Units of Which Behavior Is Composed. Knowl

JOURNAL OF THE EXPERIMENTAL ANALYSIS OF BEHAVIOR 1986, 469 219-242 NUMBER 2 (SEPrEMBER) UNITS OF ANALYSIS AND KINETIC STRUCTURE OF BEHA VIORAL REPERTOIRES TRAVIS THOMPSON AND DAVID LUBINSKI UNIVERSITY OF MINNESOTA

It is suggested that molar streams of behavior are constructed of various arrangements of three elementary constituents (elicited, evoked, and emitted response classes). An eight-cell taxonomy is elaborated as a framework for analyzing and synthesizing complex behavioral repertoires based on these functional units. It is proposed that the local force binding functional units into a smoothly articulated kinetic sequence arises from temporally arranged relative response probability relationships. Behavioral integration is thought to reflect the joint influence of the organism's hierarchy of relative response probabilities, fluctuating biological states, and the arrangement of environmental and behavioral events in time. Key words: behavioral units, units of analysis, response class, organization of behavior, behavioral taxonomy, relative response probability

PROTASIS edge of such units is not axiomatic but as in An impala gamboling across an East Afri- the case of our knowledge of other biological can Serengeti plain or a 16-month-old child units, must be arrived at empirically. Whether confronted with a novel toy presents seamless fundamental behavior units can be discovered streams of kinetic motion, free of beginnings depends largely on the criteria used for as- or endings. The scant frequency with which sessing and selecting behavior classes. Given behavioral scientists have concerned them- criteria, establishing which units satisfy those selves with identifying the fundamental units criteria is an empirical matter. Approximately of behavioral phenomena is, in part, due to three quarters of a century have been devoted the conspicuous integrity of such naturally oc- to research dealing with a range of behavioral curring episodes; indeed, at times they seem units without explicit criteria for assessing to defy scientific analysis. Of the unresolved their scientific significance. One of the pur- issues facing behavior scientists, none is more poses of this paper is to discuss the criterion central than the analysis and synthesis of the ofscientific significance as applied to the selec- fundamental units of behavior. Analyzing tion of behavioral units. Based on the stated continuously flowing behavior into basic com- criterion and available evidence, three types of ponents and then resynthesizing it is a major functional units emerge from our analysis: aim of behavioral science. This goal is predi- emitted, elicited, and evoked response classes. cated on an understanding of the fundamental The combination of these fundamental units units of which behavior is composed. Knowl- to form larger behavioral aggregates (i.e., functional behavioral composites made up of two or more fundamental units that covary Appreciation is expressed to Robert Hinde, whose collectively) is a secondary concern of this pa- helpful remarks improved the first draft of this paper, and per. It is proposed that once combined, some to John Falk, M. Jackson Marr, Paul Meehl, Richard combinations of fundamental units can no Meisch, J. Bruce Overmier, Roger Schnaitter, Murray longer be analyzed into those constituents. We Sidman, and Michael Zeiler, whose comments sizably improved the manuscript. We especially wish to express our will outline a behavioral taxonomy for ana- appreciation to our friend and teacher, Kenneth Mac- lyzing extended behavioral samples into their Corquodale. We hope the ideas expressed in this paper constituent components. are in keeping with the intellectual tradition he taught so Finally, the mechanisms involved in behav- effectively at the University of Minnesota for 37 years. Reprint requests may be sent to Travis Thompson, ioral syntax (i.e., the process by which simpler Department of Psychology, Elliott Hall, 75 E. River Road, behavioral components are integrated to form University of Minnesota, Minneapolis, Minnesota 55455. highly organized, complex behavioral reper- 219 220 TRA VIS THOMPSON and DA VID LUBINSKI toires) will be explicated. Our treatment of classification of fundamental units, behavioral these issues will include a discussion of the science has largely disregarded these issues. It apparent uniqueness of naturally occurring was not until the turn of the 20th century that behavioral episodes; some suggestions for the foundations of our current concepts of be- finding order in these phenomena are offered. havior units were laid, growing out of three traditions. Thorndike (1911) observed that behavior could be acquired by instrumental SELECTING BEHAVIORAL UNITS conditioning, and he isolated components of CRITERION OF SCIENTIFIC SIGNIFICANCE behavior by arranging some consequences of classes of movements that would control their Behavior, like any other physical phenom- future recurrence. Pavlov (1927) showed that enon, cannot be studied scientifically in its en- behavior could be conditioned by pairing a tirety. Portions of the continuous stream of an neutral stimulus with an effective eliciting organism's activity sharing certain common stimulus. His behavior units were specified as features must be abstracted for study from the whatever muscle contractions or glandular se- whole. These individual behavior instances are cretions occurred immediately following pre- important to the extent they can be shown to sentation of the eliciting stimulus. The third be members of replicable behavior classes. Al- tradition originated independently with Whit- though members of such classes are necessar- man (1899) and Heinroth (1911), who sug- ily something less than behavior in its original gested that large portions of behavior were of form, they are the basis of all scientific asser- fixed forms, elicited by complex patterns of tions concerning behavior. Clearly, the avail- stimuli, and were inherited rather than con- ability of significant behavior classes is of ditioned. The segments of behavior in the lat- prime importance, for it determines the kind ter case were less well defined, since they in- and number of lawful relations that can be volved not only simple muscle contractions, but established. It follows that the way continuous also complex patterns of coordinated move- behavioral repertoires are divided into fun- ments. Nonetheless, regular families of be- damental units is critically important in de- havior patterns could be isolated as elicited termining the success of the entire scientific by specific stimulus configurations. study of behavior. The selection of behavior The emergence of these three schools was units cannot be arbitrary, but must be based important in the development of our current on consistent criteria that "carve nature at its notions of behavior classification and units. All joints." three groups succeeded in abstracting and A common criterion applied in evaluating simplifying behavior, from a scientifically in- any unit is its scientific significance: the extent comprehensible flux to relatively simple and to which the unit enters into a wide range and observable segments, based on the conditions number of lawful relations (Sidman, 1960; controlling those behavior segments. In Spence, 1948). Theoretically rich but empir- Thorndike's case, the consequences of behav- ically unreliable measures fail to satisfy this ior that controlled recurrence of members of criterion, as do highly reliable measures with a class were the basis for identifying response extremely narrow conceptual implications. To units. Pavlov's behavioral unit was specified qualify as scientifically significant, a unit must by identifying the eliciting stimuli. Whitman reliably relate to an array of phenomena. This and Heinroth specified their units as inherited status cannot be established by argumenta- effector patterns covarying with species-spe- tion, but must be shown empirically. cific stimuli. EMERGENCE OF FUNCTIONAL RESPONSE CLASSES FUNCTIONAL BEHAVIORAL CLASSES AND Historically, behavior scientists have paid NATURAL BEHAVIOR little systematic attention to the search for sci- Behavioral units can be grouped together if entifically significant units of analysis. Indeed, they are controlled by the same variables, Loevinger (1957, p. 644) has asserted, "there characterized as functional (Skinner, 1938) or are no natural units for the study of behav- causal (Hinde, 1966) classifications of behavior." Whereas nearly every other area of bi- ior. This method of describing behavior cou- ology has placed a premium on isolation and ples the classification with its niethod of anal- KINETIC STRUCTURE 221 ysis. The success of isolating functional and specific nature. Behavior can be divided behavior units experimentally depends on the into three functionally different response diversity of the controlling conditions. If there classes according to its major controlling con- were as many uncorrelated controlling conditions: elicited behavior, emitted (or operant) ditions as there are behavior instances, we behavior, and a derivative of the former two would be at no scientific advantage. However, classes, evoked behavior. Elicited behavior is the conditions controlling behavior are distrib- identified as behavior whose probability is de- uted into several disjunctive categories. Inso- termined by antecedent eliciting stimuli, and far as each of these categories is relatively ho- conditioned via antecedent stimulus pairings; mogeneous and the number of categories is emitted behavior is defined as constituent limited, there is some possibility of studying movements of an organism, the probability of behavior systematically in less than its en- which varies as a function of a stimulus con- tirety. sequence of previous instances of members of A common complaint about functional be- that class; and evoked behavior is induced by havior classes, especially operant responses, is the temporal arrangement of current stimulus that they do not seem natural, whereas seg- events. ments of behavior selected from the behavior The defining features of these three behav- of free-moving animals outside the laboratory ior classes are inherent in the variables con- are viewed as being more natural. No class of trolling them (e.g., US, CS, SR+, Sr-, etc., and events examined experimentally is ever "nat- their temporal presentations). The more di- ural" in the sense that it is identical with dated verse these variables, the more difficult it is to historical occurrences under uncontrolled cir- specify behavioral classes. Specific responses cumstances. This criticism applies to all units within these three functional classes may be isolated by biologists inasmuch as these units further subclassified by identifying the specific are necessarily abstractions possessing certain functions of the controlling variables: (a) an- common properties but otherwise differing in tecedent conditions, (b) current stimulus cir- many respects. Such biological units as pro- cumstances, and (c) response consequences tein molecules, neurons, ganglia, or the spinal (Thompson & Schuster, 1968). cord are not natural entities existing in isolation as classes of matter. Instead, they share Elicited and Emitted Behavior common properties but differ widely among Distinctions have commonly been made be- themselves. These classes of matter are clas- tween unconditioned and conditioned elicited sified by observers in much the same sense behavior. Elicited behavior has also been sub- that functionally defined response units are so classified on the basis of the motivational sig- classified. Functional behavior units are in a nificance of controlling stimuli-that is, of ap- very real sense more natural than experi- petitive and aversive stimuli (Mackintosh, menter-selected behavior segments, for seg- 1974). Unconditioned behavior has often in- ments of behavior naturally existing in an or- volved complex response sequences and to- ganism's repertoire are enabled to emerge pographies, elicited by species-specific stimuli. rather than being imposed by the observer. Mono- and polysynaptic reflexes have been Behavioral units cannot be imposed upon an distinguished from fixed action patterns, the organism by manipulating a variable that nat- former invariate in form and physical prop- urally controls that behavior. erties, the latter more variable, depending on the controlling stimulus and state of the or- THREE FUNCTIONAL BEHAVIOR CLASSES ganism. Presenting an eliciting stimulus for a The fundamental units of behavior are all fixed action pattern produces a topographi- response classes. The response class is an in- cally complex response sequence-a stochastic herently interactive concept, defined as a dis- series of conditional response probabilities. position to exhibit responses of a given func- Typically, the first component response of the tional type when appropriate stimulus events series is highly predictable and circumscribed, are provided (Hocutt, 1967; MacCorquodale but as the fixed action pattern unfolds, the & Meehl, 1954, pp. 218-231; Skinner, 1938). terminal components are less predictable The term encompasses a heterogeneous col- (Tinbergen, 1948). Some investigators have lection of entities that display both inter- and suggested referring to FAPs as "modal action intra-species variations in strength, molarity, patterns" (MAPs) to reflect this variability 222 TRA VIS THOMPSON and DAVID LUBINSKI

