Influences of target stimulus probability and reinforcement probability on hens' performance on a signal detection task

Abstract

Hens’ signal-detection performance was examined across a range of signal-present trial probabilities and reinforcement probabilities. These variables have been studied previously with the yes/no procedure, but have not been systematically evaluated using the go/no-go procedure. This study employed natural contingencies where hits were reinforced, but all other responses had no scheduled consequences. The hens were required to discriminate between a signal-present (bright light) and signal-absent trial (dimmer light) by responding on the stimulus key or bypassing the trial by responding on a second key. The hens were exposed to reinforcement rates of 100%, 75% and 50% and signal probabilities of 50%, 25%, 12.5%, and 6%. Manipulation of reinforcement rate across the range examined did not significantly influence performance at any stimulus probability. Specificity increased over sessions, suggesting a practice effect for signal detection tasks. The go/no-go procedure should be considered for operational signal detection applications as it was found to produce robust accuracy across a range of conditions.