the hawthorne effect: a reconsideration of the ...hawthorne experiments was the insight they...

12
Journal of Applied Psychology 1984, Vol 69, No 2, 334-345 Copyright 1984 by the American Psychological Association, Inc The Hawthorne Effect: A Reconsideration of the Methodological Artifact John G. Adair University of Manitoba, Winnipeg, Manitoba, Canada The methodological Hawthorne effect, generally denned as the problem in field experiments that subjects' knowledge that they are in an experiment modifies their behavior from what it would have been without the knowledge, originated out of the classic studies at the Hawthorne Works of the Western Electric Company. Closer examination reveals that although it is a phenomenon that should be of considerable concern to researchers, the Hawthorne effect is poorly understood and, with the passage of time, has taken on several different, yet still quite imprecise meanings. The nature of the artifact is inconsistently described in contemporary psychology textbooks, and there is a lack of agreement on how the effect is mediated. A review of controls for Hawthorne effects in current field research, mostly in education, revealed that the controls took several forms, each designed for different purposes. Studies intended to produce Hawthorne effects suggested that the commonly held notion of the artifact that has evolved is not the problem that plagued the Hawthorne researchers. A remterpretation reveals a different methodological concern that re- sembles more recent research on the social psychology of the psychological ex- periment and a need to adopt different methodological control procedures than those normally used to counter the "Hawthorne effect." The Hawthorne studies, conducted more than 50 years ago, are a social science classic. Their major impact has been in industry, spe- cifically in the impetus they provided for con- sideration of psychological and social factors within the workplace. In sharp contrast to the impersonality of industrial work of the time, the studies claimed to have demonstrated that increased personal attention by supervisors led to improved productivity. In industrial soci- ology or psychology, no other theory or set of experiments has stimulated more research and controversy nor contributed more to a change in management thinking than the Hawthorne studies and the human relations movement they spawned. A second significant contribution of the Hawthorne experiments was the insight they provided into the reactivity of human subjects A version of this article was presented as an invited address to the International Association of Applied Psy- chology, Edinburgh, Scotland, July, 1982. The author's research was supported by a grant from the Social Sciences and Humanities Research Council of Canada Requests for reprints should be sent to John G. Adair, Department of Psychology, University of Manitoba, Win- nipeg, Manitoba, Canada R3T 2N2 m experiments. The observation that mea- surements of behavior in a controlled study were altered by subjects' knowledge that they were in an experiment, later to be called the Hawthorne effect, "caused some consternation in both academic and industrial circles" at the time (Dickson & Roethlisberger, 1966, p. 22). Remembering that the studies followed Wat- son's Behavionstic declaration by little more than a decade, that they were contemporary to Thurstone'sfirstmeasurements of attitudes, and that they antedated the leadership and group dynamics research of Lewin, Lippitt, and White, restores some of the pioneering flavor of this observation. In spite of its ap- plicability to research in general, the artifact became specifically identified with obtrusive field experiments and became of particular concern to educational researchers, program evaluators, and others who conduct experi- ments in field settings. As we review its ap- plication in psychology and education, we shall see that the effect is not an artifact to be dis- carded but a substantive concern of signifi- cance to both disciplines. In education the artifact has been uncriti- cally accepted, and special procedures or costly control groups have been introduced as pre- 334

Upload: others

Post on 25-Dec-2019

9 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: The Hawthorne Effect: A Reconsideration of the ...Hawthorne experiments was the insight they provided into the reactivity of human subjects A version of this article was presented

Journal of Applied Psychology1984, Vol 69, No 2, 334-345

Copyright 1984 by theAmerican Psychological Association, Inc

The Hawthorne Effect: A Reconsiderationof the Methodological Artifact

John G. AdairUniversity of Manitoba, Winnipeg, Manitoba, Canada

The methodological Hawthorne effect, generally denned as the problem in fieldexperiments that subjects' knowledge that they are in an experiment modifies theirbehavior from what it would have been without the knowledge, originated out ofthe classic studies at the Hawthorne Works of the Western Electric Company. Closerexamination reveals that although it is a phenomenon that should be of considerableconcern to researchers, the Hawthorne effect is poorly understood and, with thepassage of time, has taken on several different, yet still quite imprecise meanings.The nature of the artifact is inconsistently described in contemporary psychologytextbooks, and there is a lack of agreement on how the effect is mediated. A reviewof controls for Hawthorne effects in current field research, mostly in education,revealed that the controls took several forms, each designed for different purposes.Studies intended to produce Hawthorne effects suggested that the commonly heldnotion of the artifact that has evolved is not the problem that plagued the Hawthorneresearchers. A remterpretation reveals a different methodological concern that re-sembles more recent research on the social psychology of the psychological ex-periment and a need to adopt different methodological control procedures thanthose normally used to counter the "Hawthorne effect."

The Hawthorne studies, conducted morethan 50 years ago, are a social science classic.Their major impact has been in industry, spe-cifically in the impetus they provided for con-sideration of psychological and social factorswithin the workplace. In sharp contrast to theimpersonality of industrial work of the time,the studies claimed to have demonstrated thatincreased personal attention by supervisors ledto improved productivity. In industrial soci-ology or psychology, no other theory or set ofexperiments has stimulated more research andcontroversy nor contributed more to a changein management thinking than the Hawthornestudies and the human relations movementthey spawned.

A second significant contribution of theHawthorne experiments was the insight theyprovided into the reactivity of human subjects

A version of this article was presented as an invitedaddress to the International Association of Applied Psy-chology, Edinburgh, Scotland, July, 1982. The author'sresearch was supported by a grant from the Social Sciencesand Humanities Research Council of Canada

Requests for reprints should be sent to John G. Adair,Department of Psychology, University of Manitoba, Win-nipeg, Manitoba, Canada R3T 2N2

m experiments. The observation that mea-surements of behavior in a controlled studywere altered by subjects' knowledge that theywere in an experiment, later to be called theHawthorne effect, "caused some consternationin both academic and industrial circles" at thetime (Dickson & Roethlisberger, 1966, p. 22).Remembering that the studies followed Wat-son's Behavionstic declaration by little morethan a decade, that they were contemporaryto Thurstone's first measurements of attitudes,and that they antedated the leadership andgroup dynamics research of Lewin, Lippitt,and White, restores some of the pioneeringflavor of this observation. In spite of its ap-plicability to research in general, the artifactbecame specifically identified with obtrusivefield experiments and became of particularconcern to educational researchers, programevaluators, and others who conduct experi-ments in field settings. As we review its ap-plication in psychology and education, we shallsee that the effect is not an artifact to be dis-carded but a substantive concern of signifi-cance to both disciplines.

