Behavioural Processes 57 (2002) 65 – 69 www.elsevier.com/locate/behavproc Preface SQAB 2001: an abundance of riches Accepted 21 December 2001 Introduction Scientific research programs are not unlike organisms —they grow, develop, reproduce and eventually die Afterwards historians of science assess their contributions by tracking their offspring In Physics and Astronomy the ideas of Ptolomy, Copernicus, Galileo, Kepler, and Newton, or in Biology the ideas of Galen, Vesalius, Colombo, Cesalpino, Fabricius, Harvey, and Malpighi show the typical marks of common descent But how we assess research programs during their infancy or childhood, when growth and change are ongoing but reproduction is impossible or rare? Perhaps the best we can under the circumstances is to look for signs of healthy and vigorous growth, for these signs tend to be positively correlated with survival and future reproductive success Among these signs is the 6ariety of problems, techniques, methods, concepts, models, and theories entertained by the developing program The variety of problems expresses the researchers’ awareness of the complexity of the research object; the variety of methods and techniques expresses the researchers’ sensitivity to the specifics of the research object and the questions they ask about it; and the variety of concepts, models, and theories expresses the researchers’ current gropings, hunches, and attempts to render the research object intelligible The Society for the Quantitative Analyses of Behavior, SQAB for short1, held its 24th annual In this context, SQUAB would be a more appropriate acronym meeting in May, 2001, in the lively city of New Orleans About thirteen of the 18 talks presented at the meeting are printed in the following pages We believe the reader of these papers will readily agree that 6ariety is their distinctive mark To wit, the authors of the papers ask how animals and humans time events, generalize and discriminate temporal and numerical stimuli, choose among alternatives, vary their behavior, and integrate their learning experiences To answer the questions, the authors use pharmacological techniques and computer games, mathematical and statistical analyses, computer simulations of genetic algorithms and neural network models —not to mention the traditional operant and classical conditioning procedures Their research subjects come from three animal species: Columba li6ia, Rattus nor6egicus, and Homo sapiens sapiens, the last one under the variety sophomoricus (Parenthetically, rumor has it these are the gifts the three wise men brought to Jesus who was heard to say ‘‘More research is needed’’.) As a result of their investigations, the authors advance accounts of how memory traces are tuned to environmental periodicities, how rats behave as if they had computed expected time functions, how stimulus effects diffuse and generalize, how stimuli interfere in Pavlovian conditioning, how people choose in groups and individually, and even how behavioral analysis should relate to neuroscience There is hardly a topic on learning that is not covered somewhere, somehow in these papers Given this abundance of riches, our classification of the papers into four categories is per force arbitrary 0376-6357/02/$ - see front matter © 2002 Elsevier Science B.V All rights reserved PII: S - ( ) 0 0 - 66 Preface Timing The first three papers deal with general models or theories of timing In the first, Staddon, Chelaru, and Higa set to explain how pigeons and rats adjust to changes in interfood intervals Based on their previous work on the dynamics of habituation, the authors advance a model composed of a cascade of leaky integrators, an adjustable response threshold, and a noise source Two features of the model seem noteworthy: (a) because the integrators have different time constants, their output can be tuned to a wide range of interreinforcement intervals; and (b) because the threshold is adjusted rapidly, changes in the prevailing interreinforcement interval may cause immediate changes in behavior The authors show that this combination of a slow-acting process (the stabilization of the output of the integrators) with a fast-acting process (the changes in the threshold) accounts for a series of experimental findings that have proven difficult to assimilate by competing models of timing In the next paper Kirkpatrick develops a model of conditioning and timing Starting from the observation that behaviors such as key pecking, lever pressing, or nose poking in the feeder often occur in bouts or packets, the author suggests that these packets are emitted according to the expected time to the next food delivery That is, the probability of initiating a packet of responses at some time t since a time marker depends on the average time the animal has waited for food at time t The most important achievements of the packet model include a straightforward account of the scalar property in fixed-time schedules, the effects on responding of the overall reinforcement rate and the temporal distribution of reinforcers in time-based schedules, and the effects of the CS –US and US –US intervals on response strength in classical conditioning The paper by Odum deals with yet another model of timing, Scalar Expectancy Theory (SET) Specifically the author asks whether the neurophysiological mapping of SET proposed by Meck can account for the effects of some pharmacological manipulations (e.g increase in dopamine) In a novel conditional discrimination task, SET’s predictions are contrasted with the old empirical generalization according to which the effects of a drug depend on the rate of the behavior in the absence of the drug Stimulus control In the next