You are on page 1of 13

External Validity Is More Than

Skin Deep
Some Answers to Criticisms of ,
Laboratory Experiments

LEONARD BERKOWITZ University of Wisconsin


EDWARD DONNERSTEIN University of Wisconsin

ABSTRACT: We discuss some criticisms of laboratory miliar with the longstanding debate that has
experiments in psychology, giving special attention to swirled about the topic of experimental validity
the claim that these experiments lack external validity is well aware of the complex issues bound up in
(since it is -widely assumed that ecological validity this matter. Nonetheless, recent discussions of lab-
makes for external validity). We suggest that represen- oratory experiments vis-a-vis naturalistic investi-
tative designs are inadequate for testing causal hy- gations in the areas of industrial-organizational,
potheses, that ecological validity may facilitate the for-
developmental, and social psychology suggest that
mulation of population estimates but is not necessary
for causal hypothesis testing, and that experiments are it is again important to offer a brief defense of
not conducted to establish population estimates. More- experimental research throughout all of psychol-
over, the meaning that subjects assign to the laboratory ogy. This argument will not go into all of the con-
setting and their actions, rather than the laboratory siderations that might be mentioned. We will focus
setting's mundane realism, affects the generalizability on what we regard as some of the chief objections
of the laboratory results. We emphasize, as other writers to laboratory experiments, first examining some of
have, that whether laboratory results are generalizable the general criticisms that have been leveled
to other situations is an empirical question. Research against these studies and then applying our re-
on aggression, especially in regard to the "weapons ef- buttals to the special area of research on aggression.
. feet," is employed to illustrate the possible extension We will argue, first, that the logic of an experiment
of laboratory findings to other, more natural, situations.
does not require that the settings and subjects rep-
resent real-world conditions and, second, that the
Most criticisms of laboratory experiments in psy-
external validity of a study (in the sense of the
chology fault them for their "artificiality" or, in
generalizability of its results) is not necessarily gov-
more sophisticated terms, for their lack of external
erned by physical representativeness.
validity. As everyone knows, the great majority of
psychology's experiments employ a very limited
sample of participants (typically, college students) The Supposed Limited Validity of
placed in a fairly unique setting (a university lab- Experimental Social Psychology
oratory) and usually working on tasks bearing little
resemblance to their everyday activities. Given the Harre and Secord (1972) have provided an excel-
unrepresentativeness of these subjects and situa- lent illustration of the attacks on laboratory ex-
tions, the critics ask, how can the findings be gen- perimentation in their discussion of what they term
eralized to the "real world" of ordinary people the "information-processing model" in experimen-
engaged in their daily lives? In one form or an- tal social psychology. In this paradigm, they con-
other, this question has been raised by professionals
and outsiders, by trained social scientists, and by
Requests for reprints should be sent to Leonard Berkowitz,
persons having only a slight acquaintance with the Department of Psychology, University of Wisconsin, Madison,
field. There can be no easy answer. Anyone fa- Wisconsin 53706.

Vol. 37, No. 3, 245-257 AMERICAN PSYCHOLOGIST • MARCH 1982 • 245


Copyright 1982 by the American" Psychological Association, Inc.
0003-066X/82/3703-0245$00.75
tended, researchers neglect the social contexts ior as well as from their conceptions of the purpose
within which people are embedded in their usual of experiments. At the risk of oversimplifying, it
lives and ask their subjects to react to some stimulus will be helpful to take a brief look at some of their
object on the basis of extremely limited informa- thinking.
tion. Byrne's (1969) well-known studies of inter- Opposing the classic systematic designs (which
personal attraction are held up as particularly hor- control and manipulate a limited set of variables
rible examples of this isolation of subjects from through on,ly a restricted range of values of these
their social surroundings. The experimenter in variables), Egon Brunswik (1955) advocated rep-
these investigations told the subjects about the at- resentative designs employing a representative
titudes supposedly held by a (fictitious) stranger sampling of situations and of subjects. He insisted
and then required the participants to rate their on this "ecological validity" (the notion he intro-
liking for that individual. According to Harre and duced to psychology) because of his probabilistic
Secord view of behavior and especially of perception. For
him the classic procedures exaggerate the impact,
What is glaringly wrong here is the contrast between the
laboratory situation, in which the action of the partici- of proximal cues and unduly minimize the signif-
pants is confined to a series of dispassionate judgments icance of the surrounding distal cues. The role of
based upon written descriptions, and the face-to-face these latter cues could only be assessed adequately,
situation where a person reacts to another person with Brunswik argued, if they were allowed to vary
warmth and friendliness and a feeling of spontaneous naturally. It seems to us that Campbell and Stanley
liking . . . The dissimilarity between the life situation
and the laboratory situation is so marked that the labo- (1963) also adopted a probabilistic position in their
ratory experiment really tells us nothing about the gen- discussion of external "validity. This validity for
esis of liking and friendship among real people, (pp. 51- them was a matter of inductive inference: an in-
52, italics in original) ference could be drawn from a research sample
Here, in extreme form, is the oft-repeated claim to a larger population only to the degree that the
that laboratory experiments lack external validity: given sample was representative of that popula-
Byrne's procedure is supposedly so unrepresenta- tion. Since behavior was affected by a wide variety
tive of the development of liking in naturalistic of factors, proximal and distal, these possible in-
settings that we cannot possibly generalize from fluences ought to be present in the sample to the
this research to a "face-to-face situation where a extent that they occurred in the broader population
person reacts to another person with warmth and to which the researcher .wanted to generalize.
friendliness and a feeling of spontaneous liking." Other critics have based their objections to lab-
Other writers have voiced similar doubts. In his oratory experimentation on their conception of
survey of the history of social psychology, Gordon human nature. They remind us that people seek
Allport (1968) maintained that "many contem- meaning, govern their actions according to social
porary studies seem to shed light on nothing more rules, and strive to fulfill their inherent capacities.
than a narrow phenomenon studied under specific These proclivities are supposedly neglected or even
conditions." Even if the experiment is successfully denied by experimental research. Along with
repeated, he complained, "there is no proof that Harre and Secord (1972) and others, Gadlin and
the discovery has wider validity" (p. 68). Several Ingle (1975) held that the human subject of psy-
books that object to the strong experimental em- chology experiments is manipulated by the exper-
phasis in contemporary social psychology (e.g., imenter much as if he or she were an inert, "ma-
Gilmo.ur & Duck, 1980; Strickland, Aboud, & Ger- nipulable object." The result, according to Schultz
gen, 1976) have now been published. What is es- (1969) and others, is that psychology has developed
pecially disturbing in these attacks is the wide- an inadequate, overly mechanistic image of the
spread equation of experimental value with
ecological validity; the research supposedly is of
1
little benefit unless its conditions and subjects faith- Much of our argument pertains to other areas of psychology
as well as to social psychology. Developmental psychology in
fully mirror "real people in the real world."1 particular has seen a rapidly mounting insistence on the eco-
logical validity of research under the impetus of such influential
The Bases of the Misgivings About writers as Bronfenbrenner (Note 1). We wonder if this insistence
might have become too rigid and perhaps even somewhat dog-
Experimental Validity matic—Theeditor of one of the leading developmental journals,
in discussing criteria for the selection of manuscripts, once
The criticisms of laboratory experiments stem stated, "In the case of laboratory experimental studies . . . the
from the critics' assumptions about human behav- ecological validity of such studies will be a major consideration."

