Abstracts of 2000 Conference Papers

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

SQAB Presentations

Friday, May 26, Room Wilson A/B

Gene Fisch
Yale University

Evaluating Data from Behavioral Analysis: Visual Inspection of Statistical Models?
(Early Bird Breakfast Tutorial)

Traditional behavior analysis relies upon single-subject study designs and visual inspection of graphed data to evaluate the efficacy of experimental manipulation. Attempts to apply statistical inferential procedures to analyze data have been strenuously and successfully opposed for many decades, despite increasingly cogent arguments for the use of inferential statistics and the identification of limits to visual inspection. In a series of experiments, we showed that trained behavior analysts could identify level shifts in responding during intervention phases ('treatment effect') in autocorrelated data, but trends were either misconstrued as level treatment effects or went completely unnoticed. These errosr illustrate the liabilities of using visual inspection as the sole means by which to analyze behavioral data. The classic argument advanced opposing use of statistical inference was that it ceded experimental control to statistical control as the means by which to reduce inter-subject variability in responding. However, as behavioral techniques have graduated from laboratory experimental chambers into classrooms, group homes and other facilities for developmentally delayed individuals, experimental control over the environment has diminished. Session-to-session responding can be more variable and the effectiveness of interventions compared to baseline responding less discernable. Meanwhile, because of greatly increased computer power and advanced mathematical techniques, previously undeveloped or underutilized statistical methods have become far more sophisticated and brought to bear on a variety of problems involving longitudinal data. Nonparametric procedures and other stochastic models will be introduced and used to evaluate traditional behavioral data to augment, not replace, visual inspection procedures.

William Uttal
Arizona State University

Questions Concerning the Accessibility, Analyzability, and Reducibility of Cognitive Processes

Much recent activity in cognitive psychology is based upon a set of fundamental assumptions that are rarely made explicit. On close examination some of these assumptions appear to be either unsupportable or unresolvable. Therefore, serious questions arise about many of the current research programs, applications, and theoretical points of view to which psychological science is currently committed. It seems appropriate to consider what these assumptions and questions are and how scientific psychology should respond to a reevaluation of them. One possible conclusion is that we may have prematurely cast aside some of the important tenets and constraints that characterized behaviorism. It may be time to reconsider this point of view in a revised and revitalized form.

Allan Stubbs, Laurence Smith, & Lisa Best
University of Maine

Graphs, Tables, and Equations: Data Practices in Psychology and Other Disciplines

Scientists use graphs for data presentation, but not all scientific fields use graphs to the same degree. We examined graph use in different scientific disciplines and different speciality areas within psychology to gather information about the number and type of graphs used in different journals. Use of graphs across seven scientific disciplines correlated almost perfectly with measures of the hardness of disciplines on a continuum of hard and soft science. The same pattern held across ten speciality areas within psychology. Use of tables also varied within psychology with greater use of tables in the softer areas of psychology. Details about types of graphs will be explored as will change in graph use over time. The results raise the issue of whether different graphical practices indicate different conventions in different areas and whether graphs are an essential part of science.

William Timberlake
Indiana University

Biology and Behavior: Some Thoughts on Integration

At the end of the 19th century, psychologists and biologists used similar methods in studying behavior. Both focused on the control of everyday behavior, the potential contributions of development and learning, and mechanistic models of causation. During much of the 20th century the methods and interests of psychology and biology diverged. However, the end of the 20th century and the beginning of the 21st show increasing promise of integrating methods and approaches, including: (1) relating laboratory procedures and results to the potential ecological function and regulation of behavior. (2) relating evolution and learning. (3) integrating learning and development. (4) connecting causal analyses of behavior to physiological, neural, and genetic substrates.

Peter Balsam
Columbia University

Associators, Accumulators and Cartographers

Three classes of models have been proposed as explanations for Pavlovian conditioning. These approaches posit that either associations, quantitative accumulation of events or time, or the acquisition of temporal maps as the basis of learning. The development of these ideas during the last century and their relations to each other will be discussed.

Ben Williams
University of California at San Diego

The Critical Dimensions of the Response-Reinforcer Contingency

Two major dimensions of any contingency of reinforcement are the temporal relation between a response and its reinforcer, and the relative frequency of the reinforcer given the response versus when the response has not occurred. Various quantitative descriptions of the quantitative descriptions of the effects of the temporal dimension (e.g., Mazur's) have assumed that response strength is a hyperbolic function of the response-reinforcer delay. Data will be presented showing that time, per se, is not sufficient to explain the effects of delay-of-reinforcement procedures; needed in addition is some account of the events occurring in the delay interval. Moreover, the effects of the same absolute time values vary greatly across situations, such that any notion of a standard delay-of- reinforcement gradient is simplistic. The effects of reinforcers occurring in the absence of a response are yet more difficult to specify, because it remains unclear whether such effects depend upon the strengthening of competing behavior, or a reflection of the animal's calculus of causation. Data will be presented illustrating the importance of the stimulus conditions under which "unpaired reinforcers" are presented, which pose major difficulties for any theoretical framework that invokes contingency as a causal variable. The major challenge to future research is to reconcile models of reinforcement from the associationistic literature, which are episodic in nature, with those from the operant literature which invoke rate of reinforcement as a primitive variable. Recent timing-based accounts of operant behavior seem promising for a reconciliation of these two separate traditions, but these too seem to have major limitations.