In education the artifact has been uncriti-cally accepted, and special procedures or costlycontrol groups have been introduced as pre-

334

Page 2: The Hawthorne Effect: A Reconsideration of the ...Hawthorne experiments was the insight they provided into the reactivity of human subjects A version of this article was presented

THE HAWTHORNE EFFECT 335

cautions against its effects. In curriculum re-search, the artifact has been reined into a pre-cise effect that is even anticipated to occur inmany studies. One third to one half of subjectsm this kind of research are run as Hawthornecontrols. Considered in virtually every edu-cational research methods text, students of ed-ucation are led to expect that "many . . . ex-periments report changes and improvementsthat are due to the Hawthorne effect" (Borg& Gall, 1979) and that such biases may extendto "fully half the classroom study conclusions"(Cook, 1962).

Although less prominent in psychology,Hawthorne has taken on renewed importancewith increased emphasis on field experimen-tation as an alternative to the laboratory. Fieldresearch has been viewed by some as a panaceato the problem-plagued psychological labo-ratory, and its use has noticeably increasedover the past decade. Critics of field researchhave focused on problems of control, whereasreactivity of subjects to field studies, in contrastto the concern in education, has been over-looked or ignored in psychology. A closer ex-amination of the research within another dis-cipline that has taken the Hawthorne effectseriously may shed light on similar problemswithin psychology.

Although the foregoing sounds delightfullylogical, I must confess that my own interestin the topic began only a couple of years agowhen I was asked to write a somewhat pop-ularized account of the Hawthorne effect foran educational journal. As a social researchmethodologist this seemed like a relativelystraightforward task and, although I had notrecently given the artifact much thought, I feltthat for a popular article all I probably needdo was to freshen my memory from a coupleof secondary sources and I would be able tocomplete the project in short order. I quicklydiscovered that I was mistaken. In secondarysources I found considerable disagreement onwhat the artifact is. Intrigued and challengedby these circumstances, not to mention havingto write that article, I proceeded on my searchfor the methodological Hawthorne effect. WhatI found was a phenomenon that, although ofboth methodological and substantive impor-tance to investigators, is poorly understoodand generally ignored. It is my search that ledto these conclusions that I wish to describe.

The Hawthorne Effect m ContemporaryResearch Methods Textbooks

I initially turned to research methods text-books for a definition of the Hawthorne effect,an explanation of how it is mediated, and someinsights into how the Hawthorne researcherswere led to their conclusions. What I foundwas appalling. In the first 15 experimentalpsychology textbooks I examined with refer-ence to Hawthorne (Adair, 1981), not one haddescribed the studies accurately, let aloneshown agreement on what constituted the ar-tifact A comparable survey of educational re-search methods texts did not fare any better.Errors ranged, for example, from blurring thedistinction between the manipulated variables,as if levels of illumination, rest periods, andthe like were varied together in a single study,to introducing new variables, which, althoughplausible in today's industry, were not in factmanipulated, such as noise (Runkel &McGrath, 1972) or music (Harrison, 1979).The female workers were described as men inone text (Sax, 1968), and in another the studieswere named after the industrial city in whichthe research must have taken place: Haw-thorne, New Jersey (Arnoult, 1972).

Beyond these objective errors of fact therewas also disagreement in defining the artifact.The majority of the textbooks defined theHawthorne effect as roughly equivalent to"some behavior change as a result of awarenessof being a subject in an experiment." Althoughreflecting some common understanding aboutthe artifact, I found this definition empty ofmeaning; somewhat like one of Murphy'sLaws—it sounds good at first because it anec-dotally fits situations we recall, yet on closerexamination it is meaningless. It gives no in-dication of why the change occurs or whatmediates the effect.

The variables identified by textbook authorsas presumed mediators expanded rather thannarrowed my search. Most often cited werespecial treatment or attention accorded sub-jects by virtue of being in the experimentalgroup, followed closely by change m routineor novelty associated with the experimentalprogram. A number of texts persisted withawareness of being in an experiment as themediator or failed to go beyond the initial def-inition as if it has been sufficient to explain

Page 3: The Hawthorne Effect: A Reconsideration of the ...Hawthorne experiments was the insight they provided into the reactivity of human subjects A version of this article was presented

336 JOHN G. ADAIR

the effect. Other mediators mentioned withreasonable frequency included increased mo-tivation (although how that was to occur variedwidely), prestige from being selected, teamwork, hypothesis awareness, and, to a lesserextent, surveillance, change in task strategies,heightened attention to the task, improvedmorale, and changed interpersonal relations.Each was presented as if it explained the effect,yet how it led to the artifact was generally onlyimplied and certainly never documented. Al-though textbook authors may be faulted forflawed descriptions of the research, pseudo ex-planations of an imprecisely denned artifactbased on nothing but speculation compoundedthe bad scholarship. And these are the textson which the training of future researchers isprimarily based!

What Happened at Hawthorne?

Clearly I had to return to the original Haw-thorne studies to reconsider what the meth-odological artifact had been about. The re-search took place at the Western ElectricCompany's mammoth Hawthorne works lo-cated on the boundary between Chicago andCicero, Illinois. There were four reactive andtwo nonreactive experiments at Hawthorne;the first two reactive studies were of primaryrelevance to the methodological artifact.

The illumination studies The first of these,known as the "Illumination Studies," wereconducted by the National Research Councilbefore the more widely publicized HawthorneStudies began. Although among the first studiesto signal the importance of subject factors inbehavioral research, the data were never for-mally reported, and the best available descrip-tion is a 3-page summary in Roethlisbergerand Dickson (1939).

