Behavior Analysis And Signal-detection Theory.

Parmenter (1999) wrote a theory paper. The author asked: can we borrow math from signal-detection theory to clean up our operant data?

Signal-detection theory comes from radar research. It splits performance into two parts: how well you detect the signal and how eager you are to say "yes." The paper maps those parts onto stimulus control and reinforcement effects.

The paper does not give new data. Instead it shows formulas you can drop into any graph. One index tells you how sharp the stimulus control is. Another tells you how much the payoff bends the curve.

The math lets you say, "The child’s errors are 70 percent stimulus confusion, 30 percent payoff bias," instead of just counting errors.

Jensen et al. (2013) extends the same idea but swaps in information theory. Where Parmenter (1999) uses hit and false-alarm rates, Greg uses bits and entropy. Both want numbers that pull stimulus control away from reinforcement.

Boudreau et al. (2015) also extends the plan. They apply information bits to conditioned reinforcement. Their paper shows the math works for cues that predict delay, not just simple signals.

Cameron et al. (1996) is methodologically similar. That team built behavioral momentum to add a new number—persistence—to plain response rate. Parmenter (1999) adds sensitivity and bias. Both papers say raw counts are not enough.

Next time you run a discrimination probe, plug the hits and false alarms into the signal-detection sheet. You will see if errors come from weak stimulus control or from the kid’s bias toward saying "blue." If bias is high, tweak the payoff, not the prompt. If sensitivity is low, sharpen the stimulus. The split saves you from chasing the wrong variable.