(Barlow, 1968). In addition, attempts at clas- the availability of another class of reinforcing sically conditioning fixed action patterns (e.g., events." Zeiler (1977) proposed a distinction Thompson & Sturm, 1965) have produced between direct and indirect variables deter- new reflexes in which only a subset of the mining effects of reinforcement schedules. Di- components of the fixed action patterns are rect variables arise from formal properties of elicited. Although reflexes and fixed action schedules (i.e., the conditions that must be sat- patterns differ in complexity, they nonetheless isfied for the presentation or withdrawal of a are primarily controlled by an antecedent elic- stimulus event). "Indirect variables are those iting stimulus, and are treated here as being that are imposed without being explicitly pre- members of the same functional class (i.e., scribed by the schedule" (page 204). Direct elicited). variables engender stereotypic performance, Although in respondent conditioning no new whereas indirect variables tend to produce responses are ever conditioned (only new re- regularly fluctuating performance. Perfor- flexes), the units within operant repertoires mances maintained by ratio schedules are pri- may combine to form seemingly new response marily controlled by direct variables, whereas topographies. Elements of coordinated move- performances controlled by interval schedules ments existing prior to operant conditioning reflect the products of both direct and indirect are modified by their consequences and emerge variables. to form functional components of new oper- The causal significance of indirect variables ants. Such elements may, initially, be parts of determining the temporal structure of behav- respondents, fixed action patterns, or compo- ior can be profound, with adjunctive behavior nents of other operants unified by their dis- providing the most significant examples. Falk tinct environmental consequences. As elicited (1961) describing excessive drinking by water- behavior may be subclassified according to the satiated rats exposed to a 60-s variable-inter- nature of the controlling stimuli, emitted be- val food-reinforcement schedule, called the havior is similarly categorized. For example, phenomenon "schedule-induced polydipsia." operant behavior patterns are described as es- Rats drank as much as 10 times their normal cape or avoidance responses, depending on the daily water intake under these schedule-gen- nature of the controlling negative reinforcer. erated conditions. Originally believed limited Operants are negatively reinforced by removal to a unique food-water relation in rats, sim- of a primary negative reinforcer in the former ilar phenomena occur with other responses case, and removing a conditioned negative such as eating, fighting, escape, running, and reinforcer in the latter. across several species such as mice, pigeons, monkeys, and chimpanzees (see Falk, 1971, Adjunctive and Other Evoked Behavior 1977; Staddon, 1977). Jean-Paul Sarte (1938) wrote, "I think they do that to pass time, nothing more. But time Generator Schedules is too large, it refuses to let itself be filled up." Schedules capable of evoking adjunctive be- The terms "displacement activity," "interim havior are labeled generator schedules, and their behavior," and "schedule-induced behavior" formal properties may be defined by either have been used to refer to members of this response-contingent or noncontingent stimu- class. We view adjunctive behavior as a higher lus presentations, that is, variable-interval order behavior class because it emerges when (VI), fixed-interval (FI), fixed-time (FT), and stimulus conditions regulating the probabili- variable-time (VT) schedules. The constitu- ties of fundamental ("first-order") units are ents of the behavior patterns evoked by sched- temporally structured in a particular fashion. ule induction must previously exist at some The probabilities of these response classes are strength, but in the absence of the generator controlled by interstimulus interval parame- schedule exhibit low probability (e.g., drink- ters, not by the specific nature of the scheduled ing by a water-satiated animal). The indirect environmental events per se. effects of generator schedules arise from their Falk (1977, p. 325) has defined adjunctive increasing the probability of otherwise low- behavior as "behavior that is maintained at a probability behavioral units with no contin- high probability by stimuli which derive their gent or direct eliciting relations to the behav- exaggerated reinforcing efficacy primarily as ior in question. a function of schedule parameters governing Generator schedules are characterized by KINETIC STRUCTURE 223 two parameters, their effective range and their 1.0f optimal value. Simple stimulus-presentation schedules capable of functioning as generator schedules exhibit an effective range (i.e., a range of values capable of evoking adjunctive behavior). Within the effective range there is .8 an optimal value for engendering adjunctive behavior. The optimal value represents the vertex of a biotonic inverted U-shaped function; and, as the value of the generator sched- 4-. .6 ule deviates from this value (in either direc- ._ tion), the probability of a given adjunctive .0co behavior pattern diminishes. Although the 0 quantitative characteristics of this function L. may vary, the optimum interval for simple .4 fixed and variable schedules for several species (pigeon, rat, monkey) is 1 to 3 min. Adjunc- tive behavior in humans has received less attention; however, some aspects of human per- .2 formance seem susceptible to schedule induction (Cantor, Smith, & Bryan, 1982; Clarke, Gannon, Hughes, Keogh, Singer, & Wallace, 1977; Falk, 1986; Fallon, Allen, & Butler, 1979; Frederiksen & Peterson, 1974; 10 20 30 Kachanoff, Leveille, McLelland, & Wayner, 1973; Lasiter, 1979; Porter, Brown, & Gold- Post-food time (sec) smith, 1982; Wallace & Oei, 1981; Wallace Fig. 1. Distributions of relative response probabilities & Singer, 1976). For example, generator generated by a rat in an experimental chamber that per- schedules have enhanced the probability of mitted drinking, wheel running, and food delivery. The certain aspects of cigarette smoking (Cherek, gradients were temporally structured by an FT 30-s gen- 1982, Wallace & Singer, 1976) and "motor erator schedule of food reinforcement (from Staddon & excitability" (Muller, Crow, & Cheney, 1979; Ayres, 1975). Wallace, Singer, Wayner, & Cook, 1975). Generator schedules have highly predict- integrate evoked responses with the response able effects on the temporal distributions of patterns controlled by the formally specified the evoked responses. Staddon and Ayres direct variables. Generator schedules do not (1975; see Figure 1) studied the time rats de- evoke novel responses; rather, they shift the voted to eating, drinking, running, and hov- temporal patterns of relative response proba- ering near a food area when food was pre- bilities correlated with the stimulus presen- sented at regular fixed intervals. They, like tations regulated by the generator schedule. Falk, found that the probability of drinking increased sharply after food-pellet ingestion. Other Emergent Behavior Then drinking probability declined and run- Falk (1977) has argued that the evolution- ning increased; finally, as the probability of ary utility of adjunctive behavior lies in its running declined, hovering near the feeder maintaining an organism in a problematic but area increased. The regularity of these tem- overall favorable situation. However, other poral modulations of relative response prob- schedule-induced emergent behavior patterns abilities of adjunctive responses as a function can be traced to generator schedules regulated of a concurrent food schedule is striking (see by unconditioned aversive stimuli, and are not Figure 1). obviously "favorable" in an evolutionary sense. This property of generator schedules high- For example, Hutchinson, Renfrew, and lights their capacity to structure the temporal Young (1971) studied patterns of biting and distributions of relative response probabilities. lever manipulation by squirrel monkeys ex- In addition to enhancing certain response posed to unavoidable tail shock. Immediately probabilities, generator schedules temporally following shock, the probability of biting in- 224 TRA VIS THOMPSON and DAVID LUBINSKI creased sharply and then declined. As the time According to this formulation, elicited and for the next shock approached, the probability evoked behavior patterns are viewed as ends of manual responses progressively increased to of a single continuum of induction, that is, the time the unavoidable shock was presented. "action produced without proximal contact" That is, the temporal distribution of the two (Segal, 1972, p. 12). This dimension is an- recorded responses (biting and lever manipu- chored by phasic induction (i.e., critically lation) came under the control of the fixed damped and highly tuned systems-that is, schedule of unavoidable shock. Similarly, Kel- simple reflexes) and periodic induction (i.e., leher and Morse (1968) demonstrated that moderately damped and moderately tuned under the right schedule conditions, electric- systems-that is, adjunctive behavior). If the shock presentation can come to function as a concepts of tuning and damping function uni- maintaining event. Using squirrel monkeys as formly along a continuum, an argument can subjects, a noncontingent FI 5-min shock was be made for only two behavior classes. How- superimposed onto an established VI 2-min ever, if a discontinuity exists between phasic food baseline. At first, this manipulation sup- and periodic induction, then consideration pressed response rate as the end of the 5-min should be given to a third response class. interval drew near. After a number of ses- Although Hineline's account of evoked be- sions, the temporal distributions of the mon- havior is appealing, two distinctions between keys' responses began to change. The pattern inorganic and organic systems pose potential of responding became positively accelerated difficulties in applying an inorganic resonance (with the highest response rate preceding the model to behavioral phenomena. In inorganic shock). At this point, the VI food schedule was systems, when periodic stimulus inputs exceed discontinued, leaving response-produced FI the system's range of tuning, the system re- shock as the only consequence for responding. turns to a resting state. However, when the The monkeys continued to display positively parameters of a generator schedule exceed the accelerated responding maintained only by critical range, direct schedule effects continue shock presentations. Furthermore, these shock- to control the probability of the baseline be- maintained performances did not extinguish havior and only the indirect adjunctively in- over time. Shock-maintained performances duced behavior drops out. Moreover, in in- have been widely replicated and constitute a organic systems, periodic stimuli typically robust set of phenomena (see Branch & engage the same type of motion individually Dworkin, 1981; Hutchinson & Emley, 1977; or phasically when their intensity is increased. Malagodi, Gardner, Ward, & Magyar, 1981; In behavioral systems, generator schedules Morse & Kelleher, 1966). These findings sug- evoke different types of responses, the specific gest adjunctive behavior belongs to a broader forms of adjunctive behavior being determined class of evoked emergent behavior, which is largely by the available stimuli. Indeed, gen- produced by indirect schedule variables. They erator schedules calibrated to an optimum are otherwise low-probability behavior pat- value may not evoke any adjunctive behavior terns that are slow to develop and are schedule if the ambient stimuli within the organism's induced. purview are not germane to the dispositions Hineline (1981, 1986) proposed extending the organism brings to the stimulus situation. the concept of resonance from physics to describing phenomena we have termed evoked Summary behavior. Resonance, a dispositional charac- The foregoing three behavior classes reflect teristic of inorganic systems, refers to their re- our current knowledge regarding the funda- sponsiveness to sinusoidal or pulsed inputs. mental units of behavior. These behavioral Resonant systems have two important prop- units are defined by their relational, not their erties: (1) their degree of tuning-that is, the intrinsic, properties. The intrinsic (topo- extent to which a system responds to periodic graphical) properties of several responses may inputs that differ from the system's resonant cut across all three behavioral classes, yet their frequency (resonant frequency corresponds to functional class membership may vary. Nearly what has been called optimal value here) all of the laws or principles forming the foun- and (2) damping-that is, how quickly a sys- dation for our current knowledge base vis-a- tem returns to rest when periodic inputs are vis the behavior of infrahuman organisms (e.g., discontinued. acquisition, extinction, discrimination, gener- KINETIC STRUCTURE 225 alization, habituation, schedule control) are of behavioral components is a given, as is the based on these three fundamental units (Ca- interrelatedness of constituent cells of the co- tania, 1984; Ferster & Skinner, 1957; Hinde, lonial flagellate, Volvox, or the component be- 1970; Honig & Staddon, 1977; Locurto, Ter- havior patterns prerequisite to emergent ver- race, & Gibbon, 1981; Mackintosh, 1974; balizations in young children. One must Rachlin, 1976; Rescorla, 1980; Staddon, 1980; understand how each cell of Volvox functions Thompson & Zeiler, 1986). The conceptual in relation to its companions to grasp how this implications of these behavior classes are colonial animal navigates through its aqueous deeply rooted in 20th-century behavior the- medium, and one must understand the com- ory. Collectively, they meet the criteria of sci- ponent behavior patterns developed by a young entific significance and function as the rudi- child to understand how transitive verbal re- mentary components for synthesizing more sponses emerge without being specifically molar streams of behavior. What kinds of taught (Sidman, Rauzin, Lazar, Cunning- combinations of basic components can be con- ham, Tailby, & Carrigan, 1982; Sidman & structed, and in what arrangements can these Tailby, 1982). That an organism's constituent components combine? responses are often inextricably intertwined is cause for perspicuous curiosity, not analytic nihilism. THE COMBINATION AND We assume that much of the behavior of all ARRANGEMENT OF FUNCTIONAL but the simplest or the youngest organisms is BEHAVIOR UNITS constructed of constituents that are tena- Virtually no behavior of an adult organism ciously combined, much as protein molecules is a pure instance of a single functional unit; are composed of individual elements that can- a behavioral instance is nearly always com- not readily be recovered, once denatured. posed of multiple individual units (Hinde, However, our later discussion will show that 1966). Indeed, some early formulations of me- some behavior can be analyzed into constitu- diation theory (Mowrer, 1947; Spence, 1956) ent parts and resynthesized again, and that were based upon the presumed embedded complex behavior patterns are, at times, sub- Pavlovian relationship contained in instru- ject to partial dissolution into components. An mental learning (Rescorla & Solomon, 1967). understanding of the basic units and the ways This consideration is also central in more con- in which they combine is important to an un- temporary "two-process" behavioral accounts derstanding of complex behavior. (Overmier & Lawry, 1979). Nonetheless, it is useful to identify the minimal units (emitted, Functional Behavior Combinations elicited, and evoked) of which such behavioral Functional behavior combinations are com- compounds are composed, in the same sense posed of two or more units that collectively that it is useful to show that water is com- covary to form an aggregate response class. posed of oxygen and hydrogen atoms. A given The constituents of these combinations may segment of behavior may be constructed of be of the same type (e.g., elicited or emitted parts that are indistinguishable in the concate- responses) or of different type (e.g., elicited nated segment but that were essential in the and emitted). Combinations of the same type origin of that segment. can be called homogeneous combinations, and Some have argued that distinctions among those composed of two or more types of units behavior units are of doubtful value. Having can be called heterogeneous combinations. gotten the "learned-instinctive" arguments of Homogeneous behavior combinations in- several decades ago well behind them (Hinde, volve two or more simple components of the 1968), some writers view distinctions among same functional type. The simplest case in- behavioral constituents as anachronistically volves a single response occurring recurrently, naive (Hinde & Stevenson-Hinde, 1973). Ac- such as a bird song occurring repeatedly in cording to this reasoning, a question such as the same form (e.g., Thorpe, 1961). Such bouts "Which components of an infant's approach of fixed action patterns can be contrasted with to its mother are operant and which are fixed two or more fixed action patterns of different action patterns?" must be viewed rhetorically, forms occurring as a unit. A single operant, because it is rarely possible to adequately tease such as key pecking, can be repeated several them apart. In our view, the interrelatedness times, leading to food reinforcement. The form 226 TRA VIS THOMPSON and DAVID LUBINSKI