two papers, the focus shifts from general models of timing to time as an additional dimension of stimulus control Based on delayed matching-to-sample data, White argues that remembering is a form of discrimination that depends on two processes, diffusion and generalization; diffusion in the sense that the effect of a temporally distant event (the sample stimulus) diffuses or spreads along the temporal dimension, and generalization in the sense that the effect of training with a specific delay (the retention interval) generalizes to nearby delays When combined, the two processes predict double exponential forgetting functions—a remarkable result published earlier by White Machado and Keen examine the influence of time on the discrimination of stimulus numerosities When stimuli occur sequentially, time and number are inextricably linked and the authors’ model of the numerosity discrimination process suggests the form of the link The model accounts for a variety of findings, such as a Weber-like fraction in the number domain, recency and primacy effects, and the effects of the retention interval The authors compare their model with an alternative theory Both fit the data equally well, but rely on different assumptions concerning the underlying processes These assumptions may be tested in subsequent research Stimulus control is also central in the next two papers Williams and Braker examine the thorny issue of configural versus elemental stimulus control They ask: Is there less generalization from a new compound stimulus (e.g AB) to its elements (A and B) after configural learning than after elemental learning? Compare the reinforcement contingencies of the following two cases: C+, D+ ,CD − , which requires configural learning, and C+ ,D+,CD +, which does not The authors reason that the reinforcement contingen- Preface cies of a configural learning task may increase the salience of features specific to the compound stimulus and hence reduce generalization from that stimulus to its elements Williams and Braker address the issue with both Pavlovian and operant conditioning preparations Miller and Escobar focus on stimulus interference in Pavlovian conditioning The authors first divide stimuli into cues and outcomes and then present them together or apart The resulting 2× table provides a convenient framework to organize the research on interference For example, overshadowing and blocking happen when two cues occur together and are followed by an outcome, whereas proactive and retroactive interference happen when two cues occur separately and each is followed by the same outcome Miller and Escobar report interference on all four cells of the table and propose a qualitative mechanism to account for the data They also challenge the Rescorla and Wagner model and their cousins to account for the same findings Response strength and choice Learning is as much a matter of stimuli as of responses Like the two faces of Necker’s cube, whether we speak of reinforcement value or response strength, decay of stimulus control or fading memory depends mainly on which viewpoint we adopt In the next four papers Necker’s cube is reversed and the response face comes to the foreground Nevin continues to explore his behavioral momentum metaphor, which relates changes in response rate to the force of disruptors and to behavioral mass He asks whether differential resistance to change in the two components of a multiple schedule satisfies the conditions of a ratio scale, and answers affirmatively That is to say, if we let ‘*’ represent the operation of combining reinforcement rate, a, and reinforcement duration, b, then Nevin reports that the effect of the combination on differential resistance to change, f(a*b), equals the sum of the individual effects, f(a) and f(b) He also shows 67 that differential resistance to change can be used to scale force and mass in terms of food rate From a different aspect of Necker’s cube, Ross and Neuringer stress the lack of resistance to change, the fluidity or variability of responding and the conditions in which they occur The authors show that college students readily learn to draw rectangles that vary along three response dimensions, provided reinforcement is contingent on that variation More surprising perhaps they also show that students learn rapidly to vary the rectangles along two dimensions while repeating them along a third dimension, provided such variation and repetition are required to obtain the reinforcers It remains to be seen which other response classes show this exquisite sensitivity to contingencies of variation and repetition, and what sorts of processes underlie the effects of reinforcement on behavioral variability The variation observed by Ross and Neuringer is apparently unsystematic, random-like But in social interactions systematic variation may be at a premium When social behavior varies systematically and is goal-directed we often call it a strategy Baker and Rachlin attempt to understand how reinforcement and punishment shape the social strategies of cooperation and competition The authors asked college students to play a version of the Prisoner’s dilemma game against an opponent and then compared the students’ behaviors when the opponent followed the ‘Titfor-tat’ and the ‘Pavlov’ strategies Tit-for-tat is conceived of as a teaching strategy that rewards cooperation with cooperation and punishes competition with competition; Pavlov is conceived of as a learning strategy that repeats (rewarded) choices followed by cooperation and changes (punished) choices followed by competition Simple as it may look, the interaction that takes place during this and similar games is actually very difficult to analyze