LOGAN'S LITERATURE AND LAWS OF LEARNING 12/31/99 QUAD-L DATABASE: Preface Although it is a challenge to compile the literature and to write root-abstracts of those publications, the primary objective of the Quad-L is to organize this knowledge into some comprehensive and comprehensible system. The most familiar organizational principle involves the use of key words. This approach, combined with logical operators, enables one to locate many publications relevant to a topic of interest. The Quad-L is not intended to replace the familiar use of search engines on databases of abstracts. Indeed, key-word searching is certainly possible with the Quad-L database. For several reasons, I think the key-word approach is inadequate. First, sometimes what is basically the same operation has been called by different terms by different people. As a result, we all have failed to cite others and failed to be cited by others. Conversely, the same terms have often been used for very different operations. As a result, we may have to wade through innumerable irrelevant citations in order to locate the relevant ones. Furthermore, the key-word approach permits the un- bridled proliferation of categories without any explicit organization. As a result, the system may reflect the same disunity that plagues the literature. But most importantly, key words do not encourage looking for relationships among operations that may suggest a larger context into which to conceptualize a topic. As something of an antidote, I have made a preliminary attempt to develop a quasi-logical organization of the categories. I hope to improve upon this system myself as I expand the analysis to include other learning paradigms, but I also hope to encourage others to develop other systems. In the spirit of "carving nature at the joint," the goal is to develop for our field a system somewhat analogous to the Periodic Table of Elements in Chemistry. Toward a Taxonomy of Operations in Basic Learning, Motivation, Memory, & Cognition A systematic organization of the scientific literature, in this case primarily of the behavior of non-human animals, could have several effects: a) It could promote appreciation and comprehension of the wealth of phenomena in the domain and the context of any particular topic within that domain, b) It could encourage and facilitate integrating conceptually related topics in spite of differences in methodology and terminology, c) It could both challenge and inspire the development of general theories of behavior. Against these lofty goals, the Quad-L (Logan's Literature and Laws of Learning) is woefully inadequate. Not only is it primitive, it is haunted by prejudices and predispositions and offers no really exciting insights that "carve nature at the joint." Hence, it is at best a first approximation to a taxonomy of operations and it begs to be superceded by a better system. In order to describe the rationale of the Quad-L system, let us imagine playing a variant of the game, "Twenty Questions." In this case, the target is some topic in the aforementioned domain and the objective is to identify the target in as few binary questions as possible. The questions posed by the Quad-L are as follows: Question 1. Excitation or Inhibition? Because of my background in animal learning, I immediately divided the domain into the two basic phases of acquisition and extinction. However, it soon became apparent that the change to continuous nonreinforcement (extinction) is only one of many possible changes in the conditions or parameters of learned behavior. One class of such changes is traditionally called "generalization," and others are typically called, "transfer." Then, if one combines generalization and extinction, one has the paradigm for "differential conditioning." There is no ideal term for the combination of extinction, generalization, transfer, and differential conditioning but "Inhibition" seemed most nearly appropriate. I do not mean to imply a theoretical explanation of these phenomena, merely operations that imply some form of decrement in performance. In keeping with that decision, the early term "Acquisition" was changed to "Excitation." (This first distinction is captured in the second letter of the code preceding each root-abstract. A capital letter is used for the "excitation" topics, and a lower-case letter for "inhibition" topics.) The target topic does not fall into the four-fold categories of extinction, generalization, transfer, and differential conditioning, and hence it is classified as an Excitation topic. Question 2. Independent or Interdependent Variable? Although few experimental psychologists have made the distinction, I believe it is conceptually important to recognize whether an experimental variable is truly "independent" in the sense of occurring without regard to the behavior of the organism. Food as a US in Pavlovian conditioning is a good example of such a variable, but food as reinforcement in operant/ instrumental conditioning is an interdependent variable because it can only occur if the organism emits the designated behavior. Coping with interdependence is a conceptual challenge. There are two caveats to be made in this distinction. First, even an independent variable may require some orienting or other response to be effective. For example, in the original Pavlovian preparation, the dog had to go to the food trough and eat the food for it to serve its intended function as a US. And it is often appropriate to modify the term "stimulus" with "putative" to remind ourselves that the actual stimulus may depend importantly on how