What is the stimulus that increases the probability that a proceeding behavior will occur again?

Pitfalls of Stimulus Discrimination Training

A discriminative stimulus (SD) is a stimulus that predicts reinforcement whereas other stimuli (SΔ) do not predict reinforcement. Such stimuli are said to ‘control’ behavior because organisms behave differently in the presence of such SD stimuli compared to their absence. For example, we stop at red lights and go when the light turns green. We obey traffic lights because they promote our safety.

Differential reinforcement is a procedure that brings about stimulus control. For example, parents that provide positive consequences for compliance with their directives and withhold access to positive reinforcers or administer aversive consequences for non-compliance with their directives will find that their children ‘listen’ to them which is how parents talk about children ‘obeying’ them.

Common examples of discrimination training pitfalls can be found in parent–child interactions. For example, a parent may ask a child to remove their books from the table so that the family can have dinner. The child ignores this request and continues what they are doing; perhaps watching TV. The parent asks the child a second time to remove their books from the table. Again the child ignores this request and continues with what they are doing. Sometimes a third or fourth request is made in a louder more threatening voice, but often to no avail. Finally, the emotionally distraught parent yells a final threat and then the child removes their books. Since no further consequences follow for the child’s behavior, the child learns to wait for the final threat. In addition to experiencing these exchanges in my clinical practice, I witnessed a variation of this pitfall that took place at a community pool. An older sister told her younger brother that he should hurry up and change his clothes so that she could drive him home soon. He said that he would. But he took his very sweet time taking a shower and getting dressed. Throughout this time I could hear the sister threaten to drive home without him thus making him walk a modest distance. But she didn’t. Instead, she begged and pleaded with him to hurry up, which he didn’t. Finally, he was finished and met her outside the dressing room where she continued to scold him as they walked to the parking lot. He learned to ignore her demands because they were not correlated with any consequences.

The Importance of Developing Stimulus Control over Academic Responding

Academic tasks constitute SDs designed to evoke a response of some type. The desired response is ultimately to be brought under stimulus control of the academic task itself alone without other forms of assistance (e.g., hinting at the correct response or pointing out clues that increase the salience of some of the stimuli; Vargas, 1984). The response class in question can be defined with varying degrees of specificity, depending on one’s purpose. To establish the generality of the principles governing behavior in this case, one might refer broadly to the response class as “academic responding,” understanding that it encompasses a wide variety of topographies (e.g., reading a word on a page, answering a math problem) that do, however, retain a common distinctiveness relative to other response classes (e.g., disruptive classroom behavior). This perspective is helpful to understanding at a strategic level that the stimulus relations that should control responding for everything from reading a letter on a page to correctly performing a chemistry experiment (and all things in between) are grounded in this single basic behavioral process of stimulus control. A narrower definition of the response class in question (e.g., oral reading fluency, completing math computation problems) is more helpful when a specific problem has been identified and the basic behavioral processes have to be operationalized at a tactical level for a given student. For example, the topography of prompting methods used to increase oral reading fluency will look different from those used to increase correctly written responses on a math worksheet. Yet, each form of prompting when used appropriately works effectively for the same functional reason.

Two distinctive features of academic responding have implications for designing instructional interventions. First, responding is at low (or even zero) levels when the teacher begins instruction. Such low response levels may also occur when a consulting team is called upon to help a teacher develop an intervention for a student who is not progressing adequately following instruction. The key issue is that the teacher is dealing with a behavioral deficit, which means functionally that there are no controlling variables at the present time. By contrast, for a behavioral excess like self-injury the problem is the very existence of controlling variables that maintain the problem through evocative and reinforcing events. It is possible to conduct a functional analysis with self-injury as the target behavior because the controlling variables are discoverable. For an academic performance problem, there are no maintaining contingencies to investigate. Indeed, the problem to resolve is the lack of maintaining contingencies. Thus, the fundamental goal of instructional intervention is to establish stimulus control which is developed through DR (Cooper, Heron, & Heward, 2007).

Another distinctive feature of academic responding is that it includes a myriad of interrelated behavioral repertoires. These repertoires are specified in advance by an educational curriculum the objective of which is to prepare students to be successful in the “real world” where reinforcement schedules and other relevant variables are not tidily packaged. Helping prepare students for what comes after the educational curriculum (e.g., a job, college) is obviously a long process, given the complexity of the skills that these environments require. The curriculum specifies a continuum of behavioral repertoires that begin with the most basic skills and progress toward closer and closer approximations to real-world activities meant to prepare the student for when he or she encounters similar situations beyond school. As the student ascends through the curriculum and instructional tasks become increasingly more difficult, he or she must be able to coordinate multiple behavioral repertoires smoothly and in a timely fashion to be successful, which amounts to getting the right answer “on demand.” Teachers jump into a spot along this continuum by isolating and targeting specific subordinate or component response classes for instruction (e.g., teaching to segment phonemes versus teaching to improve oral reading fluency). This feature makes establishing both the validity and sequencing of component response classes absolutely critical. Targeting a wrong or unnecessary behavioral repertoire or incorrectly sequencing how they are targeted may slow down or stall the efficiency with which stimulus control develops while potentially having a negative cumulative effect on the terminal goal of real-world preparation. The idea of isolating a particular component response class for instruction is akin to taking a “slice” along the continuum of response repertoires, the purpose of which is to strengthen it in such a way that learning at later points in the curriculum is easier and more efficient for the student.

