Listed below are the SQAB abstracts for Presentations, Poster Presentations, and Preeminent Tutorials.

Friday, May 25

John Staddon, I. M. Chelaru, & J. J. Higa
Duke University, Texas Christian University

A Tuned-Trace Theory of Interval-Timing Dynamics

Under certain conditions, organisms on interval schedules of reinforcement will rapidly adjust a temporal dependent variable, such as wait time, to changes in the prevailing interreinforcement interval. We describe data on the effects of impulse, step, sine-cyclic and variable-interval schedules and show that they can be explained by a tuned-trace timing model with a one-back threshold-setting rule. The model can also explain standard steady-state timing properties such as proportional and Weber-law timing. The model provides a partial account for the learning of multiple intervals, but does not account for response patterns, such as the FI "scallop" and some data from "square-wave" schedules. The model assumes that reinforcers and other time markers have a decaying effect (trace) with properties derived from the rate-sensitive property of habituation (the MTS model). Response threshold is set according to the trace value at the time of the most recent reinforcement.

Amy Odum
University of New Hampshire

Behavioral Pharmacology and Timing

Behavioral pharmacology uses the methods of the experimental analysis of behavior to examine the impact of pharmacological variables on behavior. Behavioral pharmacologists often arrange situations that make a consequence available at regularly occurring times (e.g., fixed-interval schedules) because the resulting behavior is sensitive to the effects of drugs in robust and reliable ways. Drug effects on temporally patterned behavior are often described under the rubric of rate dependency: the effect of a drug on behavior is related to the rate of behavior in the absence of the drug. Specifically, drugs increase low rate behavior and decrease high rate behavior. These same types of effects are interpreted in the timing literature, however, as changes in temporal discrimination. The present series of experiments arrange situations that allow divergent predictions based on the two interpretations. Specific focus is given to a neuropharmacological information-processing model of timing. Predictions are compared for drugs that are thought to affect the clock and memory stages in the model.

Geoffrey White
University of Otago

Temporal Generalization and Diffusion in Forgetting

Two processes may contribute to the decrement in discriminability (forgetting) with the temporal distance between the occasioning event and later choice. One is the length of the interval and the other is generalization decrement. In the model described by White and Wixted (1999), choice was predicted by the ratio of reinforcers obtained for correct delayed matching responses. The reinforcer ratio was related to Thurstone-like probability distributions of the effect of the sample stimuli. In the model, discriminability decreased with increasing temporal distance because the variance of the distributions increased with time. However, White and Wixted did not specify the function relating increasing variance to temporal distance. If a diffusion process is assumed, the resulting function can be described by a negative exponential. An additional process involves exponential generalization of remembering from one time to other times. The combination of the two processes yields forgetting functions that are double-exponential in form and which appear consistent with a wide range of data.

Kim Kirkpatrick
University of York

Packet Theory of Conditioning and Timing

Packet theory is based on the assumption that the momentary probability of producing a packet of responding is controlled by the expected time function. Packets of head entry responses of rats into a food cup contain about 5 responses that are emitted over a duration of about 7 s. Characteristics of the packets such as a number of responses, rate of responding, duration of responding, and time between successive responses are invariant across a range of conditions. Thus, in Packet theory the packet is the unit of behavior that is produced as a function of the expected time until reinforcement. The expected time function is the mean expected time remaining until the next food delivery as a function of time since an event such as food or stimulus onset. Packet theory will be applied to a range of phenomena in timing and classical conditioning.

William Roberts
University of Western Ontario

Timing and Counting by Pigeons: Do they require similar or different mechanisms?

Some recent models of timing have been extended to account for counting in animals. These include scalar timing theory, Church and Broadbent's connectionist model of timing, and Staddon, Higa and Chelaru's multiple time scale model. Two findings support the hypothesis of a common mechanism. When rats or pigeons are trained to respond differently to short and long duration signals and/or small or large numbers of events, they show similar time and number psychophysical curves when tested with intermediate magnitudes, and they show similar choose-short duration and choose-small number effects when memory delays are introduced before the choice response. In recent experiments using the peak procedure, pigeons were cued to either keep track of a fixed interval or to count a fixed number of light flashes. These experiments suggest that pigeons used different mechanisms to track time and number. When cued to keep track of number, pigeons also kept track of time; however, when cued to keep track of time, pigeons did not keep track of number. Close examination of pigeons' pecking behavior during fixed interval and fixed number cued trials indicated that timing was automatic and did not depend on behavior but that counting was based on a behavioral code.

Richard Keen & Armando Machado
Indiana University

Relative Numerosity Discrimination and Short-term Memory

This study examined how pigeons discriminate the relative frequencies of events when the events occur serially. During a sample phase, pigeons were shown two different colored lights, the first for nf times and the second for nl times. Then, during a choice phase, the pigeons were rewarded for choosing the color that was shown the least number of times during the sample. Four variables were manipulated: 1) the total number of stimuli in a sample (T = nl + nf); 2) the difference between the frequencies of the two stimuli (D = nf - nl); 3) the delay between the sampling phase and the choice phase; and 4) the delay between the first and the second stimuli of the sample. The pigeons' accuracy increased with the absolute value of the difference D, and for D > 0, it decreased with T. Performance also showed clear recency and primacy effects, which were found to depend on T. These results, combined with the stimulus control decay functions obtained with the various delay manipulations, are discussed in reference to a mathematical model of the discrimination process.

Allen Neuringer
Reed College

Operant Variability and a Theory of Operant Behavior

The evidence shows that levels of response variability are controlled by reinforcing consequences and discriminative stimuli. As is the case for other operant dimensions, schedules of reinforcement influence levels of variability; reinforcement contingencies are more controlling than reinforcement frequencies; choice probabilities to vary or not match relative frequencies of reinforcement; variability transfers to novel situations; and extinction affects operant variability and operant repetition in similar ways. However, operant variability may differ from other dimensions in providing explanations for the unique characteristics of operant behavior, including its voluntary nature.