Begun in November of 1924, it was notuntil the third illumination study, later in theproject, that the first glimpse of the importantrole of human factors was provided. Placed mseparate enclosures, control subjects receiveda constant illumination of 10 foot-candleswhile the illumination for experimental sub-jects, begun at this level, was decreased in1-foot-candle steps in successive work periods.Throughout the experiment, both sets of sub-jects increased their performance, slowly butsteadily. It was not until illumination in the

experimental room reached 3 foot-candles thatsubjects started to complain that they couldhardly see what they were doing, and pro-ductivity finally started to decline. Somethingother than level of illumination was affectingproductivity.

Subsequent informal experimentationclearly indicated the contribution of psycho-logical factors. In one instance the illuminationwas reduced to .06 of a foot candle, the levelof ordinary moonlight, and yet efficiency wasmaintained! Following these studies, as Dick-son and Roethlisberger (1966) later were tonote, some experimenters became curious asto why they should have had the effect theydid upon their subjects, yet "little was said;just a few eyebrows were raised" (Dickson &Roethlisberger, 1966, p. 21).

Relay Assembly Test Room Study. TheHawthorne Effect probably owes its existenceto a second experiment, called the Relay As-sembly Test Room Study. Five female em-ployees who spent each work day assemblingrelays, were separated from their large de-partment and placed into a special test roomwhere all relevant variables could be bettercontrolled or evaluated. The study was de-signed to explore the optimal cycle of rest andwork periods. However, to make the subjectsindependent of influences from their formerdepartment and sensitive to experimental ma-nipulations, the investigators began by chang-ing the method of determining wages. Duringthe experiment the investigators also manip-ulated, on different occasions and sometimesconcurrently, the length and timing of rest pe-riods, the length of the work week, the lengthof the work day, and whether or not the com-pany provided lunch and/or beverage. Pro-ductivity seemed to increase regardless of themanipulation introduced. Finally, well into thesecond year the investigators decided to dis-continue all treatments and to return theworkers to full work days and weeks withoutbreaks or lunches. Unexpectedly, rather thandropping to preexperiment levels, productivitywas maintained. Obviously the worker's be-havior was influenced by the effects of someother variable that the investigators had un-intentionally manipulated.

Before the experiment had begun, the in-vestigators had feared that workers taken fromtheir regular work to be placed in a test room

Page 4: The Hawthorne Effect: A Reconsideration of the ...Hawthorne experiments was the insight they provided into the reactivity of human subjects A version of this article was presented

THE HAWTHORNE EFFECT 337

would be negative and resistant to the exper-iment. To overcome this anticipated negativeset supervision was removed, special privilegeswere allowed, and considerable interest andattention was expressed toward the worker, allchanges intended to provide for a controlledexperiment. However, it was these uninten-tional manipulations, researchers were forcedto conclude, that had caused the subjects toimprove their overall productivity and that hadgiven birth to the Hawthorne effect.

The methodological Hawthorne effect Theterm Hawthorne effect was not coined by anyof the original investigators (White, 1978), al-though Roethhsberger had talked about theresults as being due to these inadvertentchanges made just to create a controlled study.As near as I have been able to determine, theterm was first applied to the methodologicalartifact problem in 1950 by French. Its wide-spread acceptance undoubtedly stems from hischapter m the well-known Festinger and Katzmethodology text in 1953. French pointedspecifically to the influence in an obtrusivefield experiment of "the 'artificial' social as-pects of the experimental conditions set upfor measurement" (1953, p. 101). Exactly howthis process was to occur was not explainednor was it clear which of the Hawthorne studieswas responsible for the artifact.1

For the most part, the confusion was dueto the studies themselves. They contained somany uncontrolled variables that it becamevirtually impossible to identify any causativerelationships.2 In defense of their conclusions,Dickson and Roethlisberger, writing in 1966,conceded that there could have been as manyas 17 different possible mediators and inter-related these to show that they all had to worktogether to produce the effect. Such a broad-brush interpretation may have been sufficientfor purposes of initiating the human relationsmovement, but it certainly was not for myattempt to identify the precise nature of themethodological artifact. Although it did notseem possible to identify its precise naturefrom within Hawthorne, the studies had none-theless provided a stimulus for a concern withthe problem of reactivity m field experimentsand for controls to be routinely applied meducation. It was to these Hawthorne controlsthat I next turned in my search for the natureof the reactivity problem in the field.

Controls for Hawthorne Effects

With considerable difficulty I managed toidentify 40 studies that had explicitly employedcontrols for Hawthorne effects, including thosereviewed previously (Cook, 1967). These stud-ies, virtually all in education, were classifiedand analyzed according to the type of controlprocedure, the particular Hawthorne variablecontrolled, and whether or not the study ob-tained significant treatment effects. Aside fromthe tendency for Hawthorne effects to beclaimed only in designs with multiple controlgroups, no other systematic relationships wereobserved. As can be seen in Table 1, the Haw-thorne variables for which the controls wereemployed were special attention, awareness ofexperiment participation, and, to a much lesserextent, novelty. This narrowed considerablythe variables that were perceived to be re-sponsible. However, the control procedure foreach was quite different. Hawthorne controlsfor awareness were typically led to believe thatthey were in an experimental group or pro-gram, whereas controls for special attentionor novelty were usually given a placebo activitythat presumably had no known effects on be-havior related to the experimental treatment.The information or placebo activities were de-signed to equate the feelings or experienceswithin the study of the experimental and con-trol subjects. Less frequently used alternativecontrol procedures were unobtrusive experi-

1 French attributed the effect to an experiment on a"small group of girls wiring relays" in a "bank-wiringroom separated from the rest of the factory" The problemwith this account is that the Bank-Wiring Room was anobservational study of a group of male employees; theRelay Assembly Test Room study was an experiment ona small group of women who assembled relays with ma-chine screws, not wiring Although these inaccuracies ukelydid not contribute to the later confusions surrounding theartifact, they set the trend for those that followed

2 Together with problems of design and research practice,the studies became target for critiques, reanalyses, andreinterpretations Some have attempted to sort out alter-native angle-variable explanations, such as Parson's (1974)operant analysis attributing effects to performance feed-back, although most critics have pointed to the impos-sibility of inferring any valid conclusions from studieswith such inadequate design and procedures (Argyle, 1953,Carey, 1967) Recent application of contemporary statis-tical techniques to the original data has yielded nonsig-nificant results and additional grounds for rejecting thestudies' conclusions (Franke & Kaul, 1978; Franke, 1979).