stepping on a treadle can be conditioned with a stimulus light as reinforcer. The light elicits approach during imprinting, although these birds had not previously been imprinted. The 1. birds' approach to the light is a fixed action pattern elicited by the conspicuous stimulus (Bateson, 1966; Bateson & Reese, 1968). The resulting sequentially integrated operant and fixed action pattern comprises a new heterogeneous unit. Integrated sexual responses also involve combined operants and elicited responses (see Hinde, 1970; Masters & John- son, 1966). Human sexual behavior comprises a diverse collection of heterogeneous behavior combinations; however, although the emitted components of these combinations do vary considerably, the elicited components are rather stereotyped. Behavior Combinations in Humans 5. 6. Fig. 2. A six-component homogeneous behavioral Just as the power of a microscope must be combination composed of responses emitted by moderately adjusted as a function of the phenomenon un- retarded workers. These topographically disparate re- der study, so too does the level of behavior sponses are all members of a single functional class analysis need to be adjusted to the functional (Thompson & Grabowski, 1972). unit of behavior-environment interaction. To be specific, when order is not apparent at a molar level, a more molecular analysis may be of each response may vary but will be similar. necessary.... Conversely, if one fails to find Such a combination of discrete responses com- an immediate stimulus that controls a re- prises the simplest homogeneous operant com- sponse, perhaps the response is only an ele- bination. Homogeneous response sequences ment of a larger functional unit which is con- may also involve topographically dissimilar trolled by currently operating variables not to that element. which are under the control immediately attendant (Mor- operants, all of ris, Higgins, & Bickel, 1982, pp. 119-120) of a given consequence, such as industrial as- sembly operations (see Figure 2). Similarly, Hineline (1984) has remarked, Combinations of behavior units from dif- "In conjunction with the quantitative evalu- ferent types of response classes (heterogeneous ation, we need techniques for efficiently iden- combinations) have been studied less exten- tifying the most effective scales of analysis for sively than homogeneous combinations, and particular situations, thus establishing the are more difficult to interpret. Rats placed in complementarity of molar and molecular close proximity in an enclosed space will fight analyses" (p. 506). So far, our discussion of if presented with painful foot shock. Reflexive response classes (i.e., minimal units and be- fighting is reliably elicited by shock and can havioral combinations) has focused on dispo- be classically conditioned (Vernon & Ulrich, sitions typically studied by experimental psy- 1966). The fighting of a rat in an environment chologists and ethologists (e.g., Catania, 1984; in which shock has repeatedly elicited fighting Dewsbury, 1978; Drickamer & Vissen, 1982; will be partially a classically conditioned and Hinde, 1970; Honig & Staddon, 1977; Lo- partially an unconditioned fixed action pat- curto et al., 1981; Mackintosh, 1974). Traits tern elicited by the shock. The foregoing com- are broader response classes (Falk, 1956; bination of a fixed action pattern and a clas- Meehl, 1978, p. 808; Tellegen, 1981) consist- sically conditioned aggressive pattern is ing of emitted and/or elicited components. simpler and more stereotyped than the com- They are often referred to as "dispositional bination of fixed action patterns and operant clusters" or "response families," because the responses. Young chickens and ducklings class of behavioral components defining a par- KINETIC STRUCTURE 227

