The Straw Machine As Tar Baby

The Behavior Analyst 1993, 16, 59-65 No. I (Spring)

Contextualistic Mechanism or Mechanistic Contextualism?:The Straw Machine As Tar Baby

M. Jackson MarrGeorgia Tech

For a generation, we behaviorists havebeen chiding our cognitive friends thatthey seemed to be treating the organismas a white box in a black world. For manycognitive psychologists, at least histori-cally, human behaviors emerge likewraiths out of a mysterious cauldron ofprocessed images, thoughts, strategies,schemas, rules, and representations, allapparently disembodied from history andcurrent contingencies. I am reminded ofthe linguist Julie Andressen's address(1991) at the convention of the Associ-ation for Behavior Analysis where sheshowed de Saussure's diagram of a float-ing head spouting words into a contextualvoid. Well, cognitive psychology hasgradually discovered context as they haveother things (like implicit learning andremembering) we behavior analysts haveknown about from the beginning. Withinbehaviorism, however, "context" isgradually taking on the character of the"Man Who Came to Dinner," that is, aonce-welcomed guest is becoming, in myview, an overbearing nuisance. "Con-text" has evolved into "contextualism,"a conceptual and presumably founda-tional framework-eine Weltanschauungfor radical behaviorism. Contextualistbehaviorism had been argued by Kantorin bygone days, and more recently, withaccelerated vehemence, by Linda and

Portions ofthis paper have been presented at theconvention of the Association for Behavior Anal-ysis, 1992, and the First International Congress onBehaviorism and the Sciences ofBehavior. A some-what different version will appear in J. T. Todd andE. K. Morris (Eds.), Modernperspectives on classicaland contemporary behaviorism (in press). I wish tothank Hayne Reese and Ed Morris for theirthoughtful and stimulating comments, formal andinformal.Address correspondence to the author at the

School of Psychology, Georgia Tech, Atlanta, GA30332-0170.

Steve Hayes, Hayne Reese, Ed Morris,and others.Ed Morris has been a most scholarly,

eloquent, and persuasive spokesman forthe behaviorism-as-contextualism view,and this essay will be in response pri-marily, though not exclusively, to hispresentations of that view. My sourcesinclude his article entitled "Contextual-ism: The World View of Behavior Anal-ysis" (1988b) and two presentations, oneat the 1988 American Psychological As-sociation meeting in a symposium onSkinner's The Behavior of Organisms(1988a) and his presidential address(1991) at the meeting of the Associationfor Behavior Analysis. Those presenta-tions are encompassed in a book chapterentitled "The Behavior ofOrganisms: AContext Theory of Meaning" (in press).His thesis includes at least the followingpoints: (a) Radical behaviorism is fun-damentally committed to a contextualistposition; (b) Skinner began with this view,then subsequently either ignored orabandoned it; and (c) behavior analysishas, and indeed, must reassert this po-sition conceptually and empirically in or-der to compete effectively in the toughmarket ofpsychological theory and prac-tice.

In keeping with Morris's perspective,his work must be viewed in historicalcontext. Foremost is the strong imprintof J. R. Kantor's "integrated field ori-entation" (see, e.g., Morris, 1984, 1988b),and Pepper's World Hypotheses (Hayes,Hayes, & Reese, 1988; Pepper, 1942). TheHayes et al. paper on Pepper's WorldHy-potheses deserves special attention. Thisvery stimulating paper demonstrates, inspite of its thesis (and that of Pepper),that there is a kind oflaw ofconservationofopinion: The more defensible one viewis, the less defensible another view. Pep-per's assertion is that one worldview is

