Independence of sensitivity to relative reinforcement rate and discriminability in signal detection.

The team worked with pigeons in a signal-detection task. Birds pecked one key when they heard a tone and another key when they heard silence.

The researchers changed two things across sessions. They made the tone louder or softer to vary how easy it was to tell the signals apart. They also changed how often correct pecks earned food on each key.

Each bird served as its own control. The order of loud and soft tones was mixed within the same daily session so that only the current signal, not the past, guided the choice.

When the tone was soft, pigeons made more errors. Their discrimination accuracy dropped, but their sensitivity to reinforcement stayed the same.

No matter how clear the signal was, shifting food from one key to the other always pulled the birds’ bias the same amount. Easier signals did not make the birds more sensitive to the new payoff rates.

In short, discriminability changed accuracy, while payoff ratio changed only the birds’ willingness to pick one side.

LeBlanc et al. (2003) seems to disagree. They saw higher reinforcement rates lift both response rate and discrimination accuracy. The key difference is procedure: A et al. compared rich versus lean schedules across separate conditions, while D et al. kept overall payoff constant and only moved food between the two keys within the same condition. The first paper shows reinforcement can strengthen the behavior stream; the second shows it can’t sharpen the eye or ear once the signal is fixed.

Gulley et al. (1997) extends the bias story to humans. In a matching-to-sample game, raising the payoff for one sample made it more likely to control the final choice. Together with D et al., this tells us that payoff shifts which stimulus “wins,” not how well the person or pigeon sees the stimulus.

Rand (1977) and Rilling et al. (1969) are earlier pigeon studies that set the stage. They showed response rate follows time allocation, which in turn tracks reinforcement rate. D et al. adds the signal-detection twist: once the bird is already dividing time, payoff moves the divider, not the clock speed.

When you shape discrimination, first make the stimulus easy to see or hear. After that, don’t expect richer reinforcement to improve accuracy; it will only tilt the learner toward one response. Use payoff to balance bias, not to sharpen vision or hearing. If a client keeps choosing left even when the right stimulus is present, check your reinforcement ratio before you re-train the discrimination.