Conditioned reinforcement and information theory reconsidered.
Conditioned reinforcers work because they act like headlines that tell the client when the main reward is coming.
01Research in Context
What this study did
Boudreau et al. (2015) wrote a theory paper. They asked: what if a conditioned reinforcer is just news that food is coming sooner?
They used math from information theory. The math counts how much a stimulus cuts uncertainty about the next meal.
What they found
The new formula fixed old problems. It lined up with Delay Reduction Theory and worked for both Pavlov and operant cases.
The paper says a click, a light, or a spoken word is valuable only when it reduces the guess-work about payoff time.
How this fits with other research
Jensen et al. (2013) already said information theory can give behavior analysts new rulers. Boudreau et al. (2015) took that idea and aimed it straight at conditioned reinforcement.
NEVIN et al. (1963) showed white noise could reinforce bar pressing even when it gave no cue about what to do next. The new paper keeps that finding; it just explains the power of the noise as ‘uncertainty cut’ instead of ‘association strength.’
PLISKOFF et al. (1960) found water deprivation flipped a conditioned reinforcer from good to bad. The information view still fits: thirst changes what counts as useful news, so the same stimulus now gives less uncertainty reduction about water.
Why it matters
You can now test whether a stimulus is a conditioned reinforcer by measuring how much it tightens the client’s prediction of reinforcement time. If the stimulus gives no new timing info, swap it for one that does. This single shift can clean up your chaining and token programs without extra cost.
Want CEUs on This Topic?
The ABA Clubhouse has 60+ free CEUs — live every Wednesday. Ethics, supervision & clinical topics.
Join Free →Track how long after each SD the reinforcer arrives; keep only stimuli that shorten that wait guess.
02At a glance
03Original abstract
The idea that stimuli might function as conditioned reinforcers because of the information they convey about primary reinforcers has a long history in the study of learning. However, formal application of information theory to conditioned reinforcement has been largely abandoned in modern theorizing because of its failures with respect to observing behavior. In this paper we show how recent advances in the application of information theory to Pavlovian conditioning offer a novel approach to conditioned reinforcement. The critical feature of this approach is that calculations of information are based on reductions of uncertainty about expected time to primary reinforcement signaled by a conditioned reinforcer. Using this approach, we show that previous failures of information theory with observing behavior can be remedied, and that the resulting framework produces predictions similar to Delay Reduction Theory in both observing-response and concurrent-chains procedures. We suggest that the similarity of these predictions might offer an analytically grounded reason for why Delay Reduction Theory has been a successful theory of conditioned reinforcement. Finally, we suggest that the approach provides a formal basis for the assertion that conditioned reinforcement results from Pavlovian conditioning and may provide an integrative approach encompassing both domains.
Journal of the experimental analysis of behavior, 2015 · doi:10.1002/jeab.142