246 • MARCH 1982 • AMERICAN PSYCHOLOGIST


human as a stimulus-response machine. The "lim- between distal or proximal stimuli on one hand
itations of the experimental method," we are told, and the organism's responses on the other on the
yield "misleading findings and one-sided concep- basis of such designs, but cannot draw any definite
tualizations" (Wachtel, 1980, p. 407). conclusions from this research about the processes
that mediate between the external stimuli and the
subsequent responses. The lesson should be clear.
Some Replies If the investigator wishes to estimate the magni-
tude of these correlations in some target popula-
DOES METHODOLOGY DETERMINE CONTENT?
tion, he or she should turn to representative designs
A number of experimentally oriented psychologists high in ecological validity. But if the researcher
have answered these attacks. To cite only a few, is primarily interested in "testing the implication
Kruglanski (1976) insisted that the content of psy- of a theoretical proposition," systematic experi-
chological theorizing is not inevitably affected by ments are desirable (Postman, 1955, pp. 222-223).
experimental methodology. We can see this in the To phrase this somewhat differently, laboratory
type of theorizing that is now dominant in exper- experiments are mainly oriented toward testing
imental social psychology. Contrary to the claim some causal hypothesis (and according to Postman,
that experiments portray people as relatively un- ought to deal chiefly with mediational processes)
thinking, almost robotlike reactors to the events and are not carried out to determine the proba-
impinging on them, the most influential analyses bility that a certain event will occur in a particular
of human behavior in experimental social psy- population.2
chology today emphasize the role of the individ- Kruglanski (1975) offered much the same view
ual's thought processes and assume that he or she as to why experiments are conducted. He main-
is an active seeker after meaning and often tries tained that many objections to laboratory experi-
to control what is happening. One can even go ments actually confuse two very different types
further and argue that experiments give us a truer of psychological research: (a) particularistic in-
image of human complexity than do uncontrolled, quiries in which attention is focused mainly on the
naturalistic investigations. Where the latter can accuracy of statements about specific instances and
only hint at the richness of human nature, the rel- the intended generalizations are restricted in scope,
ative precision of laboratory studies shows us much and (b) universalistic (or theoretically oriented)
more clearly how people can be deliberate and investigations in which the researcher tentatively
thoughtful on some occasions and impulsive, non- claims a universal scope of generality for the find-
thoughtful, and even irrational at other times. Ex- ings "across all sorts and varieties of theoretically
periments do not always picture humans as invol- irrelevant conditions." In a universally oriented
untary reactors to external stimuli by any means, laboratory experiment
but people do respond in such an automatic fashion the hypothesis tested concerns the causal relations among
under some conditions, and this can be revealed general constructs . . . of which the specific manipula-
in well-controlled experiments. More will be said tions and measures constitute arbitrary operational def-
initions, and the subject sample is an arbitrary group
about this later. from the general universe (e.g., the class of all humans)
to which the hypothesis is assumed to apply. (Kruglanski,
THE MAIN PURPOSE OF AN EXPERIMENT 1975, p. 105)

Other psychologists have questioned the probabi-


2
listic approach to the study of human behavior In an interesting article published after this manuscript was
written, Henshel (1980) offered another argument in support
taken by Brunswik and others. Their argument of "artificial" laboratory experiments. Along with several phi-
must be given attention because it essentially holds losophers of science, Henshel began by differentiating between
that the main purpose of experimental research is the objective of verification (which is aimed at the testing of
some hypothesis) and the objective of discovery (which seeks
not necessarily best served by an overriding em- to uncover "new regularities under conditions that do not exist
phasis on representative designs. Postman (1955) anywhere in natural settings" but that are capable of existing).
made this point in a symposium on Brunswik's His thesis is that the experimental control occurring in labo-
ratory studies facilitates such discoveries, and thus artificiality
perspective. He noted that representative designs is no problem for this objective. Henshel points to the research
"necessarily entail the abandonment of classical on biof eedback and the possible acquisition of language by non-
notions of control; we must take the interaction human primates as only two examples of the fruitful pursuit
of discovery by laboratory experiments. These phenomena
among variables as we find it" (p. 222). This means could not have been uncovered if investigators had confined
we can only make statements about correlations themselves to naturally occurring variations in conditions.

