Opinions and Social Pressure - Conformity Study Guide Part 6

The implicit attributions of the persuasion treatments in this study were negative because the behaviors they were calling for (which the subjects were presumably not emitting) were positive. Likewise, the attribution treatments were positive because the behaviors they attributed to the subjects were positive. For practical and ethical reasons, all of the treatments in the present study were aimed at producing positive or socially desirable behaviors. It is our idea that the implicit or explicit labeling itself, and not merely the rewardingness or punishingness of the labels, made the present attribution treatments effective and the persuasion treatments ineffective. This can only be known for sure, however, by a study that aims at producing undesirable behavior, in which case attributions of the behavior would have negative implications, which persuasion to the behavior would have implicit positive implications for the present self-concept. If people were responding to the attributions in the present experiment only because they were rewarding, we would expect that attributions of an undesirable behavior would have reverse effects and, indeed, that persuasion under these circumstances might even be more effective. The sociological literature on deviance (e.g., Schur, 1971), however, suggests that negative attributions (e.g., labeling as delinquent) can indeed produce or support the attributed behavior, as does a recent experimental study of labeling “charitable” or “uncharitable” behaviors (Kraut, 1973). We may suspect, then, that the positive implications of the attributions in the present study were not the sole key to their effectiveness and that if a suitable experiment could be designed in which the target behavior were socially undesirable, attribution would continue to be more effective than persuasion in generating that behavior. Attribution and Reinforcement A straightforward reward contingency program seems to modify behavior because it makes that behavior worthwhile to the subject. However, the separation between reinforcement and attribution seems somewhat confounded. Symbolic reinforcement, as used in the present study, has some attributional aspects. Simple praise is often interpreted as a “You are X” statement. Furthermore, attribution can contain elements of reinforcement especially when socially desirable behaviors are the focus of the attributional process. Thus to some extent a reinforcement procedure that produces enduring change may require elements of attribution, while a successful attribution treatment may involve elements of reinforcement. The remaining feature distinguishing attribution from reinforcement would seem to be the noncontingent nature of the attribution. In this regard it is interesting to note that Kazdin (1973) has recently found that under circumstances where the desired behavior was emitted at a fairly high base rate and subjects believed that reinforcement was contingent, noncontingent reinforcements were as effective as contingent ones in modifying behavior. Attribution treatments, however, may have the very important further advantage over simple reinforcement of serving to elicit behavior (like modeling; see Bandura, 1969) as well as to maintain it. Practical and Ethical Implications The present study supports the idea that an effort to improve the child’s academic self-concept will help improve academic performance, if only because the improved self-image will make actual success less inconsistent, less unexpected, and less likely to be discounted or rejected (Brickman, 1972). A distinguishing feature of the present attribution treatments is that they focused on raising self-esteem in a specific area of skill rather than raising global or general self-esteem. The failure of more general “cultural enrichment” or general “self-concept enhancement” programs may be due in part to the fact that these global manipulations have only vague and diffuse implications for particular areas of academic performance (but cf. deCharms, 1972). Nonetheless, there are a number of reasons to be cautious about considering these results as the basis for a solution to any social problem. It is unlikely that long-standing individual differences in accomplishment will be overcome by short-term manipulation of motivation, incentive, or self-regard. A gain of three problems solved after a week of treatment is not very substantial in terms of life chances; and there is no reason to assume that 10 weeks of treatment will necessarily result in a gain of 30 problems. Second, the attribution treatments in Study 2 were difficult to administer and, unlike the treatments in Study 1, did not produce any immediate and visible indications of success to sustain teacher enthusiasm. Although the teachers involved in Study 1 were quite positive toward the study, those involved in Study 2 had decidedly mixed feelings about the value of the time and energy involved. More seriously, the false attributions came increasingly to be felt by at least one teacher and by the principal as an intolerable risk to their credibility. As indicated, the principal terminated her office meetings with the students after the third day of treatment mainly on these grounds. While future research could tailor attributions not to be too discrepant from individual pretest baselines, the practical and ethical difficulties involved in maintaining such attributions will not thereby be eliminated. All of these matters warn against an uncritical application of the present results to matters of educational importance. 