When Brown and Jenkins1 first described auto-shaping of pigeons' key-pecks, they introduced the experimental protocol as one in which delivery of reinforcement was related to or "conditional on stimulus values but not on responses." Thus, the close temporal relationship between back-lighting a standard pigeon key and presentation of food, which engendered various behaviors directed toward the key, including eventually pecking at it, became the basis for autoshaping.
After magazine training, pigeons were exposed to 8 s of illumination of the key-light and 4 s access to a tray of pigeon grain upon the offset of the light. Intertrial intervals (ITI) generally varied randomly between 30 to 90 s, with a mean ITI of 60 s. One additional condition (protocol) used included the termination of the keylight and the immediate presentation of 4 s access to reinforcement. The third condition included a sort of punishing operation whereby a key-peck during the ITI delayed the presentation of the lighted key for 60 s. Brown and Jenkins1 also explored the consequences of backwards conditioning (reverse pairing); if the pigeon pecked the lighted key, which was illuminated for 8 s after presenting the food tray for 4 s, the key-light was turned off and the pigeon again had access to 4 s of grain reinforcement. The forward pairings resulted in the emergence of behaviors near and, within two sessions of 80 presentations each, all 36 subjects eventually pecked the key at least once. Reverse pairing resulted in only 2 of 12 subjects pecking at the key within 160 trials.
In the discussion of this seminal paper, the authors argue in favor of a classically conditioned (first) key-peck response through either some sort of stimulus substitution, since pecking is presumably innately directed at the unconditioned stimulus (grain) used as the reinforcer or because intermittent presentation of a stimulus (keylight) followed shortly thereafter by non-contingent presentation of food (until the very first key-peck) leads to a multitude of (classically conditioned or superstitious?) motor behaviors, near and at the key, while it is lighted during an 8 s or 3 s trial. Brown and Jenkins1 finally offer the procedure as a time and labor-saving, standardized method for obviating influences brought about by "individual differences among experimenters in the art of hand-shaping." This is one of the salient characteristics of the procedure that initially drew our attention to its potential utility.
Thus, "... autoshaping paradigms provide an unusually good opportunity to obtain parametric data on variables that influence the acquisition of such behavior, an issue that Williams refers to as the psychophysics of association."2 Attempts to attain such a goal and theoretical issues and arguments related to the concept that autoshaped behavior is primarily a classically conditioned behavior, an instrumen-tally conditioned behavior, a bridge between the two (S-R vs. R-S contingencies) or a new class of behavior with components of each, can be gleaned from the many chapters in the book edited by Locurto, Terrace and Gibbon3 entitled Autoshaping and Conditioning Theory.
Although the majority of studies dealing with the phenomenon of autoshaping have used a procedure whereby delivery of reinforcers was not contingent upon a specified response, my own laboratory has taken advantage of the automaticity and lack of experimenter-induced bias conveyed by autoshaping and we too have utilized a response-contingent component whereby the subject (normally a rat but we, as others4 have, used autoshaping with chickens as well) is given the opportunity to manipulate its environment (e.g., touching the extended lever, initiating its retraction earlier than programmed to do so noncontingently) leading to the delivery of a reinforcing stimulus (e.g., food) slightly sooner than its delivery in the absence of an operant (lever touch) response. For protocols we generally use, the subject cannot earn more reinforcers during a daily session in which 12 to 30 or more trials are presented and their session lengths cannot be shortened by more than a fraction, thereby most likely obviating a shortened session length as a reinforcing outcome (i.e., trials without an autoshaped response typically are comprised of the presentation of a retractable lever for 15 s, followed by retraction and either immediate delivery of a reinforcing stimulus or delaying the reinforcer for up to 8 or 9 s) followed by a random or fixed time ITI, typically averaging 45 s. By introducing a delay of reinforcement we believe we have introduced a so-called working memory component as a variation of the task and this has enabled us to demonstrate selective and specific effects of neurotoxic insults to the hippocampus and associated structures (vide infra).