Donald Blough
Brown University

Quantitative Approaches to Stimulus Control

The paper will briefly characterize major attempts to identify key variables and processes involved in stimulus control. I will touch on relationships between associative and psychophysical approaches, especially applications of signal detection theory. This will lead to a review of newer methods, particularly search procedures and the reaction time measures, and the ways these may expand our understanding of stimulus control."

Charles Shimp
University of Utah

Historical, Theoretical, and Cultural Perspectives on 'Choice Behavior'

Ian Hacking's recent book, The Social Construction of What?, describes several levels of commitment to a constructionist philosophy of science. An intermediate level involves a process of 'unmasking,' by which he refers to a position 'which does not seek to refute ideas but to undermine them by exposing the function they serve. ...We unmask an idea ... to strip it of a false appeal or authority.' It is interesting to explore what might happen if we apply this level of constructionist analysis to the behavioral analysis of choice behavior. Scientific traditions may carry with them extra- scientific and inadequately examined assumptions imported from the surrounding culture. Several of these assumptions may be identified in the contemporary experimental analysis of choice behavior. Some of them have become laboratory traditions by which data are collected and analyzed. Others underlie the way behavioral researchers talk and how they evaluate theory. These assumptions involve the meaning and value of 'smooth curves,' 'parsimony,' mean response rate as a basic datum, 'classic' contingencies of reinforcement, the role of simple algebraic formulas as summary descriptions of behavior, the nature of scientific observation, and ultimately, 'behavior.' All these assumptions might need clarification, revision, or even rejection. A different kind of analysis of choice behavior might ensue that would promote developing alternative methods and theories for the control of the behavior of an individual organism.

Edmund Fantino
University of California at San Diego

Choice and Foraging: It's the Context, Stupid

Seminal research in several areas has underscored the central role played by context in the control of behavior. Landmark studies in classical conditioning (with both conditioned suppression and autoshaping procedures) and in conditioned reinforcement (using the observing paradigm) are reviewed. The role of context also proved central in the study of choice (including the matching law and delay-reduction theory). This latter work contributed to the development of experimental analogs to foraging behavior. Research on foraging has also highlighted the importance of context and has led to some counterintuitive predictions that are mediated by context.

Saturday Morning, May 27, Room Wilson A/B

Gene Fisch
Yale University

Evaluating Data from Behavioral Analysis: Visual Inspection of Statistical Models?
(Early Bird Breakfast Tutorial)

Traditional behavior analysis relies upon single-subject study designs and visual inspection of graphed data to evaluate the efficacy of experimental manipulation. Attempts to apply statistical inferential procedures to analyze data have been strenuously and successfully opposed for many decades, despite increasingly cogent arguments for the use of inferential statistics and the identification of limits to visual inspection. In a series of experiments, we showed that trained behavior analysts could identify level shifts in responding during intervention phases ('treatment effect') in autocorrelated data, but trends were either misconstrued as level treatment effects or went completely unnoticed. These errosr illustrate the liabilities of using visual inspection as the sole means by which to analyze behavioral data. The classic argument advanced opposing use of statistical inference was that it ceded experimental control to statistical control as the means by which to reduce inter-subject variability in responding. However, as behavioral techniques have graduated from laboratory experimental chambers into classrooms, group homes and other facilities for developmentally delayed individuals, experimental control over the environment has diminished. Session-to-session responding can be more variable and the effectiveness of interventions compared to baseline responding less discernable. Meanwhile, because of greatly increased computer power and advanced mathematical techniques, previously undeveloped or underutilized statistical methods have become far more sophisticated and brought to bear on a variety of problems involving longitudinal data. Nonparametric procedures and other stochastic models will be introduced and used to evaluate traditional behavioral data to augment, not replace, visual inspection procedures.

Thomas Zentall
University of Kentucky

Animal Cognition: Emergent Relations or Representations?

The dangers of hypothesizing cognitive mechanisms are well known to behavior analysts. The benefits may be less well appreciated. I will present two examples of how cognitive theories can broaden the scope of one's research program; research that contributes to our understanding of animal learning and which may not have been conducted in the absence of such theorizing. The first question deals with the nature of a pigeon's 'representation' of two stimuli both of which are associated with correct responding to a third (stimulus equivalence or common coding). The second has to do with the effects of prior effort on the preference for the outcome that follows, a phenomenon that social psychologists classify as 'cognitive dissonance.'

Russell Church
Brown University

Animal Cognition: 1900-2000

Early in the century, the basic problems of animal cognition had been described, and studies had been conducted that used various research methods. Progress during the century has been characterized by improvements in the apparatus, the experimental methods, and the explanations of results. Apparatus of importance includes the switch, clock, counter, relay, cumulative recorder, and computer. Experimental methods of importance include those that permit the identification of critical independent variables, and those that permit the description of functional relationships between independent and dependent variables. Explanations of importance include quantitative models with intervening variables that may have only formal properties or that may correspond to psychological or neural processes. Both probabilistic and deterministic models have provided reasonable approximations to some experimental results, but further improvements are necessary for accurate predictions to be made about the results from many different procedures.

