Note: This article originally appeared in the October 2015 issue of Behavior Analysis Quarterly, Vol. 1(3).
By Ben Witts, PhD, BCBA-D
St. Cloud State University
A student recently brought to my attention a twist to the typical token economy that she used in a clinic where she works with young children diagnosed with autism spectrum disorders. The twist, which I reveal in a moment, was so well-received by staff that it became company policy to alter the way token boards were developed at the site. Historically, this clinic had been using tokens of the same color to mark progress toward some putative reinforcement. However, it was argued that by changing the color of the final token, that final token’s effectiveness would be enhanced. For example, if a child was working for 5 tokens to earn computer time, the first four tokens would be yellow with the fifth token blue. The rationale was that the blue token would be more reinforcing than the yellow, adding an additional reward, making the token board more effective and fun for the client.
There was a major flaw with their idea, however, and it relates to work started nearly half a century ago. While the blue token might be more valuable in terms of reinforcement efficacy, this twist to the token board will actually make the yellow tokens less valuable than if all tokens were yellow! The 50-year-old idea that leads us to this conclusion is based on Edmund Fantino’s work on delay reduction, and I will recount some of his work to help illustrate why it’s better to use all yellow tokens than to switch up the colors.
Assessing Conditioned Reinforcement
There are many ways to determine if some stimulus is reinforcing. However, the practitioner, often faced with a very demanding work environment, must sacrifice months of well-controlled experimental work for a more pragmatic approach. In a sense, the practitioner just wants to know if something works as reinforcement, and this is often proven by seeing good on-task behavior and progress toward skill acquisition. In our two-colored token board, it’s likely that both criteria—good behavior and learning—are met. But we must remember that even poor reinforcement can result in good behavior and acquisition, particularly with children who have a history of responding to tutor demands. We often attribute a generally-compliant way of being to instructional control, and this might have more to do with a history of punishment for not responding or responding inappropriately than it does with the quality of reinforcement. In other words, children learn to do as they are told, or else.
While we may not have the luxury of articulate, repeated reinforcement assessments in our practice, we can draw from work that does have that luxury. Starting in 1969, Edmund Fantino published early work just what makes conditioned reinforcement reinforcing. This work culminated in the delay reduction theory (DRT). DRT states that, all things being equal, the more powerful conditioned reinforcers are those that are more highly correlated with a reduction in time to some primary reinforcement. For example, most employees prefer 4:50pm over 1:30pm as 4:50pm is more closely related to leaving (assuming a 5:00 end to the workday). Note that the clock in itself need not be reinforcing, but its value as a potential reinforcer changes with time. Consider, too, a young child’s behavior as a favorite holiday season approaches. First a change in outdoor temperature, then festive decorations go up, then holiday shows and movies come on television, and finally wrapped gifts are found to congregate around a slowly dying conifer. Each of these elements just described helps to predict that the opportunity to open gifts is approaching, and the culminating effect, here on an interval schedule, might produce some adjunctive behavior in the child. The point is clear, I hope: we tend to like those things that tell us it’s almost time for reinforcement.
Perhaps it’s best if we review what is essential about DRT through Lewis Gollub’s 1958 thesis (as cited and recounted in Fantino, 2008). In his study, Gollub arranged two equal chained schedules to obtain food. Both schedules required pigeons to complete five FI 1 minute schedules to earn access to food. One schedule was a tandem schedule, which simply means that all five FI components were accompanied with the same colored key light (discriminative stimulus). In the chain schedule each FI component had a different colored key light. So in the chain schedule the light might switch from white to red, then blue, then green, and so forth. The tandem would remain the same, perhaps just white throughout. Pigeons by far prefer the tandem schedule, even though only one conditioned reinforcer is presented. The reason? If both tandem and chain schedules start with a white key, the chain schedule’s white key essentially says, “Hey buddy, you got a ways to go to get reinforcement!” Alternatively, the tandem’s white key says, “Hey, this might be the one that leads to reinforcement!” Get it?
So what’s wrong with the Blue Token?
There’s actually nothing wrong with the blue token on its own. However, the blue token is the token that best predicts reinforcement is coming. Under this preparation, blue tokens are good, and yellow tokens are indicators that the blue token, and thus reinforcement, are a ways off. In other words, the yellow tokens predict a period void of reinforcement. Tokens should be a form of conditioned reinforcement, and here it borders on conditioned punishment!