Alexandra Logue & Yenny D. Anderson
Baruch College, City University of New York

Self-control, Impulsiveness, and Higher Education Administration

Administrators of colleges and universities must sometimes choose between a less delayed but ultimately less valued outcome (impulsiveness) and a more delayed but ultimately more valued outcome (self-control). Which choice is made can affect the long-term health of the administrator's organization. In the first empirical research concerning the relationships between administrative actions and their long-term consequences, administrators with greater experience were significantly more likely than administrators with less experience to mention long-term consequences when describing their past and future administrative actions. However, in hypothetical choice situations, the administrators with greater experience were also significantly more likely to choose smaller amounts of funds available immediately for their units versus larger amounts of promised future funds. With experience, administrators may both become more aware of their actions' long-term consequences and learn that they are unlikely to receive promised future funds. The contingencies in effect for higher education administrators may lead to their making choices that do not result in their institutions meeting the highest standards. Research and analysis from laboratory investigations of self-control and impulsiveness can be useful in understanding administrators' choices.

Howard Rachlin & Forest Baker
State University of New York at Stony Brook

Teaching and Learning in the Prisoner's Dilemma

An experimental analysis of strategies in repeated prisoner's dilemma games reveals that a tit-for-tat strategy reinforces cooperation (by cooperation) and punishes defection (by defection). Another strategy, commonly called Pavlov, earns more points than tit-for-tat in computer simulations. Pavlov is a learning strategy in the sense that a player using it repeats a prior choice (to cooperate or defect) when that choice is reinforced and changes a prior choice when it is punished. Participants cooperated more playing against tit-for-tat than against Pavlov when they believed that they were playing against the computer. Participants cooperated more playing against Pavlov than against tit-for-tat when they believed that they were playing against another participant. The results show that the context of a prisoner's dilemma game (the discriminative stimuli provided by instructions and display) may be as important as the reinforcement and punishment contingencies themselves in determining whether cooperation is obtained.

John Kraft
Armstrong Atlantic State University

Quantifying Human Social Behavior with the Ideal Free Distribution model

Group Choice is our term for describing the behavioral phenomenon in which a group of individuals chooses between two alternatives over time. We use the Ideal Free Distribution (IFD; Fretwell & Lucas, 1970) analysis from behavioral ecology to account for Group Choice. Group Choice is a social phenomenon analogous to individual choice, and the IFD is analogous to the Matching Law account of individual choice. The IFD theory suggests that the ratio of foragers at two resource sites should equal the ratio of obtained resources. In this talk, I present several experiments that investigate the usefulness of IFD analyses of human Group Choice based on a procedure reported in Sokolowski, Tonneau, & Freixa i Baque (1999). Instead of non-human animals foraging at two sites for resources, a group of humans were given the opportunity to arrange themselves into two subgroups to receive points that could earn cash prizes. The results showed that IFD matching was dependent on the method of resource distribution to subgroup members (i.e., sharing points evenly produced IFD matching, but probabilistic point distribution did not). In addition, I explored the relation between the groups' choices and the choices of individuals in the groups. I found that the IFD is a useful analysis for determining the sensitivity of a group's choices.

Anthony McLean
University of Canterbury

Extraneous Reinforcement, Response Rate and Resistance to Change

Response rates in multiple schedules were simulated using a simple hyperbola in which the efficacy-of-reinforcement term (Herrnstein's Re) undergoes systematic changes. A model was developed for changes in Re in a multiple-schedule component as a function of the difference in reinforcement between components (reallocation of extraneous reinforcers), based on data on re-distribution of alternative reinforcers in some earlier work of my own. Substituted into the hyperbola, it yields a response rate equation that predicts multiple schedule undermatching and contrast. Predictions for response rates in resistance to change tests were made by assuming global devaluation of food reinforcers as a function of food consumed, similar to the mechanism proposed recently by McSweeney and Weatherly. The equation also predicts most of the phenomena reported in the resistance to change literature, including the cross-over observed in prefeeding and satiation tests by Nevin, Tota, Torquato, and Shull.

Randolph Grace
University of Canterbury

Comparing Linear Operator and Representational Models of Choice Acquisition

Two popular classes of quantitative models are linear operator and representational. The critical difference is that linear-operator models encode reinforcement history in a single number (e.g., "associative value"), whereas representational models assume that the entire history is encoded (e.g., distribution of reinforcement delays in memory). I propose simple and generic versions of each type of model for choice between schedules of delayed reinforcement in concurrent chains, and show that the models make generally similar predictions regarding acquisition. These predictions were confirmed in two experiments in which pigeons chose between different combinations of fixed-interval and variable-interval terminal-link schedules. However, the linear-operator model correctly predicted the effects of overall schedule duration, whereas the representational model did not. Thus, the linear- operator approach is preferred for parsimony and because it gives a more complete account of the data, although it remains to be seen whether the increased complexity of representational models will be necessary to account for choice in other situations.

William Baum
University of New Hampshire

Analysis of Visits in the Dynamics of Choice

When the reinforcer ratio in a pair of concurrent variable-interval schedules changes frequently and unpredictably, control by the reinforcers' alternative sources becomes extremely local. We find that every reinforcer draws choice strongly toward its source. This results in a "pulse" of preference immediately following each reinforcer. The pulse decays gradually but rapidly, so that choice approaches indifference after about 40 s. An alternative analysis of the local variation focuses on discrete visits to the two alternatives. Immediately following a reinforcer, a long visit is almost certain to occur at that reinforcer's source. After that post-reinforcer visit, visits at both alternatives become brief. These two analyses of the same phenomenon reflect different assumptions and intuitions about choice. In particular, they raise the question of whether behavior should be thought of in terms that are continuous, discrete, or both.