Achieving a level of responding that makes it possible for the student to integrate the newly learned skill with other skills and as a part of more complex tasks with minimal instruction is an indication of stimulus control. The problem is that the student has been referred for a lack of progress following instruction with the current skill, which is often an indication that prerequisite skills have not been sufficiently developed (Howell & Nolet, 2000). Continuing instruction at the current level may create cumulative skill deficits that attenuate future instructional effects and make school tasks progressively harder (Binder, 1996; Johnson & Layng, 1992). Wolery, Bailey, and Sugai (1988) recommend either “slicing back” to an easier version of the skill (e.g., teaching addition problems with sums to 10 rather than with sums to 18) or “stepping back” to a prerequisite skill (e.g., teaching a student to blend and segment phonemes before giving phonics instruction). The reader is referred to Carnine, Silbert, Kame’enui, and Tarver (2010), Stein, Kinder, Silbert, and Carnine (2005), and Howell and Nolet (2000) for well-established sequences of basic academic skills.

VII.D Recombination of Discriminations

Novel behavior results when discriminative stimuli controlling different responses or response dimensions are combined in novel ways. Recombinant generalization underlies some instances of problem solving, instruction following, and productivity in language. Figure 21 illustrates an example taken from the literature on language training for children with mental retardation. In this study, the child was first taught to push the 12 objects when given the instructions “push the glass,” “push the scissors,” and so on for the remaining nouns. Then, the next verb, “drop,” was taught in the same way. As more verbs were taught, the child soon generalized new verbs to the objects after training with only one noun. Recombination processes have been demonstrated in a variety of species and in other contexts, including productivity and problem solving (e.g., cases in which a solution requires testing different combinations of elements).

6.3 Sequencing of training sessions

In order to establish valid measures of the discriminative stimulus effects of drugs, subjects must be conditioned to cue exclusively to drugs and not at all to other stimuli present in the environment. The temporal sequence of training sessions with drugs and vehicles is an important tool in striving for this ideal situation. The order of sessions must be such that subjects cannot obtain reinforcers by learning the sequence of training sessions. Simple alternation of drug and vehicle (or other drug) sessions is precluded because the sequence can be learned too easily. It is also important that subjects cannot detect which response is correct from cues left by previously trained subjects; olfactory cues can control the behaviour of different rats trained later in the same chambers (Extance and Goudie, 1981).

Two solutions seem to be in widespread use. One approach is to present drug and vehicle training sessions in a random order, usually with a restriction that not more than three sessions take place in succession with either drug or vehicle. Different random sequences should be used for each subject, so that the correct response in a particular session bears no relationship to the performance of subjects trained previously in the same apparatus. An alternative approach that appears equally satisfactory is to use complex sequences of alternating drug and vehicle sessions. By this means, over a period of 8 weeks, subjects can be trained with nearly all possible sequences of stimulus conditions, thus preventing the formation of expectancies associated with simpler sequences (Colpaert et al., 1976). Furthermore, at least two different sequences must be used for different subjects, so that cues left by subjects trained first do not influence the performance of later subjects; these different sequences should be interchanged after each is completed.

Drug Discrimination

The ability of a drug to serve as a discriminative stimulus is thought to result from the generation of interoceptive cues produced by the drug, which can be modeled in laboratory animals using the DD paradigm. DD serves two important functions. First, by comparing a novel drug’s discriminative stimulus effects to those produced by a known drug of abuse, DD provides a method of assessing the abuse liability of new drugs. Second, by attempting to potentiate or attenuate the discriminative stimulus effects via agonist and antagonist treatment, DD paradigms allow for the study of CNS mechanism of drug action. Further adding to its utility, human DD studies have yielded findings similar to those of non-human studies.

During quantitative generalization testing, the dose of the training compound is varied. Unlike under training conditions where only responding on the correct lever is reinforced, under testing conditions responding on either lever is reinforced. This is an important distinction between training and testing conditions since reinforcement (or lack thereof) following the first choice could serve as the cue to the correct lever choice for the remainder of the behavioral session. Test data are often represented graphically by sigmoidal curves where percent drug-appropriate responding values correspond with ordinate coordinates and drug dose values with abscissa coordinates. At very low doses, the subject usually emits the majority of the responses on the vehicle-appropriate lever, with increasing doses resulting in greater responding on the drug-appropriate lever.