Page 5: The Hawthorne Effect: A Reconsideration of the ...Hawthorne experiments was the insight they provided into the reactivity of human subjects A version of this article was presented

338 JOHN G ADAIR

ments for problems of awareness and measuresof whether the treatment effects persisted ordissipated over time for novelty effects.

Of the 40 studies, 7 made some claim ofHawthorne effects. Four of these controlledfor attention. Two controlled for awareness,and one controlled for novelty. Although itappears from this number that effects wereobtained at beyond the chance level, closerexamination of these studies indicated as muchabout the tendency of educational researchersto cry Hawthorne as it shed light on the natureof the artifact. For example, in one study ofnonsignificant results (Sabatino, Ysseldyke, &Woolston, 1973), it was speculated that theslight trend toward performance gains of ex-perimental and control subjects might be dueto Hawthorne effects, even though there wasno empirical evidence to substantiate theclaim; whereas in other studies (Fisher, 1971;Levison, 1971) no Hawthorne effects were ob-

tained on predicted dependent variables, yetthe artifact was invoked to explain unexpectedfindings on quite secondary measures. In acouple of studies (Corder, 1966; Kleinberg,1971), the pattern of Hawthorne control resultswas unclear, and Hawthorne explanations mayor may not have been appropriate.

In two studies (McCracken, 1968; Pella,Stanley, Wedemeyer, & Wittich, 1962), thereseemed to be evidence for a reverse Hawthorneeffect. In the McCracken study, for example,the control teacher, visibly bothered by theapparently greater success of the experimentalgroup, tried extra hard to teach so as to showthat she was as good as those selected for theexperimental class. This reaction has been ob-served in other studies (Zdep & Irvine, 1970).Rather than being incorporated into the Haw-thorne literature, this has become known ineducation as the "John Henry Effect." JohnHenry was the legendary railroad worker who

Table 1Educational Studies Employing Controls for Hawthorne Effects"

Hawthorne variable controlledType of control

procedure Attention Awareness Novelty

Multiple groups Lang & Tohtz, 1964*Corder, 1966

McCormick, Schnobnch, Foothk& Poetker, 1968

•Fisher, 1971•Levinson, 1971•Sabatino, Ysseldyke, & Woolston,

1973Chasey, Swartz, & Chasey, 1974Lowe & McLaughhn, 1974Gutkin, 1978

Payette, 1961Cooper, 1964

•Pella, Stanley, Wedemeyer,& Wittich, 1962

•McCracken, 1968Haddad, Nation, &

Williams, 1975Bell, 1976Campbell, 1978

Stude, 1973Potts & Leyman, 1977

•Klemberg, 1971

Single group Crosby, Fremont, & Mitzel, 1958 Sjogren, undatedRimoldi et a l , 1962Lambert et a l , 1964Downing, 1964Perhsh, 1968Serwer, Shapiro & Shapiro, 1970Bauer & Soar, 1971Bonfield, 1972Axelrod, 1973Hudgins, 1979

Hershberger & Terry, 1963Keating, 1963Ryan, 1965Higgins & Rush, 1965Good & Grouws, 1979

Other procedures Coone & White, 1968Gallegos & Phelan, 1977

Porter, 1961Popham, 1962Benedetto & Jacobs, 1978

Note Asterisked articles indicate studies that claim to have found Hawthorne effects* The complete list of references for these studies may be obtained from the author

Page 6: The Hawthorne Effect: A Reconsideration of the ...Hawthorne experiments was the insight they provided into the reactivity of human subjects A version of this article was presented

THE HAWTHORNE EFFECT 339

attempted to outperform new railroad spike-driving machinery. It would seem more par-simonious, however, to consider that the oc-casion and direction in which the Hawthorneeffect operates might be cognitively based onsubjects' perceptions of the experiment andthat the so-called John Henry effect is merelya manifestation of that process. This hypoth-esis, about the only insight derived from mysurvey of control studies, held up well as mysearch turned to studies that had been designedto produce Hawthorne effects.

Studies Designed to ProduceHawthorne Effects

Although their quality was not uniformlyhigh, I felt that a review of these studies, espe-cially of systematic differences between studiesthat intentionally produced the effect and thosethat were unsuccessful, might provide insightinto the nature of the artifact. Of the 13 studiesI was able to identify, 9 failed to show theeffect and 4 did. The differences between thetwo sets of studies, outlined in Table 2, wereclear and revealing.

Examining these studies, I saw that therewas clearly insufficient evidence to support aninterpretation of the Hawthorne effect as dueto any of the three variables for which controlswere typically employed, for example, novelty,special attention, or mere awareness of par-ticipation Alternatively, there was evidence(Hanson, 1967; Johnson & Foley, 1969) tosuggest that reactivity to experimentation oc-curred through subjects' identifying a purposefor the experimental treatment and an expec-tation for their own behavior in the researchsituation. For example, Hanson and Walker(reported m Hanson, 1967) found that fresh-man males who met three times weekly as anexperimental group to receive and discuss theapparent effects of a fructose placebo showedsignificantly greater gains in bench step en-durance than a baseline control, a differencethat they attributed to the Hawthorne effect.

The experiment conducted by Johnson andFoley (1969) provides the clearest insight intothe artifact. They found that students, removedfrom class and told that an experimentalagenda discussion method for studying Intro-ductory Psychology had been proven effective,scored significantly higher on the number of

questions discussed, on the rated degree ofcompleteness of the discussion, on achieve-ment and on rated satisfaction than either anexperimental class not given the positive ex-pectation or control students who completedthe task as a time-filling exercise in the class-room. The fact that there were no significantdifferences between the latter two groups onmost of these measures, especially achieve-ment, indicates that merely being a subject inan experiment was not sufficient to replicatethe Hawthorne effect.

