An efficient adaptive procedure for psychophysical discrimination experiments
- PDF / 850,773 Bytes
- 7 Pages / 558 x 756 pts Page_size
- 26 Downloads / 186 Views
An efficient adaptive procedure for psychophysical discrimination experiments PETER WERKHOVEN TNO Human Factors Research Institute, Soesterberg, The Netherlands and HERMAN P. SNIPPE University ofGroningen, Groningen, The Netherlands We present an adaptive procedure to conduct psychophysical discrimination experiments. In a discrimination experiment, an observer senses (sees, hears, feels, etc.) two stimuli (separated in space or time) and is asked to order these stimuli with respect to a particular parameter (say, s). Under the usual assumption of a locally linear internal representation of s, perturbed by additive Gaussian noise, the probability P(s2)-ofjudging test stimulus S2 "larger" than a reference stimulus sl-is an error function (a cumulative normal distribution). Such an error function, Erf [(s 2 - .u) fa], is parametrized by two parameters: .u and a. The parameter .u is the value for which P( s 2) = 50% and is related to possible bias effects in the internal representation of the observer. The parameter a is v2 times the standard deviation ofthe noise distribution and is generally called the discrimination threshold. In this paper, we present (I) an algorithm to estimate .u and a given the data generated by such an observer, (2) an analysis of the efficiency of a stimulus presentation with respect to the estimation of .u and a, and (3) a method that controls the specific choice of stimulus values during an experiment so that an optimal estimation of.u and a is obtained. Many current adaptive psychometric procedures are explicitly geared toward detection experiments and yield an estimate of one parameter, the detection threshold. In psychophysical discrimination experiments, however, one is often uncertain about two parameters that describe the observer's behavior: the discrimination threshold and the location on the stimulus axis that yields chance (50% correct) behavior. Both parameters have to be estimated from the data. This paper describes a method that is capable of handling this additional complexity in an efficient way. In recent years, adaptive methods for psychophysical detection experiments have become very sophisticated (see King-Smith, Grigsby, Vingrys, Benes, & Supowit, 1994, for a review). These methods feature the following: 1. Explicit incorporation of prior knowledge about the psychophysical parameters to be estimated. However, contrary to the case of(contrast) detection, where it is claimed that there is a generic prior probability distribution function for the contrast detection threshold (King-Smith et aI., 1994), for the huge variety of possible discrimination experiments there will not be one generic prior distribution function that describes the a priori probability of discrimination thresholds. Of course, the choice ofthe range of stimuli that the experimenter prepares for the discrimination experiment shows expectations about the
Correspondence should be addressed to P.Werkhoven, TNO Human Factors Research Institute, PO Box 23, 3769 ZG Soesterberg, The Netherlands (e-mail: werkh@
Data Loading...