59

60 M. JACKSON MARR

incommensurate with another, so thatstrengths in one do not imply weaknessesin another. Argument (and, I believe, re-ality) usually proceeds otherwise, how-ever. The incommensurate position canbe made defensible if the various world-views are adroitly defined. The issue aris-es, however, as to whether those defini-tions are merelyformal-that is, usuallyhighly simplified verbal abstractionsconstructed or derived from dictionariesand other sources-or functional-de-scriptive, for example, ofthe relevant be-haviors of those in fields in which theterms in question have a relatively safeand comfortable home. Part ofmy thesisis that formal-functional distinctions arebasic to arguments asserting that behav-iorism is one thing or another. If, for ex-ample, it is asserted that behaviorism isnot mechanistic, but something else, thenhow is "mechanism" to be defined (nevermind the something else)? My approachis first to look at what the theories andpractices ofmechanism are and have beenin those fields in which the term has per-sisted in one form or another (physics isa prime example, but there are many oth-ers). Second, I compare those activitieswith the interests, basic and applied, ofthose calling themselves behavior ana-lysts. There are numerous dimensions ofwhat constitutes a mechanistic approach.Space does not permit, however, any-thing like a full accounting of how allthose dimensions might be reflected inmodem behavior analysis. In this essayI want to focus on the issue of "contex-tualism" as a conceptual framework forbehaviorism, set in contrast to "mecha-nism."As a starting point, consider Pepper's

"root metaphors" for "mechanism" and"contextualism." Here, I rely on Hayeset al. (1988):The root metaphor of mechanism is the machine.Any common-sense machine is composed of dis-crete parts related to other parts in some systematicway. Relations among the parts do not change thenature ofthe parts, however, because the parts existindependently of those relations. Further, in anycommon-sense machine, some sort of force or en-ergy is exerted on or transmitted through the systemto produce predictable outcomes. (p. 98)

As for contextualism,The root metaphor ofcontextualism is the ongoingact in context. Another term might be the historicalact, but not as a dead description of a thing done.It is doing as it is being done, as in hunting, shop-ping, or making love. (p. 100)

Having planted these roots, it mightappear that very different kinds of florawould grow from them. I would like toargue, however, that distinctions be-tween those flora are based upon a kindof category error. Thus to place mecha-nism and contextualism in contrast is tomisunderstand the role ofboth in the lan-guage game of scientific inquiry. Behav-ioral contextualists have set up a strawmachine that, under examination, turnsout to be a conceptual tar baby to snarethe unwary.Problems arise immediately with the

root metaphors as quoted. The parts ofa machine are said to exist independentlyof the relations between them within themachine. Hayes et al. (1988) continue byexamining the lever as "a simple exam-ple": "The lever is composed oftwo dis-crete and independent parts-a lever anda fulcrum" (p. 98). (It is curious that theauthors came to use the same word todescribe both the machine and a part ofthe machine. To avoid some confusion,I will use the term "bar" instead of "le-ver" to describe the part.) So, then, alever consists of a bar and a fulcrum. Inwhat sense, however, can the bar and ful-crum be said to exist independently oftheir role in the machine? Certainly theyare not independent as far as their func-tion in a lever is concerned. "Fulcrum"has meaning only in the context of a le-ver. Moreover, there is no particular ob-ject to which the term "fulcrum" nec-essarily applies, independently of its rolein a lever. Try going to a hardware storeand asking for a "fulcrum"! The case of"bar" is different, but the conclusion isthe same. We could point to a bar lyingon a table, but that bar might be used formany purposes having nothing to do withlevers (e.g., construction, a weapon, aweight, a prison door, etc.). Thus, to saythat the bar can exist independently ofits role in a lever is saying very little. This

CONTEXTUALISM AND MECHANISM 61

is a kind of disembodied existence with"'potential" purpose. Again, the meaningof "bar" depends on context. With re-spect to the machine as a whole, as Greg-ory (1981), in a very valuable discussionof machines and mechanisms, puts it:

A machine is generally conceived as a functionalentity consisting of clearly defined parts, such thatthe internal functions can be understood fromknowledge of the parts and how they interact. Tounderstand its external functions-what it does-we must know about its "environment." (p. 73)

And,When we speak of the function of the entire organ-ism, rather than its organs, or ofan entire machinerather than its parts, then we have to see the or-ganism or machine within a context within whichit plays its part. (p. 83)