Peter Killeen
Arizona State University


What is being summed by my title? Can science cumulate, or just its data? Have models evolved, or just changed? What are we doing when we do QAB? Have we been in analysis too long? How do we know what serves as answer to these questions? Does it matter as long as we're reinforced -- having fun and funded? My last words on the topic in the 20th century will settle the salient issues. But, like all oracles, they conserve entropy by raising other issues in their stead.

SQAB Poster Presentations

Carlos F. Aparicio
University of Guadalajara

The Barrier Choice Paradigm: Varying the Number of Alternatives and the Size of the Space.

This poster reports recent results obtained with rats responding for food under the barrier choice paradigm. The experiment compared the performance of two groups of rats across two phases that involved different numbers of levers. The first phase involved two levers in Group 1 and four levers in Group 2; 30.5-cm barriers blocked free access to the levers. In the second phase the number of levers was increased from two to four in Group 1, the available space was reduced, and the barriers were removed -- whereas in Group 2 the number of levers was increased from four to eight, the available space was enlarged, and more barriers were added to block access to all levers. The rats' patterns of searching for food changed as a function of travel requirement and the number of available levers. The slope of the generalized matching law was above 1.0 when barriers were present. With free access to four levers, the generalized matching law showed slopes below or equal to 1.0. The role of locomotion in choice situations is discussed.

SangWeon Aum, Bruce L. Brown, Nancy S. Hemmes, & Robert N. Lanson
City University of New York and Queens College

Evaluation of Temporal Control by the Feature Stimulus in a Pavlovian Feature-positive Discrimination

Timing of the feature stimulus in Pavlovian serial feature positive discriminations was investigated in pigeons. Two different feature-conditioned stimulus (CS) intervals were signaled by different visual patterns (feature stimuli) projected on a key. The measure of conditioning was proximity to the feature stimulus (Brown, Hemmes, & Cabeza de Vaca, 1997). Relative timing of the feature-CS intervals was assessed by the superposition of functions for relative response level plotted against relative time (Gibbon, 1991). While the superposition values were consistently higher than those for the relationship between absolute response levels and absolute time, they did not differ from those for the relationship between relative response levels and absolute time. These results have implications for evaluation of scalar timing.

Munire Ozlem Cevik
Indiana University

The effects of increased physical activity and methamphetamine on temporal discrimination in rats.

Sprague Dawley rats were trained to press different levers following short and long signals. Eight different signal durations spaced at equal geometric intervals were used. Probability of a long response increased as a sigmoidal function of signal duration. Both forced running on a treadmill during the ITI and administration of a 0.75 mg/kg methamphetamine mainly decreased the slope of the psychometric function. Considerable variability in the effects of methamphetamine was observed both between subjects, and within subjects over time. A change in the speed of the internal clock cannot explain all aspects of the observed changes in behavior.

Bryce S. Cleland, T. Mary Foster, & William Temple
University of Waikato

Interactions of topographically different behaviors in multiple schedules when reinforced and in extinction

Behavior in one component of a multiple schedule has often been shown to be influenced by the schedule in effect for the behavior in another component. Not understanding the extent of interactions between behaviours in any particular multiple schedule experiment may lead to incorrect attribution^s about the independent variable of interest. In addition, identifying the role of interactions in producing the results of an multiple schedule experiment may be required if the aim is to generalise the conclusions to simple schedules. Some data suggests that the extent of interactions may be influenced by the topography of responses used. Data will be reported on the interaction between topographically different behaviors (head bobbing and door pushing) and topographically similar behaviors (left and right key pecking) in hens during periods where the reinforcement rate of one behavior varied. If interactions are evident when contingencies vary across components they may also occur when contingencies are the same across components but have previously been different. Data will also be reported from periods when the behavior in both components was under extinction following the periods where the reinforcement rate of one behavior varied.

Michael Lamport Commons
Harvard Medical School

Hierarchical complexity of contingencies of reinforcement and of stimulus discriminations

Quantitative analysis of behavior has historically focused on the reinforcement behavioral relationship. Quantitative account of choice (measure by rate or simple choice) have yielded complex and not entirely satisfying results (Davison, personal communication). Without abandoning the quest to understand operant and respondent conditioning, is there any characterization of contingencies that might help? In cognitive psychology, the amount of information in situations that are really just contingencies has been varied and performance has been tracked. These result have been largely carried out without free operant behaviors. But an entirely separate field, Precision Teaching, whose members would not support the following interpretation have yielded an account of information processing using free operant methods. They have learned that there are two types of information, the simple tradition amount that a person or animal may discriminate in a period of time as shown by the number of response (either all responses or correct responses) and hierarchical complexity, which they refer to as elements and combinations. It is suggested here the schedules of reinforcement may also be analyzed as in precision teaching, with two types of information characterizing the schedules or trial contingencies. Example are used from Rachlin's and Herrnstein's accounts of choice. With ought to attend to the order of hierarchical complexity of contingencies of reinforcement, the acquisition of stimulus control under both free-operant and trial situations.

Conover, K., Oda, K. & Shizgal, P.

How Does Cocaine Amplify The Reward Value Of Brain Stimulation? - Different Rats Tell Different Stories.