the organism perceives it. The second caveat is more subtle. Although the occurrence of reinforcement in interdependent in operant/instrumental conditioning, the conditions of reinforcement (e.g., the amount) are typically independent variables. That is because the organism usually has no control over those conditions. For a variable such as the amount of reinforcement to be interdependent, one would have to use a condition of correlated reinforcement in which not only the occurrence, but the amount of reinforcement depends on some aspect of the organism's performance (such as faster speeds earn larger rewards). The target topic would be classified as an independent variable. Questions 3 & 4. Is it a contingency, stimulus, organism, or response operation? (Four alternatives require two BITS of information.) All studies of learning require a number of decisions that can be placed in one of four general categories. "Contingency" operations are ones concerned with the S-S or R-S contingency that presumably is to be learned by the organism. "Stimulus" operations refer to the context of the study and any explicit putative stimuli involved. "Organism" operations include choice of organism (species, gender, age, etc.) and any setting operations (e.g., deprivation) involved. Finally, "Response" operations refer to decisions about which aspects of behavior are to be monitored and any associated measurement operations. Although every study involves all classes of operations, most studies focus on only one of them for experimental analysis. For example, a study of the interstimulus interval in classical conditioning is concerned with a contingency variable. A study of the CS intensity in classical conditioning is concerned with a stimulus variable. A study of the effect of deprivation on classical conditioning is concerned with an organism variable. And a study of the latency of the CR over the course of classical conditioning is concerned with a response variable. Of course, some studies include interactions across these classes, in which case I have categorized them in the one that appears to be the main objective. For example, a study of the effect of CS intensity on CR latency would be classified as a stimulus operation. The target topic would be classified as an independent stimulus variable. Question 5. Current or Historic? Some variables are present at the time behavior is being monitored but others occurred in the past. For example, the effect of partial reinforcement on performance is a current variable, but the effect of partial reinforcement on resistance to extinction is an historic variable. Early deprivation on adult learning is an historic variable; the state of deprivation at the time of observation is a current variable. The target topic would be classified as an independent, historic stimulus variable. If you are playing the game with me, the topics that are still viable range from the effects of an early enriched environment to latent inhibition, from prior electro-convulsive shock to the effect of US intensity in acquisition on classical extinction, etc. The list is still large but beginning to come into focus. Question 6. Paradigmatic or Behavioral Effect? A "paradigmatic" variable is one that is an explicit part of the behavioral paradigm; for example, the CS, US, and interstimulus interval are intrinsic to the classical conditioning paradigm. However, the variables such as the context, deprivation, and pre-conditioning experiences may have an effect on classical conditioning, but are not integral parts of the paradigm. If one studies the effect of alcohol on FI performance in operant conditioning, alcohol would have a behavioral effect; if one asks whether alcohol will reinforce FI performance, alcohol would be a paradigmatic variable. Of course, in a prolonged study, a paradigmatic variable may also have a behavioral effect. For example, if alcohol can serve as a reinforcer in operant conditioning, then the alcohol might then begin to have a direct effect on performance. This complication is a general one; any form of reinforcement may affect the organism's drive motivation. But I think it is useful to make this distinction among experimental variables. The target topic is not a part of the paradigm and hence it is an independent, historic, stimulus variable that may have a behavioral effect. Question 7. Natural or Unnatural? Although we are frequently reminded by ethologists that most laboratory studies place organisms in unnatural contexts, I consider the variable to be "natural" if it is the type of event that the organism might encounter in the wild. I have been quite generous in classifying operations as being natural, but such operations as electric shock through an electrode implanted in the septum are, although potentially invaluable for research purposes, definitely unnatural. The target topic would be classified as natural, so now we know it is an independent, historic, natural stimulus variable that may have a behavioral effect. Question 8. Social or Asocial? There is abundant evidence that the behavior of organisms may depend on the presence or other organisms in the context, especially conspecifics. Such evidence ranges from social facilitation of eating to the decrement observed if two animals are placed in the same avoidance situation. Partly in an effort to encourage more research on the effect of social stimuli, I have included this question. The target topic involves another organism and hence it is an independent, historic, natural social stimulus variable that may have a behavior effect. Questions 