Special Event

Russell Church
Brown University

Tribute to John Gibbon

Saturday Morning, May 26

Ralph Miller & Martha Escobar
State University of New York at Binghamton

Interference Between Cues and Between Outcomes Presented Together and Presented Apart

"Stimulus competition" in recent years has referred exclusively to competition for behavioral control between cues trained in compound with a common outcome (e.g., overshadowing or blocking) (1). Rarely cited is an older literature with human subjects, now complemented with data from humans and rats in Pavlovian preparations, demonstrating competition between cues trained separately with a common outcome (i.e., proactive and retroactive interference) (2). Similarly neglected is the extensive literature concerning competition between outcomes trained separately with a common cue (e.g., counterconditioning) (3). New and not-so-new animal and human data demonstrating competition between outcomes trained in compound (4) complete the four cells of a 2 x 2 matrix ([compound vs. separate training] x [competing cues vs. competing outcomes]), thereby highlighting the ubiquitous nature of stimulus interference. Most contemporary theories of acquired behavior can account for one or at most two of these interference effects, which are demonstratively due to more than the degraded contingency that inevitably occurs in these studies. Stimulus interference appears to be greatest when two associations have one (not two) element in common with a common ordinal position in the stimulus dyad of training.

Douglas Williams
University of Winnipeg

Input Coding in Animal and Human Associative Learning

Two appetitive conditioning experiments with rats investigated whether the degree of generalization between a compound and its component parts is fixed by the physical properties of the stimuli involved, or is variable and affected by past learning. The experiments used a two-stage transfer design. In Stage 1, the elemental groups learned that food pellets associated with a compound and its component parts were consistent (i.e., A+, B+, AB+), whereas the configural groups learned that food pellets associated with a compound and its component parts were inconsistent (i.e., A+, B+, AB-). In Stage 2, the rats received a common discrimination to determine if the strategy acquired in training had transferred to test. Experiment 1 found no evidence that past configural learning reduced generalization when a new set of alike-treated elements were presented in compound for the first time. Experiment 2 found no evidence that past configural learning reduced generalization when the stimuli of Stage 1 occurred in a new A-, B-, AB+ relation. In contrast to findings with humans, these results suggest that past experience plays a minor role in the coding of compound stimuli in animal conditioning.

Tony Nevin
University of New Hampshire

Behavioral Momentum: Measurement Properties of Force and Mass

Research on resistance to change shows that changes in response rate during disruption are directly related to the magnitude of the disruptor and inversely related to rate and/or amount of reinforcement. In the metaphor of behavioral momentum, disruptors are analogous to external forces and the effects of reinforcement are analogous to physical mass. Many physical attributes such as length and weight are additive, as shown by an empirical process of combination with results that are consistent with the arithmetical process of addition. Several sets of data show that separate disruptors have additive effects when combined, and that the separate effects of rate and amount of reinforcement also have additive effects when combined. Thus, the behavioral equivalents of force and mass are additive, satisfying a necessary condition for fundamental measurement.

John Donahoe
University of Massachusetts

On Neuroscience and Behavioral Analysis

Darwin proposed a powerful functional principle-natural selection-to interpret phylogenetic diversity and complexity. Nevertheless, it took some 75 years before even biologists embraced his interpretations. The ultimate triumph of natural selection required two additional factors: (a) biological mechanisms (genetics) that could implement the functional principle and (b) quantitative procedures (population genetics) that could trace its cumulative effects. Thorndike and Skinner proposed a functional principle- selection by reinforcement-to interpret ontogenetic diversity and complexity. This principle has been substantially refined by subsequent experimental and theoretical work and now provides an equally powerful functional principle. However, once again a purely functional principle has not persuaded scientists (aside from the behavior-analytic minority) that the complexity and diversity of behavior can be interpreted by a selection principle. If the history of ontogeny recapitulates the history of phylogeny, the biological mechanisms of reinforcement must be discovered and procedures must be devised to trace its effects. In short, the triumph of reinforcement awaits the integration of behavior analysis with the neurosciences and the development of coordinated quantitative procedures. The presentation identifies specific implications of neuroscience for a behavior-analytically-faithful interpretation of reinforcement using neural networks. Implications for the architecture of networks and for the learning algorithm are emphasized.

Saturday Afternoon, May 26

Michael Davison
University of Auckland

Chair: William Baum

What do reinforcers do to behaviour?

I don't know why reinforcers affect behaviour, but I do know that some behavioural consequences change the probability of behaviours that they happen to follow. I also know that the context in which these consequences occur affect behaviour change. This functional and contextual relationship has been the subject of much research over many years without, dare I say, a great deal of progress beyond the ordinal being made. I will first briefly review some of this history. Then, making the assumption that all behaviour is choice behaviour, I will discuss some recent quantitative research that shows how reinforcers that follow what animals have done affect what animals subsequently do in both the short and longer terms. I will look specifically at the effects of present and historical contexts, at environmental variability, at how stimulus control might modulate reinforcer effects, and at what these results mean for managing behaviour change.

Randolph Grace
University of Canterbury

Chair: John Nevin

From Basics to Contemporary Systems: Quantification

Quantitative methods are widely used to model phenomena. For example, the flexible, iterative, graphical approach to understanding data developed by John Tukey - exploratory data analysis - may be understood in terms of the following equation: Data = Model + Error. And inferential statistics, which has long been controversial because it is so easily misunderstood, may be construed as a method for choosing between alternative models. Building models and making decisions about models is central to how quantitative methods are employed in behavior analysis, and this tutorial will consider a range of these methods.