Psychophysical tradeoff studies provide clear evidence that cocaine increases the reward effectiveness of lateral hypothalamic (LH) stimulation. However, tradeoff methods do not reveal how cocaine affects the reward circuit subserving LH self-stimulation. Cocaine may amplify the output of the circuit, if so, then all stimulation trains will increase in reward value, including those at the peak value. Alternatively, amplification of the input to the circuit will only increase the value of trains below the peak, without changing the maximum possible reward value. Labor supply theory provides the basis for measuring reward value for rats in a lever holding task independently of the operant response rate. In this approach, rats balance the utility of time spent in leisure activities (grooming, exploring), with the utility of stimulation such that under fixed conditions the average amount of leisure time per reward (demand ratio), remains constant. The demand ratio decreases as the reward value increases or as the required lever hold time decreases. A quantitative model of these effects was then used to determine the range of pulses at which reward value peaked and the number of pulses at which the value was half-maximum. Moreover, this approach affords measurement of the time course of changes in reward value following cocaine injection. Out of five rats, cocaine (10 mg/kg, ip) amplified the output of the reward circuit in two, and amplified the input in three. These various results suggest that cocaine may act differently between individuals, possibly due to different stimulation sites.

Bill Dube & Bill McIlvane
E. K. Shriver Center for Mental Retardation

Assessing Behavioral Momentum in Humans with Mental Retardation and Unstable Baselines

Behavioral momentum is typically assessed by measuring relative disruption in the components of multiple schedules. Disrupters may include procedures such as prefeeding, concurrently available alternative reinforcers, and so forth. Behavioral mass is indexed by the degree of reduction in response rates during disruption, as a proportion of baseline rates. Thus, typical procedures require a series of baseline sessions with stable response rates, followed by one or more sessions with disrupters. In an ongoing study assessing behavioral momentum in humans with mental retardation, we have occasionally encountered cases where even liberal baseline stability criteria were not met within a practical number of sessions. This poster will describe an alternative testing procedure for such situations. The baseline schedule was multiple (VI 10-s) (VI 10-s + VT 7-s) with alternating 1-minute components. The disrupter was a concurrently available response option with a VI 7-s schedule. In every third session, this disrupter was presented during the last 1- minute period for each component. Response-rate reductions during the disrupter periods were measured relative to the earlier baseline rates within the same session. For two subjects tested thus far, the data were consistent with momentum effects.

Echenko O., Kostenkova V., & Nikolskaya K.
Department of Higher Nervous Activity
Moscow State University

A New Behavioral Model for Studying Animal Cognition

Recently, simple behavioral models have become popular in neuroscience. A battery of simple tests is widely used in psychopharmacology and also used for behavioral characterization of genetically modified animals. This approach has a serious shortcoming, as simple models do not allow analysis of behavior in details, the study of cognitive processes or individual peculiarities of learning and memory. We aimed to demonstrate the advantages of a systematic approach to study animal cognition using a model of problem solving in a multi-alternative maze, which could refer a human cognitive activity.

Rats had to form a cyclic 4-link habit in a complex T-branched maze by themselves. The rats had to enter the maze (1st link), locate and get a piece of food from one or two true feeders of four ones in the maze (2nd and 3rd links). The other two feeders never contained food (false). In order to obtain more food rats were required to leave the maze (4th link) and re-enter it again (1st link). The peculiarities of the model are the following: 1) the task can not be solved simultaneously because the input information capacity exceeds the brain capacity; 2) existence of several equivalent route realizations of the formed plan of behavior; 3) the task logic is presented in an implicit form, so a creation of a few working hypotheses is required for its determination. The main concepts of information theory, semiotic and psycholinguistic theories have been applied to the analysis of information processing. Two learning strategies have been revealed.

An exponential learning curve, slow extinction of mistakes, high associative abilities and leading role of working memory characterized the "excitatory" type of learning. The task performance was unstable. An insight-like learning curve, fast extinction of mistakes, delayed recognition of the task structure and leading role of long-term memory characterized the "inhibitory" type of learning. The habit performance was very stable and optimal. This model allows studying: 1) principles of autoshaping since a free-choice method excludes any conditioned stimuli; 2) cognitive processes (perception, estimation and prognosis) since the task logic is set in implicit form and information re-coding is needed; 3) habit optimization since there are several possible route realizations; 4) motivational state as a number of cycles per session is dependent on the animal; 5) inter-group and inter-strain differences of cognition, learning, and memory function. Various vertebrates and invertebrates were tested in this model. The model appeared to be very sensitive in detecting cognitive dysfunction after pharmacological or physical influences.

Gordon R. Foxall
Departments of Management and Psychology
Keele University, England

How And What Do Consumers Maximize?

Matching theory predicts that choices on concurrent the variable ratio schedules on which consumer brand selection occurs will show maximization via exclusive choice of the richest schedule. However, aggregate studies of consumer choice indicate two modes of consumer brand purchase within a product category: either exclusive purchase of one brand or multi-brand purchasing. This paper uses brand selection data from individual consumers to determine whether, at this level of analysis, (i) consumers' purchasing patterns can be described as matching, (ii) the observed patterns support the expectation that consumers maximize returns, and, if so, (iii) what it is that they are maximizing. Consumers of fast moving consumer goods for whom data are presented are found to exhibit both exclusive choice and multi-brand purchasing that can be described as matching. Further analysis indicates that the latter takes tow forms: for substitutes (e.g., competing brands of butter) brand selection is price sensitive, suggesting both melioration and maximization; for non-substitutes, choice is not price sensitive but still appears consistent with maximization of price- and nonprice-related sources of value. Implications are drawn for the analysis of consumer behavior in the context of marketing.