AMERICAN PSYCHOLOGIST • MARCH 1982 • 247


The experimentalist pursuing this type of re- that ordinary viewers will behave aggressively af-
search need not be especially worried about the ter they watch a violent TV program in their own
representativeness of the sample and laboratory homes—The settings, subjects, and stimulus ma-
conditions. He or she can have greater confidence terials employed in these experiments are hardly
in the validity of the causal relations that are un- representative of naturalistic TV viewing. What
covered, Kruglanski reminded us, when these re- the studies usually do indicate, however, is that
lationships are "replicated with alternative oper- observed aggression can influence people in the
ational definitions of the same constructs and with audience to display stronger aggression than they
other samples" (p. 105), but the chief question is otherwise might have exhibited. Employing care-
whether the experimental situation "captures the ful, systematic designs, the experiments have dem-
intended essence of the theoretical variables" onstrated such a causal effect far more clearly than
(p. 106). great numbers of field surveys employing repre-
Following the distinction drawn by Carlsmith, sentative designs could. Then too, this argument
Ellsworth, and Aronson (1976), we can say that also maintains that laboratory experiments in the
Kruglanski called for a greater emphasis on ex- area of aggression should not be faulted for their
perimental realism than on mundane realism. supposed failure to copy real-world conditions
Mundane realism exists when the experimental faithfully. Some of the criticisms of this line of
setting and procedure resemble things that occur research maintain that the experimental procedure
in the "real world," that is, when there is a su- is not sufficiently realistic. The critics insist on
perficial appearance of reality. Many experimental mundane realism rather than on experimental re-
social psychologists would rather engage their sub- alism, and they place too much emphasis on rep-
jects fully in the laboratory events than strive for resentativeness and ecological validity.
this surface resemblance to naturalistic settings.
They assume, with some justification, that exper- EMPIRICAL EVIDENCE
imental realism is more apt to "capture the in- REGARDING THE GENERALIZABILITY
tended essence of the theoretical variables" than OF UNREPRESENTATIVE RESEARCH
is a mere mundane realism, and they are more
interested in these variables than on how faithfully Even though laboratory experiments need not be
the laboratory conditions seem to duplicate the preoccupied with attempts to mirror naturalistic
physical and social characteristics of particular conditions, many of their findings can be gener-
real-world settings. Littman (1961) used an inter- alized to other settings. As the critics should know,
esting phrase to make the same observation when many laboratory results have been duplicated by
he said that psychology was a "socially indifferent investigations carried out in a more representative
science." Psychology should be concerned with manner. Take the Byrne research attacked by
psychological processes, Littmari argued, not with Harre and Secord as being unduly "artificial."
specific social content. Its focus should be on the These experiments have shown that attitude sim-
development of universalistic theoretical state- ilarity affects the degree of liking that people have
ments regarding these mediating processes, prop- for each other. Other studies have also found that
ositions that apply to humans everywhere, and attitude similarity predicts the development of lik-
should give little attention to particularistic reports ing among university men residing in a student
of the behavior of certain groups in limited situ- rooming house (Newcomb, 1961) and people kept
ations. Those who would adhere to this general togethervfor 10 days in a small room (Griffitt &
prescription for psychology must necessarily give Veitch, 1974), as well as the interpersonal pref-
greater weight to the adequacy of systematic ex- erences expressed by white and black job appli-
perimental designs than to the representativeness cants who did not know they were participating
of the subjects and settings in the laboratory re- in an experiment (Rokeach, 1968, pp. 68-75). The
search. . same processes that influenced the extent to which
Although such a statement might appear com- Byrne's subjects said they liked a fictitious stranger
monplace when it is phrased abstractly, our own evidently also operate, to a considerable degree at
research interests can help give this point greater least, in a wide variety of more realistic situations
impact. Consider what laboratory experiments on involving face-to-face interaction.
the effects of filmed violence can or cannot ac- Similar observations can be made from other
complish. Our discussion clearly implies that most areas of psychology. Dipboye and Flanagan (1979)
of these investigations do not tell us the likelihood examined almost 200 laboratory experiments and

248 • MARCH 1982 • AMERICAN PSYCHOLOGIST


300 field studies in industrial-organizational psy- unusual in such objective characteristics as age, sex,
chology and concluded that there probably is no education, and class background may arrive at the
difference in the degree to which we can gener- same interpretation of the events around them as
alize from either of these two kinds of investiga- those persons with more ordinary qualities.
tions. Research carried out in a supposedly artifi- This is not to say, of course, that laboratory ex-
cial setting need not have less external validity than periments are necessarily as externally valid as
research conducted under more lifelike circum- field studies conducted in naturalistic settings.
stances. As they commented later, "Too often the Rather, with Dipboye and Flanagan (1979), we
assumption is made that because a study was con- argue only that the generalizability of laboratory
ducted in a field setting, it is inherently more ex- findings in comparison with the results of natu-
ternally valid than a laboratory study" (Dipboye ralistic investigations is an empirical matter. In
& Flanagan, 1980, p. 388). some areas of psychology, such as developmental,
Bass and Firestone (1980) have offered an ex- industrial-organizational, and social psychology,
cellent addendum to this comment. It is important, there is an ever-mounting insistence on field stud-
they noted, to distinguish between the represen- ies as if these investigations are inherently more
tativeness and generalizability of psychological re- valid than laboratory experiments. No such flat
search. Many writers "imply that research which statement is warranted: "There are limits on the
is not representative in terms of subject populations generalizability of all findings, which can be re-
or setting characteristics cannot be generalized" vealed only through systematic testing with dif-
(p. 463), However, Bass and Firestone pointed out, ferent subjects, settings, and responses" (Flanagan
the psychological attributes of the research may & Dipboye, 1980, p. 465).3
affect the generalizability of the results far more We might perhaps forgive statistically oriented
than the study's objective features such as the de- researchers for their automatic equation of exter-
mographic characteristics of the 'sample or the nal validity with demographic representativeness,
physical nature of the situation the subjects are in. but more interesting is the implicit view of ex-
These attributes can include the participants' def- perimental subjects held by many of the other crit-
inition of their task and their interpretations and ics of laboratory experimentation. Although they
judgments. profess a true appreciation of the thinking human
being, in this important respect they actually seem
THE MEANING OF A SITUATION AS A DETERMINANT to give greater weight to physical characteristics
OF THE GENERALIZABILITY OF THE BEHAVIOR than to the subjects' interpretations. They may be
too similar to Watsonian behaviorists in this re-
We have now come to our central thesis: The gard.4
meaning the subjects assign to the situation they
are in and the behavior they are carrying out plays
a greater part in determining the generalizability 3
pf an experiment's outcome than does the sample's Carlsmith et al. (1976) put this very well when they observed
that we cannot be sure that the same variables were operating
demographic representativeness or the setting's in the laboratory setting and a natural situation even when the
surface realism. former is a close analogue of the latter. "One cannot guarantee
Social scientists have long recognized that con- generalizability simply by providing an experiment that has a
high degree of mundane realism. This does not increase our
sistent interpretations can lead to behavioral con- confidence in our ability to generalize from the results, for in
sistency. If an individual interprets two objectively the final analysis the question is an empirical one" (p. 86). These
different situations in much the same manner, he two sentences should be sent to every critic who has faulted an
experiment solely because it seemed to lack ecological validity.
or she will tend to act in the same way on both 4
One of Bronfenbrenner's (Note 1) criticisms of laboratory
occasions. As just one example, Dweck, Goetz, and experimentation is based on the experiment's supposed ambi-
Strauss (1980) recently argued that when individ- guity. Many laboratory studies, he suggested, are so unclear to
the participants that they do not show consistent behavior from
uals make the same causal attributions in seem- one time to another. We contend that such a high-level of
ingly disparate settings, they will generalize their ambiguity is not a necessary property of all experiments. Indeed,
experiences from one situation to the other. We many experimentalists, at least in social psychology, strive to
have their subjects formulate a very clear and definite under-
are simply extending this idea and are noting that standing of the laboratory events. The experimental setting may
two very different persons may act in the same not bear a physical resemblance to a particular natural situation
way because they have essentially the same un- (i.e., it may lack ecological validity), but it can still have a
definite meaning for the participants. In any caSe, laboratory
derstanding of what is going on and what they are behavior need not always have a lower test-retest reliability
doing. Moreover, even people who are relatively than behavior in more naturalistic situations.

