Differential reinforcement and signal detection.

Rincover et al. (1975) tested pigeons in a signal-detection task. Birds pecked one key when they saw a color and another key when no color appeared.

Sometimes the researchers gave grain after wrong pecks. They wanted to know if paying for errors would hurt the birds' ability to tell signal from no-signal.

Reinforcing errors made the pigeons worse at the task. Their perceptual sensitivity dropped.

The result broke the classic rule that signal-detection sensitivity should stay the same no matter what you reinforce.

Dougherty et al. (1996) later used the same signal-detection setup. They showed that pigeons still follow signal-detection rules when errors are not paid. Together the two studies draw a clear line: reward mistakes and vision blurs; withhold reward and vision stays sharp.

Davison et al. (1989) also ran color-detection with pigeons but only changed how often correct responses were paid. They saw no bias, matching the 1975 paper's warning that the trouble starts only when wrong responses get tickets too.

Marcucella (1974) tested signaled DRL the year before. That study found signals help only when animals have time to track them. The 1975 paper flips the coin: signals lose power the moment you reinforce the wrong read.

If you are shaping a new discrimination, never deliver reinforcement after an incorrect response. Even one or two paid errors can muddy the stimulus control you are trying to build. Instead, withhold all rewards for wrong choices and use extinction or prompt correction. This keeps the learner's sensory window clear and speeds mastery of the target distinction.