Randolph C. Grace
University of Canterbury

Pavlovian Determiners of Acquisition in Concurrent Chains

Experiment 1 compared the acquisition of initial- and terminal- link response patterns in concurrent chains. The terminal-link schedules were FI 10 s and FI 20 s, but some presentations were analogous to no-food trials in the peak procedure, lasting 60 s with no reinforcement delivery. Four pigeons completed a series of reversals in which the schedules signaled by the terminal-link stimuli (red and green on the center key) were switched. For all subjects, acquisition of temporal control of terminal-link responding (as measured by peak location on no- food trials) was more rapid than acquisition of preference in the initial links. Experiment 2 compared acquisition in concurrent chains using the typical arrangement in which the terminal-link schedules are reversed with a novel procedure in which the initial-link key assignments were switched while the terminal-link schedules remained constant. Across subjects, acquisition of preference was faster in the latter condition in which the terminal- link stimulus-reinforcer relations were preserved. Overall, both experiments suggest that although the terminal-link stimulus reinforcer relation is an important determiner of preference, it does not play a direct mediational role.

Leonard Green, Joel Myerson, & Rachel C. Schneider
Washington University

Is There A Magnitude Effect In Tipping?

The magnitude effect refers to the finding that a ratio of smaller amounts is often perceived to be less than an equal ratio of large amounts (e.g., the ratio of $1 to $10 may be judged as smaller than the ratio of $100 to $1,000). The magnitude effect is of interest because it is hypothesized to reflect the shape of the utility function for money. Assume that when deciding how much to tip, people are trying to leave an amount that is a given proportion of the bill. If this assumption is correct, then the magnitude effect predicts that the relative size of the tip (i.e., the percent tip), should be a decreasing function of the amount of the bill. In a real-world test of the magnitude effect, we collected data from restaurant diners, cab patrons, and hair salon clients. In all three tipping domains, although the absolute amount of the tip increased with the size of the bill, the percent tip decreased, as predicted based on the magnitude effect. These results are consistent with the Increasing Proportional Sensitivity hypothesis (Chapman & Winquist, 1998; Prelec & Lowenstein, 1991), proposed to explain the magnitude effect, and set constraints on theoretical accounts of the shape of the utility function for money.

Julie A. Grimes & Richard L. Shull
University of North Carolina-Greensboro

Behavioral Momentum: Free (VT) Milk Enhances Persistence of Rats' Pellet-reinforced Leverpressing

If, during training, one stimulus is correlated with a higher rate of reinforcement than another, responding is more resistant to extinction in the presence of that higher-rate signal--even if many of the reinforcers are independent of responding (i.e., free or VT deliveries) (e.g., Nevin, Tota, Torquato, & Shull, 1990). Apparently, the stimulus-reinforcer (Pavlovian) contingency affects a response's persistence. For the present study we asked if the free (VT) reinforcers could be different from the response-contingent reinforcers for this effect. Eight Long Evans Hooded rats obtained 45-mg food pellets by leverpressing (variable-interval 100-s schedules) in the presence of two stimulus conditions (blinking or steady lights) that alternated every minute during daily 40-cycle sessions. Also, one of the stimuli (counterbalanced across rats) signaled response-independent presentations of diluted sweetened condensed milk (VT schedule). Extinction sessions were identical to training except that neither pellets nor milk were delivered. Some training conditions produced no evidence of differential persistence, but others did. Leverpressing was more persistent in the presence of the milk-correlated stimulus when (a) the size of the milk deliveries during training (VT 30 s) was 0.04 ml (versus 0.01 ml) and (b) 2- or 4-minute blackouts separated components. Free reinforcers do not have to be the same as the earned reinforcers to enhance persistence.

L. J. Hayes, K. W. Hunter, N. Publicover, R. Quniones & C. Vinci
University of Nevada-Reno

Four patterns of movement to designated locations in a restricted space were shaped under different stimulating conditions, and subsequently extinguished, in sequence. Movements were recorded as digitized photographic images and shaped automatically. Following acquisition of the last pattern, an extended period of extinction for all patterns, in the absence of all previous discriminative stimuli, was implemented, and the resurgence of previously extinguished movement patterns was observed. These procedures were employed with mice subjected to diets deficient in protein, to examine the effects of malnutrition on learning and remembering.

Lauren Kettle, Lewis Bizo, & Peter Killeen
Department of Psychology, Arizona State University

The Paradoxical Incentive Effect

Previous research has suggested that rats tend to respond faster for smaller reinforcers. Why? The current study was conducted to address this issue. Six rats experienced an ascending series of variable-ratio values with reinforcement consisting of 1 sucrose pellet, 1 A/I Noyes pellet, or 2 A/I Noyes pellets. Response rates were higher, at smaller ratio values, when lever pressing for 1 pellet than 2 (the paradoxical incentive effect). The mathematical principles of reinforcement (Killeen, 1994) suggests that large reinforcers erase memory for previous behavior, with 2 pellets erasing more memory than 1 pellet. Killeen's lambda parameter captures the rate of short-term memory decay and was used to examine the memory hypothesis of the paradoxical incentive effect.