These points provide a link betweenfunction in a machine and function in anorganism. The former has an ontologicalhistory established through contingen-cies related to human need and use, andperhaps a bit of adventitious controlthrough evolutionary mechanisms. Bothare the outcome of selective contingen-cies; for example, one could trace a se-lective shaping ofdesign leading from theWright brothers' original Kitty Hawkplane to the space shuttle. Moreover, al-though the term "purpose" cannot rea-sonably be applied to the outcome of bi-ological evolution as it might to amachine, there is no clear necessity toabandon a mechanistic perspective toembrace something called contextualismwhen the former already implies the lat-ter.The retreat from mechanism often

hinges on the relations between part andwhole in the machine, or the organism,or whatever. The trick, however, is to sailcarefully between the Scylla of ghostlyemergence and the Charybdis of simplereductionism. Emergence is often citedas a counter to simple-minded mecha-nism. The whole is somehow more thanthe sum of its parts. Contextualism em-braces emergence. Context, improperlytreated, can easily become an eerie aurasurrounding a fimctional account, as wellas an infinite well of explanatory appeal.

What is contributed beyond a carefulspecification ofcontingencies is not clear.As for the whole being more than the

sum of its parts, the immediate problemis that both "parts" and "sum" are un-defined. How does one "sum up" the"parts" of a clock to yield a clock? Toquote Gregory (1981) once more:I think we can dispel Ghostly Emergence from ma-chines . . . by saying that when parts are combinedinto more complicated ... structures, these struc-tures have different interfaces with each other andwith their local environment, and so when com-bined they display different properties from the iso-lated parts, and from the parts combined in differ-ent ways. Thus, for example, two halves ofa wheelseparated are obviously functionally very differentfrom the complete wheel, just because, and with nomystery, there is not a continuous rim for half-wheels, so they will not run on a surface. In otherwords, what emerges is given simply by the differ-ences at interfaces, as the parts are added or re-moved. When the interfaces are different, the be-haviour ofthe machine is correspondingly different.(And it does not matter what is inside the parts, foronly the interfaces are significant.) (p. 87)

Naive reductionism, wherein all phe-nomena at one level of analysis are fullyaccounted for by processes at a subad-jacent level, and so forth, is not a viewheld by any modem perspective onmechanism, or, indeed, any scientific en-deavor. The issues of reductionism, likethose ofmechanism, are very much morecomplex than are usually appreciated (see,e.g., Marr, 1990; Mercer, 1981; Nagel,1979). To invoke, for example, as somecontextualists have, Skinner's (1938)treatment of the "conceptual nervoussystem" and his views on neurologizingas justifying an antimechanistic stance forradical behaviorism is to display a redherring that is, in fact, also a dead her-ring.The root metaphor ofcontextualism as

the "ongoing act in context" (Hayes etal., 1988) tells me little about what con-text means, intuitively or otherwise; in-deed, there seems to be no metaphor thereat all. Morris (in press), however, hasprovided a more useful focus:

the historical context ... establishes what func-tional relationships may occur among stimuli andresponses at present.... The current context, inturn, affects behavior in two ways: It influences

62 M. JACKSON MARR

which of these relationships can and will occur.More specifically, the form of the current context(i.e., organismic and environmental structure) in-fluences whether the historically-established rela-tionships can physically occur (i.e., be enabled), whilethe function of the current context ... influenceswhether the relationships will be instantiated. (em-phasis in original)

"Current context," in the senses thatMorris has given us, has always been anessential part ofthat most lofty represen-tation of mechanism-classical mechan-ics-Newton's "system of the world."Newton's second law of motion, F =m(dv/dt), is a differential equation that,suitably applied, can describe the motionof the moon, the fall of a sparrow, thebounce of a billiard ball, and the beat ofa drum. By itself, however, the law is notonly useless but meaningless. The law iscomprised of three terms-force, mass,and acceleration-no one of which canbe defined independently ofthe other two.Likewise, the elements comprising Skin-ner's three-term contingency cannot standalone in a contextual vacuum. In bothcases, the meaning of these relations isto be found in their application. But thecommonalities do not end there. As adifferential equation, Newton's secondlaw cannot yield useful solutions withoutspecification ofinitial, and, in many cases,boundary conditions. As an example,consider a vibrating drum membrane.This system may be described by a partialdifferential equation whose form per secan encompass an infinite number ofsys-tems and solutions. The equation takeson meaning by assignment of initial con-ditions, for example, the initial form ofdisplacement and its location on themembrane, and by boundary conditions,such as the shape of the drum. Also tobe included are properties of the drummembrane, for example, its "springi-ness." All these specifications dwindle toa precious few an infinity of possibilities.So it is that any use of Skinner's three-term contingency requires clear descrip-tions of the properties of and interrela-tions among responses, consequences,and prevailing stimuli. Otherwise, we areleft chasing our conceptual tails.Although initial and boundary condi-