Table 1 An eight-cell taxonomy based on the functional response types (e.g., emitted and elicited) and temporal combinations of functional types. The examples in each cell are discussed in the text. Temporal Response type combinations Homogeneous Heterogeneous Sequential and parallel Conjoint schedules Conflict analysis Catania (1966) Lewin (1935) Compound Behavioral trees Courtship displays Findley (1962) Hinde (1970) Complex Interlocking scKedules Shock-avoidance adjusting schedules Reynolds (1975) Sidman (1962) Superordinate Higher order reinforcement schedules Betta splendens Kelleher (1966) Higher order: operants-CS-FAPs Thompson (1966) ticular trait often contains seemingly disparate under the control of group social conse- mega-topographical constituents (Lubinski & quences. Thompson, 1986; Meehl, 1986). Although traits are much less circumscribed TEMPORAL BEHAVIORAL COMBINATIONS than fixed action patterns, monosynaptic re- Sir Charles Sherrington (1906), in attempt- flexes, and the behavioral combinations men- ing to understand integration of spinal re- tioned above, they nonetheless manifest com- flexes, noted: "The outcome of the normal re- mon dispositional properties (Carnap, 1956; flex action ... is an orderly coadjustment and Pap, 1958; Sellers, 1958). That is, in a stim- sequence of reactions....The coordination in- ulus situation of a given kind, a response of a volves orderly coadjustment of a number of given type will occur (Tellegen, 1981, pp. 218- simpler reflexes occurring simultaneously (or 219). The extent to which traits are composed in) ... orderly succession" (p. 8). The tem- of emitted and elicited responses depends on poral arrangement of behavioral units can also the response class under consideration. The be useful in characterizing behavior combi- purest examples of homogeneous behavioral nations: (a) sequential or parallel, (b) com- combinations at this level of analysis appear pounds of sequential and parallel, (c) com- to be response classes typically called skills plex, and (d) superordinate arrangements (e.g., verbal, numerical, and spatial skills), (Thompson & Grabowski, 1972). Hence, be- dispositions composed exclusively of emitted havioral complexes can be classified into eight responses. disjunctive categories according to the combi- "Extraversion" refers to the tendency to nation and arrangement of their component display members of a heterogeneous response functional units (see Table 1). class composed of emitted and elicited components. The strength of clusters of social re- Sequential and Parallel Combinations sponses are controlled by classes of social con- In sequential behavioral combinations the sequences (i.e., they are emitted), whereas completion of certain preceding behavioral affective responses are conditioned through units is a necessary condition for occurrence associative pairing (i.e., they are elicited) and of the succeeding unit. Such behavioral se- typically have characteristic interoceptive dis- quences have been discussed in detail by Hinde criminative-stimulus properties. For example, and Stevenson (1969). Chain reflexes, such as the probability that access to social gatherings the chewing movements of decerebrate cats will serve as maintaining events may be esti- (Sherrington, 1906), are among the simplest mated by indices of extraversion. People scor- homogeneous sequential arrangements. A ing above the normative mean on "extraver- piece of meat dropped in the mouth elicits sion" on certain personality inventories are contraction of the masseter muscle lifting the likely to seek out social contact, whereas in- mandible. When proprioceptive feedback from dividuals scoring at the low end of this di- the contracted muscle reaches a specific level, mension will exhibit little emitted behavior the jaw falls open. On the same continuum, 228 TRA VIS THOMPSON and DAVID LUBINSKI but far more complex, are chains of fixed ac- mon stimulus (Hinde, 1954). Similarly, dogs tion patterns, such as the sexual behavior pat- prepared with parotid fistulas were trained to terns of Stickleback fish (Tinbergen, 1951). emit an operant response leading to the pre- Operant behavior frequently involves sequen- sentation of a food reinforcer under an Fl tial combinations (Catania, 1984; Fantino & 2-min schedule. The number of drops of sa- Logan, 1979; Ferster & Skinner, 1957; Pier- liva produced by the parotid gland was re- rel & Sherman, 1963). In chained reinforce- corded as well as operant bar presses. As the ment schedules, reponding in the presence of time when reinforcement was forthcoming ap- one discriminative stimulus with a given set proached, the number of drops of saliva in- of reinforcement contingencies (e.g., FR 25) creased progressively; the salivation curve produces a different discriminative stimulus. parallels the operant responding curve (Sha- Responding in the presence of the second dis- piro, 1961). The conditions controlling sali- criminative stimulus under specific reinforce- vation (elapsed time) seem to be the same as ment contingencies (e.g., FI 60 s) produces an those controlling operant responding, al- unconditioned, reinforcer (e.g., food). The to- though there does not appear to be any direct pographies comprising successive units in interaction between the two behavior patterns. chains may be similar (e.g., lever presses) or Most parallel combinations of two or more may be quite dissimilar (e.g., cutting a piece behavioral units involve some form of inter- of meat with a knife, followed by picking up action. Chaffinches frequently fly toward a the piece of meat with a fork). Multicompo- stuffed owl while producing a characteristic nent performances under chained schedules mobbing call, or they may suddenly turn and serve as entire behavior units, the probabili- fly away from the owl. Both responses have ties of which are manipulable as a whole simultaneously high probabilities. The ob- (Findley, 1962; Gollub, 1977; Kelleher & served behavior is often some combination or Gollub, 1962). Everyday activities frequently rapid alternation of the two (Hinde, 1966). involve chains of operants, each with its own Under concurrent reinforcement schedules integrity but under the control of a single con- (Ferster & Skinner, 1957) two operants consequence (e.g., rising and preparing to go to trolled by two discriminative stimuli and two a job, completing units of work in school, pre- reinforcement schedules are in effect simul- paring a meal). taneously. Instead of generating performance Although it is common to assume that the characteristic of either individual schedule, the organization of behavior in time is intrinsic to resulting performance often reflects an inter- neural circuitry (see Church, 1978; Fodor, action of the two. Key-peck performance of 1981), in most cases the sequential organiza- pigeons maintained by a concurrent FR 100 tion of human behavior seems to be imposed FI 5-min schedule of food reinforcement re- largely by external environmental events. The veals sequences of responses at intermediate key must be inserted in the lock of the car to high rates during the interval component, door before the door can be opened, and the alternating with long pauses. This behavior key must be inserted in the ignition before the pattern is more characteristic of ratio sched- engine will start, and so on. Nothing about ules (Catania, 1966). neural circuitry requires that such behavioral The clearest case of interaction in concur- sequences be organized in such a fashion. The rent operants involves behavior combinations functional organization of behavior by extrin- engendered by conjoint reinforcement sched- sic proximal controlling variables interacting ules. Two reinforcement contingencies are in with existing dispositions and biological state effect simultaneously, but only one response modulators provides an adequate account in operandum is provided. The resulting com- most instances. bined performance has features of two indi- In parallel behavior combinations, two or vidual performances (Catania, Deegan, & more units occur concurrently. A stuffed owl Cook, 1966). In the natural environment con- was presented to a chaffinch and elicited a joint contingencies are familiar. A worker's characteristic mobbing call ("chink") and high rate of productivity may be reinforced by concurrently elicited erection of the crest an employer (ratio schedule), whereas a mod- feathers. The two responses occurred inde- est rate of output is reinforced by his co-work- pendently but were often elicited by a com- ers (e.g., a pacing schedule). The resulting KINETIC STRUCTURE 229 integrated performance reflects the combined operant 2 contingencies. Interactions between parallel conditioned operant 1 WATER 1 and unconditioned behavioral arrangements FR 80 have received less attention. Two rats that operant 3 could avoid painful shock by lever pressing FR 80 reached very high levels of successful avoid- -J FOOD 1 ance when separated by a transparent parti- FR 80 tion. When the partition separating the ani- Fig. 3. A behavioral tree composed of a single operant mals was removed, avoidance responding FR 80 leading to an option of FR 80 reinforced by food deteriorated and was partially replaced by re- or FR 80 reinforced by water. Following the completion of either food or water reinforcement, the conditions of flexive fighting. Thus, the combined behavior operant 1 are reinstated, as indicated by the numerals included components regulated by avoidance following the reinforcement designations (after Findley, contingencies and by the eliciting stimulus for 1962, Figure 39). reflexive fighting (Ulrich, Stachnik, Brierton, & Mabry, 1966). Lewin's (1935) conflict analyses (i.e., approach-approach, approach- parallel response classes (i.e., compound be- avoidance, avoidance-avoidance, and double havioral units). Bernstein and Ebbesen (1978) approach-avoidance) can be analyzed in sim- have extended the analysis of multioperant ilar ways. The construct of "conflict" seems repertoires to include Premack's formulation to be employed when these consequences of a of reinforcement, and Emurian, Emurian, particular response both involve positive and Bigelow, & Brady (1976) have employed this negative reinforcement with equal momentary design to analyze cooperation in a three-per- effectiveness (approach-avoidance), or when son environment (see also Bernstein & Brady, the consequences both involve positive rein- in press). forcement (approach-approach), or in yet Symbolic communication between animals another case, both involve negative reinforce- can be analyzed in terms of a series of inter- ment/punishment (avoidance-avoidance), locking compound arrangements (Epstein, again with equal intensity. Lanza, & Skinner, 1980; Lubinski & MacCorquodale, 1984; Savage-Rumbaugh, Compound Combinations Rumbaugh, & Boysen, 1978). In these exper- Compound combinations involve sequential iments, organisms' exchanging of discrimina- and parallel functional behavior units, yield- tive stimuli was conditioned via arbitrary ing highly complex behavior. A single behav- matching tasks when only one of the organ- ioral unit composed of a single operant lead- isms had access to the stimuli correlated with ing to an option of two or more operants has reinforced responding (hence, the options en- been called a behavioral tree (Findley, 1962). countered in certain links of these arrange- Figure 3 shows a behavioral tree composed of ments were contingent on the behavior of 80 lever presses leading to an option of 80 another organism). For example, Savage- lever presses maintained by food or 80 lever Rumbaugh et al. conditioned chimpanzees' presses maintained by water. The tree func- supplying geometric symbols that were dis- tioned as a reinforceable unit, engendering criminative for another subject's food-rein- performance similar to that maintained by a forced responses. And Epstein et al., using two two-member chained schedule. Complex and adjoining experimental chambers separated by extended segments of behavior have been a Plexiglas divider, conditioned pigeons' in- brought under control by integrating behavior teracting in a similar fashion by pecking let- units of these types. One of the more elaborate tered response keys. Lubinski and Mac- demonstrations involved bringing 90 to 95% Corquodale demonstrated that pigeons can of a normal adult human's time in an exper- perform such communicative exchanges in the imental environment (24 hr per day for 5 absence of a primary establishing operation months) under control of scheduled contin- (e.g., deprivation, aversive stimulation, or un- gencies (Findley, 1966). Twenty-four differ- conditioned reinforcement). ent operant units were combined in serial and Probably the most clear-cut examples of parallel arrangements and groups of serial and heterogeneous compound arrangements are 230 TRA VIS THOMPSON and DAVID LUBINSKI courtship and sexual behavior (Hinde, 1970). reinforcement, but it is reduced linearly, reach- Sexual behavior includes several emitted and ing 1 after 10 minutes. (1957, p. 6) elicited components temporally linked both se- This is an instance of a decreasing interlocking quentially and in parallel. Organisms may schedule (i.e., as time passes, the number of initiate such compounds in several ways (e.g., responses required for reinforcement de- "display behavior," "posture," or "increasing creases). In increasing interlocking schedules, proximity"); upon the completion of the first as time passes, the number of responses re- component, additional options ensue until the quired for reinforcement increases. In regard culmination of the series. to increasing interlocking schedules, Reynolds (1975) commented: "This is the sort of insid- Complex Combinations ious schedule that exists in cumulative edu- Complex combinations are temporally joined cational systems in which, as time passes, the by the confluence of ratio and interval contin- requirements for success become larger and gencies. There are two main subtypes of stim- larger" (p. 93). ulus arrangements that combine behavioral An adjusting schedule is one units in this fashion: (a) arrangements in in which the value of the interval or ratio is which the value of the ratio and interval con- changed in some systematic way after rein- tingencies are fixed, independent of behavior, forcement as a function of the immediately pre- namely, conjunctive and alternative, and (b) ceding performance.... For example, a fixed arrangements in which the value of the ratio ratio is increased or decreased by a small or amount after each reinforcement, depending interval values vary as a function of pre- upon whether the time from the preceding re- vious behavior, namely, adjusting and inter- inforcement to the first response is less than or locking (see Ferster & Skinner, 1957; Thomp- greater than an arbitrary value. (Ferster & son & Grabowski, 1972). Skinner, 1957, pp. 6-7) A conjunctive schedule requires both an interval and a ratio to be satisfied before rein- Adjusting schedules are characteristic of "work forcement is forthcoming. Herrnstein and deadlines," when a fixed number of responses Morse (1958) employed a conjunctive FI 15- must be completed in a fixed period of time. min FR n to investigate the variables respon- Training programs and employment situa- sible for maintaining the number of responses tions frequently employ interlocking and ad- emitted in FI performance. They found that justing schedules, as well as alternative and a response requirement as few as 10 conjunc- conjunctive schedules. Such instances are ex- tive FI 15 FR 10 reduced responding signif- amples of homogeneous complex arrange- icantly. Conjunctive arrangements are also ments. (For further discussions of these ar- seen in certain work settings, where a fixed rangements, see: for conjunctive schedules amount of work must be completed in an 8- [Katz & Barrett, 1979]; for alternative sched- hr shift. ules [Harzem, Lowe, & Spencer, 1978; Rider, In alternative schedules, reinforcement is 1980]; for interlocking schedules [Berryman contingent upon completion of either an in- & Nevin, 1962; Rider, 1977]; and for adjust- terval or ratio requirement, whichever comes ing schedules [Kelleher, Fry, & Cook, 1964; first. Elementary school teachers implement Olvera & Hake, 1976; Sidman, 1962].) alternative schedules when they announce (after assigning 20 math problems, 30 min before Superordinate Combinations recess), "Those of you who complete the as- Behavior segments may be structured su- signment before the bell may begin recess perordinately-that is, one behavior class may early"-such an arrangement would be des- incorporate several subclasses into an entire ignated alternative FI 30-min FR 20. functional unit. A phenomenon not normally Ferster and Skinner defined interlocking used to generate superordinate behavior com- schedules as contingencies binations contains the rudiment of the basic in which the organism is reinforced upon com- relation. Pavlov (1927) conditioned a dog's pletion of a number of responses; but this num- salivation in the presence of a metronome by ber changes during the interval which follows repeatedly pairing the metronome with food the previous reinforcement. For example, the on the tongue. Subsequently, a black square number may be set at 300 immediately after was repeatedly presented followed by the met- KINETIC STRUCTURE 231 ronome. Pavlov found that the dog came to food delivery). The FI 1-min component of salivate when presented with the black square, this schedule generated typical FI 1 -min scal- a procedure called higher order classical con- lops and also evoked polydipsia; excessive ditioning. Although the robustness of this drinking preceded the brief flash in all inter- phenomonon has been questioned, there ap- vals (i.e., whether it occurred alone or in con- pear to be replicable cases of higher order junction with food delivery). classical conditioning (see Razran, 1955; Res- Studies on higher order multioperants have corla, 1980). provided the most compelling evidence that The functionally equivalent procedure ap- superordination of operant units is a powerful plied to operants also occurs far more fre- method for synthesizing complex behavior quently than was initially recognized (Kelle- (Thompson & Grabowski, 1972). Using a ba- her & Gollub, 1962). These higher order boon as the subject, Findley (1962) shaped a operant behavior units are specified by the behavioral tree (a single operant leading to an controlling contingencies, or higher order re- option of two operants) composed of FR 80 inforcement schedules. A behavior unit con- units. This behavioral unit was then rein- trolled by a single specified schedule contin- forced under other schedules. In one case, suc- gency (e.g., FI 4-min) that is in turn reinforced cessive trees had to be spaced 2 min apart if on another schedule contingency (e.g., FR 15) they were to be reinforced (a DRL 2-min is called a higher order schedule. In the pre- schedule). The emission of trees came under ceding example, the FI 4-min is treated as if the control of these scheduled contingencies, it were a single response reinforced according indicating that the tree was reinforceable as a to an FR 15 schedule (Kelleher, 1966). unit. Similarly, another tree was established Kelleher (1957) studied a chimpanzee's le- that was composed of FR 80 units, and then ver pulling maintained on a schedule such that occurrences of the tree produced food or water each individual FR 125 unit was followed by on an FR 4 schedule (i.e., four successive trees token presentation and five 125-response units composed of FR 80 s were required per rein- produced access to food. That is, a 125-re- forcer). Trees occurred in runs at a high con- sponse unit functioned as if it were a single stant rate, followed by long pauses, much as operant. In experiments with hospitalized one would expect with simple FR 4 schedules. psychiatric patients and with people who are Finally, a tree composed of FR 20 units was mentally retarded, as well as with nonhandi- reinforced under one set of stimulus condi- capped children in classrooms, tokens have tions on an FR 4 schedule, and under another been used in similar ways; however, each set of stimulus conditions on a DRL 2-min component of the first-order schedule typi- schedule. The following units were synthe- cally differs topographically from other com- sized and independently manipulable: (a) in- ponents-that is, typically the responses are dividual lever presses, (b) ratios of lever different adaptive educational or clinically presses, (c) trees of ratios of lever presses, (d) relevant performances (Ayllon & Azrin, 1965; ratios of trees, or DRL 2-min controlled trees, Kazdin, 1977; McConahey, Thompson, & of ratios of lever presses, (e) multiple schedule Zimmerman, 1977). Outside of laboratory controlled ratios or DRL 2-min controlled settings, the integration of large samples of trees of ratios of lever presses (see Figure 4). human behavior is often accomplished under In the foregoing case, as with many labo- basic schedules by token (money) reinforcers, ratory operants, the basic first-order response which in turn are exchangeable for other rein- unit was topographically uninteresting (i.e., forcers when enough tokens have been accu- the lever press). However, an operant is de- mulated (Skinner, 1953). fined as the behavioral segment the probabil- The conditioned reinforcers embedded in ity of recurrence of which is controlled by a superordinate arrangements are capable of particular reinforcing consequence. Operants evoking adjunctive behavior. Rosenblith are not restricted to lever presses, chain pulls, (1970) conditioned rats' lever pressing with or key pecks. More natural heterogeneous re- deliveries produced according to an FR 3 (Fl sponse combinations have been brought under 1-min) second-order schedule (i.e., the first superordinate operant control. The behavior response emitted after 1 min produced a brief of the Siamese fighting fish (Betta splendens) flash of light and the first response after three illustrates such superordinate control. such intervals produced the brief flash and Betta splendens exhibit complex uncondi- 232 TRA VIS THOMPSON and DAVID LUBINSKI