Moreover, all studies that showed Haw-thorne effects generally tested new knowledgeacquired during the course of the experimentor behaviors that were under cognitive-mo-tivational control. Simpson (1978) found thatgraduate students in English showed improvedrecall of paired associates when the learningtask was conducted in another room as anexperiment than as part of a class exercise.However, neither behavioral science students,nor undergraduate students in English showedthe same effect Dignan (1979) found that aclass in health education following ostensiblyexperimental instruction on the use of con-traceptive devices had shown greater knowl-edge than controls on an end of unit test. Inshort, it appeared from these studies that thecontext of an experiment provides cognitivelycapable adult subjects with the necessity todevelop a purpose or meaning for the situationthat they then use to guide their behavior.

Differences between studies that did and didnot produce the Hawthorne effect were com-patible with this interpretation. Fust, all studiesthat had produced Hawthorne effects em-ployed adult subjects, whereas studies that didnot, with the sole exception of Diamond's(1974), employed child subjects. Second, noneof the dependent variables in the studies withHawthorne effects involved measurement ofwell-practiced skills, such as reading, writing,and arithmetic, that characterized virtually allof the studies without an effect. Undoubtedlyclear performance expectations and heightenedmotivation accompanying the testing situa-tions in these latter studies led to ceilings inperformance for both experimental and con-trol groups that could not be altered by anypurported Hawthorne manipulation.

Evidence from research in industrial andsocial psychology There were a couple of ex-

Page 7: The Hawthorne Effect: A Reconsideration of the ...Hawthorne experiments was the insight they provided into the reactivity of human subjects A version of this article was presented

Table 2Studies Designed to Produce Methodological Hawthorne Effects

Study

Goodwin, 1966

Cook, 1967

Welch & Walberg, 1970

Bauernfeind & Olson,1973

Diamond, Study 1, 1974

Diamond, Study 2, 1974

Rubeck, 1975

Flohr, 1977

Schneiderman, 1977

Knowledge ofparticipation

j

j

j

•/

•/

•/

•j

Hawthorne manipulations

Observeror novelperson

j

•j

•J

Other

No Hawthorne effect

Advanced notice of testing

Novel procedures

Pretesting

Early vs late knowledge of experimentparticipation

Participative decision making, interviewing,and extra experimental materials

Time limit

Novel procedures

Encouragement of success

Child

Child

Child

Child

College

College

Child

Child

Child

Subjects Dependent measure

Arithmetic achievement test

Mathematics achievement test

Physics achievement test

Reading and arithmetic test

Speed reading andcomprehension

Speed reading andcomprehension

Reading test

Arithmetic achievement test

Achievement test

JOH

N G

. A

I

Hanson, 1967

Johnson & Foley, 1969

Simpson, 1978

Hawthorne effect

Fructose placebo, group discussion andinteraction

Expectation of success

Selection of subjects to whom task wouldbe novel

College freshman Bench stepping endurance

College psychology Knowledge of psychology

Graduate and undergraduates Paired associate learning

Dignan, 1979 Unmarried undergraduates Knowledge of contraception

Page 8: The Hawthorne Effect: A Reconsideration of the ...Hawthorne experiments was the insight they provided into the reactivity of human subjects A version of this article was presented

THE HAWTHORNE EFFECT 341

periments in industrial psychology againstwhich to compare these interpretations. Rosenand Sales (1966) assessed productivity levelsin a furniture manufacturing plant before,during, and after the entry of behavioral re-searchers. Consistent with the previous review,reactions to the researchers were dependenton the perspective of the worker: Because theresearch was generally perceived as supportiveof management, older employees, those notactive m the union, and those with a ruralbackground showed increased productivity,whereas the productivity of younger, union-active, and urban employees tended to declineduring the research.

In the same factory, Rosen (1970) laterdemonstrated that changes of foremen matter-of-factly introduced by management led totemporary, indiscnminant increases in pro-ductivity even though some were consideredby the workers as changes for the better andsome were considered undesirable. Ten weekslater after behavioral researchers had assessedgroup attraction, preferences for foreman, andthe like, productivity changed precisely in ac-cord with the now obvious research hypothesis,that is, dramatic increases were obtained withdesirable foremen and declines were obtainedwith undesirable foremen. According to Rosen,the intended effects were conveyed throughthe rating scales, and the workers had com-plied. These industrial studies had reaffirmedthe importance of the meaning of the situationto the subject.

Yet another study relevant to this topic wasSilverman's (1968) experiment on reactivityeffects in studies of attitude change. Althoughnot described as a field study of the Hawthorneeffect, his design was virtually identical to thosedescribed previously. In a classroom setting,experimental and control subjects were givena persuasive communication headed "Reportof the Students Participating in Education."The report advocated the use of closed-circuittelevision tape to present lectures in largeclasses. This message was presented by thesame female assistant, introduced either as arepresentative of a student group wanting toobtain the views of fellow students or, in theexperimental condition, as Silverman's re-search assistant who presented the persuasivemessage as an experiment. Significantly greateracquiescence to the persuasive message was

found when it was presented in the context ofa psychological experiment than when it wasnot. Presumably the subjects' interpretationof the situation varied as a function of whetheror not they perceived it to be an experiment.

Meaning of the Situation to SubjectsUnfortunately the importance of subjects'

expectations and thoughts about the experi-ment have never been made as salient m dis-cussions of the Hawthorne effect as mere par-ticipation in an experiment. Yet if you returnto the original you will find that it was focalto Roethlisberger and Dickson's concerns. Re-peatedly they emphasized the importance ofthe subjects' view of "what is happeningabout him" and that "the importance of achange in a worker's situation, however insig-nificant it may appear to an outsider, can onlybe judged m terms of the meaning attachedto it by the employee" (Roethlisberger &Dickson, 1939, p. 387). In the Relay AssemblyTest Room Study "the experimentally intro-duced changes . . . had proved to be carriersof social meaning" (p. 184). In particular, thesubjects' interpretations of the supervisory andother changes that accompanied the test roomexperimentation was a significant determinantof their work performance. However, since theearliest stages of the research the Hawthorneinvestigators had "completely ignored the hu-man meaning of these changes to the peoplewho were subjected to them" (Roethlisberger,1941, p. 11).

