Experimental Methodology in Political Science
Author(s): Rose McDermott
Source: Political Analysis , Autumn 2002, Vol. 10, No. 4, Special Issue on Experimental
Methods in Political Science (Autumn 2002), pp. 325-342
Published by: Cambridge University Press on behalf of the Society for Political
Methodology
Stable URL: https://www.jstor.org/stable/25791696
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms
and Cambridge University Press are collaborating with JSTOR to digitize, preserve and extend
access to Political Analysis
This content downloaded from
147.251.110.42 on Wed, 01 Feb 2023 14:32:06 UTC
All use subject to https://about.jstor.org/terms
Political Analysis, 10:4
Experimental Methodology
in Political Science
Rose McDermott
Department of Government, McGraw Hall,
Cornell University, Ithaca, NY 14850
e-mail: rmm21@cornell.edu
Experiments offer a useful methodological tool to examine issues of importance to polit
ical scientists. The historical and cultural differences between experiments in behavioral
economics and social psychology are discussed. Issues of central concern to experimen
talists are covered, including impact versus control, mundane versus experimental real
ism, internal versus external validity, deception, and laboratory versus field experiments.
Advantages and disadvantages of experimentation are summarized.
1 Introduction
A few years ago, several physicians associated with the National Bowel and Breast Cancer
multisite trials lost their jobs and reputations as well as millions of dollars in research fund
ing because of their failure to adhere properly to the experimental protocols of a study on
lumpectomy versus mastectomy in breast cancer patients. In particular, one of the partic
ipating physicians in Toronto was admitting patients whose tumors were larger than the
accepted protocol for study and treatment. When asked about this behavior, the physician
replied that he was simply trying to help his patients by giving them access to state-of-the
art treatment. Yet his violation of protocol, along with that of others, forced the complete
reanalysis of data that had previously indicated that lumpectomy with radiation was as
effective as mastectomy in treating certain breast cancers. Although the overall recommen
dations did not shift as a result of this reexamination, thousands of women were unneces
sarily terrified and many more people lost faith in the reliability and honesty of medical
research. Why did this happen? Because a few smart and medically well-trained doctors
failed to understand the important methodological requirements of proper experimental
procedure.
This article addresses the proper purpose and use of experimental methods within polit
ical science. It begins with a discussion of the differences among experiments in behavioral
economics, social psychology, and political science. It then addresses some of the main
methodological issues that arise in connection with the use of experimental procedure. A
brief discussion of the advantages and disadvantages of using experiments follows.
Author's note: The author thanks Arthur Lupia, an anonymous reviewer, and James Alt and the participants at
the Experimental Conferences at the Center for Basic Research in the Social Sciences at Harvard University for
very helpful suggestions and comments throughout. I also thank the psychologists who helped to train me in
experimental methods and design: Philip Zimbardo, Lee Ross, and the late Amos Tversky.
Copyright 2002 by the Society for Political Methodology
325
This content downloaded from
147.251.110.42 on Wed, 01 Feb 2023 14:32:06 UTC
All use subject to https://about.jstor.org/terms
326 Rose McDermott
I do not argue that experimentation is the only, or even the best, methodological form
of investigation, but rather that it is one valuable tool among many possible methods,
and a strategy that should not be neglected out of hand. Experiments may not be useful
for all research questions or problems, but can be especially helpful to investigators who
have had difficulty achieving consensus using other methods, seek a stronger basis for
causal argumentation, or are interested in developing and testing theoretical models in
a direct, empirically grounded manner. Also, because of the unique benefits offered by
experimentation, failure to use this strategy may impose unwarranted limitations on the
state of knowledge and the speed of its development in our field.
1.1 History
Informal experimentation goes back at least as far as the court members surrounding French
King Louis XIV, who began to develop probability theory in an attempt to improve their
winnings in the gambling games that consumed much of their free time. Probably the earliest
published example of an experimental study reported Bernoulli's (1738) investigations of
the St. Petersburg paradox, which demonstrated certain limitations of expected utility as a
universal normative model.
Systematic experimentation has produced a long and impressive series of findings in
most disciplines in the hard sciences, including biology, chemistry, and physics. This is true
in the social and behavioral sciences as well. Experimentation in psychology goes back at
least to the end of the 19th century. Since the mid-1980s, in particular, work in behavioral
economics has transformed the broader field of economics through the systematic use of
experimentation. Experiments in social psychology and behavioral economics often focus
on issues of concern to political scientists. However, the history and purposes of each
of these disciplines differ enough that they have produced divergent styles and goals of
experimentation that often lead to confusion when experimentally naive observers attempt
to interpret or integrate them. It is thus worth reviewing briefly the different traditions
of each field before proceeding to a discussion of the relevant methodological issues that
challenge any experimenter.
1.1.1 Behavioral Economics1
One of the most diagnostic features of economic experiments is the monetary incentives
that are used to induce and validate subjects' behavior. In some ways, this makes sense
given the economic topics under investigation. Yet in another way, there is no empirical
reason to assume a priori that money ensures that subjects will behave more closely in
line with experimental instructions simply because there is a fiscal reward for behaving
rationally in the experiment. Individuals can have other, particularly powerful social, goals
as well. Experimental biases that plague noneconomic experiments, including demand
characteristics, social desirability biases, and so on, might still operate even in the face of
monetary incentives.
The reason for the widespread use of monetary incentives in economic experiments is
at least partly historical as much as theoretical. An early experiment by Thurstone (1931),
designed to develop individual indifference curves, was criticized by W. Allen Wallis and
Milton Friedman (1942) for the hypothetical nature of the choices presented to the subjects.
^is section draws on the first part of the introduction chapter by Alvin Roth of John Hagel and Alvin Roth
(Eds.), The Handbook of Experimental Economics, 1995.
This content downloaded from
147.251.110.42 on Wed, 01 Feb 2023 14:32:06 UTC
All use subject to https://about.jstor.org/terms
Experimental Methodology 327
Wallis and Friedman argued that in order to learn anything useful from an experiment,
subjects needed to react to "actual stimuli" and not merely to their conjectures of how they
might behave if they were confronted by a real-life situation such as the one described
experimentally. This criticism gained widespread credence among future experimentalists,
and monetary payoffs became the sine qua non of economic experimentation. Monetary
rewards allowed economists to assume that subjects were equalized by their striving to
make the most money possible in all experimental situations. In addition, transparency
became the name of the game as subjects were expected to be able to see and understand
how the monetary rewards worked in order for the monetary incentives to have an effect.
Experimenters wanted subjects to realize that their reward was contingent on their ostensibly
rational behavior.
Early economic theorists such as Adam Smith and John Maynard Keynes understood
that humans were psychologically complex and incorporated this understanding into their
models. However, by the 1920s economists were striving to mold themselves into a "real"
science, like physics. This required complex mathematical models dependent on assump
tions of people as rational, self-interested actors who respond to price changes but not to
emotions in making their economic decisions (Uchitelle 2001). By the 1940s, this notion
of "rational man" formed the orthodoxy of economics. This model was advanced by the
increasing use of mathematics in economics. Of course, mathematical models are much
easier to formulate when people are assumed to be rational (Economist 2000).