tations contingent on swimming, stimuli paired A,L4WLLL with mirror presentation elicit components of aggressive display through classical condition- FR ing (Adler & Hogan, 1963; Thompson, 1966; DRL Thompson & Sturm, 1965). If the entire behavior sequence of a male Betta splendens is observed under these circumstances, the fish Fig. 4. Cumulative records of chain pulls emitted by first emits the swimming operant, then begins a baboon on a higher order multiple schedule in which a to display aggressively to the classically con- tree composed of an FR 20 leading to an option of FR ditioned eliciting stimuli, and finally exhibits 20 food or water was reinforced on an FR 4 schedule a fixed action pattern elicited by the presence alternating with a DRL 2-min schedule. The event pin was down during the FR schedule and up during the of the mirror image itself. A portion of this DRL schedule. The schedule control exerted by the two sequence can be brought under superordinate contingencies is apparent from the spacing of successive control. The swimming operant and classi- trees (after Findley, 1962, Figure 43). cally conditioned display can be reinforced under a fixed-ratio schedule in which the mir- tioned aggressive displays when confronted ror is produced by several transits through the with its own image or another Betta splendens channel. On each transit, the subject exhibits (Fantino, Weigele, & Lancy, 1972; Lissman, classically conditioned aggressive display, and 1932; Simpson, 1968). A common naturally on the final transit all three components occur. occurring type of behavior such as swimming This heterogeneous behavior combination through a channel suspended in the aquarium closely resembles behavior of the Betta splen- can be strengthened if emission of that re- dens in a more natural setting (see Figure 5). sponse leads to presentation of the mirror im- It is possible to construct complex reper- age or a model of another Betta splendens toires involving serial arrangements and (Fantino et al., 1972; Thompson, 1963). The choices of activities involving not only operant mirror image elicits an unconditioned aggres- behavior or fixed action patterns, but hetero- sive display. After many such mirror presen- geneous combinations as well. A key integrat-

D(blue light) Swim r -D(yelw light) Swim r- D(green light) Swim r -R (red light & mirror) s -0------0-S R...... 1Rm~

CS-*CR CS-_CR CS US -CR *UR (aggressive (aggressive (aggressive display) display) display)

C >. classically classically unconditioned a) -conditioned conditioned rs . FAP a: t aggression_ r aggression_I \.

\ I-r-~~\wiswim I/

Time Fig. 5. A heterogeneous behavior combination exhibited by male Siamese fighting fish (Betta splendens) maintained and integrated by access to a mirror image that elicits an unconditioned aggressive display (UR). The sequence involves swimming through a channel in the presence of blue, yellow, then red lights (presented one following the others). The onset of the discriminative stimuli for succeeding components in the series (e.g., yellow light) served as a conditioned reinforcer, a discriminative stimulus, and also elicited dassically conditioned aggressive display. The lower curves indicated the relative probabilities of the swimming operant and the conditioned and unconditioned display that merged to form a smoothly coordinated response sequence (after Thompson, 1966). KINETIC STRUCTURE 233 ing variable combining behavior under super- classic observations of egg-retrieval by the ordinate control is the arrangement of grey-lag goose (Tinbergen, 1953) revealed that consequences for groups of units. The extent although the sequence typically occurred in a to which the kinds of units determine the de- smoothly integrated pattern, the retrieval re- gree and kind of control is unclear. sponse is composed of two independent components-the side-to-side orienting response Analysis and Synthesis of Complex Repertoires and the pulling toward the breast with the The proposed taxonomy is intended to pro- beak-controlled by distinctive variables. vide a conceptual framework for analyzing Similarly, the nest-building sequence of the complex systems of behavioral flux into mu- female canary is composed of independent be- tually exclusive and exhaustive categories. The havioral units, some under the control of evoc- rudimentary units of such temporal combi- ative stimulus antecedents, others maintained nations may be identified by manipulating the by stimulus consequences, and yet others reg- parameters of the component schedules regu- ulated largely by hormonal state variables lating the synthesized repertoire. The com- (Hinde, 1958). Some brain-damaged patients ponent schedules are used to tease apart ad- suffering from apraxia (the inability to per- jacent components. By varying the stimulus form skilled movements, typically resulting presentations that temporally link individual from a lesion on the sensorimotor cortex) dis- components, the constituent units may be de- pray fragmented behavior composed of nu- marcated within the concatenated repertoire. merous minimal behavioral units out of syn- The degree to which an analysis can be chrony. Individuals suffering from apraxia achieved varies with the extent to which the provide rare instances for observing rudimen- minimal components are integrated. Certain tary behavioral units of previously synthesized behavioral combinations, once synthesized, skilled performances. These fundamental units cannot be readily decomposed into their basic may appear aberrant, because the mecha- components again. Complex skills (such as nisms for maintaining their internal consis- riding a bicycle or playing a musical instru- tency are impaired. However, emergence of ment) are compound combinations consisting these individual units provides a means of cor- of several sequential and parallel components. roborating the hypothesis that molar behav- Once acquired, the possibility of experimen- ioral streams are aggregate entities composed tally breaking down such behavioral combi- of several temporally combined fundamental nations into fundamental units is question- units, even in highly overlearned skilled per- able. formances. As more highly integrated performances are Analyzing individual components of highly achieved, the control of certain facets of the integrated performances in the natural envi- individual's repertoire may shift from extero- ronment may be no more possible than resyn- ceptive or verbal discriminative stimuli to thesizing a denatured protein molecule. Un- more subtle interoceptive kinesthetic or pro- der such instances an interpretation is the only prioceptive stimulation. Because many of these feasible objective, rather than analysis controlling stimulus events are not accessible, (Schnaitter, 1978; Skinner, 1969). A behav- manipulation of individual components is often ioral interpretation involves inferring a plau- impossible. For example, it is often very dif- sible causal analysis from incomplete data on ficult for a skilled musician to teach a novice the system under consideration combined with beginning fingering technique, because the generalization from similar cases studied in minimal units no longer exist individually in isolation or partial combination. the experienced musician's repertoire. If such highly coordinated performances cannot be analyzed into minimal units, per- NEXUS: BEHAVIORAL haps behavioral integration is a unidirectional INTEGRATION process-that is, once integrated, may analy- Sherrington, in The Integration of the Ner- sis into minimal units be no longer possible? vous System (1906), wrote, "The main secret In fact, behavioral sequences that appear ir- of nervous co-ordination lies ... in the com- reversibly integrated can, at times, be shown pounding of reflexes" (p. 8). To solve this to be constructed of independent units. The problem, Sherrington proposed "the principle 234 TRA VIS THOMPSON and DA VID LUBINSKI of the common path" (p. 1 17), the notion that HIERARCHIES OF RELATIVE two or more independently elicited reflexes RESPONSE PROBABILITIES converge on a common efferent output. Sher- An organism's hierarchy of relative re- rington suggested that "since each instance of sponse probabilities represents the nature and convergence of two or more afferent neurones strength of the sum total of behavioral dispo- upon a third ... affords ... an opportunity sitions within its repertoire. The constituents for coalition or interference of their actions, of these hierarchies range from individual each structure at which it occurs is a mecha- fundamental units to functional behavioral nism for co-ordination" (p. 147). This paper combinations (i.e., concatenated behavioral has suggested a functional taxonomy of be- units combined collaterally and/or tempo- havioral units and has explored several types rally). The strength of a disposition refers to of combinations and arrangements of such the probability of a given functional unit when units. How fundamental units and their var- appropriate environmental events are pro- ious combinations become integrated remains vided. to be explicated. Although no structural ana- log provides a behavioral final common path, functional mechanisms by which behavioral Enduring Dispositions sequences and patterns become integrated can Enduring dispositions are dimensions of re- be identified. The most basic phenomenon in- sponse classes, as is true of other response volved in behavioral integration appears to be classes. Such a disposition is a persistent readi- temporally structured relative response proba- ness to exhibit responses of a given functional bility, a cardinal concept in modern behavior type when appropriate environmental events theory. are provided. These response classes are long- Skinner (1953) suggested that response term (often life-time) characteristics of an or- probability is an estimate of response strength ganism. For example, the probability that a and that response frequency is an estimate of healthy adult male Siamese fighting fish will response probability. The notion that relative exhibit gill-cover erection, fin spreading, and response probabilities (and by illation, re- tail beating when visually confronted with a sponse strength) are critical to reinforcement conspecific male approaches unity under a concepts is embedded in the fabric of current wide array of circumstances (Lissman, 1932; learning theory (see Premack, 1965, 1971; Simpson, 1968). This genetically determined Timberlake & Allison, 1974). In this final aggressive display is extremely high in the hi- peroration we shall argue that the temporal erarchy of response probabilities of Betta structure of relative response probabilities splendens. Eating is among the most probable binds functionally defined short-term behav- responses displayed by people suffering from ioral units into larger combinations. The tem- Prader-Willi Syndrome (Prader, Labhart, & poral structure of behavior of an individual Willi, 1956). This disorder, associated with organism is a function of three factors: relative an insatiable appetite and life-threatening response probability hierarchies, fluctuating obesity, is apparently regulated by a neuro- states, and the scheduled presentation of en- logical defect similar to hypothalamic hyper- vironmental events. Hierarchies of relative re- phagia in laboratory rats (Hetherington & sponse probability reflect the nature and Ranson, 1942; Teitelbaum, 1955). These bi- strength of all behavioral dispositions within ologically determined dispositions have en- the organism's repertoire. Fluctuating states duringly high probabilities within the behav- are organismic conditions enhancing or atten- ioral repertoire of the respective subjects. uating the strength of these response classes Weiner (1964, 1969) found exposure to (e.g., fatigue, illness), and the scheduled ar- fixed-ratio or DRL schedules had lasting ef- rangement of environmental events with re- fects on the operant performance of humans spect to recurrent behavioral events shifts the subsequently exposed to fixed-interval rein- local probabilities of those response classes. forcement schedules. Such enduring historical The idiosyncratic flow of natural behavior of influences can also determine the behavioral each individual depends on these three sets of actions of drugs (Barrett, 1977; Urbain, Pol- behavioral factors. ing, Millam, & Thompson, 1978). When we KINETIC STRUCTURE 235 say that someone is "extraverted," we refer in 0 3 part to enduring response classes under the C control of social consequences having major 0 genetic determinants (see Eysenck, 1956; 2- L4D/C Gottesman, 1963, 1966; Scarr, 1969). In con- 00 trast, when someone frequently displays a particular religious ritual, we are presumably referring to an enduring effect of reinforcement history (Herrnstein, 1966). Distinctions that are E LL200 between dispositions o 0. primarily genetic in origin, rather than estab- 0 5 10 15 20 25 lished by the organism's experience, are familiar; "constitutional" versus "environ- Days mental-mould" traits (Cattell, 1950) and Fig. 6. Cumulative point (token) earning by an adult psychiatric patient suffering from an affective disorder "phylogenic" versus "ontogenic" distinctions (mania) treated with lithium carbonate, over 19 days of (Skinner, 1969) are examples. In humans, hospitalization. Daily lithium dosages are given in total most molar response classes are products of milligrams and the day on which a significant disconti- both biological and environmental influences nuity occurred in rate of point earnings is marked with re- an arrow (DD). Lithium blood-level data indicate that (in varying proportions, depending on the the day of discontinuity corresponds to a lithium blood sponse class under consideration). For our level of 1.25 mEq/liter of plasma (from Thompson, purposes, estimates of enduring dispositions Golden, & Heston, 1979). need only reflect the current strength of the response class in question, not the relative contributions of the etiological sources. This tients in an acute unit were exposed to a token is not to deny the utility of understanding the reinforcement program in which tokens were relative contributions of genetic and historical administered contingent on a variety of so- factors. Indeed, for therapeutic intervention cially appropriate responses (Pickens & such understanding may be crucial. However, Thompson, 1984). Tokens were exchangeable there is no a priori reason to suppose that for supplementary goods and services. The rate knowledge of the relative contributions of bi- of token earning was negligible until the pa- ological and environmental sources for a par- tients were administered lithium. After 5 to 6 ticular enduring disposition will be helpful in days of lithium (at which point blood level of predicting the moment-to-moment temporal lithium reached 0.75 to 1.25 mEq/liter of structure of behavior. blood), there was a statistically significant discontinuity in the rates of daily point earning Fluctuating States (see Figure 6). Hence, the disposition associ- The strengths of behavioral dispositions are ated with mania was manipulable by a state modulated by state variables fluctuating rel- variable that manifested itself in recovery of atively rapidly over time (e.g., hours, days). control by normal environmental contingen- Such variables are often of biological origin cies. (e.g., deprivation, drug, fatigue, illness, and hormonal states or opponent processes [Solo- SCHEDULED ENVIRONMENTAL EVENTS mon, 1980; Solomon & Corbit, 1974]). These The constellation of an organism's relative fluctuating states modulate the level of a dis- response probabilities and fluctuating states position across a range of values, which de- constitutes a hierarchical substrate upon which termines the degree to which an environmen- stimuli act to determine the momentary prob- tal event can engender behavioral changes. In ability of a given response class (see Figure psychiatric manic patients, for example, an 7). The foregoing hierarchies of response array of response probabilities shift, and nor- probabilities are modulated by short-term mal control by socially mediated reinforce- scheduling of environmental events. Recur- ment and punishment contingencies dimin- ring environmental events shift the temporal ishes or stops. In a study by Thompson, distributions of response probabilities in the Golden, and Heston (1979), several manic pa- behavioral repertoires of organisms. Behav- 236 TRA VIS THOMPSON and DAVID LUBINSKI

