Daniel L. Weber
Stephen P. Sanger
Dept. of Psychol., Wright State Univ., Dayton, OH 45435
In separate conditions, four listeners discriminated samples from distributions of the duration of a 1-kHz sinusoid and from distributions of the frequency of a 100-ms sinusoid in a 2IFC, sample-discrimination procedure. For duration discrimination, the ``standard'' distribution (100-ms mean) was discriminated from an ``easy'' comparison (140-ms mean) distribution and from a ``hard'' comparison (110-ms mean) distribution. For frequency discrimination, the ``standard'' distribution (1000-Hz mean) was discriminated from an ``easy'' comparison (1400-Hz mean) distribution and a ``hard'' comparison (1010-Hz mean) distribution. Standard and comparison distributions were normal with standard deviations set to the difference in their means so that all conditions yielded a d' of 1.41 for an ideal observer. Pooled (all n samples from a given distribution presented in succession) and paired (one sample from each distribution presented in n successive pairs) presentation structures were tested for these four combinations of standard and comparison distributions. Information integration, represented by the increase in d' as a function of the number of samples (n=1, 2, 3, 4, 5, 6, 8, 12, and 16), was examined in all eight conditions. Pooled presentations uniformly led to better performance. The partitioned-variance model and the decision-combination model were used to represent different strategies for information integration in these data. [Research supported by a grant from AFOSR through WPAFB AL/CFBA.]