tions are not part of nature's laws, they

are obviously important variables tostudy because they define and set the lim-its under which a given law will apply.Thus, variables such as species, age, re-inforcement history, deprivation, and soforth may modulate the results of apply-ing a behavioral law, just as the materialof which a drum membrane is made, itsshape, the way it is clamped, and how itis struck will determine the pattern ofvibrations produced. It is important toemphasize that initial and boundary con-ditions are not part of the laws to whichthey apply; rather they provide a frameofreference to assess those laws. They arethe grounds upon which laws of natureare seen as figures. Perhaps this reflectsMorris's assertion that context givesmeaning to behavior. But it could just aswell be said the other way around. Con-text, by itself, is as empty and meaning-less as a frame without a picture.There are, of course, significant sub-

tleties here. One is the boundary betweencontext and the laws describing the eventsunder study. What, at first glance, seemsto be part of the context may disappearwhen a more general law is derived. Forexample, Galileo's law of falling bodiesdoes not take into account the effect ofthe medium through which a body is fall-ing (e.g., air resistance). The law reallyapplies only in a vacuum. Does adding aretarding medium such as the atmo-sphere change the context, or, in Morris'sterms, meaning of the law? No. Both areexpressible by Newton's second law ofmotion. How, then, do we take this newsituation into account? It is dealt with byderiving a new, more general law withthe vacuum as a limiting condition. Aparameter of the general law is changedaccordingly.

Perhaps one of the reasons that con-textualism claims to be the proper"worldview" ofbehavioral analysis is thatboth behavioral laws and their relevantcontexts are only vaguely discerned. Un-der these conditions, the boundaries be-tween the two must be fuzzy indeed. Butever since Skinner (I 938), we have at leastthe possibility of clarifying the bound-aries by distinguishing between prevail-ing contingencies and what Skinner called

CONTEXTUALISM AND MECHANISM 63

"third variables" (e.g., those character-izing motivational conditions). As men-tioned previously, such variables set upor establish conditions within whichfunctional relations may be assessed; indoing so, they constrain behavioral vari-ability. We typically, however, isolatethese conditions from contingency spec-ifications, such as arrangements of mul-tiple or concurrent sources of reinforce-ment for certain classes of emittedbehaviors under certain schedules andstimulus conditions. Given all these ini-tial and boundary conditions, doting ona vague entity like "current context" is asuperfluous and unfocused affectation.Moreover, different fields ofscientific en-deavor focus on different classes of de-pendent, independent, and controlledvariables- one investigator's context isanother's career.

Returning to Morris's dichotomy, whatabout "historical context"? As behavioranalysts are quick to point out, "history,"or, if you like, "historical context" maycarry a considerable explanatory burdenin an interpretive account. Moreover, therole of history is commonly emphasizedas differentiating contextualistic (and or-ganismic) from mechanistic accounts.Newtonian classical mechanics possessesno arrow of time; Newton's world runbackwards would look the same as it doesgoing forwards, a perfect temporal sym-metry. The same may be said of the twogreat pillars of 20th century physics, rel-ativity and quantum mechanics. How-ever, physics has long dealt with systemsin which there is a direction to time, fromthermodynamics and fluid mechanics tomodern nonlinear dynamical systems(see, e.g., Coveney & Highfield, 1990;Gardner, 1990; Gleick, 1987; Nicolis &Prigogine, 1989; Prigogine, 1980; Pri-gogine & Stengers, 1984; Ruelle, 1991).The issues raised by consideration ofwhatmight be called "neoclassical" mechanicsdeserve more space than is reasonable ina presentation of this sort, but suffice itto say, for example, that seemingly "sim-ple" mechanical, electrical, and chemicalsystems are capable ofconsiderable com-plexity-evolving over time in irrevers-ible and unpredictable ways; the slightest