Stimulus Event Variation .0 a.. 0 ~~~~~~SaeVariation 0 0.> : '| g 0 8oMean Probability 0 ~~~~~~ofa Given R-Class

R1 R2 R3 R...4 Rfn

Response Classes Fig. 7. The hierarchy of an organism's relative response probabilities. R, R2,... Rn = the organism's dispositional profile (each "R" represents a specific disposition, the average strength of which is illustrated by a black circle). The short horizontal lines extending from each circle illustrate short-term fluctuations due to state variables. Dispositions vary in the extent to which state fluctuations moderate their strength, which is highlighted by the larger variance of RI compared to Rn. Strength modulations due to variations in the intensity, configuration, and temporal structure of extroceptive stimulus events are shown by the narrow vertical lines extending from the bars marking variations due to state fluctuations.

ioral units occurring temporally antecedent tation may produce climbing to a crotch in a and contiguous to such scheduled events are tree where rain water collects. Although an regulated largely by the law of effect. Re- organism's hierarchy of relative response sponse classes regularly following such events probabilities determines the most likely classes are elicited (as in classical conditioning) or of behavior, exteroceptive environmental evoked (as in the case of adjunctive behavior). stimuli determine the specific (idiosyncratic) Falk (1961) demonstrated that the induction topographical patterns. of high-probability drinking between succes- Such adjunctive responses linked by idio- sive food pellets presented under a variable- syncratic operants occur in humans as well. interval schedule established access to drink- One person who sips a cocktail and then ing as a reinforcer for a second operant. The reaches for a cigarette in his pocket may differ specific operant that bridges eating to drink- from another individual in the topography of ing is important insofar as it characterizes the the operant binding drinking and cigarette idiosyncratic features of a given organism's smoking. The second person may instead ask behavior in its environment. For one animal, a companion for a cigarette (Falk, 1984). consuming a small amount of food may lead Thus, the effect of recurring scheduled events to a running operant that brings the organism is to cause regular shifts in the temporal struc- to a nearby stream; for another, food presen- ture of the probability of various response KINETIC STRUCTURE 237

Food Presentation Food ! Presentation

4._

0 ._0 C0 I coa) -._ 0 :)

Observed A B 2 Minutes c r Behavioral EatEa-~LeverLePresse- Drink- Climb -Dig -Locomote-_ press r_ Eat Sequence i Social,_s

!t~.- Stimulus

ccI- (L 0co a) r a)

20 Minutes O Observed A B Behavioral reet -RsDrinkt Drink- Request__ Smoke -Greet Sequence ) Drink ~Cigarette

Fig. 8. Hypothetical response probability relationships in a schedule-induced polydipsia situation (top) and in a schedule-induced cigarette-smoking situation (bottom). Evoking stimulus events are indicated by the vertical dashed lines (food-top; social stimulus-bottom). The evoked response probability shifts (e.g., drinking and smoking) are linked to the preceding moderate to high-probability response by arbitrary operants (e.g., lever pressing-top; requesting a cigarette-bottom). When the probability of a given evoked response (e.g., smoking) is higher than the preceding response (e.g., drinking), access to the higher probability response will strengthen an arbitrary operant (e.g., requesting a cigarette [B]). The resulting behavioral sequences are observed as a series of events flowing one into the next, rather than as a chained set of stimulus response units. classes (see Figure 8). Organisms differ in the fluctuating states, relative response probabil- specific relative response probabilities upon ities, the scheduled presentation of environ- which these short-term influences are super- mental events, including the available oper- imposed and in the specific operants available ants (Rachlin & Burkhard, 1978; Rachlin, to bridge schedule-modulated responses of Kagel, & Battalio, 1980), and the dynamic varying probabilities. These idiosyncratic re- interplay between these variables. Lacking in- sponse topographies lend the apparent unpre- formation on any one of these three classes of dictability to schedule-modulated response variables reduces the possibility of predicting patterns and impress us with the uniqueness behavior on a moment-to-moment basis. For of an individual's behavior. Uniqueness re- example, given precise information concern- sults from the idiosyncratic constellation of ing the scheduled presentation of environmen- 238 TRA VIS THOMPSON and DA VID LUBINSKI tal events, the accuracy of point predictions ability of response N, it can be said that re- concerning response classes within the kinetic sponse N has ended and response N + 1 has structure of behavior is increased, just as pre- begun. This is a communicative convention, dicting in detail the motion of a leaf dropped which bears no meaning when applied to a from a tree is improved by knowing relevant particular instance, since in reality compo- meterological information (e.g., wind velocity, nents of the two response classes merge at their temperature, and humidity). interface, producing an integrated performance. Only in the foregoing statistical sense of relative response probability can one speak KINETIC ENODATION of the beginning and end points of members Relative probabilities of members of func- of successive response classes. tional response classes emanating from diverse Temporally arranged relative response biological and historical sources are directly probabilities provide the mechanism and force measurable in infrahuman animals and fre- behind behavioral syntax. Arrangements of quently can be assessed in humans through momentary response probability are products verbal behavior. The momentary probabilities of recurring environmental events acting on of members of several response classes are de- the substrate of behavioral dispositions and termined by the temporal arrangement of fluctuating states. The relational power of such stimulus events that shift the response-prob- sequentially arranged response probabilities is ability hierarchies. The specific manner in indirectly observable in the same sense that which the probabilities of members of various the osmotic gradient along which a solvent classes are regulated in time (i.e., elicited, moves across a semipermeable membrane that evoked, or emitted) is less important than the separates solutions of two concentrations is in- quantitative relations among instances of directly observable. By using a dye to track members of temporally contiguous response migration of solvent across the membrane, the classes. The time (t) between successive re- osmotic gradient becomes observable. Simi- sponses of two classes and the differences in larly, the operant lever press linking food pre- the relative response probabilities (PrA-PrB) sentation and eating to drinking makes the determines the degree to which access to one response-probability relations observable in the response (RA) serves as a maintaining event schedule-regulated eating-drinking situation. for emission of another response (RB). Be- The relative response strength relations cre- cause the form of response (RB) that may be ated by temporally structured relative re- strengthened in this fashion will vary widely, sponse probabilities are revealed by the op- the specific resulting sequence of behavior will erants linking successive lower and higher vary within and across individuals over time. probability responses. Operant responses dif- Although the product may appear inchoate, fer from the dye used to identify osmotic con- the sequence of events is very orderly and, in centration gradients in that the operants are principle, predictable (Lubinski & Thomp- active components in the kinetics of behavioral son, 1986). syntax, whereas the dye is purely a tool re- The conspicuous continuity of successive vealing the basic mechanism underlying os- instances of response classes, with no obvious motic equilibrium. Formally specified rein- boundaries demarcating the sequenced perfor- forcement schedules are abstractions of mances, puzzles the observer expecting to find naturally occurring environmental events and discrete behavioral monads. The problem response-contingency relations regulating these arises because the observer attempts to apply kinetic processes and, as such, are useful in the statistical concept, "response class," to a understanding behavioral syntax. particular instance. Given that many instances of response N are measured in time, and comparable sampling of instances of re- REFERENCES sponse N + 1 are recorded, such data can be used to estimate the probability of those two Adler, N., & Hogan, J. A. (1963). Classical condition- response classes in successive temporal slices. ing and punishment of an instinctive response in Betta splendens. Animal Behaviour, 11, 351-354. When the probability of response N + 1 (typ- Ayllon, T., & Azrin, N. H. (1965). The measurement ically measured in terms of relative frequency and reinforcement of behavior of psychotics. Journal of of occurrence) significantly exceeds the prob- the Experimental Analysis ofBehavior, 8,. 357-383. KINETIC STRUCTURE 239