AMERICAN PSYCHOLOGIST • MARCH 1982 • 249


Experimental Artifacts great majority of laboratory subjects obviously do
follow the experimental rules spelled out to them.
Some criticisms of laboratory research recognize We cannot be sure, however, that this compliance
the importance of the participants' perceptions but stems from the participants' identification with sci-
contend that subjects frequently do not look at the ence, as Orne claimed, rather than from other
laboratory situation the way the researcher intends. motives such as an inclination to obey the dictates
Indeed, it is sometimes suggested that the subjects of legitimate authority (such as the experimenter),
are apt to view the experiment in such a different the desire for the experimenter's approval or other
manner that their reactions may not be governed benefits that might be obtained, and even intel-
by the variables of interest to the researcher but lectual curiosity regarding the experiment itself.
could actually be produced by other factors. These Whatever the reason for the subjects' acquiescence
factors, which (in the Campbell and Stanley^ 1963, to the explicit requests made of them, this com-
terminology) are said to lessen the experiment's pliance does not in itself necessarily imply that
internal validity, include demand characteristics they will strive to seek out the experiment's "true
and the participant's evaluation apprehension. We purpose" or that they will be especially astute in
will briefly discuss these matters because critics appertaining the hypothesis or that they will be
often refer to these presumed artifacts as reasons particularly intent on confirming the hypothesis.
for dismissing laboratory findings. (For example, assuming they are committed to sci-
ence, could they not try to be "good subjects" by
DEMANP CHARACTERISTICS displaying their true dispositions rather than by
behaving "falsely" in order to do what the re-
Starting with a very influential paper first pub- searcher expected?) In bur view, Orne's conception
lished almost a generation ago, Martin Orne (1962) of demand characteristics is based on a chain of
has advanced a motivational conception that has speculation that does not withstand close exami-
been more widely accepted and exposed to less nation.
critical scrutiny than probably any other motiva- Nor, for that matter, is there much empirical
tional analysis in contemporary psychology. Orne support for Orne's contention that subjects typi-
was struck by the extent to which laboratory sub- cally try hard to confirm the experimenter's hy-
jects are willing to comply with the experimenter's pothesis. According to a number of studies now,
requests: They act in all sorts of seemingly silly or as reviews by Weber and Cook (1972), Kruglanski
meaningless ways and sometimes even agree to do (1975), and Silverman (1977) have shown, there
things that otherwise are repugnant to them is little reason to believe that the subjects' behavior
merely because the experimenter asks them to do in the laboratory is governed by a pervasive and
so. The reason for this acquiescence, according to powerful motive to prove the experimenter's hy-
Orne, is that the subjects "share (with the exper^ pothesis. First of all, our own experience and other
imenter) the hope and expectation that the study researchers' reports indicate that there is not as
in which they are participating will in some ma- much awareness of the research hypothesis in
terial way contribute to science and perhaps ulti- many experiments as the critics have claimed. (Of
mately to human welfare in general" (p. 778). course, the experimenter can produce this aware-
They supposedly go along with the experimenter's ness in his or her subjects at the end of the session
explicit requests because they identify with the by insistently asking them about the deceptions
goals of science and want the experiment to be that had been practiced and what the experiment
successful. Moreover, Orne insisted that out of this was "really" about.) Then tod, in a number of stud-
concern for science, the subjects try to determine ies, when the participants were deliberately made
the true purpose of the experiment so that they •aware of the experiment's purpose at the start of
can respond in a way that will help to validate the the session, many of these people subsequently
experimental hypothesis. They then react to cues "leaned over backwards" to avoid confirming the
in the situation that indicate what the hypothesis supposed research expectation (see Kruglanski,
is—the demand characteristics of the situation— 1975, p. 117). They apparently were more inclined
rather than to the experimental variables manip- to be negatiyistic than conformistic. And further-
ulated by the investigator. more, as Weber and Cook (1972) and Silverman
Future historians of social science surely will (1977) have pointed out, at least some of the pub-
marvel at how this reasoning has been accepted lished demonstrations of subject cooperativeness
in the absence of clear evidence supporting it. The could have grown out of the participants' desire

250 • MARCH 1982 • AMERICAN PSYCHOLOGIST


to present themselves in a healthy light and not Smith, & Brown, 1974). Additionally, a number
from their supposed desire to confirm the re- of psychologists who are ordinarily quite sympa-
searcher's hypothesis. In sum, while some subjects thetic to laboratory research have expressed serious
may occasionally try to live up to the research misgivings about the experiments on aggression.
expectation, as they understand it, we should not They accept many of the pro-experiment argu-
automatically assume that any given laboratory ments that have been offered here in the abstract
finding is necessarily produced by demand char- but seem to regard the standard laboratory studies
acteristics; the motive to confirm the experi- of aggression as beyond the pale and lacking in
menter's hypothesis is probably too rare and too external validity. In the remainder of this article
weak to be counted on as a ready determinant of we will attempt to demonstrate that the points we
the experimental outcome. made earlier regarding the external validity of lab-
oratory experiments in general also apply to many
experiments on aggression in particular.
EVALUATION APPREHENSION

THE CONSTRUCT VALIDITY OF THE


Even though very few subjects have any particular
LABORATORY MEASURES
interest in confirming the investigator's expecta-
tions, most of them probably want to look good to Many of the criticisms of experiments on aggres-
the experimenter. In Rosenberg's (1969) phrase, sion focus on the laboratory measures of aggression.
their behavior is dictated to some degree by eval- These objections generally insist that the measures
uation apprehension. Rosenberg noted that many do not really reflect the subjects' aggressive incli-
people believe psychologists are generally able to nations, either because (a) the participants are only
assess the intelligence, competence, and emotional complying to the demand characteristics in the
adjustment of those with whom they come into situation or (b) the measures do not capture what
contact. Undergraduate students face to face with the critics regard as the most important aspects of
a psychologist in a laboratory experiment might aggressive behavior. Let us consider each of these
well think that they are being-evaluated and, to charges in turn,
alleviate their apprehension, try to impress the Demand characteristics. The procedures fol-
experimenter by appearing smart, sophisticated, lowed in most aggression experiments are now so
or healthy. familiar that few readers of the laboratory reports
We have little doubt that this type of desire is are surprised by the experimental manipulations.
very widespread. The Weber and Cook (1972) sur- The comparatively sophisticated audience is gen-
vey of the research into subject artifacts reported erally aware of what is done in this research and
clear evidence that this motivation is prevalent why. Quite a few members of this audience then
throughout the usual subject population. Moreover, assume that the laboratory subjects also must have
when this motive is pitted against the supposed
caught on to the nature and purpose of the de-
wish to confirm the researcher's hypothesis, most
ceptions that had been practiced. We will not ar-
participants prefer to look good, as Carlsmith et
gue about how many participants are fooled, al-
al. have commented (1976, p. 282).5 This is an
though in our experience there are not as many
important point and we will return to it later in
"aware" subjects in the typical experiment as the
our discussion of the research on aggression. For
critics suppose. (University students often know
our present purposes, however, we will observe
about the possibility of deception in experiments,
only that the subjects' evaluation apprehension
but if the cover story is adequate, they are usually
should diminish as they learn that every experi-
not quite sure what, if any, deceptions might have
menter (or every psychologist for that matter) is
been used in any given study.) What is more im-
not interested in assessing their personality or com-
portant to us here is the inference that is then
petence.
frequently drawn from this supposedly high level