However, following the publication of von Neumann and Morganstern's (1944) sem
inal work Theory of Games and Economic Behavior, experimental economics began to
focus more on individual choice, expected utility, and game theory. These topics seemed
ideal for experimental investigation because they allowed scholars to explore individual
behavior under controlled conditions that were unaffected by extraneous contaminants.
Even Thomas Schelling wrote in 1958 that "the principles relevant to successful play, the
strategic principles, the propositions of normative theory, cannot be derived by purely an
alytic means from a priori considerations" (p. 257). Among the early enthusiasts for this
method were Melvin Dresher and Merrill Flood, who began a series of experiments at the
Rand Corporation in 1950 that subsequently came to be known as the Prisoner's Dilemma
paradigm.
By the early 1950s, economists had begun to concentrate on using experiments to test
theories, especially game theories, in controlled environments. Economists sought to make
clear predictions and test them in a variety of stripped-down settings. When the predictions
did not work out, economists tested alternative hypotheses in subsequent work. Anonymity
of subjects came to be important as investigators sought to remove extraneous factors that
might contaminate the experimental findings. Building on the insights gained from suc
cesses, failures, and anomalies in earlier experimental work, economists began an ongoing
interaction between theoretical advances and experimental empirical investigation designed
to test and develop formal models of human behavior.
Experimental work continued to expand throughout the 1960s. The first large fund
ing sources for experimental economics began in the 1970s, when the National Science
Foundation (NSF) began to provide systematic and ongoing support to a few laboratories
throughout the country. Importantly, Eric Wanner at the Russell Sage Foundation began to
fund Richard Thaler's work on anomalies in human economic behavior, work that threat
ened the dominant economic model of man as a rational actor (Lowenstein 2001). It was
at this time that experimental economics began to diverge increasingly from experimental
psychology, although there continues to be a great deal of overlap, especially in work on
individual choice behavior.
This content downloaded from
147.251.110.42 on Wed, 01 Feb 2023 14:32:06 UTC
All use subject to https://about.jstor.org/terms
328 Rose McDermott
Experimental economics really arrived when, in 1988, Maurice Allais, who had discov
ered an important paradox that revealed a way in which individuals systematically deviate
from rational behavior in choice, won the Nobel Prize in economics. By this time, several
researchers in behavioral economics, most notably Richard Thaler and George Akerlof,
began to establish systematic programs of research in behavioral economics that led to the
replication and cumulation of knowledge around several specific substantive areas. Inter
estingly, both men were heavily influenced by psychologists Amos Tversky and Daniel
Kahneman, who themselves pioneered work on anomalies in human judgment and decision
making.
Two other major breakthroughs in experimental economics occurred around this time as
well. First, the stock market crashed 23% in one day on October 19, 1987, destroying the
beliefs of many economists in the triumph of rational actors operating in efficient markets
(Lowenstein 2001). Interest strengthened and developed in the work of Thaler and others
whose experimental work demonstrated ways in which individual behavior systematically
deviated from rationalist expectations. A second breakthrough in the history of behavioral
economics occurred when Harvard hired David Laibson in 1994, signaling the first time
that a major institution had deliberately hired a behavioral economist (Uchitelle 2001).
Currently, experiments in behavioral economics deemed to be successful share cer
tain core characteristics with precedents that should now be obvious: clear instructions;
absolutely no deception; stylized stripped down settings; anonymity; cash incentives; an
experimental test of a formal model, which can include a comparison with rational models;
and, ideally, subsequent validation of the findings with supportive field data (Laibson 2000).
The substantive topics that have most preoccupied behavioral economics research include
public goods, coordination problems, bargaining, experimental market behavior, auctions,
and individual decision making (Kagel and Roth 1995). Clearly, these issues overlap with
issues that preoccupy political scientists. In particular, many of these topics relate closely to
topics investigated by political economists experimentally: strategic behavior, incomplete
and/or asymmetric information, and the dynamics of change (Palfrey 1991). These areas of
overlap might easily be exploited experimentally for the mutual benefit of both disciplines.
In 1959, economists Siegel and Fouraker (1960) won the Monograph Prize of the
American Academy of Arts and Sciences with an experiment investigating levels of as
piration. Their article, as noted by Kagel and Roth (1995), makes an argument for the
virtues of the experimental method that continues to hold true today:
Our data have been observations made specifically to meet the purposes of this research. We have
not turned to preexisting data_
It is when a science starts to go out and ask exactly what it wants to know, rather than relying
on information collected for other purposes, that it begins to obtain control of its own growth,
(pp. 72-73)
1.1.2 Psychology
Psychology also possesses a long history of investigating phenomena of interest through
the use of experimentation. Although there may appear to be less obvious overlap between
the topics of interest in psychology and political science than there is between economics
and political science, most of the methodological developments used by economists study
ing individual choice behavior were pioneered by psychologists working in the same area.
Therefore, the historical development of psychological experimentation remains method
ologically primary. Furthermore, the impact and scope of experimentation in psychology
remains far greater than that in behavioral economics.
This content downloaded from
147.251.110.42 on Wed, 01 Feb 2023 14:32:06 UTC
All use subject to https://about.jstor.org/terms
Experimental Methodology 329
The scientific study of psychology began late in the 19th century, focusing primarily on
the study of human consciousness through processes of introspection. Among the earliest
laboratories designed expressly for use in experimentation of psychological phenomena
were those built by Wilhelm Wundt in Germany, his student Edward Titchner at Cornell,
Edward Thorndike at Columbia, and William James at Harvard. Thorndike (1898) began his
laboratory work around the turn of the 20th century by studying learning behavior in cats;
this was around the same time that Pavlov developed his theory of classical conditioning
by inducing his famous Russian dogs to salivate in response to a bell that originally rang in
conjunction with feeding. B. F. Skinner (1966) later built on these behavioral experimental
beginnings using mostly pigeons to construct his systematic program of research, which he
labeled the experimental analysis of behavior.
The history of social psychology informs some of the differences between experiments
in behavioral economics and those in social psychology. The most important of these
differences lies in attitudes and practices toward the use of deception in experiments. In
economics, it is virtually impossible to get an experiment involving deception published in a
respectable journal. This is because deception is understood to be antithetical to the purpose
of most economics experiments, which are designed to involve transparent instructions with
clear payoffs. In social psychology, however, more than half of the studies published in the
two most prominent journals (Journal of Personality and Social Psychology and Journal
of Abnormal Social Psychology) involved the use of some sort of deception (Adair et al.
1985). Although this last study dates from the mid-1980s, there has been no change in the
use of deception in experimental social psychology since about that time.
In psychology, the history and goals of the discipline have led to different practices.
In fact, several of the most important and influential experiments ever conducted involved
deception. Social psychologists believe that much of what they know about human behavior
has come through the use of deception, at least in part, and that deception can be very useful
as long as it is used in an ethically responsible fashion.
Two examples suffice to make the point about the importance of deception in psycho
logical experiments. One of the most famous experiments in social psychology is the Asch
(1952) experiment on conformity effects. Subjects in the Asch experiment were told that
they were about to participate in an experiment about perception, and were then asked to
match the length of one line to one of three other lines that differed in length from one
another. As in the familiar Sesame Street games, only one line matched, and the match
was perceptually obvious. In this study, Asch demonstrated that most people change their
behavior in order to conform to the group, even when this means going against their own
perceptions. Obviously, subjects in this experiment might have proved much less likely to
conform if they were told that the experiment was about conformity.