changes in initial conditions or other pa-rameters result in totally different out-comes. Such systems may evolve in wayspictured only by actually letting themevolve; that is, the outcomes and all theintermediate states can be known onlyby seeing them happen, just as in bio-logical evolution. (An expert on evolu-tion, knowing all there was to know aboutthe Jurassic period, could never predictthat something we call a giraffe wouldappear in the Miocene.) Unlike the es-sentially stochastic mechanisms under-lying biological evolution, however, so-called chaotic systems may display anastonishing underlying order expressibleby properties of deterministic (i.e., non-stochastic) equations. The principal pointhere is that immense complexity may beencompassed within a mechanisticframework that is potentially capable ofcapturing even the deep mysteries of be-havior dynamics (e.g., Hoyert, 1992;Marr, 1992; Palya, 1992).

In this context, I feel compelled to ad-dress, at least briefly, the argument thateven physics itselfis not mechanistic, and,in fact, has not been since the 19th cen-tury. Hence, if behavior analysts lookedto physics as a model ofwhat mechanismcan achieve, then they should also lookto what mechanism did not achieve. Ihave already mentioned temporal irre-versibility-the arrow of time-or, as ithas been elegantly expressed by Prigogine(1980), "from being to becoming." Mod-em nonlinear dynamical systems theo-ry-a mechanistic system-deals withphenomena exhibiting "becoming." Butperhaps the most often cited "nonmecha-nistic" aspect ofphysics is a "field." Thelogic of contextualists and other field-theoretic psychologists seems to be this:Physics successfully invokes the fieldconcept, and the field is nonmechanistic;we contextualists invoke an "integratedfield perspective," so we are not mech-anists either. Furthermore, we gain sci-entific respectability by being (or perhapsI should say "becoming") part of the re-treat from mere mechanism. This iden-tification with that most esteemed of sci-ences -physics -might be worthy iftherewere some clear correspondence between

64 M. JACKSON MARR

the term "field" as it functions in physicsand "field" as it appears in, say, contex-tualism. In physics, the concept of fieldhelped deal with uncomfortable feelingsabout action-at-a-distance and, muchmore importantly, provided an accountof properties of, for example, electro-magnetic radiation. Since at least Max-well's time, the term has been essentiallya mathematical abstraction wherein eachpoint in a space is assigned a magnitudeand possibly a direction.On the other hand, I will freely admit

that as much as I have been exposed tothe term "field" in psychology, and re-cently in contextualism, I have never un-derstood it. I have calculated and mea-sured electric and magnetic fields, but Ihaven't a clue about how to do either inthe analysis ofbehavior, or what I wouldhave if I did. Perhaps "field" is simplyone ofthose kinds ofmetaphors we chideour cognitive psychology colleaguesabout. If it is, then it would be a wispycriterion indeed for claiming a nonmech-anistic stance.My inability to appreciate the value of

the term in a behavioral context aside,"field" in physics (e.g., the electromag-netic field) possesses properties that canbe described as "mechanical." That is, ithas velocity, momentum, and energy. Infact, the notion of a field can virtually bedispensed with altogether and replacedby particle interaction, as it is in Feyn-man's (e.g., 1985) formulation of quan-tum electrodynamics. Thus, contextual-ists will have to choose another modelfor their fields. The one in physics is toomechanical. Perhaps they should look toagronomy.The experimental analysis of prevail-

ing stimulus contingencies, the influencesof differential histories, and the modu-lating effects ofwhat Skinner called "thirdvariables" has and continues to be themajor, indeed, the accelerating focus ofinterest in our science. There is nothingin modem behavior analysis to discour-age such efforts. It should be emphasized,too, that, for the most part, the experi-mental analysis of behavior offers themost effective methods and principles forassessing those classes of variables and

more. As I have tried to emphasize, theoverall approach has, and continues tofit into, an already established mecha-nistic framework with a long evolution-ary history (Killeen, 1992; Marr, 1989,1992). This framework has had from thebeginning an inherent "contextualism."Not, to be sure, the vague contextualismof, for example, Kantor and his followers,but one based upon a careful experimen-tal and analytic approach to behavioralphenomena to seek the original goals ofa behavioristic formulation -predictionand control.