William Timberlake
Indiana University

Chair: Donald Patterson

From Basics to Contemporary Systems: Behavior Systems

Ethologists introduced a behavior system analysis to account for the perceptual-motor and temporal-sequential structure of functional behavior. For example, a cat in a general predatory state is sensitized to cues related to prey, and its potential behavioral repertoire is shifted toward responses that facilitate successful hunting. Further, a series of motivational substates rather than a fixed sequence of responses appears to underlie the hunting stages of search, pursuit, capture and consumption. At a more encompassing level, the daily and seasonal behavior of a cat appears related to interacting systems of fear, aggression, and reproduction. Applying a behavior system analysis to laboratory learning research appears to contribute to the prediction and understanding of common phenomena, including: (1) the design and tuning of experimental apparatus, stimuli, manipulanda, and procedures to produce vigorous and reliable behavior; (2) constraints and predispositions in operant conditioning; (3) "irrational" operant behavior such as misbehavior, adjunctive behavior, and superstition; and (4) the form of responses and stimulus interactions in Pavlovian conditioning. In turn, laboratory learning research provides information about the nature of a behavior system. A point of current interest is how to combine the specificity of a behavior systems analysis with the seeming generality of laboratory learning effects.

Leonard Green
Washington University in St. Louis

Chair: Michael Davison

Self control, choice, and the (non-rational) form of the discounting function

A behavioral account conceives of self control as something one does, not something one has; that is to say, the determinants of choice lie more in the choice situation than within the individual. Consistent with this view, an individual's preference for smaller-sooner versus larger-later rewards changes as time to the rewards is varied. Rats, pigeons, and humans choosing the smaller-sooner reward tend to reverse their preference as a constant amount of time is added prior to the receipt of both reward options. Such preference reversals also demonstrate that the subjective value of rewards changes as time to their receipt is varied --- that is, value is discounted with time. We have attempted to determine whether there is a general mathematical form to the discounting function. Based on animal and cross-cultural human data, the form of the function is hyperbola-like, and not exponential as argued by "rational" accounts of choice in the economics literature. A similar, hyperbola-like discounting function describes the relation between subjective value and the odds against receiving a reward. Importantly, amount has opposite effects on the discounting of delayed and probabilistic rewards, although there is no consensus as to the theoretical interpretation of this reliable finding.

Friday Evening, May 25 7:00 - 9:00 pm

Carlos F. Aparicio
University of Guadalajara, Mexico

Choice Situations with Multiple Alternatives: How to Apply the Generalized Matching Law?

Applying the generalized matching law with two alternatives is straightforward. But when more than two response alternatives are part of a choice situation, it is harder to determine to which alternative the numerator of the matching equation should correspond. Three different methods to estimate the organism's sensitivity to changes in reinforcement are reported in this poster: 1) One alternative was randomly selected, the calculations of responses and obtained reinforcers in that alternative entered into the matching equation as the numerator, and the geometric means of responses and obtained reinforcers for all other available alternatives entered into the matching equation as the denominator. 2) For each alternative, the calculations of responses and obtained reinforcers within and between conditions took a turn in the numerator, and the geometric means of responses and obtained reinforcers for all other available alternatives entered into the matching equation as the denominator. 3) The sensitivity to reinforcement was separately calculated for each alternative across conditions. The calculations of responses and obtained reinforcer, calculated separately for each alternative, took a turn in the numerator, and the geometric means of responses and obtained reinforcers entered into the matching equation as the denominator. The posters shows that each method describes different aspects of choice behavior.

Orn Bragason
University of Canterbury, New Zealand

Choice and Distribution of Terminal Link Schedules

This study examined the effects on choice of delay distributions in the terminal links of the concurrent-chain procedure. Four pigeons were exposed to two pairs of terminal-link schedules that had the same arithmetic means but different harmonic means. One terminal link was paired with an exponential progression schedule, while the other was paired with an arithmetic progression schedule. All pigeons had the same initial link schedule. The results showed that the harmonic mean difference of the terminal link schedules was positively related to preference. Several different models for quantifying terminal-link value are assessed against the data.

Glenn S. Brown & K. Geoffrey White
University of Otago, New Zealand

Measuring Discriminability: Correcting for Extreme and Multidimensionally-Biased Performance

Ratio-based discriminability measures such as log d and d¢ become infinite when performance is extreme. We can apply different arbitrary corrections to our data to produce finite values, but how well do these values estimate true performance? To answer this question, we analysed the effects of different arbitrary corrections on the sampling distributions of log d and d¢. Rather than doing this via simulation, we did it via direct computation using binomial probabilities. We found that these arbitrary mathematical corrections are superior to the intuitively-plausible technique of re-running problem conditions. Further investigation allowed us to develop a superior - and less arbitrary - correction procedure. It also helped us to produce extended versions of log d and d¢ that are immune to the effects of multiple sources of bias (e.g., both colour bias and position bias), in addition to a mathematical procedure for estimating them without requiring larger datum sets.

José E. Burgos
University of Guadalajara, Mexico

A Truly US-Alone Procedure Causes Learning in Selection Neural Networks

A truly US-alone, response-independent procedure was given to a feedforward, fully-connected neural network whose processing elements functioned according to the Donahoe-Burgos-Palmer model. The network's US input element was maximally activated and none of its sensory input elements was activated, which simulated an organism in a sensory vacuum. As a result, weights in the motor-association connections increased up to roughly 30%, thus showing a substantial amount of learning. This result makes the present model a nonassociative one, since standard associative models assume the presence of at least two stimuli (either phasic or tonic) for any learning to occur.

David A. Case
University of California, San Diego

Quantitative Biased Choice Analysis

Implications of bias in choice require renewed scrutiny in advancing a quantitative model agenda. Sometimes experimenters assume that bias is non-existent or reduced relative to that which may be found using other methods, so they do not assess or report it. Sometimes reported bias is non-trivial but marginalized, for example in focusing upon a predicted trend in a parametric study accounting for little of the variance. Since trivial sources of bias are routinely controlled, remaining bias may have crucial theoretical implications. I argue that non-trivial bias should be spotlighted and ideally manipulated in evaluating vying models. A novel bias assessment is proposed for adjusting procedures, with implications shown for the function describing reinforcement weakening due to delay.

