown in Figure 5) were similar to those of AL, with a similar mild bias to the “dimmer” response. ER showed an average polarity-detection difference of 0.124 composed of 0.05 for decrements and 0.24 for increments. For LJ however, a very pronounced response bias to “brighter” was noted in Experiment 1 and shown in Figure 9. For the Pulsed-Pedestal Paradigm, increment and decrement thresholds were the same and were averaged. Contrast discrimination and polarity identification were also the same. All the data followed the two rising arms of the PC-pathway contrast response. For the Steady-Pedestal Paradigm, the position judgments were the same for increment and decrements and were averaged. However, the polarity identification staircases were different for increments and decrements. The increment polarity identification was the same as the detection. However, the decrement polarity identification thresholds fell on the data for the Pulsed-Pedestal Paradigm. In the Pedestal-Δ-Pedestal Paradigm, the decrement contrast discrimination was in the Off-pathway but there was still a deficit in polarity identification. These data were consistent with a pure response bias. The observer told us that she was trying to guess when she could not tell the polarity, however her behavior did not match her words. 

Figure 9

 

Comparison of the Pulsed-Pedestal (open symbols) and Steady-Pedestal (closed symbols) paradigm for observer LJ. Position (circles) and polarity (squares) identification is shown for the Pulsed-Pedestal. Polarity identification for the Steady-Pedestal is separated by staircase with increments shown by upright triangles and decrements shown by inverted triangles.