Barlow, G. W. (1968). Ethological units of behavior. Emurian, H. H., Emurian, C. S., Bigelow, G. E., & In D. Ingle (Ed.), The central nervous system and fish Brady, J. V. (1976). The effects of a cooperation behavior (pp. 217-232). Chicago: University of Chi- contingency on behavior in a continuous three-person cago Press. environment. Journal of the Exeprimental Analysis of Barrett, J. E. (1977). Behavioral history as a determi- Behavior, 25, 293-302. nant of the effects of d-amphetamine on punished be- Epstein, R., Lanza, R. P., & Skinner, B. F. (1980). havior. Science, 198, 67-69. Symbolic communication between two pigeons (Co- Bateson, P. P. G. (1966). The characteristics and con- lumba livia domestica). Science, 207, 543-545. text of imprinting. Biological Reviews, 41, 177-220. Eysenck, H. J. (1956). The inheritance of extraver- Bateson, P. P. G., & Reese, E. P. (1968). Reinforcing sion-introversion. Acta Psychologica, 12, 95-110. properties of conspicuous objects before imprinting has Falk, J. L. (1956). Issues distinguishing idiographic from occurred. Psychonomic Science, 10, 379-380. nomothetic approaches to personality theory. Psycho- Bernstein, D. J., & Brady, J. V. (in press). The utility logical Review, 63, 53-62. of continuous programmed environments in the ex- Falk, J. L. (1961). Production of polydipsia in normal perimental analysis of human behavior. In H. W. rats by an intermittent food schedule. Science, 133, Reese & L. J. Parrott (Eds.), Advances in behavior 195-196. science. Hillsdale, NJ: Erlbaum. Falk, J. L. (1971). The nature and determinants of Bernstein, D. J., & Ebbesen, E. B. (1978). Reinforce- adjunctive behavior. Physiology 6 Behavior, 6, 577- ment and substitution in humans: A multiple-response 588. analysis. Journal ofthe Experimental Analysis ofBehav- Falk, J. L. (1977). The origin and functions of adjunc- ior, 30, 243-253. tive behavior. Animal Learning a Behavior, 5, 325- Berryman, R., & Nevin, J. A. (1962). Interlocking 335. schedules of reinforcement. Journal ofthe Experimental Falk, J. L. (1984). Excessive behavior and drug-taking: Analysis of Behavior, 5, 213-223. Environmental generation and self-control. In P. K. Branch, M. N., & Dworkin, S. I. (1981). Effects of Levison (Ed.), Substance abuse, habitual behavior, and ratio contingencies on responding maintained by self-control (pp. 81-123). Boulder, CO: Westview Press. schedules of electric-shock presentation (response-pro- Falk, J. L. (1986). The formation of ritual behavior. duced shock). Journal of the Experimental Analysis of In T. Thompson & M. D. Zeiler (Eds.), Analysis and Behavior, 36, 191-205. integration ofbehavioral units (pp. 335-355). Hillsdale, Cantor, M. B., Smith, S. E., & Bryan, B. R. (1982). NJ: Erlbaum. Induced bad habits: Adjunctive ingestion and groom- Fallon, J. H., Jr., Allen, J. D., & Butler, J. A. (1979). ing in human subjects. Appetite: Journal for Intake Re- Assessment of adjunctive behaviors in humans using a search, 3, 13-22. stringent control procedure. Physiology 6 Behavior, 22, Carnap, R. (1956). The methodological character of 1089-1092. theoretical concepts. In H. Feigl & M. Scriven (Eds.), Fantino, E., & Logan, C. A. (1979). The experimental Minnesota studies in the philosophy ofscience: Vol. 1. The analysis of behavior. San Francisco: Freeman. foundations of science and the concepts ofpsychology and Fantino, E., Weigele, S., & Lancy, D. (1972). Aggres- psychoanalysis (pp. 38-76). Minneapolis: University of sive display in the Siamese fighting fish (Betta splen- Minnesota Press. dens). Learning and Motivation, 3, 457-468. Catania, A. C. (1966). Concurrent operants. In W. K. Ferster, C. B., & Skinner, B. F. (1957). Schedules of Honig (Ed.), Operant behavior: Areas of research and reinforcement. New York: Appleton-Century-Crofts. application (pp. 213-270). New York: Appleton-Cen- Findley, J. D. (1962). An experimental outline for tury-Crofts. building and exploring multi-operant behavior rep- Catania, A. C. (1984). Learning (2nd ed.). Englewood ertoires. Journal of the Experimental Analysis of Behav- Cliffs, NJ: Prentice-Hall. ior, 5, 113-166. Catania, A. C., Deegan, J. F., & Cook, L. (1966). Con- Findley, J. D. (1966). Programmed environments for current fixed-ratio and avoidance responding in the the experimental analysis of human behavior. In W. squirrel monkey. Journal of the Experimental Analysis K. Honig (Ed.), Operant behavior: Areas ofresearch and of Behavior, 9, 227-231. application (pp. 827-848). New York: Appleton-Cen- Cattell, R. B. (1950). Personality: A systematic theoretical tury-Crofts. andfactual study. New York: McGraw-Hill. Fodor, J. A. (1981, January). The mind-body problem. Cherek, D. R. (1982). Schedule-induced cigarette self- Scientific American, 244(1), 114-123. administration. Pharmacology Biochemistry 6& Behavior, Frederiksen, L. W., & Peterson, G. L. (1974). Sched- 17, 523-527. ule-induced aggression in nursery school children. Church, R. (1978). The internal clock. In S. H. Hulse, Psychological Record, 24, 343-351. H. Fowler, & W. K. Honig (Eds.), Cognitive processes Gardner, M. L., & Malagodi, E. F. (1981). Respond- in animal behavior (pp. 277-310). Hillsdale, NJ: Erl- ing under sequence schedules of electric shock presen- baum. tation. Journal ofthe Experimental Analysis ofBehavior, Clarke, J., Gannon, M., Hughes, I., Keogh, C., Singer, 35, 323-334. G., & Wallace, M. (1977). Adjunctive behavior in Gollub, L. (1977). Conditioned reinforcement: Schedule humans in a group gamboling situation. Phsyiology 6r effects. In W. K. Honig & J. E. R. Staddon (Eds.), Behavior, 18, 159-161. Handbook of operant behavior (pp. 288-312). Engle- Dewsbury, D. A. (1978). Comparative animal behavior. wood Cliffs, NJ: Prentice-Hall. New York: McGraw-Hill. Gottesman, I. I. (1963). Heritability of personality: A Drickamer, L. C., & Vissen, S. H. (1982). Animal be- demonstration. Psychological Monographs, 77(9, Whole havior: Concepts, processes, and methods. Boston: W. No. 572). Grant Press. Gottesman, I. I. (1966). Genetic variance in adaptive 240 TRA VIS THOMPSON and DAVID LUBINSKI

personality traits. Journal of Child Psychology and Psy- (1971). Effects of long-term shock and associated chiatry and Allied Disciplines, 7, 199-208. stimuli on aggressive and manual responses. Journal of Harzem, P., Lowe, C. F., & Spencer, P. T. (1978). the Experimental Analysis of Behavior, 15, 141-166. Temporal control of behavior: Schedule interactions. Kachanoff, R., Leveille, R., McLelland, J. P., & Wayner, Journal of the Experimental Analysis of Behavior, 30, M. J. (1973). Schedule induced behavior in humans. 255-270. Physiology a Behavior, 11, 395-398. Heinroth, 0. (1911). Beitrage zur biologie, namentlich Katz, J. L., & Barrett, J. E. (1979). Conjunctive sched- ethologie und psychologie der anatiden. Verhand- ules of reinforcement. IV: Effects on the pattern of lungen des V Internationalen Ornithologen-Kongresses, responding of changes in requirement at reinforce- 589-708. ment. Animal Learning a Behavior, 7, 483-488. Herrnstein, R. J. (1966). Superstition: A corollary of Kazdin, A. E. (1977). The token economy. New York: the principles of operant conditioning. In W. K. Honig Plenum Press. (Ed.), Operant behavior: Areas of research and applica- Kelleher, R. T. (1957). A multiple schedule of condition (pp. 33-51). New York: Appleton-Century-Crofts. tioned reinforcement with chimpanzees. Psychological Herrnstein, R. J., & Morse, W. H. (1958). A conjunc- Reports, 3, 485-491. tive schedule of reinforcement. Journal of the Experi- Kelleher, R. T. (1966). Chaining and conditioned re- mental Analysis of Behavior, 1, 15-24. inforcement. In W. K. Honig (Ed.), Operant behavior: Hetherington, A. W., & Ranson, S. W. (1942). The Areas of research and application (pp. 160-212). New spontaneous activity and food intake of rats with hy- York: Appleton-Century-Crofts. pothalamic lesions. American Journal ofPhysiology, 136, Kelleher, R. T., Fry, W., & Cook, L. (1964). Adjusting 609-617. fixed-ratio schedules in the squirrel monkey. Journal Hinde, R. A. (1954). Factors governing the changes in of the Experimental Analysis of Behavior, 7, 69-77. strength of a partially inborn response, as shown by Kelleher, R. T., & Gollub, L. R. (1962). A review of the mobbing behaviour of the chaffinch (Fringilla coe- positive conditioned reinforcement. Journal of the Ex- lebs): The nature of the response and an examination perimental Analysis of Behavior, 5, 543-597. of its course. Proceedings of the Royal Society, 142, Se- Kelleher, R. T., & Morse, W. H. (1968). Schedules ries B, 306-331. using noxious stimuli: III. Responding maintained with Hinde, R. A. (1958). The nest building behaviour of response-produced electric shocks. Journal of the Ex- domesticated canaries. Proceedings of the Zoological So- perimental Analysis of Behavior, 11, 819-838. ciety of London, 131, 1-48. Lasiter, P. S. (1979). Influence of contingent responding Hinde, R. A. (1966). Animal behaviour. New York: on schedule-induced activity in human subjects. Phys- McGraw-Hill. iology 6 Behavior, 22, 239-243. Hinde, R. A. (1968). Dichotomies in the study of de- Lewin, K. (1935). A dynamic theory of personality (D. velopment. In J. M. Thoday & A. S. Parkes (Eds.), K. Adams & K. E. Zener, Trans.). New York: Genetic and environmental influences on behaviour. New McGraw-Hill. York: Plenum Press. Lissman, H. W. (1932). Die umwelt des Kampfisches Hinde, R. A. (1970). Animal behavior: A synthesis of (Betta splendens Regan). Zeitschrift fir Vergleichende ethology and comparative psychology (2nd ed.). New Physiologie, 18, 65-111. York: McGraw-Hill. Locurto, C. M., Terrace, H. S., & Gibbon, J. (Eds.). Hinde, R. A., & Stevenson, J. G. (1969). Integration (1981). Autoshapingandconditioning theory. New York: of response sequences. In D. S. Lehrman, R. A. Hinde, Academic Press. & E. Shaw (Eds.), Advances in the study of behavior Loevinger, J. (1957). Objective tests as instruments of (Vol. 2, pp. 267-296). New York: Academic Press. psychological theory. Psychological Reports, 3, 635-694. Hinde, R. A., & Stevenson-Hinde, J. (Eds.). (1973). Lubinski, D., & MacCorquodale, K. (1984). "Symbolic Constraints on learning. New York: Academic Press. communication" between two pigeons (Columba livia) Hineline, P. N. (1981). Constraints, competing behav- without unconditioned reinforcement. Journal of Com- ior, and the principle of resonance. In C. M. Brad- parative Psychology, 98, 372-380. shaw, E. Szabadi, & C. F. Lowe (Eds.), Quantification Lubinski, D., & Thompson, T. (1986). Functional units of steady-state operant behaviour (pp. 153-164). Am- of human behavior and their integration: A disposi- sterdam: Elsevier/North Holland Biomedical Press. tional analysis. In T. Thompson & M. D. Zeiler Hineline, P. N. (1984). Aversive control: A separate (Eds.), Analysis and integration of behavioral units (pp. domain? Journal of the Experimental Analysis ofBehav- 275-314). Hillsdale, NJ: Erlbaum. ior, 42, 495-509. MacCorquodale, K., & Meehl, P. E. (1954). Edward Hineline, P. N. (1986). Retuning the operant-respon- C. Tolman. In W. K. Estes, S. Koch, K. Mac- dent distinction. In T. Thompson & M. D. Zeiler Corquodale, P. E. Meehl, C. G. Muller, Jr., W. N. (Eds.), Analysis and integration of behavioral units (pp. Schoenfeld, & W. S. Verplanck (Eds.), Modern learn- 55-79). Hillsdale, NJ: Erlbaum. ing theory: A critical analysis offive examples (pp. 177- Hocutt, M. (1967). On the alleged circularity of Skin- 266). New York: Appleton-Century-Crofts. ner's concept of stimulus. Psychological Review, 74, 530- Mackintosh, N. J. (1974). The psychology of animal 532. learning. New York: Academic Press. Honig, W. K., & Staddon, J. E. R. (Eds.) (1977). Malagodi, E. F., Gardner, M. L., Ward, S. E., & Mag- Handbook of operant behavior. Englewood Cliffs, NJ: yar, R. L. (1981). Responding maintained under Prentice-Hall. intermittent schedules of electric-shock presentation: Hutchinson, R. R., & Emley, G. S. (1977). Electric "Safety" or schedule effects? Journal ofthe Experimen- shock produced drinking in the squirrel monkey. Jour- tal Analysis of Behavior, 36, 171-190. nal of the Experimental Analysis of Behavior, 28, 1-12. Masters, W. H., & Johnson, V. E. (1966). Human sex- Hutchinson, R. R., Renfrew, J. W., & Young, G. A. ual response. Boston: Little, Brown. KINETIC STRUCTURE 241