Similarly, Milgram's (1974) famous experiment on obedience was billed as a study on
learning. Subjects were told that they had to shock a "learner" (in fact, a confederate)
when that person failed to answer a paired word association correctly. When the "learner"
began to complain about the shocks, mention a heart problem, scream, and then eventually
go silent, most subjects became incredibly uncomfortable. Yet most subjects continued to
shock their partner up to the maximum amount of shock after the experimenter told the
subject that the experiment must continue, and that he would take responsibility. Clearly,
subjects' responses might have been quite different if they knew that the experiment was
really about obedience and not about learning.
To contrast the central characteristics of experiments in behavioral economics with
those in psychology, some overlap exists, but some strong differences emerge. Although
economists insist on clear instructions, for example, psychologists often purposefully give
This content downloaded from
147.251.110.42 on Wed, 01 Feb 2023 14:32:06 UTC
All use subject to https://about.jstor.org/terms
330 Rose McDermott
ambiguous directions, especially if they want to study various forms of information
processing. As noted, psychologists are more accepting of deception. Both disciplines tend
to respect the importance of subject anonymity. However, psychologists are not as prone to
offer cash in return for participation in experimentation. For instance, oftentimes students
are simply given course credit for their participation. This is because psychologists often
feel that monetary rewards impose an artificial reward system that might skew the processes
they seek to investigate. Natural social processes, for example, might be hidden by cash
incentives. Finally, although behavioral economists often use experiments in order to test
formal models, psychologists are more often interested in less mathematical, if no less
rigorous, theories of human behavior. In psychology, experiments stand on their own more
often than they might in experimental economics, where they are often used in conjunction
with more formal models.
Some of the different emphases on deception between economics and psychology de
rive from the different natures of the questions they investigate. Although experimental
economists focus on the topics mentioned previously, social psychologists study such topics
as person perception; stereotyping and prejudice; social identity; attitudes, attitude change,
and behavior; group dynamics; relationships; and decision making. Some of these topics
would be impossible to investigate accurately without the use of deception. For example,
topics such as racial prejudice would be difficult for a psychologist to study explicitly
because many subjects might lie about their true attitudes and beliefs in order to present
themselves as more acceptable to the experimenter. Note that an economist would be very
unlikely to be interested in prejudice as a research topic, and therefore less likely to need
deception for such purposes. Thus, some of the differences between economics and psy
chology surrounding the use of deception, as well as other things, derive from the different
histories and cultures of experimentation in each discipline.
1.1.3 Political Science
Perhaps the first foray into experimental work in political science was published by Harold
Gosnell in the American Political Science Review in 1926. This "experiment" on an attempt
to stimulate voting really presented what would now be called a field experiment or a
field study, discussed at greater length below. With a few exceptions in the early 1950s,
more experimental work in political science waited for the pioneering contributions of
William Riker (1967), Riker and Zavonia (1970), and Charles Plott (1967, 1979). What
is interesting about much of the early work that used experiments to investigate political
phenomena is that it appeared in economics and psychology journals much more than in
political science journals. In fact, because many believed that their experimental work had
been unfairly rejected from the established political science journals (McConahay 1973),
the Experimental Study of Politics, a new journal expressly founded to redress this difficulty,
was established in February 1971. The broader field was not able to sustain enough interest,
and the journal lasted a little over four years before being discontinued.
Currently, experimentation is not a major methodology used by political scientists. A
comprehensive overview of experiments published in mainstream political science journals
reveals a total of 105 articles since 1926. About 80% of these articles have been published
in one of five journals, and most of the experiments relate to American voting behavior. The
second most popular topic concerned bargaining (13), with games (10) and international
relations (10) tying for the third most popular topic. Other topics that attracted experimental
interest included committee work (8), experimental bias (6), race (6), field experiments (5),
media (4), leadership (4), and experiments embedded in surveys (3).
This content downloaded from
147.251.110.42 on Wed, 01 Feb 2023 14:32:06 UTC
All use subject to https://about.jstor.org/terms
Experimental Methodology 331
It is clear that many of these topics in political science overlap with those investigated
by behavioral economists and social psychologists. Although the venues and contexts for
investigation may differ, a great potential for substantive collaboration exists and might be
exploited for greater cross-disciplinary and interdisciplinary work among political science,
psychology, and economics. Current areas of research in psychology and economics that
offer particularly promising opportunities for collaboration with political scientists include
social preferences, including investigations of norms, social networks, altruism, status and
trust; bounded rationality, involving decision making in complex environments; learning
and expectation formation; attitudes toward risk; and cognitive biases (Laibson 2000).
However, significant differences between the disciplines also exist. Psychology and be
havioral economics share some core assumptions about human nature and experimentation
that many political scientists do not: a deep skepticism toward notions of rationality in hu
man behavior; emphasis on experimental validation of modeling assumptions; integration
of micro level data; and adoption of the lessons about human cognitive processing, from
experimental social and cognitive psychology to models of human behavior (Laibson 2000).
2 Methodological Issues in Experimentation2
It may prove difficult for psychologists or behavioral economists to work easily with political
scientists who do not share some of these same starting fundamental assumptions about
human nature. However, some divides are the result of a lack of training and experience
and a failure by political scientists to fully understand the methodology of experimentation
and its interpretation. As some of this confusion abates, more productive conversations and
collaborations can be achieved. In that spirit, this section discusses some of the more
difficult and contentious aspects of experimentation for all its practitioners: impact versus
control; mundane versus experimental realism; internal versus external validity; deception;
and laboratory versus field experiments.
2.1 Impact versus Control
Aronson et al. (1995) offer a nice summary of the problem of impact versus control in
experimentation when they write:
On the one hand, the experimenter wants the experimental situation to be meaningful and involving
to the subject?in a word, the treatments should have impact. On the other hand, a situation which
the subject finds meaningful and involving is also likely to trigger a wealth of memories and
influences from the past which can affect the subject's interpretation of present circumstances. From
the experimenter's point of view, these memories and influences constitute extraneous variability
and jeopardize control over the effects of the independent variable. The choice of an empirical
realization of one's conceptual variable?and an appropriate stimulus situation in which to apply
it?always represents a compromise between impact and control, in which case a little of each is
sacrificed, (pp. 44-45)
In other words, an experimenter wants an involved subject, but not one who brings so much
extraneous "baggage" to the task that his or her response is overly biased. This can be hard
to establish in practice within a restricted experimental protocol.
Probably the most important advantage attributed to experiments is the ability of the
investigator to control the experimental situation. Indeed, the entire premise of experiments
is built on the notion that everything that can be controlled should be controlled so that the
2This entire section draws heavily upon, and greatly benefits from, discussion and commentary on these issues in
Aronson et al. (1995).
This content downloaded from
147.251.110.42 on Wed, 01 Feb 2023 14:32:06 UTC
All use subject to https://about.jstor.org/terms
332 Rose McDermott
environment is identical except for the independent variable, which differs across condi
tions. In this way, any differences in the outcome measure can confidently be attributed to
variations in the independent variable.