Gregory J. Madden, Andrea Gantz, Bethany Raiff, & Lana Kastern
University of Wisconsin-Eau Claire

Temporal Discounting in Humans: Real vs. Hypothetical Rewards

Experiments examining the degree to which humans discount the value of delayed rewards nearly always employ hypothetical rewards. That is, participants are asked to repeatedly imagine what they would do if given choices between small amounts of money now and larger amounts of money later. Some researchers randomly select one of the participants' choices and award him/her the amount of money selected in that choice trial. We examined whether participants discounted delayed rewards more or less if all of the rewards were hypothetical or if one of the rewards would be randomly selected and given at the end of the session. Using a between-subjects design, participants chose between $10 delivered after delays ranging from 6 hours to 1 year and an adjusting amount of money delivered at the end of the session. Discounting functions across this delay range were plotted and fit with hyperbolic and exponential discounting functions. Participants in the real rewards group discounted the value of the delayed rewards significantly less than participants in the hypothetical group. The hyperbolic function provided a better fit of the discounting functions in both the real and hypothetical groups.

James E. Mazur
Southern Connecticut State University

A test of the constant-difference effect in concurrent-chain schedules

Using concurrent-chain schedules, Savastano and Fantino (1996) found that preference (initial-link response ratios) remained constant as long as the subtractive difference between the lengths of the two terminal-link schedules remained constant. This result is predicted by delay-reduction theory, but not by several alternative mathematical models of choice. This experiment tested the constant- difference effect with four pigeons responding on concurrent-chain schedules, with variable-interval schedules as initial links and fixed delays as terminal links. The key colors and positions of the different terminal links were counterbalanced across conditions. The pigeons tended to show more extreme preference for 2-s versus 12-s terminal links than for 40-s versus 50-s terminal links, thus calling into question the generality of the constant-difference effect.

Francis Mechner & Laurilyn D. Jones
The Mechner Foundation

Resurgence and Performance Errors

Resurgence is defined as the reappearance of antiquated behavior patterns (those observed earlier in a subject's learning history). In a number of separate experiments, human subjects typed non-word sequences of letters on a computer. Each sequence contained both criterial (mandated) and non-criterial (optional) keystrokes. Subjects learned nine different criterial keystroke patterns. Experiments consisted of several 'history-building' sessions and a final 'test' session. The objective was to determine the effects of certain variables during the learning history on final session performance, and to study the characteristics of resurgence during the final session. Of the different history session variables, the number of repetitions for each of the nine criterial patterns had the clearest effect, measured by noting how often each pattern was chosen during the final session. Data showed that the steeper the ratio of relative numbers of repetitions during history sessions, the stronger, and more uniform, the preference for higher-repetition patterns during the final session. Resurgence was measured by noting each non-criterial keystroke sequence typed during the final session and determining when it had been previously used during the subject's history sessions. Higher levels of resurgence were strongly associated with invalid operants. There was also an association between higher resurgence levels and the use of criterial patterns that had been repeated less often during the history sessions.

Joel Myerson, Leonard Green, Missaka Warusawitharana, Wanjiang Du, & David R. Adams
Washington University

Area Under The Curve As A Measure Of Reward Discounting

Efforts to evaluate individual and group differences in discounting are complicated by the absence of a single measure of discounting that is consistent with two-parameter models (e.g., Green et al., 1999; Myerson & Green, 1995). The two parameters of such models are typically not independent, thereby limiting the statistical comparisons that can be conducted. With respect to our own hyperbola-like model, moreover, neither parameter is normally distributed (i.e., both are highly skewed and show kurtosis) and thus, parametric statistical analyses are not appropriate for testing differences in these parameters. As a result, we have typically relied on non-parametric techniques, which are generally less powerful than their parametric counterparts. To deal with these problems, we propose using the area under the discounting function as a measure of the extent to which an individual discounts delayed or probabilistic rewards. We compared two indices, one based on integrating the best-fitting hyperbola-like function and the other calculated by summing the area under each of the observed data points. Importantly, the latter measure is purely descriptive and does not depend on assumptions that may underlie specific theoretical models. Both of these methods appear to result in approximately normal distributions of individual discounting indices as required by parametric statistics. We report the results of re-analyses of previously published data that demonstrate the utility of these measures.