The Validity of Research on Aggression 5


In his summary of the relevant literature, Silyerman (1977)
suggested that demand characteristics are most likely to operate
Recent years have seen a rising tide of doubt re- (if at all) when the subjects are not sure what response will place
garding the validity of the procedures typically them in a favorable light. On the other hand, when evaluation
apprehension and demand cues are in opposition, Silverman
followed in laboratory experiments on aggression concluded that subjects are apt to try to enhance their own
(e.g., Kane, Joseph, & Tedeschi, 1976; Tedeschi, image rather than the experimenter's.

AMERICAN PSYCHOLOGIST * MARCH 1982 • 251


of subject awareness: It is often claimed that once could now believe the notion implanted in them
the participants realize what the experiment is by the questioning because it explained and ex-
really about, they become motivated to act the cused what they had done.
way they think the experimenter desires because If the claims made for demand characteristics
they want to help confirm the experimenter's hy- are somewhat dubious, however, other possible
pothesis. artifacts-such as evaluation apprehension might
This demand characteristics argument can be still operate. Indeed, along with Weber and Cook
readily illustrated by the research on the "weapons (1972) and Carlsmith et.al. (1976), our experience
effect" initiated by Berkowitz and LePage (1967). suggests that many of the students seem to be much
Assuming that weapons have an aggressive mean- more intent on looking "good"-—on appearing
ing for many people in our society, Berkowitz and "healthy" and well-adjusted—than on confirming
LePage hypothesized that the mere sight of weap- the experimenter's hypothesis. This is shown by
ons would function like classically conditioned the Turner and Simons (1974) study of how subject
stimuli to elicit involuntary, aggression-facilitating sophistication and evaluation apprehension com-
reactions. In keeping with their analysis, deliber- bine to affect the participants' reactions to the sight
ately provoked subjects administered more electric of weapons. In this experiment the investigators
shocks to their frustrater when they were in the established different levels of sophistication in their
presence of guns than when only neutral objects subjects, by having a confederate give each naive
were nearby. Other researchers held, however, participant certain information before the session
that the Berkowitz-LePage findings were artifac- began. Cross-cutting this manipulation, some of
tual and arose because of the demand cues in their the men were also induced to have'a high level of
study (e.g., Page & Scheldt, 1971). The experi- evaluation apprehension by telling them the re-
mejntal provocation supposedly informed the sub- search was concerned with psychological adjust-
jects that they were expected to "get even" with ment, while the others were made to be less ap-
their tormentor, and more important, the presence prehensive about the purpose of the study. All of
of the weapons was presumably a clear commu- the subjects were deliberately provoked by a part-
nication telling the participants that they should ner and then each participant had an opportunity
deliver many shocks to the other person. The Ber- to shock this person while two guns were visible
kowitz-LePage subjects had only given the re- on a nearby table. According to Turner and Si-
searchers the kind of behavior they wanted. In mons, both experimental variations had a signifi-
spite of Berkowitz's (1971) attempted rebuttal, it cant effect on the number of shocks that were ad-
now seems to be widely believed that the weapons- ministered. The greater the subjects' awareness
effect studies (and other aggression experiments that the investigator was interested in their re-
as well) are greatly contaminated by the subjects' sponse to the weapons, the fewer the shocks they
compliance to demand cues. gave. Rather than being more compliant to the
As we commented earlier, much of the empir- experimental demands, the sophisticated subjects
ical support for this conception is questionable and tended to hold back on the shocks they delivered.
rests on ambiguous correlational observations. As The reason for this restraint is obvious: Aggression
an example, in the Page and Scheidt (1971) re- is frowned on in our society and the subjects did
search attacking the weapons effect, many of the not want to exhibit "bad" behavior. We can also
participants who had delivered a high number of see this in the evaluation apprehension findings:
shocks in the presence of the guns verbalized the The subjects who were made to be concerned
weapons hypothesis when they were questioned at about their psychological adjustment gave fewer
the end of the session. This correlation of behavior shocks than their less apprehensive counterparts.
with awareness does not prove that the awareness At the very least, these findings (and others that
had preceded and caused the action. Many persons might also be mentioned) clearly indicate that sub-
seek to justify their aggression by inventing suit- ject compliance to demand cues is nowhere near
able reasons after they have attacked their victim as serious a problem in aggression experiments as
(Brock & Pallak, 1969). This justification could some critics have supposed. The available evidence
conceivably have been provided when the exper- suggests that many of the people taking part in
imenter interviewed the subjects at the end of the psychological experiments are inclined to be re-
session, asking them about the possible reasons for strained in attacking their victims if they suspect
the presence of the guns and their suspicions. They that the researcher is interested in their aggressive