But the problem was more than merelyknowing the subjects' view of the experiment.The real difficulty at Hawthorne was that thesubjects' view of the experiment was not thesame as the investigators. As Roethlisberger(1941) noted, "the trouble was not so muchwith the results or with the subjects as it waswith their (the experimenters') notion regard-ing the way their subjects were supposed tobehave" (p. 11). The experimenters had an-ticipated that their subjects would be negativetoward the research and felt that they neededto counter with special efforts to preserve acooperative set. In the process they had sototally altered the social situation of the testroom that "the experiment they had plannedto conduct was quite different from the ex-periment they had actually performed"(Roethhsberger & Dickson, 1939, p. 183).

Page 9: The Hawthorne Effect: A Reconsideration of the ...Hawthorne experiments was the insight they provided into the reactivity of human subjects A version of this article was presented

342 JOHN G. ADAIR

Orne's two-experiment version of demandcharacteristics theory. Viewed in this light theHawthorne effect appears very similar toOrne's (1973) most recent statement of de-mand characteristics theory. According to thisview, in any study there are potentially twoexperiments: the study planned by the inves-tigator and the study perceived and partici-pated in by the subjects. Because it is the sub-jects' view rather than the experimenter's thatwill determine how the subject responds, it isessential for the investigator to understand howthe situation is perceived "in order to draw asensible inference from the subject's respon-ses" (Orne, 1973, p. 158). Unfortunately thiscan be quite difficult to correctly anticipatebecause subjects will see the experiment froman entirely different perspective. Obvious fea-tures of the study may be completely over-looked. Subjects may react to an experimentthat is entirely different from the one the ex-perimenter intended. On the other hand, if theinvestigator had queried the meaning of thesituation to the subjects, as was eventually doneat Hawthorne, the conclusions that might havebeen drawn would have to be reconsidered inlight of the plausible alternative interpretationfrom the subject's perceptions.

Orne's conceptualization assumes that de-mand characteristics in experiments, whetherin the laboratory or in field settings, are socompelling that a substantial number of sub-jects will systematically perceive the experi-ment similarly. There may also be individu-alistic interpretations, however, this is not theartifactual problem that Orne has identified.3

Rather the more typical situation occurs whena sizeable number of susbjects perceive andinterpret the experiment in a comparablefashion. Results are so systematic that theyare often mistakenly interpreted as evidenceof a treatment effect. In such circumstances,artifact results and erroneous conclusions willbe drawn. The subjects' experiment then be-comes the rival or alternative hypothesis that,if the experimenter were aware of it, shouldbe considered in place of the hypothesizedtreatment effect.

There are numerous illustrations of how re-search conclusions may be dramatically alteredby a two-experiment analysis. Orne and Hol-land (1968), for example, reinterpret Mil-gram's obedience research from the perspective

of subjects' view of that procedure. The Nisbettand Wilson (1977) research used to demon-strate their thesis that subjects cannot accessand veridically report on their cognitive pro-cesses is interpreted differently when it is sub-jected to a two-experiment analysis (Adair &Spinner, 1981). The two-experiment problemhas frequently plagued subject-role research(Adair & Schachter, 1972; Fillenbaum & Frey,1970; Sigall, Aronson, & Van Hoose, 1970;Spinner, Adair, & Barnes, 1977) and two recentstudies in that tradition may serve as an il-lustration. Carlston and Cohen (1980) con-cluded that because uninstructed control sub-jects who were told the hypothesis did notbehave comparably to subjects instructed toplay the role of a "good" subject in an ex-periment, that bias in experiments from sub-ject roles was likely a less significant problemthan we had been led to believe in previousresearch. Through questioning of subjects'perceptions of a replication of their experiment(Carlopio, Adair, Lindsay, & Spinner, 1983),it was found that some did not believe thehypothesis they had been told and that thedata of disbelieving subjects were inconsistentwith the experimenters' conclusions. When theexperiment was repeated without blatantlydisclosing the hypothesis, uninstructed con-trols and subjects instructed to play the roleof good subjects m experiments behaved sim-ilarly.

Subjects' understandings of the specifictreatments may be as important as their viewof the meaning of the experiment as a whole.For example, Nicholson and Wright (1977)demonstrated, through questioning subjects'perceptions of aspects of the New Jersey-Pennsylvania Negative Tax Experiment, thatthe negative income tax treatment was notwell understood by participants. Previousstudies that did not take into account this poorlevel of understanding of the treatments gen-erally underestimated work-disincentive ef-

3 If the experimental situation is so ambiguous that anumber of cues are available to subjects, multiple inter-pretations of the experimental situation would result. Avariety of individualistic interpretations would have thenet effect of adding to the diversity of subjects' responses,thereby increasing the error variance within the experimentand making it exceedingly difficult if not impossible todemonstrate any treatment effect

Page 10: The Hawthorne Effect: A Reconsideration of the ...Hawthorne experiments was the insight they provided into the reactivity of human subjects A version of this article was presented

THE HAWTHORNE EFFECT 343

fects. This kind of problem is serious for manylarge-scale experiments evaluating contem-porary social policy programs on housing, ed-ucation, or health insurance. Often they havecomplicated treatments that may be poorlyunderstood or differentially perceived by par-ticipants. This could result in what Cook andCampbell (1975) have called a nonstandard-lzed treatment implementation, yet it is alsoanother instance in which assessment of par-ticipants' understanding of the experimentaltreatments must be made.