Tatsuya Ohyama, Javier Medina, & Michael D. Mauk
Department of Neurobiology and Anatomy
University of Texas Medical School at Houston

Latent acquisition of CS-US timing in classical conditioning

A model of Pavlovian conditioning based on the neurobiology of the cerebellum suggests that there are at least two sites at which synaptic changes occur (Mauk & Donegan, 1997). One site is the anterior lobe of the cerebellar cortex, which is involved in the timing of the conditioned eyelid response in relation to the interstimulus interval or ISI (Garcia & Mauk, 1998; Medina & Mauk, 1999; Perrett, Ruiz, & Mauk, 1993). A second site is the anterior interpositus nucleus, and learning at this site is assumed to play a role in the expression of the conditioned response (CR). Since changes in the interpositus are dependent on changes in the cerebellar cortex, learning at the interpositus occurs after learning at the cerebellar cortex. To test this hypothesis, we gave one group of rabbits (PP) threshold-level training in phase I by pairing a 750 ms tone with periorbital shock at a 700 ms ISI, while group UP received unpaired training and group NP received no training. In phase II, all groups received asymptotic training with a 250 ms tone and a 200 ms ISI, and were tested with the 750 ms tone presented alone. Group PP showed double-peaked responses to the 750 ms tone, with a sharp second peak occurring near the ISI in phase I. Group UP and NP showed only single-peaked responses. The data are consistent with the notion that timing was learned in phase I but remained latent in the cerebellar cortex, and this learning was expressed by subsequently inducing synaptic change in the interpositus in phase II. Although traditional learning theories may provide ad hoc explanations, the findings attest to the utility of biologically constrained models of conditioning in generating novel behavioral predictions.

Cynthia J. Pietras & Timothy D. Hackenberg
University of Florida

Risky Choice in Humans Under Positive and Negative Energy Budgets

The energy-budget rule, a model of risky choice developed by behavioral ecologists, predicts that risk sensitivity will vary as a function of rate of energy gain, daily energy requirements, and energy reserves. Under positive energy-budget conditions, in which selections of the certain option will fulfill daily energy requirements and selections of the variable option will do so only probabilisitically, the model states that choice should be risk averse. Under negative energy-budget conditions, in which selections of the certain option will not fulfill daily energy requirements, choice should be risk prone. The aim of the present experiment was to assess these predictions with adult humans under laboratory conditions. Positive and negative energy-budgets were arranged by manipulating the number of points that had to be accumulated within a block of trials (a criterion analogous to an energy requirement) for points to be exchanged for money. Three participants were given repeated choices between fixed and variable point amounts. Sessions consisted of twelve blocks of five trials. The first six blocks were forced-choice trials, in which only the fixed or variable option was presented on all five trials of a block, and the second six blocks were free-choice trials, in which both options were presented concurrently. Exclusive preference for the fixed option could meet the criterion under positive, but not negative, energy-budget conditions. Consistent with the predictions of the model, choice was risk-averse under positive energy- budget conditions and risk-prone under negative energy-budget conditions.

Diana Posadas-Sanchez, Thomas B. DeMarse, & Peter R. Killeen
Department of Psychology, Arizona State University
Division of Biology, California Institute of Technology

Will responding be maintained to a greater extent on a VI schedule that is proximate to reinforcement or to one that is distal to reinforcement?

Pigeons were trained to peck on a left and right key in two different multiple schedules. One was a (VT-VI), the other was a (VI-VT), and extinction sessions were introduced after completion of each schedule. In this experiment, the rate of responding on extinction was compared during the two VI schedules on a left and right key, when one VI schedule was either followed immediately with reinforcement (i.e., VT-VI) or by a VT schedule followed by reinforcement (i.e., VI-VT). Responding should be greater during the VI schedule closest to reinforcement. However, one possibility is that the rate at which responding decreases during extinction for each VI schedule will be the same (i.e. position does not matter). Another option is that responding during the VI scheduled furthest from reinforcement changed the slowest during extinction. Momentum Theory assumes that the ratio of resistance to change depends on the ratio of reinforcer rates obtained in the presence of the component stimuli. This prediction is tested in the present study.

C. Phillip Powell, Nagendra Tripathi, & William L. Palya
Jacksonville State University

Linear Analysis of Pulse Schedules with Concurrent VI

Linear analysis was used to study behavioral dynamics. Reinforcement rate changes and the organism's response rate changes were used as the input and the output, respectively. After converting each to the frequency domain, the output was divided by the input, producing a transfer function. Patterns of reinforcement were converted to the frequency domain and then multiplied by the transfer function to predict the behavioral output. 15 pigeons were first exposed to a single pulse trial procedure for 105 sessions. This was followed by 90 sessions of a phase which implemented a VI 240-s schedule concurrent to the pulse trial in order to increase the signal-to-noise ratio of the behavior controlled by the pulse. The third phase implemented a two-pulse procedure with a concurrent VI 240-s schedule for 115 sessions. The last phase was 90 sessions of only the standard two-pulse procedure which was used to determine the transfer function. The concurrent schedule abolished a rise in responding observed in the latter portion of the extinction period of the pulse trials. Predictions made from the concurrent schedules were superior, indicating that the concurrent schedules may lessen or eliminate any error attributable to changes in the pattern of reinforcement between the determination and test phase.

Carlos Santoyo V.
Universidad Nacional Autonoma de Mexico

A quantitative analysis of equity exchanges

This paper deals with the quantitative "Aristotelian" formulation of equity: Oi/Oi+Oj = Ii/Ii+Ij, where Oi and Oj being outcomes (points obtained in the exchange) and Ii and Ij being the inputs (responses in a variable ratio [VR] schedule of reinforcement) for subjects i and j, respectively. Elementary school Mexican children were exposed to a computer game situation where, in an alternate trials, subjects allocate to a partner both: the task which must be done by the other (a game associated to a VR 20, 40 or 80) and the points which the partner must obtain in the game situation. The formulation of equity describes properly the data. Some different configuration of strategies are exposed and discussed.