A. Charles Catania
University of Maryland, Baltimore County

Delay of reinforcement and the operant reserve

Skinner's reflex reserve assumed that only the relation of the reinforced response to the reinforcer contributed to the reserve and it was therefore unable to deal with partial reinforcement in schedules. The concept can be modified to allow each response leading up to a single reinforcer to contribute to an operant reserve weighted by an empirically derived delay-of-reinforcement gradient. Response rate at any moment is then a function of the size of the reserve. Many properties of the responding maintained by reinforcement schedules follow from this model.

J.M. Cleaveland & Juan Delius
Universität Konstanz, Germany

Can a Molecular Choice Variable Account for the Belke (1992) Results?

The experiments presented in this poster replicate Belke (1992) with attention paid to the conditional probability of a changeover given the most recent interresponse time (i.e., p(CO | IRT)). In Belke's experiment pigeons pecked within two pairs of concurrent variable-interval, variable-interval (VI-VI) schedules which were arranged in a multiple schedule. Each pair contained a common VI schedule -- in one pair relatively poor, while in the other relatively rich. During the test phase pigeons were permitted to choose between the two identical VI schedules. All showed a strong preference for the "relatively rich" alternative. Gibbon (1995) accounted for these findings using Scalar Expectancy Theory. However, another potentially more parsimonious explanation lies in the form of the changeover functions. If a stimulus associated with a rich schedule conditions a changeover function that decreases as IRTs increase, while a stimulus associated with a poor schedule conditions a changeover function that increases as IRTs increase --- results consistent with Cleaveland (1999) ---, then test results similar to Belke's would naturally emerge.

Darlene E. Crone-Todd , Barrie J. Todd, & Joseph J. Pear
University of Manitoba, Canada

The Use of a Video Tracking System in Studying Response Variability in Human Limb Movement

The present study used a real - time 3D video tracking system to examine the effect of various shaping parameters on human limb movement response variability. The following conditions were examined: (a) no shaping parameters used; (b) three levels of a forward step size (FSS) toward the levels of target; and (c) three levels of back step rate (BSR). Two female undergraduate students served as subjects in sessions that were conducted up to three times per week, over a period of nine months, during which the individuals moved a white ball at the end of a stylus within a metal cube. Computer-emitted tones served as reinforcement in all trials. Measures of variability include the relative distance traveled (i.e., total distance covered, relative to the initial distance to target, per trial) and the rate of sub-cubes entered during 1-second periods across the trial. Results indicate that when the BSR is zero or relatively moderate an inverse relationship exists between FSS and variability measures. Further, variability appears to decrease when a BSR value is increased, but sometimes increases when at a relatively high rate. The findings indicate that the VTS system has major advantages over traditional methods of studying shaping parameters.

Matthew O. Dailey, David A. Case, & Edmund Fantino
University of California, San Diego

Risk Aversion as a Function of Reinforcement Delivery in Pigeons

Five pigeons were assessed in a two choice concurrent-chains procedure whereby they were presented with a choice between a constant amount of water and a variable amount of water. Each pigeon experienced three conditions the order of which was counterbalanced: in Condition 1 the water dipper located below the selected key was operated; in Condition 2 a single water dipper was operational regardless of which key was selected; and in Condition 3 a randomly selected dipper was presented regardless of which key was selected. In each condition the pigeons preferred the constant outcome (i.e., risk aversion). However, the degree of risk aversion was dependent on how the reinforcement was delivered. Results showed that the greatest risk aversion occurred with Condition 1 and the lowest risk aversion with Condition 2. Condition 3 had intermediate risk aversion. These results suggest that the manner in which reinforcement is delivered has an impact on choice proportion with risky amounts. These results are consistent with many of the divergent findings on risky choice for amounts.

William V. Dube & William J. McIlvane
University of Massachusetts Medical School - Shriver Center

Momentum of Stimulus Control in Discrimination Reversal Learning

Behavioral momentum describes a relation between behavioral persistence and the rates of reinforcement signaled by discriminative stimuli. This experiment was conducted to test a momentum-related prediction for teaching discrete-trial discrimination performances with stimulus control shaping procedures. Nine subjects with mental retardation were trained with a fading procedure to perform a two-choice simple-discrimination task. Immediately after meeting a learning criterion, a discrimination reversal was trained within the same session, also with the fading procedure. Two conditions were compared within subjects: High, in which all correct responses for the initial discrimination were followed by reinforcers, and Low, in which the initial discrimination was trained with an intermittent reinforcement schedule that produced a lower reinforcer rate. In both conditions, all correct responses during reversal training were followed by reinforcers. If the rate of reinforcement during the initial discrimination training were related to the persistence of stimulus control, then one would predict more selections of the initially positive stimulus during reversal training in the High condition. The results were consistent with this prediction. Reversal learning error rates were higher in the High condition for eight subjects, and there was a negligible difference for one subject. These results indicate that (a) discrimination reversal procedures may be useful for momentum research. (b) Teaching procedures may sometimes benefit from tactical reductions in local reinforcement density to encourage changes or transfers of stimulus control.

Paulo Guilhardi & Russell M. Church
Brown University

Patterns of Responding on a Cued Fixed Interval Procedure

The problem was to develop a model of fixed interval performance that would account for both asymptotic and dynamic results. In a cued fixed-interval procedure, in which the interval randomly changes each session, rats readily develop a temporal gradient of responding that is appropriate for the fixed interval used on each session. Twenty-four rats were trained for 30 sessions with 30-, 60-, and 120-s intervals differentially signaled by white noise, light or clicker. Testing sessions consisted of random presentations of the same three cues. The intervals for the 30- and 120-s cues remained the same as in training; but on each session the other interval (that was 60-s during training) was one of nine different intervals ranging from 30 to 120 s. The mean response rate as a function of time since stimulus onset of individual rats was well described by a 3-parameter ogive with parameter values related to interval duration. Each session was characterized by a rapid adjustment to the session-specific middle interval. Responses occurred in packets, and an analysis of the probability of being in the packet as a function of the expected time to reinforcement described many of the properties of the response pattern of rats.