Yet there are certainly limitations to total control. Subjects come in with a host of prior
experiences, beliefs, expectations, and skills that might have an impact on the outcome of any
given experiment, regardless of the intended manipulation. Most of the time, experimenters
hope that the random assignment of subjects to conditions eliminates any systematic source
of bias by ensuring that it is randomly distributed, and does not systematically effect one
condition or another disproportionately. However, total control of a given experimental
situation might not always be attainable, and it is possible that certain extraneous factors
might still influence the outcomes. This can happen, for example, if the random error serves
to drown the experimental effects, or if the independent variable is correlated with another
unintended and unidentified aspect of the situation or subject. Sometimes running more
subjects can ameliorate this problem, but sometimes it cannot.
However, lack of experimental control might occur precisely because of the way subjects
interact with the experimental manipulation. Obviously, an experiment of an hour or so
cannot erase a lifetime of experiences and attitudes. Subjects bring very different baggage
to a given experiment and different individuals can react to the same experimental situation
quite differently; they can literally see and experience a different experiment than each other
because of the cues the experiment provides to events in a subject's past. Other limitations
to an experimental situation arise as a result of ethical constraints, time limitations, and
experimental biases such as demand characteristics. By and large, the experimenter is aware
of these concerns and does the best job possible to achieve as much experimental control
as is feasible.
The conflict arises when an experimenter must create a situation that is sufficiently involv
ing and engaging to subjects to generate real attention and impact, and must do so generally
in under an hour. This can often be difficult to do in a practical way. Economists often try to
get around this problem by offering money, assuming that this incentive alone guarantees
engagement. Often, this works, but those who have performed experiments in certain envi
ronments in which the average subject is much wealthier than the experimenter understand
the futility of this option. Also, as noted, money alone does not eliminate other sources of
bias, particularly social desirability biases. Furthermore, for psychologists studying topics
that involve internal sources of motivation, money in fact detracts from investigating the
topic of interest by providing an extrinsic reward that may serve only to lower intrinsic
motivation for the task at hand (Lepper and Green 1978).
The goal of clever experimentation is to increase the realism and interest of the task
without sacrificing control. In theory, it should be possible to maximize both goals simulta
neously. In practice, increasing impact often means losing some element of control. Those
who do field experiments, for instance, lose a lot of control in order to study real-life situa
tions. However, many paper and pencil surveys in the classroom that possess a high degree
of control fail to engage the subjects in even the most minimal ways. Subject boredom can
render an experiment as fatal as any other loss of control.
2.2 Mundane versus Experimental Realism
Some skeptics' concern about experimentation derives from a misunderstanding of the
difference between mundane and experimental realism in experiments. However, when
critics argue that an experimental situation is not "realistic" because it does not look like a
real-world situation on the surface, they are failing to grasp that in experimental settings, the
This content downloaded from
147.251.110.42 on Wed, 01 Feb 2023 14:32:06 UTC
All use subject to https://about.jstor.org/terms
Experimental Methodology 333
subjects' engagement with the underlying process is much more important for experimental
control, and thus internal validity, than the external trappings of the situation.
Experimental realism is crucial to experimental success; mundane realism is not. Experi
mental realism refers to impact in its most important sense: Do subjects believe the situation,
problem, or issue they confront? Does it engage and interest them? Does it capture their
attention? If so, then experimental realism has been achieved; if not, then it does not matter
how much the trappings of the environment resemble a real-life situation, indeed, it does
not matter if the experiment takes place in real life, like a field experiment, because the
experiment failed to achieve the experimental realism necessary to obtain accurate results.
Mundane realism refers to the similarity between the experimental situation and the real
world. Yet just because a situation appears to look like a real-world event or problem on
the surface does not mean that it elicits any meaningful behavior worthy of investigation or
consideration. One could conduct an experiment on individuals sitting on a couch watching
sporting events on television, eating chips and drinking beer, and it would have a great deal
of mundane realism because many people spend countless hours engaging in this real-life
behavior. However, this study may not engage subjects in any meaningful phenomena of
interest because it does not possess experimental realism.
Iyengar (2000) tells a great story about how when he first set up an experimental living
room in a local mall complete with couch and television in order to study the effect of
campaign ads, subjects did what they do at home: they picked up the remote control and
changed the channel when the ads came on. This experiment contained so much mundane
realism that the experimental manipulation was avoided altogether. Once the remote was
removed from the situation, making it less "realistic" in the mundane sense, the experiment
became much more realistic in the experimental sense because subjects now were forced
to sit through watching the ads of primary interest to experimenters. This slight reduction
in mundane realism in fact allowed and enhanced the more important experimental realism
in the study.
Obviously, as the Iyengar study illustrates, experiments can possess both high levels
of mundane and experimental realism. Yet many critics assume that because laboratory
experiments take place in stripped-down, stylized settings, they cannot tell us anything
meaningful about the real world, and this is a mistake. One of the reasons that experimen
talists try to achieve high levels of both impact and control in their studies is because of
the need to overcome the preexisting beliefs, memories, expectations, and experiences with
which subjects enter a study. Experimenters have only a brief period of time to engage the
subject in an experience that is sufficiently interesting and meaningful that it overrides these
potentially confounding extraneous biases. However, experiments that involve subjects in
a meaningful process should be able to overcome the seeming sterility of the environment.
As subjects become sufficiently interested in the task or process, their focus remains on the
manipulation itself, and not on distractions in the environment that may merely serve to
arouse or cue triggers to unrelated processes or events from their past.
One additional point should be considered in assessing the value of experiments for real
world events. Real-world events would be interpreted differently if these events took place
in the laboratory, just as laboratory events that take place in the real world appear differently
as well. In this way, the criticism of external validity goes both ways. Experiments that are
high in mundane realism do not necessarily increase the generalizability of the findings if
they do not achieve internal validity and control over the variables of interest first. Rather,
experiments in the laboratory that prove meaningful for subjects need not take place in a
real-world situation to provide valid results. The laboratory allows for control over potential
biases that might interfere with the process in the real world, and allows investigators to
This content downloaded from
147.251.110.42 on Wed, 01 Feb 2023 14:32:06 UTC
All use subject to https://about.jstor.org/terms
334 Rose McDermott
isolate causation, test theories, and generate new hypotheses about human behavior. Once
these pieces are isolated and examined, additional complicating factors can be added under
some kind of controlled order to begin establishing how these forces might interact in the real
world. In this way, the microfoundational basis of larger processes can be uncovered. Then,
slowly and through replication, similar studies can be aggregated into a more comprehensive
understanding of broader topics and issues.
2.3 Internal versus External Validity
External validity is another area that tends to preoccupy critics of experiments. This near
obsession remains particularly problematic because issues of internal validity are of much
greater concern to any sophisticated experimentalist than are questions of external valid
ity. This causes at least two important points of confusion. First, questions surrounding
external validity tend to be used to dismiss experiments that in fact provide a valid basis
for generating hypotheses, testing theories, and building models. Second, this confusion
results in many critics of experiments being more willing to accept deeply flawed exper
iments. Some experiments have the surface appearance of external validity because they
appear high in mundane realism. The seemingly obvious generalizability often rests on a
foundation wholly without merit because the actual experiment offers no internal validity.
