Abstract
In recent years, several investigators have noted that the presumption that the upper asymptote of the psychometric function is 1.0 may result in substantial errors in estimation of both the slope and location parameters of the function. For some psychophysical tasks, maximum asymptotic performance is clearly less than 1.0, due to what has been variously labeled misreporting error, extraneous noise, or “button” error. Such error is the result of factors extraneous to stimulus magnitude. Because lapses are infrequent, however, it is difficult to estimate empirically their rate, no less to distinguish them from the occasional incorrect response that is due to sensory noise. Still, the common belief is that this lapse rate is substantial enough (up to 0.05), especially with untrained observers, to affect psychometric function parameter estimation. The presumption of a substantial lapse rate has also affected the design of at least one clinical test (Pelli, Robson & Wilkins, 1988, Clinical Vision Sciences, 2, 187–199). Here I report on one task in which the lapse rate appears to be negligible. I assessed lapse rate using untrained naïve observers with healthy vision, who verbally identified, large, randomly selected, high contrast Sloan letters printed on paper in a fashion analogous to, and with similar properties of, chart-based clinical visual acuity or letter contrast sensitivity tests. Subjects read a minimum of 1536 consecutive letters. Subjects occasionally corrected themselves spontaneously during the test, but lapse rate was easily less than 0.001. I conclude that in a verbal letter reporting task where observers set their reporting pace, and where they may change their responses if they so wish, lapse rate is negligible. This result has implications for both clinical vision testing and basic psychophysical tests that use verbally reported letters.