Testing a stochastic foraging model in an operant simulation: Agreement with qualitative but not quantitative predictions.

Researchers set up an operant chamber like a video game for pigeons. The birds pecked keys to earn grain in two patches. One patch always gave food on the same schedule. The other patch sometimes ran out.

The team tracked when the birds switched patches. They compared the moves to a math model that predicts the best time to leave. The model comes from biology, not behavior analysis.

The pigeons got the big picture right. They left a patch more often when it was empty. But they stayed too long in dry patches and checked them again too soon.

The numbers did not fit the model. The birds over-sampled empty spots. They acted like something besides pure payoff guided their choices.

Bachman et al. (1988) saw a similar mismatch. Their pigeons also violated strict optimality rules when search time changed. Both labs show that animals use rules of thumb, not perfect math.

Grosch et al. (1981) and Cullinan et al. (2001) found that adding a signal during a delay keeps pigeons loyal to one key. The foraging study had no signals, so the birds kept sampling. The papers together say: when cues are weak, animals explore more.

Lattal (1984) showed that blackout signals protect response rates during delays. The foraging task had no such signals, perhaps why birds re-checked empty patches. The gap suggests signaling could tighten patch choices.

Your clients also over-explore when the payoff rules are unclear. If a task suddenly stops producing reinforcement, learners may keep trying or return after a break. Add clear signals — a timer, a color change, a countdown — to mark when reinforcement is no longer available. This can cut wasted responses and speed up the shift to a new task or location.