Psychonomic Bulletin & Review 1994, 1 (4), 405-420 Behavior systems, associationism, and Pavlovian conditioning WILLIAM TIMBERLAKE Indiana University, Bloomington, Indiana Associative and behavior systems accounts of Pavlovian conditioning have different emphases. The traditional associative account has focused on the role of the unconditional stimulus (US) in strengthening stimulus associations according to a set of general laws. The behavior systems account has focused on the relation of conditional responding to the preorganized perceptual, motor, and mo­ tivational organization engaged by the US. Knowledge of a behavior system enables successful pre­ diction of the form and ease of conditioning as a function of the type of conditional stimulus (CS), US, and the CS-US relation. At the same time, Pavlovian manipulations act as a window on how a behavior system works. Both associative and behavior systems accounts can be criticized as in­ complete and idiosyncratic. A comprehensive account of Pavlovian conditioning could profit from their integration. Laboratory-associationist and adaptive-evolutionary zation and motivational processes underlying behavior is accounts of Pavlovian conditioning differ markedly in a critical step in understanding conditioning. Presenta­ their focus. The former has emphasized the association tion of the US is viewed as engaging and constraining between the conditional stimulus (CS) and the uncondi­ preorganized mechanisms underlying behavior, rather tional stimulus (US), and the role of the US in strength­ than as a simple, direct cause ofconditional responding ening it (e.g., Hearst, 1988; Kimble, 1961; Mackintosh, (Timberlake, 1993b). 1974; Pavlov, 1927). Conditioning procedures have been The present paper outlines the behavior systems ap­ refined and standardized to isolate the study ofassocia­ proach to Pavlovian conditioning-an approach that is tions from the influence of instincts, nonassociative related to the adaptive-evolutionary account and that at­ changes, and response-contingent reward (Hilgard & tempts to combine structural and functional aspects of Marquis, 1940; Kimble, 1961). More recent models have learning and behavior. The basic premise ofthe behav­ expanded the types ofassociation and the role ofcontext, ior systems approach is that the determinants ofbehav­ multiple CSs, the CS-US contingency, and stimulus rep­ ior have been organized by evolutionary selection pres­ resentations in determining behavior (e.g., Miller & sure, development, and learning into functional systems, Matzel, 1988; Rescorla & Holland, 1982). However, the such as feeding, mating, parenting, and defense (Baer­ focus has remained on associations among stimuli and ends, 1976, 1988; Fanselow, 1991; Fanselow & Lester, the simple causal role of the US. Within this approach, 1988; Hogan, 1989; Scott, 1958; Timberlake, 1983a, behavior has been treated primarily as an index of the 1983b; Timberlake & Lucas, 1989; Tinbergen, 1951). strength of associations. The mechanisms and processes of a behavior system In contrast, adaptive-evolutionary accounts ofPavlov­ produce stimulus filtering and integration, timing and ian conditioning have emphasized behavior, especially memory functions, motor components and programs, the relation ofthe conditional response to the form and motivational states and hierarchies, and their neuro­ orientation of naturally occurring functional behavior physiological substrates (Domjan, 1994; Fanselow, 1994; and relevant underlying structure and processes (e.g., Hogan, 1994; Timberlake, 1993b; Timberlake & Lucas, Davey, 1989; Fanselow & Lester, 1988; Gardner & Gard­ 1989). Behavior systems are species typical in that they ner, 1988; Holland, 1984; Hollis, 1982, 1990; Konorski, are related to characteristic processes, structures, and 1967; Rozin & Schull, 1988; Timberlake, 1983b; Tim­ environments. At the same time, individual differences berlake & Lucas, 1989; Timberlake & Silva, in press). In are also expected on the basis ofgenetic, developmental, this approach, establishing the perceptual-motor organi- and experiential variation. Similarities and differences are also presumed to exist among species, based on the extent ofcommon versus divergent phylogeny, selection Completion of this manuscript was facilitated by NIMH Grant 37892 pressures, and experience (Timberlake, 1993a). and NSF Grant IBN 91 21647. The manuscript is based on a sympo­ The next section describes some distinguishing char­ sium presentation at the meetings of the American Psychological As­ acteristics ofthe behavior systems approach to condition­ sociation, Toronto, 1993. I am indebted to Cynthia Langley, Fran Silva, ing. I will then consider a specific behavior system-the and Kathleen Silva for their