Abstract

Psychophysical measurements of visual thresholds frequently involve large numbers of trials under a range of stimulus conditions, and data analysis which defines threshold as a point on a psychometric function. The slope of the psychometric function is commonly interpreted as reflecting stimulus-related, noise, due to variability in the stimulus and in the responses of the visual system to it. However, for inexperienced observers it is also necessary to consider the effects of extraneous noise, such as lapses in attention or cooperation (McKee, Klein & Teller, 1985; Teller, Mar & Preston, 1986; Klein & Manny, 1989). Stimulus-related noise affects the slope of the psychometric function without altering the upper asymptote, while extraneous noise can reduce the upper asymptote of the psychometric function below 100% correct. When extraneous noise causes the upper asymptote to be less than 100%, standard probit analysis (using a cumulative normal distribution from 50% to 100% correct) yields a shallow slope and considerable uncertainty in the threshold estimate. In the present study, we evaluated the effects of extraneous noise on analysis of data from inexperienced observers by analyzing two large sets of preferential-looking acuity data for normal infants and toddlers (data from Birch, 1985 and Birch & Hale, 1988).

© 1990 Optical Society of America