In short, without internal validity, there can be no external validity.
Prior to concerns about internal or external validity, experimenters must first be careful
that their measures possess construct validity. Both the independent and dependent variables
should be measured in ways that correspond to the theoretical concepts under investigation.
At least two problems arise with this concern. First, experimenters may use manipulations
that inadvertently affect other concepts, and this introduces measurement error that could
bias results in systematic ways. In other words, an experiment might be examining more
than is intended in ways that might compromise the findings that are emphasized.
Second, experimenters may use measures that actually do not assess the concepts being
investigated. Demand characteristics and social desirability responses within the experiment
itself can exacerbate these problems. In other words, an experimenter might not actually be
studying the phenomenon of central interest, but rather something else entirely. The best
strategies for avoiding these concerns are to use the best and most precise manipulations and
measurements possible, and to use multiple manipulations and measures wherever possible.
This latter issue is also important in assessing external validity (Smith and Mackie 1995).
Internal validity refers to whether you are studying what you think you are studying. Can
an experimenter be confident that changes in the independent variable caused changes in the
dependent variable? This is determined primarily by the experimental protocol itself, which
specifies how subjects are recruited, manipulated, and assessed. The goal is to make all the
conditions of an experiment as similar as possible in every way except the manipulation
itself. To the extent that this is possible, the experimenter can be confident that the manip
ulation of the independent variable is the cause of any differences seen in the dependent
variables. This ability to get at causation is, after all, the main advantage of experimentation,
and an advantage offered by no other methodology. Thus internal validity tells us whether
we can be confident that the effects are a result of the putative causes. Without high levels
of internal validity, it becomes impossible to render accurate causal claims. Internal validity
is aided by high levels of control, impact, and experimental realism.
Measuring effects is not the same as manipulating effects because other external factors
might be the true cause of a perceived effect. Therefore, voting studies, particularly those
that rely on public data on outcomes, cannot confidently attribute changes in voting to field
This content downloaded from
147.251.110.42 on Wed, 01 Feb 2023 14:32:06 UTC
All use subject to https://about.jstor.org/terms
Experimental Methodology 335
interventions alone, for example, because too many other factors not under experimental
control might have affected the outcomes. For example, a study that examines population
demographics and then relates those findings to voting patterns may have established a
correlational pattern, but it has not conducted an experiment because there was no random
assignments to treatment conditions or manipulations involved in associating the public
records data. Obviously, this kind of research is fine; it is just not experimental in nature.
Random assignment of subjects to conditions is the crucial best defense against threats to
internal validity.
Concerns about internal validity should always take place prior to concerns about external
validity. To the degree that an experiment is lacking in internal validity, external validity
should be a moot point. If we are not sure that A caused B because of some flaw in the
experimental research design, does it really matter whether B is representative of, or can be
generalized to, the real world? Aronson et al. (1995, p. 75) emphasize the primacy of internal
validity when they state, "But of the two, internal validity is, of course, the more important,
for if random or systematic error makes it impossible for the experimenter to draw any
conclusions from the experiment, the question of the generality of these conclusions never
arises."
Regardless, questions concerning external validity continue to elicit much greater con
cern within political science. This belief merely betrays, inadvertently, a failure to under
stand the methodological nature of experimentation itself. There are at least two important
sources of confusion surrounding this topic. First, many critics seem to think that experi
menters are trying to say something about the real world from a single study with college
undergraduates. This is almost never the case. Psychologists and behavioral economists
recognize that external validity is established over time, across a series of experiments that
demonstrate similar phenomena using different populations, manipulations, and measures.
External validity occurs through replication. Aronson et al. (1995, p. 82) note that "[b]ringing
the research out of the laboratory does not necessarily make it more generalizable or 'true';
it simply makes it different."
Sometimes more limited subject populations or manipulations are sufficient to establish
external validity because the issue of concern is limited to that population. This might occur,
for example, in studies concerning patient compliance with physician instructions. Studies
limited to medical patients are entirely sufficient for investigating this issue because the
findings are not intended to be generalized beyond this population. However, in our work
on sex differences in aggression, for example, we report the first of a series of experiments
that eventually use different populations and different measures to try to further investigate
the many issues involved in this topic (McDermott and Cowden 2001). Indeed, it would be
irresponsible and inappropriate to draw preliminary conclusions about all world leaders, or
all female leaders, for example, from a limited study.
This raises the second important area of confusion. For most experimentalists in psychol
ogy or behavioral science, whether a particular study generalizes to a larger audience is really
not the most important concern. Rather, experimental studies are intended to test theories
and build hypotheses about larger issues involving human behavior. So, for example, I might
not expect to find the same behavior in college undergraduates as I would in world leaders,
but I would hope that my experimental studies help illuminate more universal properties
related to aggression within and across the sexes that might prove more universal in nature.
Therefore, the idea of experimentation is not primarily that a specific set of experimental
results apply directly to broader conclusions about human behavior in other settings, but
rather that these results develop and test theories that then, in turn, and in aggregation, help
explain and predict certain underlying causal mechanisms in more universal issues (Smith
This content downloaded from
147.251.110.42 on Wed, 01 Feb 2023 14:32:06 UTC
All use subject to https://about.jstor.org/terms
336 Rose McDermott
and Mackie 1995). Thus, experiments should not be evaluated based on appearances of
broader generalizability, but rather on the merit of their underlying theoretical constructs.
Aronson et al. (1995, p. 81) summarize the issue as follows:
Is the ability to generalize to other people, places, and times in the real world?to ensure that
our findings are "externally valid' as the term is traditionally used?necessarily the standard
against which the value of experiments... should be assessed? Not always. Numerous important
experiments... have been conducted by researchers whose goals were not to generalize to other
subjects, operationalizations, or other settings. And this does not diminish the magnitude of their
contributions to understanding behavior, for their goals were not to make generalizations, but rather
to test them.
There are several criticisms that seem to characterize arguments about external validity
within experiments. Typically, these relate to the triviality or artificiality of the experimental
situation. Indeed, many experimental settings are artificially sterile and unrepresentative of
the environments in which subjects might normally perform the task required. This occurs
at least partly as a result of the artificial nature of the experiment itself: time is restricted,
conditions are controlled, and extraneous variables are eliminated. The artificial nature of
the experimental setting may make it difficult for subjects to act as they would in a more
natural setting. Yet environments are partly artificial because this also serves as a way to
reduce bias and engage subjects in an experimentally, and not mundanely, realistic way.
However, many aspects of real-world complexity are difficult to simulate in the labo
ratory. Cultural norms, relationships of authority, and the multitask nature of work itself
might invalidate any results that emerge from an experiment that does not, or cannot, fully
incorporate these features into the environment or manipulation (Walker 1976). In partic
ular, subjects may behave one way in the relative freedom of an experiment, where there
are no countervailing pressures acting on them, but quite another when acting within the
constrained organizational or bureaucratic environments that operate at their political jobs.
Material and professional incentives can easily and quickly override more natural psycho
logical or ethical concerns in the real world. Yet the same forces can manifest themselves
more readily in the unconstrained environment of the laboratory. Failure to mimic or incor
porate these constraints into experiments, and difficulty in making these constraints realistic,
might restrict the applicability of experimental results to the real political world.