Linda Hayes, Nelson Publicover, Kate Kellum, Aki Masuda, Jennifer Thomas, Margaret Heaton, & Ken Hunter
University of Nevada, Reno

A Comparison of Response Pattern Acquisition by Means of Percentile Shaping Schedules and Shaping by Successive Approximations in Mice

Complex patterns of locomotion, starting from the center of the front panel of the chamber, moving to the left front corner, down the left side, and across the back of the chamber to the back right corner, were shaped in 9 Balb C mice in daily 15-minute sessions, conducted 7 days a week. Three subjects were shaped by the method of successive approximations, wherein 13 steps to the target response were identified in advance, and progression through the steps required a minimum of 4 correct responses per step. Additional trials, to a maximum of 7, were required if the time of the last correct trial was greater than that of the first correct trial on a given step. Correct responding was followed by a brief tone and the delivery of a food pellet. There were no consequences for errors. The same pattern of locomotion was trained in 6 mice by means of a percentile shaping schedule, in which the response requirements for reinforcement were adjusted continuously and progressively in accordance with the subjects' performances. Digital imaging technologies were used to detect the precise locations of the animals, and all program changes (e.g., food deliveries, step changes), contingent on the animals' positions and patterns of movement in the chamber, were computer controlled. The percentile shaping schedules produced more rapid acquisition of the target response and a greater proportion of session time in reinforced activity than the successive approximation shaping procedure.

Miguel Herrera & Arturo Bouzas
Universidad Nacional Autonoma de Mexico, Mexico

Intertemporal Choice of Sequences of Outcomes With Addicts and Matching Controls

Currently there are a lot of investigations that show that alcoholics and heroin addicts show higher rates of discounting for delayed rewards than matched controls (Kirby, Petry and Bickel 1999, Vuchinini and Simpson. In Press, Petry y Casarrella 1999). However, this line of research has been limited because it has dealt mostly with single, discrete outcomes. Recent research in intertemporal choice has begun to work with sequences of outcomes. The present investigation joins these two lines of research. Specifically, we want to know if alcoholics and addicts show different rates of discount than matched controls, when choosing between sequences of outcomes. Results are interpreted using Loewenstein and Prelec's model.

Jennifer J. Higa & Karen Jennings
Texas Christian University

Temporal Performance During Massed and Alternating Conditions

Previous studies suggest that temporal performance in a session may be affected by the interval duration in the preceding session. For example, performance on a cyclic interval schedule depends on whether conditions are conducted in blocks (massed) or alternated. These results suggest that the window over which intervals have their effect appears to extend across sessions. The cause of such effects is not clear. One idea is that learning to pause (wait) on long intervals may take longer because of the residual tendency to pause short from exposure to a shorter interval schedule. We present an analysis of results from an experiment in which we presented to rats across-session changes in fixed-interval (FI). We compared wait times from massed and alternating conditions, and across transitions, and determined to what extent prior exposure to a short (or long) FI schedule affects current performance. The results are discussed in terms of models of timing, with respect to the time window in which intervals are supposed to have their effect on performance.

Dan Holt, Leonard Green, & Joel Myerson
Washington University

Delay and Probability Discounting in Gambling and Non-Gambling Young Adults

This study sought to determine if delay and probability discounting of young adults who gamble frequently (3 or more times a month) differs from that of their non-gambling counterparts. Individuals were asked to choose between hypothetical monetary amounts. In the delay discounting task, they chose between immediate and delayed rewards; in the probability discounting task, they chose between certain and probabilistic rewards. Some individuals participated in both discounting tasks whereas other individuals participated in only one. Self-report measures of the frequency and type of gambling behaviors were also obtained. The same hyperbolic function accurately described the discounting of delayed and probabilistic rewards for both the gambling (N=8) and non-gambling individuals (N=21). For the probability discounting task, the areas under the discounting curve suggest that participants who reported higher rates of gambling behavior were more "risk-seeking" (i.e., discounted probabilistic amounts less steeply) than their non-gambling counterparts. On the delay discounting task, no differences were found between the areas under the curve for gamblers and non-gamblers. The preliminary finding that gamblers and non-gamblers differ in probability discounting but not in delay discounting suggests that different processes may be involved in making decisions about delayed and probabilistic rewards.

Ken Hunter, Linda Hayes, Rene Quinones, Fernando Guerrero, Joe Rodrigues, & Nelson Publicover
University of Nevada, Reno

A Demonstration of the Utility of Operant Methods for the Early Detection of Joint Dysfunction in an Animal Model of Arthritis

A complex pattern of locomotion was shaped by the method of successive approximations in 20 DBS/1LacJ mice. Stable performances of the target response were achieved in an average of 10 15-minute sessions, conducted once per day. Following acquisition of the target response, daily sessions continued, with reinforcement available for only the target response, throughout the remainder of the experiment. Upon acquisition of stable responding at the target level, arthritis was induced in these animals, plus an additional 20, by injection of bovine type II collagen in Freund's complete adjuvant. The subjects were then divided into 4 groups of 10, in which half were exposed to the shaping protocol. Three of these groups were then treated with anti-inflamatory compounds, while the third group was given a placebo. Routine clinical and histopathological assessments were conducted on all of the subjects. In addition, changes in the locomotor responses of the trained animals were assessed against their baseline performances, and compared to ongoing clinical measures. Changes in locomotor measures, consistent with joint dysfunction, were observed prior to standard clinical measures of joint swelling and redness. The study demonstrates the utility of operant methods for the early detection of joint dysfunction in degenerative diseases.

