The material to follow may be unfamiliar to some readers.

Most of the introductory conditioning or animal learning literature has been written from the behaviorists’ perspective, leaving the more contemporary cognitive achievements largely undocumented at that level (1). However, the behaviorists have failed to delineate sufficiently the determining factors that produce learning, the essence of that learning, and the ways in which that learning influences behavior.

Early behaviorists based psychology on observable events and made the simplified assumption that the unobservable underlying psychological and neurological processes of learning corresponded isomorphically with observable events. For example, if in a particular stimulus situation, the data showed a change in behavior, behaviorists assumed that it was a corresponding stimulus-response connection that was learned. If the data showed a gradual strengthening of a particular response in a particular stimulus situation, it was assumed, based on the gradual strengthening, that subjects did not know they were learning. Learning was thought to be automatic and unconscious, simply consisting of a gradual strengthening of stimulus-response connections (unconscious response habits that were triggered by the associated cue). Additionally, since it was a change in behavior that was observed, it was assumed that it was a change in behavior that was learned. They supposed that stimulus-response connections were automatically and directly established either by simple contiguity between a stimulus and a response (in classical conditioning) or by a combination of contiguity and the strengthening of a response through reinforcement and repetition (in instrumental conditioning). Furthermore, it was assumed the presentation of an effective outcome that followed the stimulus and response served only to reinforce the connection between the stimulus and response. The reinforcer supposedly did not enter into the associative structure; it functioned only as a catalyst to unite the link between the stimulus and response. Traditionally, behaviorists thought that all learning, for all animals, under all circumstances involved the direct establishment of stimulus-response connections and that all instances of intelligent behavior could be explained based on conditioned responses to specific stimuli; output generated by a particular input. All behavior, including the workings of the brain, could be reduced to a set of muscular movements or reactions to a set of stimuli. They argued that behavior could be understood without appeal to concepts such as knowledge, beliefs, intentions, or other cognitive influences. Subjects did not know why they were performing the response or what would happen when they responded. They did not acquire knowledge that a particular stimulus signaled an increase in the probability of a particular outcome or that a particular response would cause the occurrence of a particular outcome.

Fortunately, it is now recognized that associative learning is comprised of a much richer array of associative processes than textbook descriptions typically offered of classical and instrumental conditioning. There is now a good body of evidence indicating that conditioning does involve the acquisition of knowledge and beliefs about the world, which forms the basis of behavior. Animals use their acquired knowledge to guide their choices and actions. During conditioning, involving a reinforcement contingency, animals are not learning a new response; they are learning that a particular response causes a particular outcome or that a particular stimulus signals an increase in the probability of a particular outcome. The response is a manifestation of their knowledge about the relationship between the events the trainer has contingently arranged, the cause and effect relationship. Moreover, the establishment of a causal relationship or successful conditioning is not simply dependent on the number of times a particular stimulus or response and reinforcer have been paired together in a timely fashion. The course of conditioning is also affected by the subject’s prior learning experience with that particular event and reinforcer, by the presence of other events signaling that reinforcer, and by the subject’s past experience with those other events. Although contiguity is important to help the subject detect a relationship, it is only an indicator for the s of a probable causal relationship between the to be conditioned event and reinforcer. Even if the reinforcer follows the to be conditioned event immediately after its occurrence, the subject will not necessarily associate the to be conditioned event with the reinforcer. For successful conditioning to occur, the to be conditioned stimulus or response must provide information about the occurrence of the reinforcer better than other events. Conditioning cannot be reduced to a simple affair in which the subject can only register contiguously conjoined events. If this were the case, animals would not be capable of distinguishing the events that are better correlated with a valued outcome from those that are less correlated. Their associative system would be at the mercy of the coincidental conjunction of events and they would not be able to distinguish the true cause or predictor of an outcome from other environmental events that are only fortuitously related to that outcome. As such, their view of the world would be very distorted. Luckily for them, animals are considerably more sophisticated than what was once envisioned.

Leading scholars in the field of animal learning and conditioning now believe that animals register the antecedent events of an outcome and then, in an effort to reduce discrepancies, evaluate their predictive value relative to the other events that have been contiguously paired with that outcome, and then select the event that better predicts that outcome. In other words, animals attribute events of consequence to their most probable causes by associating selectively the better predictors of an outcome, at the expense of poorer predictors. Selective or discriminative knowledge, comprising a broad range of environmental relations, is thought to be one of the main ways animals represent the structure of the world (Dickinson, 1980; Mackintosh, 1983; Rescorla, 1988).

In nature, it is often important for animals to differentiate or discriminate the environmental stimuli that are the better predictors or signals of biologically important events from those that are not. When animals learn to discriminate the events that better signal important outcomes from those that do not they are learning to attend selectively to certain environmental stimuli. This voluntary selective attention is the result of the learned predictive relationship between the stimuli and their outcomes (reinforcers). That is, animals learn to attend selectively to the stimulus events that predict or inform them of future events of importance and ignore less informative, redundant, or uninformative events. They learn which environmental stimulus events better inform them of a valued outcome and which are less or uninformative. Selective attention is therefore an important determinant of discrimination learning.

Now that animal learning is viewed as the acquisition of knowledge about the world, the arrangement of training procedures can be regarded as a way of imparting information to the subject that they can use to influence their decisions and work out effective actions. Therefore, an important prerequisite to discrimination training is an understanding of how animals learn about the relations between events and the conditions in which animals select the environmental stimuli that are better sources of information from those that are not. With this knowledge, trainers will be in a better position to arrange training and solve problems that can ultimately result in a better service dog.

This website is a work in progress and is subject to change, but my intention is to outline first the ways in which knowledge about the relations between events is gained. Then explain what discrimination training really involves, what animals are learning during the course of discrimination training, and offer a user friendly, applied cognitive science curriculum for scent discrimination training that will gain and maintain attention to human scent and increase accuracy.

Note:

(1).  By and large, detailed reviews of contemporary research and theories can only be found by searching through research or review articles. There are some books on the topic but most of them have not been written at an introductory level. Even at higher levels, texts reviewing the various learning theories tend to extensively cover the traditional theories while only minimally reviewing the disputing evidence and more contemporary theories.

References:

Dickinson, A. (1980).  Contemporary animal learning theory.  Cambridge University Press, Cambridge.

Mackintosh, N.J. (1983).  Conditioning and associative learning.  Oxford University Press, Oxford.

Rescorla, R.A. (1988).  Pavlovian conditioning: It's not what you think it is.  American Psychologist, 43, 151-160.