Nestor Schmajuk & Horatiu Voicu
Duke University

Three-Dimensional Cognitive Mapping With A Neural Network

We apply a model for spatial navigation that includes an Action System capable of guiding the search for specific goals with the assistance of a Cognitive Map. A three-dimensional unexplored environment is represented in the Cognitive Map as a set of linked, unvisited places. Different types of terrain are also represented in the Cognitive Map. During exploration, these unvisited places constitute the goals the Action System should reach. The system efficiently and completely explores the environment. Once the environment is mapped, the Action System can plan the best route to navigate between places.

Matthew T. Sitomer & Robert W. Allan
Lafayette College

Stimulus-reinforcer and response-reinforcer effects on continuous clock-supported behavior

Previous work has shown that pigeons' pecking behavior may be auto-shaped and auto-maintained when a continuous clock is added to fixed- or variable-time schedules. The rates of responding supported by these schedules are often very high thereby resulting in many non-contingent but coincident peck- reinforcer sequences. In an attempt to examine the effect of these temporal coincidences a fixed-interval (operant) response- reinforcer requirement was added to ongoing fixed-time + added- clock schedule. Results suggests that there are separable stimulus- reinforcer and response-reinforcer schedule effects. A potential model of these time and contingency-based effects is also presented.

Paul Soto & Jack J McDowell
Emory University

The effect of an alternative reinforcement source on Herrnstein's single alternative equation.

Herrnstein's (1970) account of the relationship between reinforcement rate and response rate on variable-interval (VI) schedules predicts that the parameter re will vary directly with the rate of reinforcement delivered from a secondary source and that the parameter k will remain constant as secondary reinforcement rate is varied. The present experiment constituted a parametric test of this prediction. The experiment consisted of 7 conditions during which, ten rats lever-pressed for food on concurrent VI schedules. During each condition, the schedules on the left lever varied between 10 values (6 sec, 10 sec, 14 sec, 20 sec, 45 sec, 55 sec, 100 sec, 200 sec, 350 sec, and 450 sec) and the VI value on the right lever remained constant. The right-lever VI value varied across conditions (10 sec, 17 sec, 50 sec, 75 sec, 150 sec, 350 sec, and EXT). Data were analyzed using two methods of calculating response rate. The first method of calculating response rate used a time base of the component length minus time spent in the reinforcer blackout. The median k decreased with increasing secondary reinforcement rate and re remained approximately constant. The second method of calculating response rate used a time base of the component length minus both the post reinforcement pause time and reinforcer blackout time. Fitting Herrnstein's equation to these response rate data produced a decreasing median k and a linearly increasing re with a slope less than 1. The results of both methods are at odds with the predictions of Herrnstein's accounts. It is suggested that the effects of bias and sensitivity might be the cause of the results.

SQAB-Invited Preeminent Tutorials

Gordon Burghardt
University of Tennessee

Chair: William Timberlake

Evolution, Behavioral Variation and Plasticity

There is great behavioral diversity both among and within species in natural populations. Much of this diversity has been ignored in most traditional experimental and animal psychology beyond acknowledging the obvious constraints due to sensory and response abilities. However, the variation seen is not 'noise' and is often remarkably consistent over time. The application of behavioral principles, in spite of their often great power, ultimately needs to include detailed knowledge of the biology and evolution of the organisms under study. In this tutorial the basics of measuring behavioral variation will be outlined including the use of phylogenies, the information obtained from molecular genetics, the study of modes of natural selection (e.g., directional, stabilizing, disruptive), and quantitative approaches to isolating and measuring population level parameters from individual animals and their parents.

Peter Balsam
Columbia University

Chair: William Timberlake

Pavlovian Conditioning

Pavlovian conditioning plays an essential role in many behaviors including feeding, reproduction, drug addiction and defense. An understanding of this learning process requires analyses of the necessary conditions for learning, the contents of that learning, and how learning is expressed in performance. The lecture will serve as an introduction to research and theory related to these areas of study.

Jennifer Higa
Texas Christian University

Chair: Peter Killeen


Most species of animals - from mammals and birds to some fish - show a learned behavioral sensitivity to the timed-occurrence of events. This ability to detect, integrate, and use temporal information (to "time") is a basic and essential process of behavior and learning: It plays a fundamental role in what is learned and how learning progresses. The goals of the tutorial are: 1) to present a survey of the methods for studying timing and the data generated by these procedures; 2) to review the dominant theories of interval timing; and 3) to describe a core set of issues that the data and theories raise about the timing mechanism. Some of the problems for timing that I plan to focus on include whether or not there is a "clock," time-window effects, the role of memory, asymmetries in timing short and long intervals, and the effects of interrupting the to-be-timed interval.

Marc Branch
University of Florida

Chair: Stephen Dworkin

Behavioral pharmacology is a discipline that emphasizes the interrelationships and interactions between behavioral variables and the actions of drugs. The tutorial will first outline the history and origins of behavioral pharmacology, and also the logic behind the field. Examples of how behavioral variables can modify drug effects, and how such phenomena can be studied systematically, will be presented. Finally, illustrations of how examination of drug effects can illuminate behavioral processes will be provided.

SQAB main page

Date Updated : April 25, 2000