Lauren Kettle & Peter Killeen
Arizona State University

Event- vs. Time-Based Memory Decay: A Correlational Analysis

Does memory of behavior decay with intervening events or the passage of time? Eight pigeons were exposed to a conjunctive fixed-time 30 s, fixed-ratio 4 schedule of reinforcement, which encourages variability in response rate. On each trial inferred memory was calculated with exponentially-weighted moving averages of inter-response times (IRTs). These equations weight IRTs as a function of a) time between the IRT and reinforcement; or b) the number of responses intervening between the IRT and reinforcement. Correlations between inferred memory and obtained response rate on the following trial were ascertained. The calculations were iterated over a range of memory decay rates to find the best-case predictions for event and time models. Will temporal or event decay provide the best prediction?

Eric Macaux, Armando Machado, & Rich Keen
Indiana University

Development of Preference in a Choice Situation

Compared to the vast literature on steady-state behavior, relatively few studies have sought to understand how an organism comes to prefer the richer of two probabilistic alternatives. This is quite surprising because although every experiment includes a period of acquisition of preference the majority of studies have not analyzed the data from this early stage. In the present research, we examined how two variables affect the rate at which pigeons learn to prefer the richer of two probabilistic alternatives. Experiment 1 varied the overall probability of reward, T, while keeping the difference in reinforcement probability, D, constant. Experiment 2 varied the difference in reinforcement probability, D, while keeping the overall probability of reinforcement, T, constant. We propose a theory based on a Weber-like ratio D/T (perhaps with different weights, e.g., D^a/T^b) that determines the rate at which the optimal patch is learned such that larger ratios result in faster acquisition. In general, the theory accounted for the results well.

James S. MacDonall
Fordham University

Stay and Switch Scheduling in Concurrent Chains

My recent research focused on examining the effects of stay and switch scheduling of reinforcements in concurrent schedules using two pairs of stay and switch schedules, and in isolation using one pair of stay and switch schedules. I found that in a concurrent procedure exchanging the switch schedules in each pair produces large changes in slopes of the generalized matching law. The present research extends this approach by viewing concurrent chain schedules as consisting of two pairs of terminal link schedules, each pair consists of a terminal link schedule that arranges reinforcers for staying and a terminal link that arranges reinforcers for switching. The present experiment exchanged the switch terminal link schedules which produced increases in slope of the matching line, which corresponds to the change seen in concurrent schedules.

Francis Mechner & Laurilyn D. Jones
The Mechner Foundation

Effects Of A Single Reinforcer Presentation On A Behavior Stream

The presentation of a single presumptive reinforcer within a stream of operant responses can have many different effects, of which the reinforcement of the immediately preceding response is only one. Our purpose was to determine the effect of a non-contingent presentation of $5 (a stimulus that could have reinforcing properties) on a stream of operant responses that consisted of keystroke sequences typed on a computer keyboard. The first three sessions were divided into "blocks" within each of which only three keystroke patterns were used as part of the responses, i.e., either patterns 1, 2, and 3; 4, 5, and 6; or 7, 8, and 9. Only operants that included one of the patterns 4, 5, or 6 were followed by $5 presentations, on a VR5 schedule. The other six patterns were never followed by money. In the fourth (final) session, all patterns except 4, 5, and 6 were used. At four points in that session, $5 non-contingent presentations followed operant occurrences randomly. For a few subjects, the presentation resulted in the immediate resurgence of one of the three patterns previously associated with monetary reinforcement, i.e., patterns 4, 5, and 6. In addition, three other possible effects were observed.

Marina Menez & Florente López
Universidad Nacional Autonoma de Mexico, Mexico

The role of reinforcement delay and interreinforcement interval in waiting time

Dynamically oriented proposals of timing suggest that waiting time in temporal schedules is the result of the delay from the first response within an interval to the subsequent reinforcer. To analize this posibility we exposed rats to chained Fixed-Ratio 1 Fixed-Time(x) schedules in an effort to directly manipulate delay time. To partial out the effects of the interreinforcement interval, a group of yoked subjects received reinforcement at the same times as the subjects in the chained schedule, provided they had emitted at least one response within the interval. The results indicated that waiting time was a linear function of the interreiforcement interval in both kinds of schedules. Waiting time also showed dynamic effects that are analysed in the context of current theories of timing.

Le'Ann Milinder, Peg Chaffee, Mike Raimon, Jenna Phillips, Jen Hale, & Donna Lee
New England College & Institute of Professional Practice

A Quantitative Program Evaluation: Preliminary Assessment of the Behavioral Impact of Individualized Settings

In the past two decades, behavioral interventions for adults with developmental disabilities have increasingly been conducted in community rather than institutional settings, with some states closing their institutions for the developmentally disabled. The last ten years have also seen a shift toward provision of treatment in individualized rather than group settings. The current study examined quantitatively the impact of a shift in setting on 40 adults with developmental disabilities and behavioral challenges who moved from group residential programs providing behavioral interventions to individualized community services. The individualized service options (ISO) all provided behaviorally oriented treatment to a single individual living with an individual, couple, or family. A study of archival data utilized a multiple baseline design across subjects. Rates of target behavior showed consistent, long-lasting reductions after change in treatment model. Representative data from 7 individuals are presented.

Cynthia J. Pietras, Nicole Dorey, & Timothy D. Hackenberg
University of Florida

Do obtained equality points improve the predictions of two choice models?