McConahey, 0. L., Thompson, T., & Zimmerman, R. Rachlin, H. (1976). Introduction to modern behaviorism (1977). A token system for retarded women: Behavior (2nd ed.). San Francisco: Freeman. therapy, drug administration, and their combination. Rachlin, H., & Burkhard, B. (1978). The temporal In T. Thompson & J. Grabowski (Eds.), Behavior triangle: Response substitution in instrumental con- modification ofthe mentally retarded (pp. 167-234). New ditioning. Psychological Review, 85, 22-47. York: Oxford University Press. Rachlin, H., Kagel, J. H., & Battalio, R. C. (1980). Meehl, P. E. (1978). Theoretical risks and tabular as- Substitutability in time allocation. Psychological Re- terisks: Sir Karl, Sir Ronald, and the slow progress of view, 87, 355-374. soft psychology. Journal of Consulting and Clinical Psy- Razran, G. (1955). A note on second-order condition- chology, 46, 806-834. ing-and secondary reinforcement. Psychological Re- Meehl, P. E. (1986). Trait language and behaviorese. view, 62, 327-332. In T. Thompson & M. D. Zeiler (Eds.), Analysis and Rescorla, R. A. (1980). Pavlovian second-order condition- integration ofbehavioral units (pp. 315-334). Hillsdale, ing: Studies in associative learning. Hillsdale, NJ: Erl- NJ: Erlbaum. baum. Morris, E. K., Higgins, S. T., & Bickel, W. K. (1982). Rescorla, R. A., & Solomon, R. L. (1967). Two-process Comments on cognitive science in the experimental learning theory: Relationships between Pavlovian con- analysis of behavior. Behavior Analyst, 5, 109-125. ditioning and instrumental learning. Psychological Re- Morse, W. H., & Kelleher, R. T. (1966). Schedules view, 74, 151-182. using noxious stimuli: I. Multiple fixed-ratio and fixed- Reynolds, G. S. (1975). A primer ofoperant conditioning interval termination of schedule complexes. Journal of (rev. ed.). Glenview, IL: Scott, Foresman. the Experimental Analysis of Behavior, 9, 267-290. Rider, D. P. (1977). Interlocking schedules: The rela- Mowrer, 0. H. (1947). On the dual nature of learn- tionship between response and time requirements. ing-a re-interpretation of "conditioning" and "prob- Journal of the Experimental Analysis of Behavior, 28, lem-solving." Harvard Educational Review, 17, 102- 41-46. 148. Rider, D. P. (1980). Alternative fixed-ratio fixed-inter- Muller, P. G., Crow, R. E., & Cheney, C. D. (1979). val schedules of reinforcement. Journal of the Experi- Schedule-induced locomotor activity in humans. Jour- mental Analysis of Behavior, 33, 243-252. nal ofthe Experimental Analysis ofBehavior, 31, 83-90. Rosenblith, J. Z. (1970). Polydipsia induced in the rat Olvera, D. R., & Hake, D. F. (1976). Producing a by a second-order schedule. Journal ofthe Experimental change from completion to sharing: Effects of large Analysis of Behavior, 14, 139-144. and adjusting response requirements. Journal of the Sarte, J. (1938). Nausea (L. Alexander, Trans.). Cam- Experimental Analysis of Behavior, 26, 321-333. bridge, MA: R. Bentley. Overmier, J. B., & Lawry, J. A. (1979). Pavlovian Savage-Rumbaugh, E. S., Rumbaugh, D. M., & Boysen, conditioning and the mediation of behavior. In G. H. S. (1978). Symbolic communication between two Bower (Ed.), The psychology oflearning and motivation chimpanzees (Pan troglodytes). Science, 201, 641-644. (Vol. 13, pp. 1-55). New York: Academic Press. Scamf, S. (1969). Social introversion-extraversion as a Pap, A. (1958). Semantics and necessary truth. New Ha- heritable response. Child Development, 40, 823-832. ven, CT: Yale University Press. Schnaitter, R. (1978). Private causes. Behaviorism, 6, Pavlov, I. P. (1927). Conditioned reflexes (G. V. Anrep, 1-12. Trans.). London: Oxford University Press. Segal, E. F. (1972). Induction and the provenance of Pickens, R. W., & Thompson, T. (1984). Behavioral operants. In R. M. Gilbert & J. R. Millenson (Eds.), treatment of drug dependence. In J. Grabowski, M. Reinforcement: Behavioral analyses (pp. 1-34). New L. Stitzer, & J. E. Henningfield (Eds.), Behavioral York: Academic Press. intervention techniques in drug abuse treatment (NIDA Sellers, W. (1958). Counterfactuals, dispositions, and Research Monograph Series 46, pp. 53-67; DHHS the causal modalities. In H. Feigl, M. Scriven, & G. Publication No. ADM 84-1282). Washington, DC: Maxwell (Eds.), Minnesota studies in the philosophy of U.S. Government Printing Office. science: Vol. 2. Concepts, theories, and the mind-body Pierrel, R., & Sherman, J. G. (1963, February). Bar- problem (pp. 225-308). Minneapolis: University of nabus, a rat with college training. Brown Alumni Minnesota Press. Monthly, pp. 8-14. Shapiro, M. M. (1961). Salivary conditioning in dogs Porter, J. H., Brown, R. T., & Goldsmith, P. A. (1982). during fixed-interval reinforcement contingent upon Adjunctive behavior in children on fixed interval food lever pressing. Journal of the Experimental Analysis of reinforcement schedules. Physiology & Behavior, 28, Behavior, 4, 361-364. 609-612. Sherrington, C. (1906). The integrative action of the ner- Prader, A., Labhart, A., & Willi, H. (1956). Ein syn- vous systems. Forge Village, MA: Murray Printing drom von adipositas, kleinwuchs, kypotorchismus und Company. oligorphrenie nach myatonleartigem zustand in neu- Sidman, M. (1960). Tactics of scientific research. New geborenenaltr. Schweizerische Medizinische Wochen- York: Basic Books. schrift, 86, 1260. Sidman, M. (1962). An adjusting avoidance schedule. Premack, D. (1965). Reinforcement theory. In D. Le- Journal ofthe Experimental Analysis ofBehavior, 5, 271- vine (Ed.), Nebraska Symposium on Motivation (Vol. 277. 13, pp. 123-180). Lincoln: University of Nebraska Sidman, M., Rauzin, R., Lazar, R., Cunningham, S., Press. Tailby, W., & Carrigan, P. (1982). A search for Premack, D. (1971). Catching up with common sense symmetry in the conditional discriminations of rhesus or two sides of a generalization: Reinforcement and monkeys, baboons, and children. Journal of the Exper- punishment. In R. Glaser (Ed.), The nature of rein- imental Analysis of Behavior, 37, 23-44. forcement (pp. 121-150). New York: Academic Press. Sidman, M., & Tailby, W. (1982). Conditional dis- 242 TRA VIS THOMPSON and DA VID LUBINSKI

crimination vs. matching to sample: An expansion of Thompson, T., & Sturm, T. (1965). Classical condi- the testing paradigm. Journal ofthe Experimental Anal- tioning of aggressive display in Siamese fighting fish. ysis of Behavior, 37, 5-22. Journal of the Experimental Analysis ofBehavior, 8, 397- Simpson, M. J. A. (1968). The display of the Siamese 403. fighting fish, Betta splendens. Animal Behaviour Mono- Thompson, T., & Zeiler, M. D. (Eds.). (1986). Analysis graphs, l(Whole Pt. 1). and integration of behavioral units. Hillsdale, NJ: Erl- Skinner, B. F. (1938). The behavior of organisms. New baum. York: Appleton-Century-Crofts. Thorndike, E. L. (1911). Animal intelligence. New York: Skinner, B. F. (1953). Science and human behavior. New Macmillan. York: Macmillan. Thorpe, W. H. (1961). Bird-song. Cambridge: Cam- Skinner, B. F. (1969). Contingencies of reinforcement: A bridge University Press. theoretical analysis. New York: Appleton-Century- Timberlake, W., & Allison, J. (1974). Response depri- Crofts. vation: An empirical approach to instrumental perfor- Solomon, R. L. (1980). The opponent-process theory of mance. Psychological Review, 81, 146-164. acquired motivation: The costs of pleasure and the Tinbergen, N. (1948). Social releasers and the experi- benefits of pain. American Psychologist, 35, 691-712. mental method required for their study. Wilson Bul- Solomon, R. L., & Corbit, J. D. (1974). An opponent- letin, 60, 6-51. process theory of motivation: I. Temporal dynamics of Tinbergen, N. (1951). The study of instinct. Oxford: affect. Psychological Review, 81, 119-145. Clarendon Press. Spence, K. W. (1948). The postulates and methods of Tinbergen, N. (1953). The Herring Gull's world. Lon- 'behaviorism.' Psychological Review, 55, 67-78. don: Collins. Spence, K. W. (1956). Behavior theory and conditioning. Ulrich, R. E., Stachnik, T. J., Brierton, G. R., & Mabry, New Haven, CT: Yale University Press. J. H. (1966). Fighting and avoidance in response to Staddon, J. E. R. (1977). Schedule-induced behavior. aversive stimulation. Behaviour, 26, 124-129. In W. K. Honig & J. E. R. Staddon (Eds.), Handbook Urbain, C., Poling, A., Millam, J., & Thompson, T. of operant behavior (pp. 125-152). Englewood Cliffs, (1978). d-Amphetamine and fixed-interval perfor- NJ: Prentice-Hall. mance: Effects of operant history. Journal of the Ex- Staddon, J. E. R. (Ed.). (1980). Limits to action: The perimental Analysis of Behavior, 29, 385-392. allocation of individual behavior. New York: Academic Vernon, W., & Ulrich, R. (1966). Classical condition- Press. ing of pain-elicited aggression. Science, 152, 668-669. Staddon, J. E. R., & Ayres, S. L. (1975). Sequential Wallace, M., & Oei, T. P. S. (1981). Differences in and temporal properties of behavior induced by a scheduled induced behaviour as a function of reinforc- schedule of periodic food delivery. Behaviour, 54, 26- er in humans. Physiology & Behavior, 27, 1027-1030. 49. Wallace, M., & Singer, G. (1976). Adjunctive behavior Teitelbaum, P. (1955). Sensory control of hypothalamic and smoking induced by a maze solving schedule in hyperphagia. Journal of Comparative and Physiological humans. Physiology & Behavior, 17, 849-852. Psychology, 48, 156-163. Wallace, M., Singer, G., Wayner, M. J., & Cook, P. Tellegen, A. (1981). Practicing the two disciplines for (1975). Adjunctive behavior in humans during game relaxation and enlightenment: Comment on "Role of playing. Physiology & Behavior, 14, 651-654. the feedback signal in electromyographic feedback: The Weiner, H. (1964). Conditioning history and human relevance of attention," by Qualls and Sheehan. Jour- fixed-interval performance. Journal ofthe Experimental nal ofExperimental Psychology: General, 110, 217-226. Analysis of Behavior, 7, 383-385. Thompson, T. (1963). Visual reinforcement in Siamese Weiner, H. (1969). Controlling human fixed-interval fighting fish. Science, 141, 55-57. performance. Journal of the Experimental Analysis of Thompson, T. (1966). Operant and classically-condi- Behavior, 12, 349-373. tioned aggressive behavior in Siamese fighting fish. Whitman, C. 0. (1899). Animal behavior. Biological American Zoologist, 6, 629-641. Lectures, Marine Biological Laboratory, Wood's Hole, Thompson, T., Golden, R., & Heston, L. (1979, May). Boston, Massachusetts. Point-earning in a token economy as an index of lithium Zeiler, M. D. (1977). Schedules of reinforcement: The effects in mania. Paper presented at the annual meeting controlling variables. In W. K. Honig & J. E. R. of the Association for Behavior Analysis, Milwaukee, Staddon (Eds.), Handbook ofoperant behavior (pp. 201- WI. 232). Englewood Cliffs, NJ: Prentice-Hall. Thompson, T., & Grabowski, J. G. (1972). Reinforce- ment schedules and multioperant analysis. New York: Appleton-Century-Crofts. Thompson, T., & Schuster, C. R. (1968). Behavioral Received April 4, 1986 pharmacology. Englewood Cliffs, NJ: Prentice-Hall. Final acceptance May 25, 1986