comments. Please address correspondence to W. Timberlake, Department of Psychology, Indiana University, feeding system ofdomesticated rats-and how it relates Bloomington, IN 47405 (e-mail: [email protected]). to the study of Pavlovian conditioning. I will conclude 405 Copyright 1994 Psychonomic Society, Inc. 406 TIMBERLAKE by considering the prospects for an integration ofasso­ simple stimulus exposure to reciprocal response contin­ ciative and behavior systems accounts of conditioning. gencies. It includes the conditioning of motivational states to external and internal stimuli; the alteration, in­ BEHAVIOR SYSTEMS AND LEARNING tegration, and recombination ofmotor programs; the dif­ ferentiation and combination of stimulus classes; and The behavior systems approach is not primarily a the­ the integration and sequential combination of percep­ ory oflearning. It is, though, a theory that deals with the tual, motor, and motivational determinants (Timberlake, mechanisms and processes underlying functional be­ in press; Timberlake & Lucas, 1989). havior. Learning is embedded within a behavior system It follows that the traditional procedural distinctions that inevitably influences the form, circumstances, between Pavlovian and operant conditioning do not de­ speed, and maintenance of acquisition. The role of fine critical differences in basic learning processes, but learning is assumed to have evolved in conjunction with rather specify differences in how a system is engaged the system and environment to facilitate the local fit of and measured (Gormezano & Kehoe, 1975; Pear & El­ behavior and environment (Johnston, 1981; Nottebohm, dridge, 1984). Because the underlying system is the same, 1972; Timberlake & Lucas, 1989; Tinbergen, 1951). A the results ofPavlovian and operant conditioning should behavior system provides the initial perceptual, motor, overlap as well. Thus, if food is presented to a hungry and motivational conditions and operating characteris­ rat, the feeding system will influence the nature oflearn­ tics that organize and direct learning and in turn are re­ ing whether the experimenter is using the procedures of organized and modified by learning. From these as­ operant conditioning, Pavlovian conditioning, habitua­ sumptions flow several general characteristics of a tion, or insight learning. For example, the use ofa Pavlov­ behavior systems approach. ian procedure to pair a moving stimulus with food and the use ofan operant procedure to require the rat to con­ Learned Behavior is Complexly Caused tact the moving stimulus to obtain food have produced In the traditional associative approach, learning is de­ similar behavior (Timberlake, Wahl, & King, 1982). fined by and referenced to experimental paradigms, such In short, in the behavior systems approach, specific as Pavlovian and operant conditioning. By defining learn­ learning paradigms can be viewed profitably as tools for ing in terms of the manipulations of the experimenter, investigating the nature and operation ofthe system and the associative approach has encouraged the view that the role of learning in its expression. This is not to say learning is caused in a simple, direct way by the presen­ that a particular learning paradigm cannot be used to in­ tation of the US (Timberlake, 1993b). For example, an vestigate a specific form of learning. Just as a simple important part of the carefully crafted distinction be­ wiping reflex can be studied in a spinal frog, the condi­ tween operant and Pavlovian conditioning is based on tioning ofa simple salivary reflex can be studied by using procedural differences in how the reward is presented Pavlovian procedures and a specially prepared and con­ (Hilgard & Marquis, 1940). strained dog. However, a similar difficulty is raised in The behavior systems approach references learning to both cases. As pointed out by many investigators, there a functional context rather than to an experimental pro­ is no simple spinal reflex in intact animals (Fearing, cedure. A critical attribute ofa behavior system is that it 1930; Sherrington, 1906). The spinal preparation only produces strings of responses that lead toward or away hints at how a reflex works in more typical circum­ from a particular stimulus or condition (Timberlake & stances. Similarly, there are no simple learned responses Silva, in press). These seemingly purposive strings are in animals that are free to move about the environment. often probabilistic and fragmentary. Whether they actu­ As Domjan observes about his work on sexual condi­ ally achieve a particular outcome depends on the fit be­ tioning (Domjan, 1994), a common US does not guar­ tween the system and the environment, a fit that depends antee a uniform set oflearning effects. Instead,