for at least three reasons First, closed feedback loops (i.e what I influences what you which in turn influences what I …) are intrinsically hard to understand Second, strategies vary with contextual factors (e.g playing against a computer versus playing against a person) And third, not only 68 Preface does each choice have an immediate consequence, but also strategies have temporally distributed consequences Barker and Rachlin coordinate the concepts of reinforcement and punishment, immediate and delayed reinforcement, molecular and molar response patterns, and teaching and learning strategies to shed light on their findings Kraft, Baum and Burge also analyze choice in a social setting, but their reference is not game theory but behavioral ecology If a group of foragers know the distribution of food resources in two patches, choose freely between them, divide the food equally when in the same patch, and attempt to maximize net energy gain, then the foragers will distribute themselves between the patches in the same ratio as the food resources delivered by the patches Known as the ideal free distribution in behavioral ecology, this rule is mathematically equivalent to Herrnstein’s matching law More accurate empirically is a generalized version of the ideal free distribution, a rule that is mathematically equivalent to the generalized matching law The problem Kraft, Baum, and Burge address is how group choice ratios relate to individual choice ratios In a series of studies with college students the authors varied the response type and the method of resource distribution and found that regularities at the group level are not easily explained by regularities in the behavior of the individual members of the group Interestingly, this finding has parallels at the level of individual choice, for we still not know how to interrelate the molar and molecular regularities of an individual’s choice behavior Integration and a tribute When a response causes a stimulus change and because of that it becomes more likely to occur, we speak of reinforcement Conceptually, the reinforcement relation is simple and straightforward, but empirically it is not As any introductory textbook readily confirms, some key issues surrounding the concept of reinforcement have stubbornly resisted one hundred years of intense research Part of the difficulty is that reinforcers are continuously recurrent events whose effects cumulate throughout the organism’s history A few cycles, as it were, of this recurrence and extremely complex behavioral forms may emerge The issue then is how best to characterize the cumulative effects of reinforcement For reinforcers can affect temporally brief and temporally extended response patterns; they can engender repetitive or variable behaviors; they can promote contextual sensitivity or insensitivity; and through them complex topographies can be literally built or shaped In his paper, Donahoe argues that to understand the workings of reinforcement psychologists need to look at neuroscience— which promises to unravel the biological mechanisms of reinforcement—and to biologically plausible neural network models of the brain—which promise to elucidate the complex, cumulative effects of simple recurrent causes Through genetic algorithms we may come to understand the evolution of the architecture of neural networks; and through analysis of biologically plausible learning rules we may come to understand how adaptive behavior is generated The last paper of the Proceedings is to some extent also on integration—not integration in the abstract though, but integration as it was sought and accomplished by a good friend of SQAB and a regular contributor to its annual meeting, the late John Gibbon Russell Church, possibly Gibbon’s closest collaborator, reminds us of the amazing 6ariety of our colleague’s research: His topics included avoidance, punishment, autoshaping, human and animal timing, circadian rhythms, and foraging; his experimental subjects ranged from human adults and children to rats to pigeons, ring doves, and starlings; his methods came from psychophysics, animal learning, neuropsychology, and mathematical psychology Gibbon’s brainchild, Scalar Expectancy Theory, is currently the most influential theory of timing Whether researchers test or entertain, improve or reject Gibbon’s ideas, Preface one thing is clear, they not ignore them The recognition of a scientist’s intellectual offspring is perhaps the highest tribute one can pay to him So, there you have it Around the table, ready to be relished, are 13 papers—plus this Preface, for the superstitious—on a broad variety of issues, using a variety of methods and techniques, proposing a variety of ideas, models, and theories, and raising a variety of new questions SQAB and the research program of its members are alive and well Come visit us next year! 69 Armando Machado, Uni6ersidade Minho, Instituto de Educac¸a˜o e Psicologia, 4710 Braga, Portugal E-mail: armandom@iep.uminho.pt Peter Killeen Department of Psychology, Arizona State Uni6ersity, Tempe, AZ 85287, USA, E-mail: killeen@asu.edu ... variety of issues, using a variety of methods and techniques, proposing a variety of ideas, models, and theories, and raising a variety of new questions SQAB and the research program of its members... 6ariety of our colleague’s research: His topics included avoidance, punishment, autoshaping, human and animal timing, circadian rhythms, and foraging; his experimental subjects ranged from human adults... scale force and mass in terms of food rate From a different aspect of Necker’s cube, Ross and Neuringer stress the lack of resistance to change, the fluidity or variability of responding and the