In a study that compared possible route-related differences in the discriminative stimulus effects of cocaine, the generalization gradients of cocaine were examined under three different routes of administration. Rhesus monkeys were initially trained to discriminate intramuscular (IM) injections of cocaine from saline and, when stable, test sessions were conducted with different doses of cocaine administered using the IM, intravenous (IV), or intragastric (IG) route. Interestingly, while the shapes of each generalization gradient were comparable – suggesting shared discriminative stimulus effects regardless of route, the IV and IM curves nearly fell on top of each other, but the IG curve was shifted to the right, reflecting a 40-fold decrease in potency compared to the other two routes. One important caveat demonstrated by this study is that the time of the onset of the drug effect must be taken into consideration. For example, in humans it was reported that the subjective effects of oral cocaine were not present until at least 30 min after delivery. This is in contrast to the subjective effects of IV and intranasal cocaine, which peak within 5–10 min after administration. Had the investigators studying the discriminative stimulus effects of cocaine in rhesus monkeys not waited 60 min post-IG administration to conduct test sessions, they would not have reported that orally administered cocaine shared discriminative stimulus effects with cocaine delivered by IM and IV routes. A similar study in squirrel monkeys trained to discriminate IV cocaine from saline found that the IV, IM, and inhalation routes all produced similar generalization gradients. However, this study demonstrated that IM cocaine was approximately half as potent as IV cocaine. While this finding might be partially attributed to subtle species differences, it is more likely a result of the different training conditions.

In addition to quantitative generalization determinations as described above, several studies have utilized DD to examine the qualitative nature of the discriminative stimulus effects of cocaine. In general, test compounds pharmacologically related to the training drug will substitute for the training drug and compounds from different pharmacological classes will occasion vehicle-appropriate responding. Using DA compounds as an example, the indirect-acting DA agonists amphetamine, GBR 12909, and 2-β-propanoyl-3-β-(4-tolyl)-tropane (PTT), and various direct-acting DA D1- and D2-like receptor agonists will substitute for cocaine. Similarly, giving DA antagonists in combination with cocaine should attenuate the cocaine-like discriminative stimulus effects resulting in a rightward shift in the cocaine dose–response curve. DA D1- and D2-like receptor antagonists robustly produce these shifts.

One caveat in interpreting generalization gradients is that the magnitude of the training dose itself can affect the shape and position of the dose–response curve. For example, in rats trained to discriminate d-amphetamine, it has been demonstrated that lowering the training dose resulted in a parallel shift in the discrimination gradient to the left. In addition, training dose also affected the time course of drug effects. While lowering the training dose seems to increase sensitivity to the training cue, there is also evidence for accompanying decreases in specificity. In one compelling study, rats trained to discriminate cocaine (10 mg kg−1) from saline in a two-lever food-reinforced DD paradigm were then subsequently retrained to discriminate progressively smaller training doses of cocaine (0.16–5.0 mg kg−1) from saline. At each training dose reduction, the number of sessions required to meet the discrimination criterion increased, and the number of rats acquiring discrimination decreased. However, those rats that acquired discrimination at the lower training doses exhibited increased sensitivity to the cocaine cue as demonstrated by leftward shifts in dose–response curves. Additionally, the apomorphine generalization gradient was shifted up and to the left when the cocaine-training dose was reduced from 10 to 2.5 mg kg−1.

While human studies of cocaine discrimination are beyond the scope of this chapter, it is important to note that there are cases when the subjective effects of cocaine and the reinforcing effects do not match. The relationship between the discriminative stimulus effects and the reinforcing stimulus effects of cocaine was recently addressed in monkeys. In that study, experimentally naïve male monkeys were first trained to discriminate self-administered cocaine from saline under a chain schedule of reinforcement. That is, the first link in the chain involved an FR contingency leading to an injection of either saline or cocaine, followed by a typical discrimination component (the second link of the chain schedule). Monkeys could self-administer the injection they just discriminated in the third link of the chain schedule; injections were available under a PR schedule of reinforcement. In all three monkeys, there was at least one cocaine dose in which the animal discriminated it as “saline-like” yet functioned as a reinforcer when self-administered under the PR schedule, suggesting that discriminative stimulus and reinforcing stimulus effects do not provide similar behavioral or pharmacological information.

Although the process of training cocaine discrimination often requires weeks to months, neither tolerance nor sensitization to the interoceptive cues produced by cocaine appears. However, procedures that modify the experimental design by suspending discrimination sessions and administering drugs chronically have resulted in apparent tolerance. In one experiment, rats were first trained to discriminate IP injections of cocaine from saline. After acquisition, animals were subjected to a 7-day period of three daily injections of cocaine in the absence of daily sessions. Following chronic cocaine administration, cocaine dose–response curves were re-determined and there was a shift to the right in the dose–response curve, suggestive of tolerance to the discriminative stimulus effects of cocaine. Importantly, subjects eventually recovered their original sensitivity once chronic administration had been discontinued. In fact, the investigators demonstrated that tolerance was lost at the same rate it was acquired.

Is any stimulus that increases the probability that a proceeding behavior will occur again?

What increases the probability that a behavior will be repeated?

What is a stimulus that precedes a behavior?