17 You Are What You Do: Implicating the Self to Influence Behavior Christopher J. Bryan University of Chicago, Booth School of Business It probably happens to you many times a week: you find yourself carrying an unneeded piece of paper or an empty soda can, looking for a place to get rid of it. You walk by a garbage can but there is no recycling bin in sight. So, do you drop the recyclable item in the trash or do you hold onto it until you can find someplace to recycle it? In this chapter, I will argue that a critical factor in determining this and similar decisions is how you construe recycling the can — is it just something you do or is it a reflection of the kind of person you are? We make many decisions like this. Do we donate to charity? Help out a stranger who is struggling with an unwieldy piece of furniture as we walk by? Correct a cashier who accidentally gave us too much change? Go to the gym when we’re feeling tired or lazy? Do we do the thing we know we should do — the thing that is good for us or good for society, even though it entails some hassle or expense? Or do we sit back and opt for ease, comfort, and convenience instead? I will describe a program of research that suggests that an important reason why we sometimes choose to do the hard, good thing is that we like the way doing that kind of thing allows us to see ourselves. We want to think of ourselves as charitable, giving people, as good Samaritans, as physically and mentally strong. And so many of the things we do — especially the sort of thing that involves some personal sacrifice but is widely approved of — we do because we want to think of ourselves as the kind of person who does those things. Ultimately, this is because we are social animals and we care a great deal about being able to think of ourselves as worthy of admiration and approval from others (Aronson, 1969; Dunning, 2002; Leary, 1999; Markus & Nurius, 1986; Sherman & Cohen, 2006; Steele, 1988). And our behavior feeds into our self-concept — what we do shapes how we see ourselves (e.g., Bem, 1972). So, when we construe behavior as having implications for our self-concept, this provides an opportunity to curate that selfconcept. We pursue the goal of being a person worthy of social approval by behaving in ways that are consistent with that view of ourselves. Of course, we do a lot of things and not all of them are spontaneously construed as having implications for our self-concepts. So my collaborators and I have conducted a series of experiments in which we framed behavior as relevant to the self-concept — who you are or who you could become — expecting that this would cause people to behave more like “worthy” people: engage in behavior that is effortful but beneficial and socially approved, and avoid behavior that is tempting but harmful and socially disapproved. In crafting these experiments, we drew on past research investigating how subtle differences in language can shape people’s perceptions of themselves and of others. In particular, describing people with nouns can signal that an attribute or behavior is representative of their essential, underlying character — who they really are (Carnaghi et al., 2008; Gelman & Heyman, 1999; Gelman, Hollander, Star, & Heyman, 2000; Markman 1989; Walton & Banaji, 2004). For example, even young children think another child described as a “carrot eater” has a stronger, more enduring preference for carrots than one who “eats carrots whenever she can” (Gelman & Heyman, 1999). And adults who read Shakespeare and are induced to describe themselves as “a Shakespeare reader” subsequently report having a stronger and more enduring appreciation for the English playwright’s work than those induced to indicate that they “read Shakespeare a lot” (Walton & Banaji, 2004). In our work, we took an approach that was superficially similar to this but with a twist that changed the psychology considerably: Instead of using nouns to describe what a person already is (like a “Shakespeare reader” if they already read Shakespeare), we used them to refer to prospective behavior — things people are considering doing in the future. So, rather than putting labels on people, we offered them the opportunity to claim an identity. That is, we used noun wording to turn an easily trivialized decision about whether or not to engage in a given behavior into a seemingly more meaningful one about whether or not to be a particular kind of person. That is, we framed the behavior as relevant to the self-concept. And this subtle linguistic tweak can have startlingly powerful effects on how people behave. Motivating Voter Turnout by Implicating the Self In a first set of experiments testing this idea, we looked at voting. Voting, of course, is a critically important behavior; democracies cannot function properly if citizens do not vote. But voting involves a personal sacrifice. It’s a hassle and, at the end of the day, you know your individual vote will not meaningfully affect the election result. In fact, from the standpoint of individual self-interest, voting is irrational — the costs (in terms of effort, time, etc.) far outweigh any potential individual material benefits (Downs, 1957). But, what is important is that voting is widely approved of. Think how popular those “I voted” stickers are that they distribute at polling places, or of how many of your friends announce on Facebook that they voted. People want others to know that they voted. Perhaps being the kind of person who votes — “a voter” — is seen as a way to be a good or valued person in our society. In two experiments, we tested whether framing voting as relevant to the self-concept would