252 • MARCH 1982 • AMERICAN PSYCHOLOGIST


behavior, probably because of evaluation appre- dale, 1974). Angered subjects often try to hurt the
hension. This inhibiting influence is yet another one they believe had provoked them. Seeking this
restriction on the representativeness of the labo- injury, moreover, they also are especially gratified
ratory setting, but it also means that whatever high by information about their tormentor's suffering.
levels of aggression are obtained come about in Under suitable conditions, such "pain cues" can
spite of the subjects' awareness and not because stimulate heightened aggression. Thus, according
they know the study's true purpose. to Baron (1977), strongly angered subjects are more
We believe that some of the objections to the punitive than they otherwise would have been on
usual laboratory measures of aggression are also learning that their tormentor is in pain. Since the
unwarranted, The most common criticism raises suitable injury of the tormentor is the provoked
the matter of the supposed unrepresentativeness person's goal, neutral stimuli previously associated
of" the laboratory responses ("Who gives electric with the provocateur's suffering can acquire the
shocks in real life?"), but other critics have ques- capacity to evoke impulsive, aggressive reactions.
tioned the conception of aggression that is implied In an experiment by Swart and Berkowitz (1976),
in these responses. In our view both of these points angered subjects displayed the strongest impulsive
can be answered with essentially the same argu- aggression when they encountered stimuli that had
ment: The essential feature of the laboratory be- earlier been paired with information about their
havior for the subjects, the meaning their actions tormentor's injury. Furthermore, findings pub-
have for them, is that they intentionally are hurting lished by Feshbach, Stiles, and Bitter (1967) in-
their victims. This intentional harm fits what we dicate that this rewarding pain feedback infor-
regard as the best definition of "aggression" and mation can also reinforce nonaggressive responses
also provides the measure with a fair degree of as well. \
external validity. Taken together, these results show that angry
Injury as an aggressive goal. In saying that aggressors are motivated to hurt the person who
aggression should be defined as the deliberate in- had offended them. Their attacks in response to
jury of another we are of course also maintaining the provocation are not only influence attempts,
that the aggressively instigated individual has a nor for that matter, are they only efforts at re-
particular goal (to hurt someone) and is reinforced storing self-esteem. Whatever other outcome may
when this objective is attained. For some investi- also be pursued, these aggressive reactions are at
gators (e.g., Patterson & Cobb, 1973; Tedeschi et least partly oriented toward inflicting injury. We
al., 1974), the aggressors mainly desire to coerce conclude from all this that laboratory measures
their victims. They may hurt their opponents phys- involving the deliberate harm of another individ-
ically or only threaten to do so, insult them ver- ual have construct validity as aggressive responses.
bally, or attempt to deprive them of some valued Correlations of laboratory measures with other
resource. Moreover, although the aggressors may measures of aggression. Further attesting to the
do these things either in the heat of anger or in validity of the laboratory measures of aggression,
an unemotional pursuit of some other goal, what- there is also evidence that these scores are often
ever they do and in whatever way, their funda- associated with other indexes of aggressiveness.
mental objective supposedly is to influence other Williams, Meyerson, and Eron (1967) provided
people's behavior. This may be too narrow a con- their third-grade boys and girls with an opportu-
ception. Along with other writers (e.g., Feshbach, nity to punish a peer for mistakes made on a joint
1964; Moyer, 1976; Rule & Nesdale, 1974), we sug- task, using a slightly modified Buss procedure, and
gest that there are different kinds of aggression then related the laboratory scores to an indepen-
and that they are not all equally reinforced by the dently validated index of each child's customary
same outcomes. All aggression is not carried out aggressiveness as rated by the children's classmates.
simply to coerce someone. Furthermore, it also The youngsters of both sexes who were seen by
appears that humans who have been angered by their peers as being most aggressive in their daily
some aversive event are apt to be especially mo- social encounters administered significantly more
tivated to inflict harm so that knowledge of the intense punishment on the laboratory apparatus,
intended target's appropriate injury is particularly for longer durations, with shorter latencies, and
rewarding. with greater frequency than did their less aggres-
Laboratory experiments have yielded observa- sive counterparts. Shemberg, Leventhal, and All-
tions in accord with this thesis (e.g., Rule & Nes- man (1968) reported similar findings in a study

AMERICAN PSYCHOLOGIST • MARCH 1982 • 253


with high School boys and girls. Again using the 1967), several investigators attempted to replicate
Buss procedure, those teenagers who were rated the initial results, while still others sought to show,
by their counselors as being most aggressive gen- as we have seen, that these findings arose because
erally gave their victims the most intense punish- of the subjects' compliance to demand cues. A sur-
ment. A more recent unpublished investigation by vey of this literature (conveniently summarized by
Leyens (Note 2) in Belgium demonstrates that Turner, Simons, Berkowitz, & Frodi, 1977) tends
these results are also obtained when the partici- to confirm Berkowitz's initial contention; although
pants are clearly told that their responses will harm there have been some failures to obtain the weap-
the other person. The school children in this study ons effect (e.g., Buss, Booker, & Buss, 1972), other
were informed that their high-level responses researchers have found that the sight of weapons
would prevent their partners from doing what they can lead to heightened aggression. Moreover, these
had wanted. After each youngster had been frus- positive results have been obtained in Europe as
trated by the other child, those who were rated by well as in the United States, with young children
their peers as being the most aggressive were the as well as with adults, in naturalistic situations as
most intense in their aggressive reactions with the well as in the experimental laboratory, and with
laboratory apparatus. Finally, Malamuth and Check a number of different measures of aggression.
(Note 3) found that male subjects with the highest Charles Turner has carried out the most com-
scores on a scale measuring acceptance of violence prehensive series of studies of the weapons effect.
against women were later most aggressive in the Two of his experiments utilizing nursery school
Buss procedure toward a woman who had insulted children (cited in Turner et al., 1977, p. 361) dem-
them. These same men also considered themselves onstrated that play with toy guns stimulated the
angrier with this woman and expressed a stronger youngsters to high levels of verbal and physical
"desire to hurt her with punishment." antisocial behavior. (The "antisocial behavior" re-
All of these observations seem to form a nice, corded in these studies excluded actions that are
coherent whole. Taken together, they suggest that generally appropriate for these toys and had more
aggression should be defined as the deliberate in- to do with such acts as pushing and shoving.)
jury of another and that most laboratory measures Turner and Goldsmith noted that other studies of
purporting to assess the subject's aggressiveness children's aggressive play have yielded very sim-
actually do fit this definition, at least in part. We ilar results, so the findings seem to be quite reliable.
would go even further. Not only do we regard the Yet another investigation, employing adults who
laboratory responses as "aggression" but, we be- did not realize they were taking part in an exper-
lieve, so do most of the subjects. Although the re- iment, also showed that the sight of a gun can
sponses they make in the experiment certainly do produce increased aggressiveness even in a fantasy
not physically resemble the attacks they display setting (cited in Turner et al., 1977, p. 360). Young
in their daily lives, all of these actions appear to meri participating in spring carnivals at a western
have much the saine meaning to the participants; university threw more wet sponges at a clown-tar-
they know they are intentionally hurting someone get when a gun was visible in the booth than when
else. This common meaning relates the laboratory no weapon was present. But it is not only playful
behavior to the conduct shown in more naturalistic aggression that can be affected by the sight of a
settings. weapon, according to still another field study with
adults. In this experiment (cited in Turner et al.,
THE GENERALITY OF THE WEAPONS EFFECT pp. 359-360) the aggressive responses were the
blasts of an auto horn sounded by impatient mo-
Another way to support our thesis is to demonstrate torists as they waited for a (deliberately) stalled
that the findings obtained in the university labo- pickup truck to get out of their way at an inter-
ratory are paralleled by observations made in very section. Turner and his associates hypothesized that
different settings with a variety of behavioral mea- these aggressive reactions would be most likely to
sures. We can do this with the weapons-effect re- occur when the frustrating truck bore an aggres-
search mentioned earlier dealing with the aggres- sive stimulus. The stimulus they placed on the
sion-facilitating influence of the mere sight of truck in some instances was the combination of
weapons. (a) a rifle prominently displayed on a gun rack at
After the publication of the first experiment re- the back of a truck and (b) a bumper sticker pro-
porting such an effect (Berkowitz & LePage, claiming Vengeance! The investigators assumed

