The problem of contrast metric for reaction time to aperiodic stimuli Angel Vassilev 1, 2 Adrian Murzac 2, Margarita B. Zlatkova 3 & Roger S. Anderson 3 1 Institute of Neurobiology, Bulgarian Academy of Sciences, Sofia, Bulgaria, 2 New Bulgarian University, Sofia, Bulgaria 3 Vision Science Reasearch Group, School of Biomedical Sciences, University of Ulster, Coleraine, Northern Ireland, UK

Aim We react faster to a strong stimulus than to a weak one. But how to measure stimulus strength? The aim of the present talk is to give an example how the choice of stimulus metric affects the conclusions drawn from a reaction time study.

A typical reaction time experiment A stimulus is flashed briefly and the observer has to press on a key (or release a key) as soon as possible. The metric of stimulus strength most commonly used is Weber contrast,  I/I: the change in luminance relative to the background.. Background Fixation mark Background, I Stimulus,  I Stimulus Luminance Distance

Is always Weber contrast the appropriate measure of stimulus strength? Difficulties are faced when comparing performances to suprathreshold stimuli that share a common physical metric, but give rise to different threshold sensitivities. Such is the case with luminance increments and decrements.. Background, I For equal  Is, increments and decrements are of equal Weber contrast yet the thresholds might differ. Usually, the threshold of decrements is lower. Stimulus,  I Luminance Distance

The work of Pokorny’s group Using an unique technique, Cao, Zele and Pokorny (2007) provided a rich RT data set. They measured cone and rod RTs to stimuli presented within either a Rapid-On or Rapid-Off ramp temporal window (fast phase:luminance increment or decrement). 1 sec They were able to selectively stimulate the rods over a 5 log units range of retinal illumination. Part of their data are presented in the next slide. Background Luminance Time

Cao, Zele & Pokorny (Vision Res. 2007): Cone and rod RTs compared Weber contrast (%) Reaction time (ms) Note that the cone RTs to Rapid-ON and Rapid-OFF stimuli are similar, while the rod RTs differ systematically Observer DCObserver AJZ

Cao et al.:Rod RT: Observer DC Weber contrast (%) Reaction Time (ms)

Cao et al.:Rod RT:Observer AJZ Reaction Time (ms) Weber contrast (%)

The work of Pokorny’s group In a parallel paper, Zele, Cao & Pokorny (Vision Research, 2007) posed the question regarding the metric of stimulus strength used to compare performance to suprathreshold stimuli.

The work of Pokorny’s group Contrast sensitivity (expressed in Weber contrast units) to rod Rapid-Off stimuli was about two times higher than to rod Rapid-On stimuli. As expected in view of the difference in sensitivity, reaction time to Rapid-Off stimuli was shorter than to Rapid-On stimuli over a range of Weber contrasts. However, expressing stimulus strength in multiples of threshold did not equate incremental and decremental RTs. Instead, for stimuli at the same suprathreshold level, RT to increments turned out shorter than RT to decrements.

The work of Pokorny’s group Two more contrast metrics, tested by them also failed to account for the differences between increment and decrement RTs. Zele at al. (2007) assumed that the only meaningful comparison of reaction times is the comparison of asymptotic RTs. Here we show that a simple contrast metric equates their rod reaction times and allows inferences about the mechanisms of stimulus detection.

Two cues, two types of detection: two contrast metrics The stimulus generates a temporal gradient  L of the background illumination L b as well as a spatial gradient, (  L + L b ) against L b. The stimulus might be detected by temporal (successive) luminance discrimination or by spatial (simultaneous) luminance discrimination (Sperling & Sondhi, 1968). Weber contrast =  I/I Spatial luminance ratio = L max /L min, the larger and the smaller of L s and L b Background, I Stimulus,  I Ls (Lb) Luminance Distance

Two cues, two types of detection: two contrast metrics Weber fraction  L/ L b captures the temporal change of luminance relative to the background. We assume that spatial discrimination should depend on the ratio between background luminance and stimulus luminance, Lb and Ls. We calculate it as Lmax/Lmin, where Lmax and Lmin are the larger and smaller of Lb and Ls. The next two figures show the results of transforming Weber contrast into spatial luminance ratio. The rod reaction time data of Cao et al. (2007) are presented as functions of Weber contrast and the spatial luminance ratio.

Spatial luminance ratio Rod RT: Observer DC Weber contrast (%) Reaction Time (ms)

Spatial luminance ratio Reaction Time (ms) Weber contrast (%) Rod RT: Observer AJZ

Spatial luminance ratio Weber contrast (%) S-cone selective increments and decrements (Murzac, 2004; Murzac & Vassilev, 2004) Reaction Time (ms)

S-cone selective increments and decrements Perceptual differences supporting the assumption of two types of stimulus detection: Some observers reported that perception of S-cone selective stimuli differs from the perception of ordinary achromatic stimuli. The sense of winking accompanies the onset of ordinary stimuli but is absent with S-cone selective stimuli. Two kinds of perception: Perception of stimulus onset Perception of stimulus presence

Summary of results Reaction times to luminance increments and decrements are compared under several experimental conditions. Cao, Zele & Pokorny’s (2007) data: A. Photopic (all-cone) reaction times cluster around a single RT/Weber-contrast function regardless of the stimulus sign, increment or decrement. B. Rod incremental and decremental reaction times form two distinct RT/Weber-contrast functions but cluster around a single function when plotted against the spatial- luminance ratio. Murzac (2004), Murzac & Vassilev’s (2004) data: S-cone reaction time tends to behave like rod reaction time

Interpretation Physiological and psychophysical data show that photopic cone vision is faster and predominantly transient while scotopic rod vision is slower and predominantly sustained (Pepperberg, 2001). Also this seems to be the case for S- cone vision (Reid & Shapley, 2002). Weber contrast is a measure of the transient stimulus component while the spatial luminance ratio is a steady- state measure. The fit of both incremental and decremental RTs by a single Weber-contrast function or by a single spatial- contrast function parallels the properties of the systems involved in stimulus detection.

Conclusions We assume that the type of neural activity, predominantly transient or sustained, and, respectively, the type of stimulus detection by temporal (successive) luminance discrimination or by spatial (simultaneous) luminance discrimination determines the appropriateness of Weber contrast or spatial luminance contrast metric for reaction time.