However, much of this artificiality is purposeful because it helps isolate the variables
of interest. As noted, it is more important to create a research design high in experimental
realism than to mimic every aspect of a real-world situation. It is, in fact, experimentally
useless to conduct such mundane mirroring. Oftentimes arguments concerning triviality
or artificiality are based on the false assumption that similar variables should behave in
identical ways in real-world and laboratory settings. Clearly, however, this is not the case
because so many other factors besides those variables that can be isolated in the laboratory
influence real-life events. Any experimenter interested in a particular human behavior or
process must first isolate and study various aspects and features of that behavior under
controlled conditions before being able to extrapolate or expand those findings into broader
contexts. This control allows for the crucial element of internal validity that establishes the
viability of causal claims. Layers of complexity and interaction can thus be added over time
in a sequence of experiments in order to reach higher levels of theoretical development and
external validity.
2.4 Deception
Issues concerning deception have been mentioned in terms of their differing histories
in behavioral economics and social psychology. As previously mentioned, deception is
This content downloaded from
147.251.110.42 on Wed, 01 Feb 2023 14:32:06 UTC
All use subject to https://about.jstor.org/terms
Experimental Methodology 337
essentially never used in experiments in experimental economics, at least partly because
it is not necessary in order to study the phenomena of interest to economists, which are
mostly transparent economic transactions.
Yet many topics that psychologists and, potentially, political scientists may be interested
in studying might be impossible to investigate accurately without the use of deception. The
purpose of deception is simply to keep the subjects from being aware of the actual focus of
study until after the experiment is over. Typically, this is because the experimenter believes
that if subjects were aware of the true nature of the study, they would change their behavior
in ways that would bias the results of the experiment in misleading ways. Often deception
is elaborate, as in the case of the Milgram experiment, but at other times deception exists as
a sin of omission, whereby the experimenter simply fails to inform the subject of the real
purpose of the experiment.
Many observers tend to find deception distasteful almost on principle. However, there
is no empirical evidence that experiments involving deception are more harmful to subjects
than experiments that do not involve deception (Christensen 1988). Furthermore, subjects
who participate in experiments involving deception report enjoying the experience more,
and learning more from it, than subjects who participated in nondeceptive experiments
(Smith and Richardson 1983). As Aronson et al. (1995) note, this may simply be because
experiments involving deception may be more engaging and interesting to subjects than
more boring, albeit transparent paper-and-pencil survey-based experiments. In addition,
subjects appear to resist telling their friends about the true nature of an experiment after
debriefing when asked to do so (Aronson 1966).
The main concerns surrounding the use of deception have to do with invading a sub
ject's privacy, but there are practical methodological concerns with deception as well.
In general, subjects rarely correctly discern the true purpose of an experiment, unless
the so-called deception is in reality pretty transparent (Brock and Becker 1966; Fillen
baum 1966; Striker et al. 1969; Kruglanski 1975). However, the more deception is used
by experimenters, the more suspicious the subject pool as a whole can become, and
the more likely an individual subject might be to guess the real purpose of the exper
iment. Research shows that general suspicion on the part of subjects does increase the
variability of results (Kelman 1968; Striker et al. 1969). In addition, participation in an
experiment involving deception increases subjects' desire to present themselves as
desirable in the next experiment. It also makes subjects less compliant with subsequent
experimental demands (Silverman et al. 1970). Even the mere distraction of subjects
trying to discern the real reason for the experiment might bias the results in a system
atic way. Certainly the worst subjects in this regard are those who participate repeat
edly in experiments for money; such subjects become experienced and can become more
suspicious and sophisticated about hypotheses. Although economic experiments demand
monetary compensation to ensure the rational behavior of subjects, psychologists have
reason to believe that paying subjects serves only to undermine crucial naivete in certain
populations.
2.5 Laboratory versus Field Experiments
As noted, the earliest experiment conducted in political science was a field experiment
by Gosnell. Some practitioners continue to have more enthusiasm for field experiments
than psychologists, mostly because, as noted previously, they tend to be more concerned
with mundane as opposed to experimental realism, and questions of external as opposed to
internal validity, although this is methodologically backward.
This content downloaded from
147.251.110.42 on Wed, 01 Feb 2023 14:32:06 UTC
All use subject to https://about.jstor.org/terms
338 Rose McDermott
Field experiments definitely have their place in experimental research, however. First,
field experiments can be the only alternative available in some situations in which the
experimenter either cannot or should not create the desired situation within a laboratory
setting. Sometimes this is because field experiments are the only way to investigate certain
events, such as the impact of a natural disaster such as an earthquake or tornado; no laboratory
simulation could ever come close to the impact and realism of the real thing. Indeed, it would
be unethical to create such conditions experimentally; an experimenter would not be able
to study the effects of a life-threatening illness by causing disease in a subject, for example.
Second, it may be very hard to simulate accurately many all-encompassing phenomena of
interest (e.g., an election, a war, an economic recession). Skilled and clever experimentalists
excel at being able to distill the basic processes under investigation into an experimental
manipulation, but the task is neither easy nor straightforward.
However, a field experiment is not, and should not be, simply a laboratory experiment
conducted in the real world. In general, field experimenters must sacrifice a great deal of
experimental control in order to gain some advantage in either mundane realism or external
validity. This lack of control requires that many important accommodations need to take
place in field experiments in order to garner any findings that have the necessary internal
validity to assess causality confidently.
First, it is not enough simply to measure preexisting phenomena and call that an ex
periment because certain factors, such as demographics, correlate with other factors, such
as voting patterns. Second, researchers must be acutely sensitive to the many extraneous
factors over which they retain no control, but that can have an impact on their findings in the
field. Third, and perhaps most important, it is often impossible to assign subjects randomly
to conditions in the field. Sometimes this is ethically impossible, as it would be to study the
occurrence of deadly illness by infecting healthy people randomly. Yet often such failure
serves to limit the very generalizability of results in ways that experimenters may fail to
recognize or acknowledge.
Effective laboratory situations strive to be coherent, simple, and engaging. Control over
the random assignment of subjects to conditions ensures that all subjects start from the
same baseline as much as possible (Aronson et al. 1995). Similarly, experimenters in the
field try to choose a situation that exists naturally but that mirrors as closely as possible the
controlled conditions that can be created in a laboratory. Obviously this is harder to achieve
in the field where there are many more extraneous factors that might affect the recruitment,
manipulation, or measurement of subjects and outcomes. In evaluating the findings reported
by field studies, readers must be careful to assess the internal validity of the study before
buying into the generalizability of the results.
Finally, experimenters in the field face more challenging ethical concerns because often
times their subjects are unaware that they are being studied, and have not given informed
consent to be investigated. Although this may aid in achieving subject nonreactivity, lack of
awareness of participation in an experiment constitutes a serious invasion of subject privacy.
Under such circumstances, experimenters are especially obliged to impose as little stress
as possible on subjects, and to cease immediately any experimental procedure that appears
to cause any harm to subjects.