254 • MARCH 1982 • AMERICAN PSYCHOLOGIST


that this latter hostile message would enhance the resemble the way they interpret supposedly similar
rifle's aggressive meaning and found, as they had Occurrences in other settings. They may assign
predicted, that this combination drew the most much the same meanings to both sets of stimuli
horn blasts from the thwarted motorists behind the or they may not. Many people may ordinarily
truck. think of a gun only as an object used for hurting
Space limitations keep us from saying very much others, but filled with evaluation apprehension in
more about this line of research. However, to add the laboratory (because they believe their adjust-
just one observation, we can also note that the ex- ment is being tested), they may regard a nearby
ternal validity of the weapons-effect studies is also pistol in a different manner—for example, as
corroborated by Turner's experiments on the role wicked and dangerous. They may therefore re-
of memory in this phenomenon. Extending Ber- spond quite differently in these two situations.
kowitz's initial analysis, he argued that besides Thus, we are not insisting that the laboratory find-
having an aggressive meaning, weapons can also ings are necessarily generalizable to the world out-
serve as "retrieval cues for earlier aggressive epi- side. No blanket statement can be made one way
sodes." If this is the case, he reasoned, aggressive or another-about experiments as a whole. Just as
stimuli that are most easily imagined (i.e., that it is incorrect to say laboratory behavior will not
have the greatest imagery value) should be most arise in other settings that are physically very dif-
likely to remind people of earlier aggressive epi- ferent, so it is equally wrong to hold that the sub-
sodes in their lives and therefore should be the best jects' laboratory reactions will definitely occur on
elicitors of aggressive responses from those who are more naturalistic occasions. Appropriate question-
disposed to be aggressive. An experiment by ing is vital to ensure that the participants have
Turner and Layton (1976) yielded supporting ev- interpreted the experimental treatments in the
idence for this hypothesis. desired way.
Even if the crucial aspects of the experiment are
understood as the investigators wanted, the highly
Conclusions controlled nature of the laboratory procedures usu-
ally rules out the operation of other factors that
We should be very clear as to just what we are, can govern the specific behavior in the outside
or are not, claiming for laboratory experiments in world. Experiments in the area of aggression usu-
general and aggression experiments in particular. ally attempt to minimize the subjects' inhibitions
Our central argument has to do with the external against aggression, whereas these restraints are
validity of their findings—the extent to which the often quite strong in natural social settings. Be-
laboratory results can be generalized to other sam- cause of the lessened inhibitions in the former
ples, settings, and specific behaviors. External va- cases, it may actually be somewhat easier to see
lidity, we have argued, does not necessarily re- the weapons effect in the laboratory than in most
quire ecological validity. The experimental results social situations (assuming the participants do not
may tell us something about the conduct of a broad believe that the experimenter is assessing their per-
range of people in natural situations even though sonality). In other words, even if we confine our-
the subjects and laboratory settings are not phys- selves to psychological influences, the laboratory
ically representative of this population or the real- setting is not necessarily representative of the social
world situations in which they are embedded. The world within which many people act. As a con-
meaning the laboratory participants impart to the sequence, we cannot use our laboratory findings
experimental stimulation and the responses they to estimate the likelihood that a certain class of
make determines, at least in large part, the extent responses will occur in naturalistic situations. Sup-
to which their laboratory reactions parallel the pose that 60% of the subjects in a sample exhibit
behavior they exhibit in other settings. heightened aggressiveness over some baseline level
The experimenter would do well to strive for when a weapon is present. Even if these people
experimental rather than mundane realism (to use were representative of the persons in a larger pop-
the Carlsmith et al., 1976, terminology) to facilitate ulation, we could not say, of course, that 60% of
the desired interpretation of the laboratory events. this broader group would react in the same way
However, no one can guarantee at the outset that in a more realistic situation. Experiments are not
the subjects' understanding of what is happening conducted to yield such an estimate.
to them in the social psychological laboratory will Here we have to emphasize the point made by

