Assessment and choice: an operant simulation of foraging in patches.
Local odds of the next reinforcer, not overall rate, control when learners switch away from a fading patch.
01Research in Context
What this study did
Researchers watched pigeons peck keys in a patch that paid off less and less. The birds could stay or hop to a fresh patch at any time.
The team tracked how long each bird stayed before giving up. They compared these times to a new math model that weighs current payoff odds against the next patch.
What they found
The pigeons' quit times almost perfectly matched the model's guesses. Birds seemed to track the chance of the next reward right now versus later, not the overall rate.
Tiny shifts in local probability pushed them to leave earlier or stay longer.
How this fits with other research
Dougherty et al. (1994) ran a near-copy setup but added 'noisy' time memory. Their tweak fit the data better than the old perfect-memory model; Cameron et al. (1996) kept the noise and showed the same birds follow capture odds, closing the loop.
Iwata (1993) found pigeons still care about the big scheduled rates when both choices sit in view. Cameron et al. (1996) show that when one patch fades behind a key, local moment odds take over—context decides which rule wins.
Malone (1999) later moved the local rule to rats choosing between two levers. The same stay/switch math worked, proving the idea travels past pigeons and single patches.
Why it matters
When you set up concurrent schedules, think moment-to-moment, not just average. If a reinforcer thins during DRL or a token board dries up, the learner may quit early even if the overall rate looks fine. Watch local probability dips and be ready to prime or thicken the patch before the client walks away.
Want CEUs on This Topic?
The ABA Clubhouse has 60+ free CEUs — live every Wednesday. Ethics, supervision & clinical topics.
Join Free →Plot each client's last five reinforcers per minute; if the line dips for two minutes, insert an extra reinforcer before the client bolts.
02At a glance
03Original abstract
Pigeons were presented with an operant simulation of two prey patches using concurrent random‐ratio schedules of reinforcement. An unstable patch offered a higher initial reinforcement probability, which then declined unpredictably to a zero reinforcement probability in each session. A stable patch offered a low but unvarying reinforcement probability. When the reinforcement probability declined to zero in a single step, the birds displayed shorter giving‐up times in the unstable patch when the ratio between the initial reinforcement probabilities in the unstable and stable patches was greater and when the combined magnitude of the reinforcement probabilities in the two patches was greater. When the unstable patch declined in two steps, the birds behaved as if their giving‐up times were influenced heavily by events encountered during the most recent step of the double‐step change. This effect was observed, however, only when the reinforcement probability in that step was .04, not when it was .06. All of these data agree with the predictions of a capture‐probability model based on a comparison of the estimated probability of receiving a reinforcer in the current patch with that in alternative patches.
Journal of the experimental analysis of behavior, 1996 · doi:10.1901/jeab.1996.66-327