Detection and recognition of temporal patterns

Abstract

The detection/recognition theorem [S. J. Starr, C. E. Metz, L. B. Lusted, and D. J. Goodenough, Radiology 116, 533–538 (1975)] was used to predict a listener's ability to recognize (identify) one of a pair of temporal patterns based upon the listener's detection performance. The six experimental conditions compared all possible pairs of four 860‐ms patterns. The four patterns consisted of either two or four 1‐kHz sinusoidal components of either 20‐ or 200‐ms duration. Specifically, the four patterns were: two 20‐ms components separated by 820 ms; two 200‐ms components separated by 460 ms; four 20‐ms components separated by 260 ms; and four 200‐ms components separated by 20 ms. All components were square gated on and offer zero crossings. In four of these comparisons, stimuli differed on “one” dimension (either number or duration of components), whereas the other two comparisons involved differences in “two” dimensions (both number and duration of components). (Interstimulus gaps were necessarily confounded with number of components and component duration in all conditions.) Although the results were not uniform for all 11 listeners, comparisons that involve two dimensions generally led to better recognition performance than comparisons that involved one dimension. More surprising was the apparent unimportance of temporal structure for the encoding of these complex stimuli inferred from the general failure to reach predicted recognition performance. [Research supported by AFOSR through WPAFB AAMRL/BBA.]