9 & 10. Emotionally positive, neutral, or negative? Many terms have been used in the literature for this dimension: reinforcing/punishing, beneceptive/nociceptive, pleasant/unpleasant, approach/avoidance, etc. The circularity of these terms has also been widely discussed. When in doubt, however, or when the emotional significance of the event is not germane to its possible effect, I have categorized as emotionally neutral. This includes events that may acquire emotional significance by virtue of association with emotionally significant events. The target topic is emotionally neutral in the sense described. So now let us take stock of the possible topics after ten questions. We know it involves another organism, and since it is a stimulus operation, it involves seeing another organism. The experience occurred in the past (although perhaps by only a few minutes) independently of the organism's behavior. It was not intrinsically either emotionally positive or negative and is the type of experience organisms might encounter in their natural habitat. So what is it? The topic is "observational learning" but it is not the only possibility. Imprinting experience on adult sexual preference could have been the topic. It could have been the size of the litter during rearing. It might have been habituation to an organism of another species. These and any other alternatives could presumably be eliminated by asking additional questions, but I have decided that the purpose of the Quad-L is sufficiently illustrated by the above ten questions. A better system would probably have some kind of hierarchical arrangement of distinctions in order to avoid some of the intercorrelation among the ones in the Quad-L. (For example, all setting operations are concurrently organism and behavioral effect variables.) Insofar as any system is better than none, the Quad-L has merit but it is still a work in progress. The Quad-L Root-Abstract Codes Althought it is not necessary for the user of the Quad-L to learn the details of the coding system, it might be of interest for those who would develop alternative systems. Hence, the specific letter codes I have used in column 2 of the root abstracts are as follows: Excit Inhib A a - Independent Current Contingency Variables B b - Interdependent Current Contingency Variables C c - Independent Historical Contingency Variables D d - Interdependent Historical Contingency Variables E e - Independent Current Stimulus Variables F f - Interdependent Current Stimulus Variables G g - Independent Historical Stimulus Variables H g - Interdependent Historical Stimulus Variables L l - Independent Current Organism Variables M m - Interdependent Current Organism Variables N n - Independent Historical Organism Variables O o - Interdependent Historical Organism Variables P p - Independent Current Reesponse Variables Q q - Interdependent Current Response Variables R r - Independent Historical Response Variables S s - Interdependent Historical Response Variables (Not all of the above categories are relevant to all of the behavioral paradigsm.) To systematize the root abstracts according to the remaining aspects of the taxonomy described above, I have used the following codes in column 7 of the root abstracts. Emotionally-positive a Asocial Emotionally-neutral b Emotionally-negative c Natural ---------------------------------------- Emotionally-positive d Social Emotionally-neutral e Emotionally-negative f Paradigmatic -------------------------------------------------- Emotionally-positive g Asocial Emotionally-neutral h Emotionally-negative i UnNatural ----------------------------------------- Emotionally-positive j Social Emotionally-neutral k Emotionally-negative l -------------------------------------------------------------------- Emotionally-positive m Asocial Emotionally-neutral n Emotionally-negative o Natural ------------------------------------------ Emotionally-positive p Social Emotionally-neutral q Emotionally-negative r Behavioral ---------------------------------------------------- Emotionally-positive s Asocial Emotionally-neutral t Emotionally-negative u UnNatural ------------------------------------------ Emotionally-positive v Social Emotionally-neutral w Emotionally-negative x The goal of a systematic analysis is to organize the operations into categories that identify other conceptually related operations without regard to the terminology used in the publications. I have not yet discovered any heuristics that can guide categorical distinctions so as to maximize the likelihood of the intended result. Accordingly, these categories are being used in an exploratory manner in order to determine what kinds of relations result from their use and all of the quasi-logical categories should be treated as first approximations to a more refined system in which there are clear and unequivocal boundaries defining each category. How much, if any, of this scheme will prove to be useful remains to be determined from its use over a wide range of relevant experimental situations. My principal purpose in displaying it is to illustrate the basic methodology of the Quad-L approach to a systematic analysis. One long-range hope would be that searching the data-base over particular sub-sets of code letters will expose relationships that might otherwise not be readily apparent. Recognizing such relationships may, in turn, encourage conceptual integration of the field of animal learning and motivation. My undaunted belief remains that a unified theory of animal learning will be a cornerstone in any comprehensive analysis of human behavior.