REFERENCESAndressen, J. (1991). Recommendationsfrom thefield of linguistics. President's Invited Sympo-sium, Association for Behavior Analysis, Atlan-ta, GA.

Coveney, P., & Highfield, R. (1990). The arrowoftime. New York: Fawcett.

Feynman, R. P. (1985). QED: The strange theoryof light and matter. Princeton, NJ: PrincetonUniversity Press.

Gardner, M. (1990). The new ambidextrous uni-verse (3rd ed.). New York: Freeman.

Gleick, J. (1987). Chaos: Making a new science.New York: Viking.

Gregory, R. L. (1981). Mind in science. Cam-bridge: Cambridge University Press.

Hayes, S. C., Hayes, L. J., & Reese, H. W. (1988).Finding the philosophical core: A review of Ste-phen C. Pepper's World Hypotheses: A Study inEvidence. Journal of the Experimental AnalysisofBehavior, 50, 97-111.

Hoyert, M. (1992). Order and chaos in fixed-in-terval schedules of reinforcement. Journal oftheExperimentalAnalysis ofBehavior, 57, 339-363.

Killeen, P. R. (1992). Mechanics of the animate.Journal oftheExperimentalAnalysis ofBehavior,57, 429-463.

Marr, M. J. (1989). Some remarks on the quan-titative analysis of behavior. The Behavior An-alyst, 12, 143-151.

Marr, M. J. (1990). Behavioral pharmacology: Is-sues of reductionism and causality. In J. E. Bar-rett, T. Thompson, & P. B. Dews (Eds.), Advancesin behavioral pharmacology (Vol. 7, pp. 1-12).Hillsdale, NJ: Erlbaum.

Marr, M. J. (1992). Behavior dynamics: One per-spective. Journal ofthe Experimental Analysis ofBehavior, 57, 249-266.

Mercer, E. H. (1981). Thefoundations of biolog-ical theory. New York: Wiley.

Morris, E. K. (1984). Interbehavioral psychologyand radical behaviorism: Some similarities anddifferences. The Behavior Analyst, 7, 197-204.

Morris, E. K. (1988a). The Behavior of Organ-isms: A context theory of meaning. Symposiumon The Behavior ofOrganisms at fifty. AmericanPsychological Association, Atlanta, GA.

CONTEXTUALISM AND MECHANISM 65

Morris, E. K. (1988b). Contextualism: The worldview of behavior analysis. Journal of Experi-mental Child Psychology, 46, 289-323.

Morris, E. K. (1991). The aim and progress ofbehavior analysis. Presidential Address, Associ-ation for Behavior Analysis, Atlanta, GA.

Morris, E. K. (in press). The Behavior ofOrgan-isms: A context theory ofmeaning. In J. T. Todd& E. K Morris (Eds.), Modern perspectives onclassical and contemporary behaviorism. West-port, CT: Greenwood Press.

Nagel, E. (1979). The structure ofscience. India-napolis, IN: Hackett.

Nicolis, G., & Prigogine, I. (1989). Exploringcomplexity: An introduction. New York: Free-man.

Palya, W. (1992). Dynamics in the fine structureofbehavior. Journal ofthe ExperimentalAnalysisofBehavior, 57, 267-287.

Pepper, S. C. (1942). World hypotheses: A studyin evidence. Berkeley, CA: University of Califor-nia Press.

Prigogine, I. (1980). From being to becoming. SanFrancisco, CA: Freeman.

Prigogine, I., & Stengers, I. (1984). Order out ofchaos. New York: Bantam Books.

Ruelle, D. (1991). Chance and chaos. Princeton,NJ: Princeton University Press.

Skinner, B. F. (1938). Thebehavioroforganisms.New York: Appleton-Century-Crofts.