These examples suggest that the phenom-enon Orne and the Hawthorne researchersidentified may not be limited to experimentalbehavior. Most persons in any clearly identifiedsituation define the context for their behaviorand respond accordingly. Their behavior takeson meaning when it is viewed in terms of their"definition of the situation" or their "situatedidentity" (Alexander & Knight, 1971). Thissociological approach, called symbolic inter-actiomsm (Mead, 1934), views individual be-havior from the subjective interpretation ormeaning of the context in which it occurs.From this perspective the Hawthorne effect,previously regarded as artifact to be controlledor removed from the experiment, may take onnew meaning. Rather than being an artifact,it may form the substance of behavior thatdirects responses and thereby may demand amore phenomenological approach than theprevailing behavoristic or objectivist philos-ophy has encouraged (Adair, 1983).

It should also be apparent that the necessityto ascertain the subjects' view of the experi-ment requires strikingly different proceduresthan those typically used to control for Haw-thorne effects in the past. Some means of ac-cessing subjects' understanding of experiments,such as postexperimental interviews or ques-tionnaires, must be routinely included. I amnot referring to simple rating scales or to whatcommonly pass for manipulation checks;rather the questions must be more probingsearches of the subjects' perceptions and ofthe understandings and meanings they applyto experimental situations. In spite of criti-cisms and challenges to the ability of subjectsto access their cognitions and to report verid-ically (Nisbett & Wilson, 1978), there is sub-stantial evidence, which I have summarizedelsewhere (Adair & Spinner, 1981), that post-

experimental questionnaires will yield verid-ical and useful reports under proper conditions.Proper conditions include appropriate lengthand depth of inquiry, careful ordering andphrasing of questions, and instructions to mo-tivate subjects and to provide them with a bet-ter set for introspective reporting. Throughresearch on awareness in verbal conditioning,for example, these methods have been devel-oped and are available.

It may well be that there remain method-ological problems of novelty and special at-tention accompanying experimental curricu-lum changes, for which appropriate controlsneed to be introduced. However, the presentreview did not provide evidence to substantiateroutine inclusion of the procedures that havebeen used to date. On the other hand, sensitiveassessment of the subjects' view of the exper-iment would appear to be a profitable meth-odological strategy. Yet in laboratory experi-mentation, where the reactivity problem hasbeen clearly identified and recognized for sometime, such controls have been slow to beadopted (Suls & Gastorf, 1980). Many re-searchers, trained from an almost opposingframework, find it difficult to view their ex-periment from another perspective and arethus unable or unwilling to apply the phe-nomenologically oriented two-experiment ap-proach. Better articulation of how to adaptpostexperimental questioning procedures to adiversity of experimental settings is clearlyneeded.

Before concluding, let us return briefly toHawthorne and a last look at how those re-searchers, without the guidance from past ex-perience or this review, coped with the artifactthey created.When a study based upon laboratory method fails, orpartially fails, because some essential factor has been un-knowingly and arbitrarily excluded, the investigator, if heis wise, returns to clinical study of the entire situation toget some hint as to the nature of the excluded determinant.The members of the research division at Hawthorne, afterthe twelfth experimental period in the test room, werefaced by just such a situation and knew it (Mayo, 1945,P 76)

The "clinical study" they returned to was tointerview the workers! In many respects after50 years it does not seem that our knowledgeof the reactivity problem and our ability tocope with it have advanced to a higher levelof sophistication. Yet, if we were wise. . . .

Page 11: The Hawthorne Effect: A Reconsideration of the ...Hawthorne experiments was the insight they provided into the reactivity of human subjects A version of this article was presented

344 JOHN G. ADAIR

References

Adair,J G (1981, November) The Hawthorne effect Aremterpretation of the methodological artifact Paperpresented at the meeting of the Society for the SocialStudies of Science, Atlanta

Adair, J G (July, 1983) Phenomenological aspects of re-search In E Sanchez (Chair), Current trends in appliedresearch XIX Interamencan Congress of Psychology,Quito, Ecuador

Adair, J G., & Schachter, B S. (1972) To cooperate orto look good1? Subjects' and experimenters' perceptionsof each others' intentions Journal of Experimental So-cial Psychology, 8, 74-85

Adair, J G , & Spinner, B (1981) Subjects' access tocognitive processes Demand characteristics and verbalreport Journal for the Theory of Social Behavior, 11,31-52

Alexander, C N Jr, & Knight, G W (1971) Situatedidentities and social psychological experimentation So-cwmetry, 34, 65-82

Argyle, M (1953) The Relay Assembly Test Room inretrospect Occupational Psychology, 27, 98-103

Arnoult, M D (1972) Fundamentals of scientific methodin psychology Dubuque, IA Brown

Bauernfemd,R.H,& Olson, C J (1973) Is the Hawthorneeffect in educational experiments a chimera'' Phi DeltaKappan, 55, 271-273

Borg, W R , & Gall, M. D (1979) Educational researchAn introduction (3rd ed) New "Vbrk. Longman

Carlopio, J., Adair, J G , Lindsay, R C L , & Spinner,B (1983) Avoiding artifact in the search for bias Theimportance of assessing subjects' perceptions of the ex-periment. Journal of Personality and Social Psychology,44, 693-701

Carlston, D E., & Cohen, J L (1980) A closer exami-nation of subject roles. Journal of Personality and SocialPsychology, 38, 857-870

Carey, A (1967) The Hawthorne studies A radical crit-icism. American Sociological Review, 32, 403-416

Cook, D L (1962) The Hawthorne effect m educationalresearch. Phi Delta Kappan, 44, 116-122

Cook, D. (1967) The impact of the Hawthorne effect inexperimental designs in educational research (ReportNo 0726) Washington, DC U.S Office of Education

Cook, T D., & Campbell, D T (1975) The design andconduct of quasi experiments and true experiments infield settings In M D Dunnette (Ed), Handbook ofIndustrial and Organizational Research (pp 223-326)New York Rand McNally

Corder, W O. (1966) Effects of physical education on theintellectual, physical, and social development of educablementally retarded boys Exceptional Children, 32, 357-364.

Diamond, S S. (1974) Hawthorne effects Another lookUnpublished manuscript, University of Illinois, ChicagoCircle

Dickson, W J , & Roethlisberger, F J (1966) Counsellingin an organization A sequel to the Hawthorne studiesBoston Harvard University.