3 Advantages and Disadvantages of Experiments
Because most readers are well aware of the threats and potential responses to issues
surrounding internal and external validity, and, now the central issues of concern to ex
perimentalists, how can we evaluate the relative merits of experimental methodology as
This content downloaded from
147.251.110.42 on Wed, 01 Feb 2023 14:32:06 UTC
All use subject to https://about.jstor.org/terms
Experimental Methodology 339
opposed to other research strategies, whether they be case work or formal modeling? As
an advocate for experimentation, I place more emphasis below on advantages over disad
vantages. Part of my argument is that experimentation is underused in political science and
might work to the advantage of practitioners who are seeking a more definitive form of
causal inquiry, a useful methodology for the cumulation of knowledge, or an additional tool
in their methodological arsenal for triangulating on problems that appear intractable when
using other forms of methodological inquiry.
3.1 Advantages
Many of the issues discussed in this section overlap with some of the previous discussion,
but should be mentioned independently for their methodological advantages. There are at
least four such advantages that experiments offer: (1) the ability to derive causal inferences;
(2) experimental control; (3) precise measurement; and (4) the ability to explore the details
of given processes (McConahay 1973).
The primary comparative advantage of experiments lies in their potentially high de
gree of internal validity. Correlational studies, for example, obviously do not establish
causation. Because a laboratory setting allows investigators to control all aspects of the
environment so that only the independent variables differ, any differences in the dependent
variable can be attributed to the manipulation, and thus offer support for causal infer
ences. No other methodology can offer such strong support for the causal inferences that
experiments allow. Thus, "the major advantage of laboratory experiments is in its abil
ity to provide us with unambiguous evidence about causation" (Aronson and Carlsmith
1968, p. 10). Experiments can prove quite useful in helping to develop and test theories
to explore whether hypothesized relationships hold and under what conditions they are
operative.
Second, experiments allow the investigator to have unparalleled control over the test
ing environment. Experimenters not only create the experimental protocol and control the
experimental environment, but most importantly, they also control the random assignment
of subjects to conditions. This randomization process creates the pretreatment similarity of
the groups, so that observed differences in outcome can be attributed confidently only to
manipulations, which take place within the experiment itself, and not to preexisting differ
ences within the subject pool. Moreover, the experimenter can decide which independent
variables to manipulate, and can vary the timing and intensity of those variables at will.
This process allows the examiner to test for alternative explanations for the hypothesized
relationships as well.
Third, laboratory experiments allow for the precise measurement of the variables of
interest. Experimenters need not overly concern themselves with improper measurement
because they design and implement the desired measures themselves. This allows experi
menters to eliminate extraneous factors that might contaminate a study by inducing spurious
results. Particularly with the increase in the use of computers to run experiments, even more
precision is possible because potential errors in coding, transcription, or data entry are
substantially reduced as well.
Fourth, experiments offer the opportunity to explore the phenomenon of interest in great
detail. Complex relationships can be broken down, isolated, and investigated in smaller units
in order to see which part of the process results in the phenomenon of interest. In addition,
experiments allow particular relationships to be explored in the presence or absence of other
variables, so that the conditions under which certain relationships hold can be examined as
well. In this way, process can be studied as well as outcome.
This content downloaded from
147.251.110.42 on Wed, 01 Feb 2023 14:32:06 UTC
All use subject to https://about.jstor.org/terms
340 Rose McDermott
Finally, another advantage of experiments results from the scientific rigor designed into
the process. Experimenters must be aware of, and control, the independent variables of
interest. They must be careful in recording results. Statistical analysis allows for the detailed
testing of the relationships between these variables and any interactions among them. With
this process, results and insights that might not have been obvious to less systematic or
larger scale analysis become prominent. Experiments allow causal inference, control of
variables, precise measurement, and clarity of detail.
3.2 Disadvantages
Clearly, however, experiments are not a panacea for all methodological concerns.
Experiments are not the only, or even the best, form of investigation for all research ques
tions. Indeed, experiments are only one of many methods that can be used to examine
political phenomena of interest. As noted, most of the popular concerns about the disad
vantages of experiments revolve around questions of external validity and how widely the
findings of the laboratory apply to real-world actors and phenomena. Many of these concerns
about experimentation, including the artificial nature of the laboratory, the unrepresentative
nature of the subject pool, and questions surrounding external validity, have been discussed
in detail above.
From the perspective of an experimenter, the main concern with experiments revolves
around problems posed by experimental bias of one sort or another. Clearly, an experi
ment must consider alternative explanations for whatever results are discovered, but exper
imenters must also be concerned about the confounding variables they bring to the experi
mental situation that might affect certain findings. The most important of these artifacts are
experimenter bias and demand characteristics.
Experiments can be biased by an experimenter setting up an experiment in such a way
as to validate prior expectation and beliefs, often unconsciously. Experimenters or research
assistants who are privy to the experimental hypotheses might induce subjects to react in
ways that confirm these expectations without even meaning to do so (Rosenthal 1966,1969).
Such experimenter bias presents a potentially debilitating confound that can threaten the
internal validity of a study.
Subjects themselves also tend to be exquisitely sensitive to implicit experimenter de
mands and expectations. Note that this is different from subjects being able to discern
the relevant hypotheses being investigated in the study. Such demand characteristics and
expectancy effects threaten the construct validity of the experiment (Orne 1962; Smith
and Mackie 1995). In addition, subjects want to put themselves in a good light, creating
a very strong social desirability bias (Rosenberg 1966, 1969). There is also the infamous
"Hawthorne effect," whereby subjects change their behavior based solely on the recognition
that they are being observed, which can limit the relevance, generalizability, or accuracy of
certain experimental results as well (Roethlisberger and Dickson 1939).
The best antidote to these concerns, if possible, is to keep those who run the experiment, or
who code the data, blind to the condition of each subject and to the experimental hypotheses.
In this regard, the use of electronic collection, such as afforded by computers, can be of
immense help in reducing experimental bias.
4 Conclusions
Experiments have a long and distinguished history of effective use in other disciplines, in
cluding medicine and psychology. They have recently acquired great cachet in the field
of experimental economics as well. They have been slower to acquire a following of
This content downloaded from
147.251.110.42 on Wed, 01 Feb 2023 14:32:06 UTC
All use subject to https://about.jstor.org/terms
Experimental Methodology 341
practitioners in political science, mostly due to concerns surrounding the external validity
of these kind of studies.
Experiments offer a clear advantage over other methods in particular areas of investi
gation. In behavioral economics, they are used to validate theories developed by formal
modeling. In psychology, they uncover basic processes in human judgment and decision
making, among other things. In political science, they can help in triangulating on a process
that has proved intractable, or presented contradictory findings, using methods that allow
for less clear causal inference. In all areas, they can support theory development, testing,
and refinement. In particular, experiments can offer many useful insights in work that in
vestigates the underlying process of a particular phenomenon as opposed to its outcome.
Experimentation may not prove as useful for single case examples, but it can serve a useful
purpose, as it has in behavioral economics, to help advance and cumulate knowledge more
quickly and carefully.
References
Adair, J., T. Dushenki, and R. Lindsay. 1985. "Ethical Regulations and Their Impact on Research Practice."
American Psychologist 40:59-72.
Aronson, E., and M. Carlsmith. 1968. "Experimentation in Social Psychology." In The Handbook of Social
Psychology (rev. ed.), eds. G. Lindzey and E. Aronson. Reading, MA: Addison-Wesley, pp. 1-79.
