Varying feedback to evaluate detection strategies: the detection of a tone added to noise

The effectiveness of rewarding strategies associated with different models of detection was evaluated for a tone-in-noise detection task. In three experiments, a tone was added to a narrowband noise masker in a two-interval, forced-choice procedure with feedback. In the first experiment the observer's task was to detect which interval had the added tone. In the second experiment the task was to indicate which interval was more intense. In the third experiment the task was to indicate which interval had the sound with the less modulated envelope. In terms of correct signal detection, thresholds were nearly the same for all three experiments. However, for three of the observers the psychometric functions indicate that observers altered their responses in accordance with the changes in feedback contingencies. This indicates that neither changes in level nor changes in envelope pattern alone can account for the detection of a tone added to noise for these observers. The fourth observer was sufficiently insensitive that the pattern of feedback associated with the level and envelope models did not differ substantially, and for her feedback had little impact on behavior. The results indicate that changes in feedback provide a means of testing alterative psychophysical models. Moreover, the results suggest that it is difficult to model the detection of a tone added to noise in terms of a single cue or a simple combination of cues. Observers are free to choose from at least two cues, and can do so.