Dignan, M B (1979) Hawthorne effect in learning byintact classes Psychological Reports, 44, 1222

FiUenbaum, S., & Frey, R (1970) More on the faithfulbehavior of suspicious subjects Journal of Personality,38, 43-51

Fisher, K L (1971). Effects of perceptual-motor trainingon the educable mentally retarded Exceptional Children,38, 264-266

Flohr, J C (1977) A failure to demonstrate the efficacyof the Hawthorne effect in an educational setting Dis-sertation Abstracts International, 37, 7036-A (UniversityMicrofilms No 77-9663).

Franke, R H (1979) The Hawthorne experiments Re-view American Sociological Review, 44, 861-867

Franke, R H , & Kaul, J D (1978) The Hawthorneexperiments First statistical interpretation AmericanSociological Review, 43, 623-643

French, J R P (1950) Field experiments Changing groupproductivity In J G Miller (Ed), Experiments in socialprocess A symposium on social psychology, (pp 79-96) New York McGraw Hill

French, J R P (1953) Experiments in field settings InL Festinger&D Katz (Eds), Research Methods in theBehavioral Sciences (pp 98-135) New York Holt,Rinehart and Winston

Hanson, D L (1967) Influence of the Hawthorne effectupon physical education research Research Quarterlyof the American Association of Health and Physical Ed-ucation, 38, 723-724

Harrison, N S (1979) Understanding behavioral researchBelmont, CA Wadsworth

Johnson, H H , & Foley, J M (1969) Some effects ofplacebo and experiment conditions in research onmethods of teaching Journal of Educational Psychology,60, 6-10

Kleinberg,N M (1971) Tachistoscopic vs pseudotachis-toscopic training and the Hawthorne effect in improvingreading achievement Dissertation Abstracts Interna-tional, 32, 1339-A, (University Microfilms No 71-24,151)

Levmson, E J (1971) The modification of intelligenceby training in the verbalization of word definitions andsimple concepts Child Development, 42, 1361-1380

Lewm, K , Lippitt, R , & White, R K (1939) Patternsof aggressive behavior in experimentally created socialclimates Journal of Social Psychology, 10, 271-299

Mayo, E (1945) The social problems of an industrialcivilization Cambridge, MA Harvard University Press

McCracken, R A (1968) An observation of Hawthorneeffect in an experiment in the teaching of reading infirst grade A hypothesis InJ A Figurel (Ed), Readingand realism (Vol 13, pp 582-585) Newark, NJ In-ternational Reading Assoc

Mead, G H (1934) Mind, self, and society ChicagoUniversity of Chicago Press

Nicholson, W, & Wright, S R (1977) Participants' un-derstanding of the treatment in policy experimentationEvaluation Quarterly, 2, 245-268

Nisbett, R. E., & Wilson, T. D (1977). Telling more thanwe can know Verbal reports on mental processes Psy-chological Review, 84, 231-259

Orne, M T (1973) Communication by the total exper-imental situation Why is it important, how it is eval-uated, and its significance for the ecological validity offindings In P Phner, L Krames & T Alloway (Eds),Communication and affect (pp 157-191) New 'YorkAcademic Press

Orne, M T, & Holland, C (1968) On the ecological

Page 12: The Hawthorne Effect: A Reconsideration of the ...Hawthorne experiments was the insight they provided into the reactivity of human subjects A version of this article was presented

THE HAWTHORNE EFFECT 345

validity of laboratory deceptions. International Journalof Psychiatry, 6, 282-293.

Parsons, H M (1974) What happened at Hawthorne''Science, 183, 922-932

Pella, M O, Stanley, J., Wedemeyer, C A , & Wittich,W A. (1962) The use of the White films in the teachingof physics Science Education, 46, 6-21.

Roethlisberger, F J (1941) Management and moraleCambridge, MA Harvard University Press

Roethlisberger, F. J , & Dickson, W J. (1939) Managementand the worker New York Wiley

Rosen, N A (1970) Demand characteristics in a fieldexperiment Journal of Applied Psychology, 54, 163—168

Rosen, N A , & Sales, S M (1966) Behavior in a non-expenment Journal of Applied Psychology, 50, 165-171

Rubeck, P A (1975) Hawthorne concept—does it affectreading progress"' The Reading Teacher, 28, 375-379

Runkel,P.J.,&McGrath,J E (1972) Research on humanbehavior A systematic guide to method New York Holt,Rinehart & Winston

Sabatino, D A , Ysseldyke, J E , & Woolston, J. (1973)Diagnostic-prescriptive perceptual training with men-tally retarded children American Journal of Mental De-ficiency, 78, 7-14

Sax, G (1968) Empirical foundations of educational re-search Englewood Cliffs, NJ Prentice-Hall

Schneiderman, M H. (1977) Hawthorne effects in a re-medial program for low achieving and learning disabled

children Dissertation Abstracts International, 37,4121-B (University Microfilms No 77-4175).

Sigall, H , Aronson, E., & Van Hoose, T. (1970) The co-operative subject Myth or reality Journal of Experi-mental Social Psychology, 6, 1-10

Silverman, I (1968) Role-related behavior of subjects inlaboratory studies of attitude change. Journal of Per-sonality and Social Psychology, 8, 343-348.

Simpson, B L (1978) An examination of the Hawthorneeffect in a verbal learning situation in an educationalsetting Dissertation Abstracts International, 38,7242-A (University Microfilms No 78-7846)

Suls, J , & Gastorf, J (1980) Has the social psychologyof the experiment influenced how research is conducted9

European Journal of Social Psychology, 10, 291-294Spinner, B , Adair, J G., & Barnes, G. E (1977) A reex-

amination of the faithful subject role Journal of Ex-perimental Social Psychology, 13, 543-551

Welch, W, & Walberg, H. (1970) Pretest and sensitizationeffects in curriculum evaluation. American EducationalResearch Journal, 7, 605-614

Whyte, W F (1978) Review of the Elusive Phenomenaby F J Roethlisberger Human Organization, 37,412-420

Zdep, S. M , & Irvine, S H (1970) A reverse Hawthorneeffect in educational evaluation Journal of School Psy-chology, 8, 89-95

Received July 11, 1983Revision received December 13, 1983 •