As is often the case, behavioral toxicologists, behavioral pharmacologists and psychopharmacologists borrow procedures and protocols from the experimental and cognitive psychologists and because of their training and reliance upon concepts like dose-response relationships, often discover a need to systematically manipulate antecedent and consequential variables which had not previously been carried out adequately by others. Likewise, it may be necessary to modify the typical physical environment used by others so that drug-behavior studies can be carried out in such a manner that ancillary measures of unconditioned behaviors can also be recorded concurrently, either automatically or via closed circuit television monitors and video recorders for later off-line analyses.
Because this contribution is for a book recommending and describing methods, I have chosen to discuss the advantages and disadvantages of the variations of autoshaping procedures we have used over the past 25 years. I have also included comments on theoretical and philosophical issues, where apparently appropriate, and discussions or comments about the technology used and choice of data collection, reduction, and statistical analyses.
The last points are not especially restricted to autoshaping as a behavioral paradigm, or to behavioral neuroscience in particular, but are important enough for a few comments at the outset because it has been my experience, in common with some colleagues, that as technological advances have enabled us to collect more and more data, at faster rates, not enough attention has been paid to the importance of experimental design, how the voluminous data will be handled once they are collected and the validity or appropriateness of whichever statistical analytical procedures are used. In my laboratory this is a continuing, evolutionary process and we continue attempting to devise/design experiments which enable us to rely upon preplanned statistical contrasts. To the credit of most psychologists and behavioral neuroscientists, at least reasonable attempts at dealing with these issues are more commonplace than has been my experience when attempting to interact or collaborate with colleagues of a more reductionist persuasion. I have, on more than a few occasions, asked a colleague who points to the obviously more intense band or spot on a gel or chromatogram, derived from a single experiment (single subject's or pooled tissue) that it is also obvious (at least living in the midwestern plains of the U.S.) that the world is flat. All one has to do is to look out of the window to confirm it. Of equally great concern (frustration?) is trying to convince a colleague that 5 to 10 experiment(s) carried out with xenopus oocytes expressing one or another receptor, but upon oocytes derived from a single female frog, is nothing more than one experiment with 5 to 10 replications and that standard parametric or nonparametric statistical analytical procedures designed for independent observations from individuals (or parts thereof), randomly assigned to one or another treatment condition, are not appropriate, even if he/she is willing to consider using such statistical contrasts.
Over the years we have used variations of autoshaped behavior acquisition and/or maintenance to study the actions of drugs or toxins upon this class of behavior(s) in mature experimental subjects, either exposed to a drug or toxin insult during development (e.g., in ovo [chick] or in utero [rat] ) or after exposure to drugs, withdrawal from them or exposure to toxins postnatally, at a more mature age. Other laboratories have followed our lead and have used identical or similar protocols to study the autoshaped acquisition of intravenous cocaine self-admin-istration,5 to study the effects of experimenter-administered cocaine upon autoshaped behavior acquisition,6 and to study effects of pyrogens/cytokines, which reportedly specifically disrupt or interfere with acquisition of food-reinforced autoshaped behavior based upon the fact that the same behavior, once established, was not affected by the same treatment.7
It has been the policy of my laboratory to attempt to give something back to the discipline from which we have borrowed theoretical and/or methodological underpinnings. I believe our contributions in this regard can be gleaned to some extent from the publication list appended to this chapter. Thus, I have chosen to emphasize some of the ways we have devised to maximize the use of autoshaped behavior in order to control for as many potential confounding variables which can interfere with design, implementation, and, most importantly, the interpretation of data derived from such studies. Almost of necessity, it is a quasi-chronology of the evolution of protocols currently in use in my laboratory (and those of colleagues who have asked for advice or patterned their work from one or more of our publications in which autoshaped behavior was one of the protocols we utilized).
Was this article helpful?
Being addicted to drugs is a complicated matter condition that's been specified as a disorder that evidences in the obsessional thinking about and utilization of drugs. It's a matter that might continue to get worse and become disastrous and deadly if left untreated.