Aronson, Eliot. 1966. "Avoidance of Inter-Subject Communication." Psychological Reports 19:238.
Aronson, Eliot, Pheobe Ellsworth, J. Merrill Carlsmith, and M. H. Gonzales. 1995. Methods of Research in Social
Psychology, 2nd ed. New York: McGraw-Hill.
Asch, Splomon. 1952. "Effects of Group Pressure upon the Modification and Distortion of Judgments." In Groups,
Leadership, and Men, ed. H. Guetzkow. Pittsburgh: Carnegie Press, pp. 177-190.
Bernoulli, D. (1738). "Specimin Theoriae Novae De Mensurea Sortis." Commentarii Academiae Scientiarum
Imperialis Petropolitanae 5:175-192. (English translation in Econometrica, 1954, 22:23-36.)
Brock, T, and L. Becker. 1966. "'Debriefing' and Susceptibility to Subsequent Experimental Manipulation."
Journal of Experimental Social Psychology 2:314-323.
Christensen, L. 1988. "Deception in Psychological Research: When Is Its Use Justified?" Personality and Social
Psychology Bulletin 14:664-675.
Economist. 2000. "Rethinking Thinking." December 16.
Fillenbaum, S. 1966. "Prior Deception and Subsequent Experimental Performance: The 'Faithful' Subject." Journal
of Personality and Social Psychology 4:532-537.
Gosnell, H. 1926. "An Experiment in the Stimulation of Voting." American Political Science Review 20:869
874.
Iyengar, S. 2000. "Presentation of Experimental Designs for Political Communication Research: From Shopping
Malls to the Internet." Experimental Methods Conference, Cambridge, MA.
Kagel, Jack, and Alvin Roth. 1995. Handbook of Experimental Economics. Princeton, NJ: Princeton University
Press.
Kelman, Herbert. 1968. A Time to Speak Out: On Human Values and Social Research. San Francisco: Jossey-Bass.
Kruglanski, A. 1975. "The Human Subject in the Psychology Experiment." In Advances in Experimental Social
Psychology, vol. 8, ed. L. Berkowitz. New York: Academic Press, pp. 101-147.
Laibson, D. 2000. Presentation to the Experimental Methods Conference, Cambridge, MA.
Lepper, Mark, and D. Greene. 1978. "Overjustification Research and Beyond: Toward a Means-End Analysis of
Intrinsic and Extrinsic Motivation." In The Hidden Costs of Reward: New Perspectives on the Psychology of
Human Motivation, eds. M. R. Lepper and D. Greene. Hillsdale, NJ: Erlbaum, pp. 109-148.
Lowenstein, Roger. February 11, 2001. Exuberance Is Rational. New York Times Magazine, 68.
McConahay, J. 1973. "Experimental Research." In Handbook of Political Psychology, ed. J. N. Knutson. San
Francisco: Jossey-Bass, pp. 356-382.
McDermott, Rose, and Jonathan Cowden. 2001. "The Impact of Uncertainty and Sex in a Crisis Simulation Game."
International Interactions 27:353-380.
Milgram, S. 1974. Obedience to Authority. New York: Harper & Row.
Orne, Martin. 1962. "On the Social Psychology of the Psychological Experiment with Particular Reference to
Demand Characteristics and Their Implications." American Psychologist 17:776-783.
Palfrey, T., ed. 1991. Laboratory Research in Political Economy. Ann Arbor: University of Michigan Press.
This content downloaded from
147.251.110.42 on Wed, 01 Feb 2023 14:32:06 UTC
All use subject to https://about.jstor.org/terms
342 Rose McDermott
Plott, C. 1967. "A Notion of Equilibrium and Its Possibility Under Majority Rule." American Economic Review
57:787-806.
Plott, C. 1979. "The Application of Laboratory Experimental Methods to Public Choice." In Collective Decision
Making: Applications from Public Choice Theory, ed. C. S. Russell. Dordrecht, Holland: D. Reidel, pp. 137-160.
Riker, W. 1967. "Bargaining in a Three-Person Game" American Political Science Review 61:642-656.
Riker, W., and W. J. Zavonia. 1970. "Rational Behavior in Politics: Evidence from a Three Person Game." American
Political Science Review 64:48-60.
Roethlisberger, F., and W. Dickson. 1939. Management and the Worker. Cambridge, MA: Harvard University
Press.
Rosenberg, M. 1966. "Some Limits of Dissonance." In Cognitive Consistency, ed. S. Feldman. New York:
Academic Press, pp. 135-170.
Rosenberg, M. 1969. "The Conditions and Consequences of Evaluation Apprehension." In Artifact in Behavioral
Research, eds. R. Rosenthal and R. Rosnow. New York: Academic Press, pp. 249-279.
Rosenthal, R. 1966. Experimenter Effects in Behavioral Research. New York: Appleton-Century-Crofts.
Rosenthal, R. 1969. "Interpersonal Expectations: Effects of the Experimenter's Hypothesis." In Artifact in Behav
ioral Research, eds. R. Rosenthal and R. Rosnow. New York: Academic Press, pp. 181-277.
Schelling, T. 1958. "The Strategy of Conflict: Prospectus for the Reorientation of Game Theory." Journal of
Conflict Resolution 2:203-264.
Siegel, S., and L. Fouraker. 1960. Bargaining and Group Decision Making: Experiments in Bilateral Monopoly.
New York: McGraw-Hill.
Silverman, I., A. Shulman, and D. Wiesenthal. 1970. "Effects of Deceiving and Debriefing Psychological Subjects
on Performance in Later Experiments." Journal of Personality and Social Psychology 14:203-212.
Skinner, B. F. 1966. "What Is the Experimental Analysis of Behavior?" Journal of Experimental Analysis of
Behavior 9:213-218.
Smith, Eliot, and Diano Mackie. 1995. Social Psychology. New York: Worth Publishers.
Smith, S., and D. Richardson. 1983. "Amelioration of Deception and Harm in Psychological Research: The
Important Role of Debriefing." Journal of Personality and Social Psychology 44:1075-1082.
Striker, L., S. Messick, and D. Jackson. 1969. "Evaluating Deception in Psychological Research." Psychological
Bulletin 11:343-351.
Thorndike, E. 1898. "Animal Intelligence." Psychology Review of Monograph Supplement 2 (4, Whole No. 8).
Thurstone, L. L. 1931. "The Indifference Function." Journal of Social Psychology 2:139-167.
Uchitelle, Louis. 2001. "Some Economists Call Behavior a Key." The New York Times, February 11, section 3,
P. i.
von Neumann, J., and O. Morganstern. 1944. Theory of Games and Economic Behavior. Princeton, NJ: Princeton
University Press.
Walker, T. 1976. "Microanalytic Approaches to Political Decision-Making." American Behavioral Science 20:93
110.
Wallis, W. Allen, and Milton Friedman. 1942. "The Empirical Derivation of Indifference Functions." In Studies
in Mathematical Economics and Econometrics in Memory of Henry Schultz, eds. O. Lange, F. Mclntyre, and
T. O. Yntema. Chicago: University of Chicago Press, pp. 175-189.
This content downloaded from
147.251.110.42 on Wed, 01 Feb 2023 14:32:06 UTC
All use subject to https://about.jstor.org/terms