AMERICAN PSYCHOLOGIST • MARCH 1982 • 255


virtually every defender of laboratory experimen- Dipboye, R. L., & Flanagan, M. F. Research settings in indus-
tation. These investigations basically seek to test trial and organizational psychology: Are findings in the field
more generalizable than in the laboratory? American Psy-
causal hypotheses: Can alterations in Variable X chologist, 1979, 34, 141-150.
lead to changes in Variable Y? Can the mere pres- Dipboye, R. L., & Flanagan, M. F. Reply to Willems and How-
ard. American Psychologist, 1980, 35, 388-390. (Comment)
ence (versus the absence) of a weapon produce Dweck, C. S., Goetz, T. E., & Strauss, N. L. Sex differences in
stronger displays of open aggression than there learned helplessness: IV. An experimental and naturalistic
otherwise would have been? Laboratory experi- study of failure generalization and its mediators. Journal of
Personality and Social Psychology, 1980, 38, 441-452.
ments cannot be truly representative designs and Feshbach, S. The function of aggression and the regulation of
must be artificial in nature. Because of this lack aggressive drive. Psychological Review, 1964, 71, 257-272.
of representativeness, they cannot be used to for- Feshbach, S., Stiles, W. B., & Bitter, E. The reinforcing effect
of witnessing aggression. Journal of Experimental Research
mulate population estimates. However, to the ex- in Personality, 1967, 2, 133-139.
tent that this artificiality stems from control over Flanagan, M. F., if Dipboye, R. L. Representativeness does have
irrelevant variables; the artificiality is the strength implications for the generalizability of laboratory and field
research findings. American Psychologist, 1980,35,464-467.
and not the weakness of experiments. To some (Comment)
degree, then, the attacks on laboratory experimen- Gadlin, H., & Ingle, G. Through the one-way mirror. The limits
tation reveal a misunderstanding of just why ex- of experimental self-reflection. American Psychologist, 1975,
30, 1003-1010.
periments are carried out. Gilmour, R., & Duck, S. The development of social psychology.
London: Academic Press, 1980.
Griffitt, W., & Veitch, R. Preacquaintance attitude similarity
REFERENCE NOTES
and attraction revisited: Ten days in a fall-out shelter. So-
1. Bronfenbrenner, U. Ecological validity in research on human ciometry, 1974, 37, 163-173.
development. In, External Validity in the Study of Human Harre, R., & Secord, P. F. The explanation of social behaviour.
Development. Symposium presented at the meeting of the Oxford, England: Blackwell, 1972.
American Psychological Association, Washington, D.C., Sep- Henshel, R. L. The purposes of laboratory experimentation and
tember 1976. the virtues of deliberate artificiality. Journal of Experimental
2. Leyens, J-P Personal communication, summer, 1980. Social Psychology, 1980, 16, 466-478.
3. Malamuth, N., & Check, J. V. The effects of mass media Kane, T. R.,-Joseph, J. M., & Tedeschi, J. T. Person perception
exposure on acceptance of violence against women: A field and the Berkowitz paradigm for the study of aggression. Jour-
experiment. Paper presented at the meeting of the American nal of Personality and Social Psychology, 1976,33,663-673.
Psychological Association, Montreal, September 1980. Kruglanski, A. W. The human subject in the psychology ex-
periment: Fact and artifact. In L. Berkowitz (Ed.), Advances
in experimental social psychology (Vol. 8). New York: Ac-
REFERENCES ademic Press, 1975.
Kruglanski, A. W. On the paradigmatic objections to experi-
Allport, G. W. The historical background of modern psychol- mental psychology. American Psychologist, 1976, 31, 655-
ogy. In G. Lindzey & E. Aronson (Eds.), The handbook of 663.
social psychology (Vol. 1). Reading, Mass.: Addison-Wesley, Littman, R. A. Psychology: The socially indifferent science.
1968. American Psychologist, 1961, 16, 232-236.
Baron, R. A. Human aggression. New York: Plenum, 1977. Moyer, K. E. The psychobiology of aggression. New York:
Bass, A. R., & Firestone, I. J. Implications of representativeness Harper & Row, 1976.
for generalizability of field and laboratory research, findings. Newcomb, T. M. The acquaintance process. New York: Holt,
American Psychologist, 1980, 35, 463-464. (Comment) Rinehart & Winston, 1961.
Berkowitz, L. The "weapons effect," demand characteristics, Orne, M. T. On the social psychology of the psychological ex-
and the myth of the compliant subject. Journal of Personality periment: With particular reference to demand characteris-
and Social Psychology, 1971, 20, 332-338. ' tics and their implications. American Psychologist, 1962, 17,
Berkowitz, L., & LePage, A. Weapons as aggression-eliciting 776-783.
stimuli. Journal of Personality and Social Psychology, 1967,
Page, M., & Scheidt, R. The elusive weapons effect: Demand
7, 202-207.
Brock, T. C., & Pallak, M. S. The consequences of choosing to awareness, evaluation apprehension, and slightly sophisti-
cated subjects. Journal of Personality and Social Psychology,
be aggressive. In P. G. Zimbardo (Ed.), The cognitive control
of motivation. Glenview, 111.: Scott, Foresman, 1969. 1971, 20, 304-318.
Brunswik, E. Representative design and probabilistic theory in Patterson, G. R , & Cobb, J. A. Stimulus control for classes of
a functional psychology. Psychological Review, 1955, 62, noxious behaviors. In J. F. Knutson (Ed.), The control of
193-217. aggression: Implications from basic research. Chicago: Al-
Buss, A., Booker, A., & Buss, E. Firing a weapon and aggression. dine, 1973.
Journal of Personality arid Social Psychology, 1972,22,296- Postman, L. The probability approach and nomothetic theory.
302. Psychological Review, 1955, 62, 218-225.
Byrne, D. Attitudes and attraction. In L. Berkowitz (Ed.), Ad- Rokeach, M. Beliefs, attitudes and values. San Francisco: Jos-
vances in experimental social psychology (Vol. 4). New York: . sey-Bass, 1968.
Academic Press, 1969. Rosenberg, M. The conditions and consequences of evaluation
Campbell, D. T., & Stanley, J. C. Experimental and quasi- apprehension. In R. Rosenthal & R. Rosnow (Eds.), Artifact
experimental designs for research. Chicago: Rand-McNally, in behavioral research. New York: Academic Press, 1969.
1963. Rule, B. G., & Nesdale, A. Differing functions of aggression.
Carlsmith, J. M., Ellsworth, P. C., & Aronson, E. Methods of Journal of Personality, 1974, 42, 467-481.
research in social psychology, Reading, Mass.: Addison-Wes- Schultz, D. P. The human subject in psychological research.
ley, 1976. Psychological Bulletin, 1969, 72, 214-228.

256 • MARCH 1982 • AMERICAN PSYCHOLOGIST


Shemberg, K. M., Leventhal, D. B., & Allman, L. Aggression Turner, C. W., & Simons, L. A. Effects of subject sophistication
machine performance and rated aggression. Journal of Ex- and evaluation/apprehension on aggressive responses to weap-
perimental Research in Personality, 1968; 3, 117-119. ons. Journal of Personality and Social Psychology, 1974, 30,
Silverman, I. The human subject in the psychological labo- 341-348.
ratory. New York: Pergamon, 1977. Turner, C. W., Simons, L. S., Berkowitz, L., & Frodi, A. The
Strickland, L. J., Aboud, F. E., & Gefgen, K. J. (Eds.). Social stimulating and inhibiting effects of weapons on aggressive
psychology in transition. New York: Plenum Press, 1976. behavior. Aggressive Behavior, 1977, 3, 355-378.
Wachtel, P. L. Investigation and its discontents: Some con-
Swart, C., & Berkowitz, L. The effect of a stimulus associated
straints on progress in psychological research. American Psy-
with a victim's pain on later aggression. Journal of Person- chologist, 1980, 35, 399-408.
ality and Social Psychology, 1976, 33, 623-631.
Weber, S. J., & Cook, T. D. Subject effects in laboratory re-
Tedeschi, J. T., Smith, R. B., Ill, & Brown, R. C., Jr. A rein- search: An examination of subject roles, demand character-
terpretation of research on aggression. Psychological Bulletin, istics, and valid inferences. Psychological Bulletin, 1972, 77,
1974, 81, 540-562. 273-295.
Turner, C.,W., & Layton, J. F. Verbal imagery and connotation Williams, J., Meyerson, L., & Eron, L. Peer-rated aggression
as memory-induced mediators of aggressive behavior. Jour- and aggressive responses elicited in an experimental situation.
nal of Personality and Social Psychology, 1976, 33, 755-763. Child Development, 1967, 38, 181-189.

AMERICAN PSYCHOLOGIST • MARCH 1982 • 257

You might also like