Predictable Peckers

The entertaining article, “The Many Lives of Superstition” (Observer, October 2008), promotes an erroneous interpretation of superstitious behavior by attributing it to accidental pairings of a response with a reinforcing stimulus (i.e., response-reinforcer contiguity). This is the explanation of superstition proposed by Skinner (1948) to account for the idiosyncratic behavior of his pigeons in a now-classic experiment. Although this explanation has intuitive appeal, it is incorrect. As subsequent research has shown, contiguity is not sufficient for learning, and the behavior contiguous with reinforcement isn’t idiosyncratic after all. Contiguity, accidental pairings of a response with a reinforcer, is not sufficient for superstitious behavior to develop. This has been demonstrated by numerous experiments that have inserted “free” reinforcers between pairings of response and reinforcer (e.g., Hammond, 1980). That is, if the probability of the reinforcer is equally likely following a response as when no response occurs, then superstitious responding does not occur. Why should it? Because, according to the contiguity view, the number of response-reinforcer coincidences remains the same. However, this view overlooks the fact that, in the “free” reinforcer experiment, the “contingency” between response and reinforcer has been eliminated. That is, learning only occurs when a response increases the probability of reinforcement relative to that probability when no response occurs. If key pecking or lever pressing or turning in a circle or wing flapping doesn’t increase the probability of food delivery compared to when these responses are not made, then these responses don’t occur reliably.

What about the idiosyncratic nature of the responses Skinner observed? It turns out that, upon a close examination of the pigeons’ behavior, the responses weren’t very idiosyncratic after all. In Skinner’s experiment, food was programmed to occur at regular intervals without regard to the animal’s behavior. Under these conditions, Skinner observed that each pigeon developed a different response in the chamber, which he attributed to adventitious reinforcement of those responses. However, replications of the experiment by other investigators (e.g., Staddon & Simmelhag, 1971) who kept detailed records of the pigeon throughout the experiment observed that, at the time of food presentation, all pigeons behaved in the same way – not idiosyncratically at all. That is, around the time of programmed food delivery, all pigeons pecked in anticipation of food delivery. Why? Because that is what pigeons are biologically predisposed to do when food is imminent. The birds had learned to expect food because it was presented at regular intervals (i.e., Pavlovian temporal conditioning) and so they pecked. However, in the intervals between food deliveries, each bird engaged in idiosyncratic “interim behavior.” Interim behavior was not the result of the pairing of food with responding since food and the interim behaviors were not coincident. Rather these idiosyncratic behaviors occurred at times when the probability of reinforcement was low (e.g., just after a reinforcer) and pecking was not occurring. Although we don’t know precisely why pigeons engage in interim activities (cf., Staddon and Simmelhag 1971), we do know that it’s not the result of adventitious reinforcement. In other words, it’s not “superstitious” behavior.