Four pigeons were given repeated choices between fixed-interval and progressive-interval schedules of food delivery in a two-component multiple schedule. In both components, completing the progressive-interval schedule requirement produced food and increased the schedule value. Completing the fixed-interval schedule requirement produced food (No Reset component) or also reset the progressive-interval schedule to its minimum value (Reset component). The value of the fixed-interval schedule was manipulated across conditions. Consistent with previous research in which reset and no-reset conditions alternated across conditions, the progressive-interval schedule value at which responding switched to the fixed-interval schedule was lower during the Reset component than during the No-Reset component. Switch points during the Reset component were analyzed in accordance with the predictions of two choice models, molar optimization and the summed-immediacy model, using either the programmed fixed-interval schedule values or the obtained switch points during No-Reset components (obtained equality points) as the fixed-interval schedule values. Although using obtained delays improved the predictions of the optimality model, the summed-immediacy model provided a more accurate description of switch points.

Bertram O. Ploog
College of Staten Island and The Graduate School & University Center of the City University of New York

The effect of differential food amounts on choice as a function of initial-link and terminal-link durations in a concurrent-chains schedule with nondifferential terminal links

In a previous study, primary reinforcement (different amounts of food) was shown to affect initial-link responding (choice) when the terminal-links were nondifferential, which eliminated conditioned reinforcement as a cause for the reported effect. In the present study, using the same general paradigm, the effect of variations in initial-link and terminal-link durations were studied in 12 pigeons. Preliminary data suggest that an increase in terminal-link durations (with VI 30 s or VI 120 s initial links) and an increase in initial-link durations (with VI 5 s or VI 20 s terminal links) decreased preferences for the large reinforcer. These findings are noteworthy because they were obtained without differential conditioned reinforcement, which is a critical factor in several choice models.

Diana Posadas-Sánchez & Peter R. Killeen
Arizona State University

Closing the Open Economy

Under open economies initial pauses increase with longer fixed-intervals, whereas in closed economies pauses do not change. In this experiment we generated responding consistent with both open and closed economies within a single long session. Ten pigeons were exposed to an ascending series of fixed-interval schedules, with two ten-hour sessions of exposure and with one or two days between sessions at each interval value. Responding at the beginning of the session (when subjects were more hungry) resembled behavior in an open economy, and responding at the end of the session (when subjects were less hungry) resembled behavior in a closed economy. In our view, motivational differences provide a parsimonious alternative account of the discrepancies between open and closed economies.

Mark P. Reilly & Diana Posadas-Sánchez
Arizona State University

Amphetamine Decreases Keypecking as Predicted by Momentum Theory

Drugs, like other disruptors such as extinction and prefeeding, should reduce responding more in the presence of a stimulus correlated with a low rate of reinforcement than in the presence of a stimulus correlated with a high rate. Five pigeons were trained to peck for food reinforcement under a multiple schedule with five fixed-ratio schedule components (4, 10, 25, 50 and 100). Components were in effect for two reinforcers or until a time limit expired. Each ratio schedule occurred five times per session, however the order was randomized with the constraint that the entire series had to be presented once before a ratio schedule was repeated. d-Amphetamine (0, 1.0, 1.8, 3.2 and 5.6 mg/kg) was administered every fourth session as long as response rates in each of the five ratio schedules were stable. As predicted by behavioral momentum theory, responding controlled by the smaller ratio schedules was more resistant to the rate-decreasing effects of d-amphetamine (ED50 values were systematically higher in small ratio schedule). This differential resistance-to-change following drug administration also was observed following exposure to extinction. These data support the viability of using momentum theory to illuminate behavioral mechanisms of drug effects.

Matthew T. Sitomer, Armando Machado, & Richard Keen
Indiana University

Quantitative Predictions of Two Timing Theories

The predictions of two competing interval timing theories, Learning to Time (LeT; Machado, 1997) and Scalar Expectancy Theory (SET; Gibbon, 1977), were tested in a time-place learning task. Pigeons learned to respond at one of two spatially disparate locations depending on the duration of a diffuse sample stimulus. LeT assumes that animals use overt behavioral patterns to time intervals. In the present experiment, the birds exhibited consistent locomotor patterns during the interval to be timed. Pivoting floor panels recorded a bird's position in the chamber throughout the session, enabling an analysis of arrival and departure times at each patch. The two theories make contrasting predictions about both the relationships between these measures and the bisection points for each pair of durations. The correlation results generally supported LeT; but, overall, departure times were considerably earlier than the predictions made by both theories. The quantitative predictions made by each theory are presented graphically and evaluated with respect to these new data.

Francois Tonneau, Maria Eugenia Soto, & Americo Rios
University of Guadalajara - Mexico

Within- and Inter-Session Resistance to Change

The dependent variable of resistance-to-change research is usually response rate at the level of an entire session. However, if the concept of resistance to change actually identifies a general property of behavior, this property should be evident at more local levels, for example from minute to minute. Considerable evidence shows that response rate decreases toward the end of a regular operant session, suggesting the presence of a disrupting agent such as satiation or habituation. Here we report the results of an experiment that compares resistance to change at the within- and inter-session levels, with rats as subjects and water as a reinforcer. We hope that our data will promote theoretical connections between the literature on resistance to change, on the one hand, and the literature on within-session effects, on the other hand.

Oscar Zamora-Arvalo, A. Bouzas-Riao, & R. Gonzalez
Universidad Nacional Autonoma de Mexico, Mexico

Learning and Development of Preferences in Multiple Schedules

In two experiments we analyzed the behavioral adjustment to transitions in multiple schedules of reinforcement. The general procedure was similar to that used by Mazur (1992, 1995 & 1996) with concurrent schedules. In Experiment 1, eight pigeons were exposed to five different conditions, which consisted of three days of training and two days of transition. During training, keypecking was reinforced according to a variable ratio schedule (VR) for both components; during transition, keypecking was reinforced according to two different variable ratio schedules (VRx VRy) for each component. Every condition was repeated four times. In Experiment 2, eight pigeons were exposed to the same five different conditions of training and transition except that an alternate concurrent component (VI 30 s) was included and the number of sessions of transition equalled three. The main finding was that keeping the difference between two reinforcement probabilities constant, the development of preference increased with the ratio of the probabilities.