motivate people to go to the polls on Election Day (Bryan, Walton, Rogers, & Dweck, 2011). Participants in both experiments completed what, from their perspective, was a standard pre-election survey the day before or morning of a major statewide election — the 2008 U.S. presidential election in California and the 2009 New Jersey gubernatorial election. Participants1 were recruited for a survey of their beliefs and attitudes about voting in the upcoming election. The survey was brief and the content of the questions was identical for all participants. But the specific wording of 10 questions depended on which of two groups participants were randomly assigned to. For one group, questions referred to voting with a verb (e.g., “How important is it to you to vote in tomorrow’s election?”); for the other, questions used a predicate noun (e.g., “How important is it to you to be a voter in tomorrow’s election?” [underlining not in original]). After participants had completed the survey, their active participation in the experiment was complete; we used state records, after the elections were over, to look up whether or not each participant voted. That small difference in the wording of 10 survey questions had a big effect on whether or not people showed up to vote. In Experiment 1, the “to be a voter” wording increased turnout by more than 13 percentage points compared with the “to vote” wording. In Experiment 2, with a larger and more representative sample, the increase was similar in size: 10.9 percentage points (see Figure 17.1). To understand the magnitude of this effect, consider a comparison: In roughly the past 20 years, there have been two major policy initiatives to increase voter turnout in the United States — “motor voter” laws, which make it easier to register to vote, and “vote-by- mail” laws, which, as the name suggests, allow people simply to mail a ballot in rather than standing in line at a polling place. Both of those policies involved significant and costly changes in infrastructure. And the most optimistic estimates of their effect on turnout have been in the range of 6 to 7 percentage points. Motivating Voter Turnout by Implicating the Self In a first set of experiments testing this idea, we looked at voting. Voting, of course, is a critically important behavior; democracies cannot function properly if citizens do not vote. But voting involves a personal sacrifice. It’s a hassle and, at the end of the day, you know your individual vote will not meaningfully affect the election result. In fact, from the standpoint of individual self-interest, voting is irrational — the costs (in terms of effort, time, etc.) far outweigh any potential individual material benefits (Downs, 1957). But, what is important is that voting is widely approved of. Think how popular those “I voted” stickers are that they distribute at polling places, or of how many of your friends announce on Facebook that they voted. People want others to know that they voted. Perhaps being the kind of person who votes — “a voter” — is seen as a way to be a good or valued person in our society. In two experiments, we tested whether framing voting as relevant to the self-concept would motivate people to go to the polls on Election Day (Bryan, Walton, Rogers, & Dweck, 2011). Participants in both experiments completed what, from their perspective, was a standard pre-election survey the day before or morning of a major statewide election — the 2008 U.S. presidential election in California and the 2009 New Jersey gubernatorial election. Participants1 were recruited for a survey of their beliefs and attitudes about voting in the upcoming election. The survey was brief and the content of the questions was identical for all participants. But the specific wording of 10 questions depended on which of two groups participants were randomly assigned to. For one group, questions referred to voting with a verb (e.g., “How important is it to you to vote in tomorrow’s election?”); for the other, questions used a predicate noun (e.g., “How important is it to you to be a voter in tomorrow’s election?” [underlining not in original]). After participants had completed the survey, their active participation in the experiment was complete; we used state records, after the elections were over, to look up whether or not each participant voted. That small difference in the wording of 10 survey questions had a big effect on whether or not people showed up to vote. In Experiment 1, the “to be a voter” wording increased turnout by more than 13 percentage points compared with the “to vote” wording. In Experiment 2, with a larger and more representative sample, the increase was similar in size: 10.9 percentage points (see Figure 17.1). To understand the magnitude of this effect, consider a comparison: In roughly the past 20 years, there have been two major policy initiatives to increase voter turnout in the United States — “motor voter” laws, which make it easier to register to vote, and “vote-by- mail” laws, which, as the name suggests, allow people simply to mail a ballot in rather than standing in line at a polling place. Both of those policies involved significant and costly changes in infrastructure. And the most optimistic estimates of their effect on turnout have been in the range of 6 to 7 percentage points. So the effect of self-concept-relevance seems to generalize quite broadly, both to another type of behavior that is associated — or can be associated — with a positive self-image, and to children as young as 4 and 5. And this suggests that children that young are already thinking, on some level, about the kind of person they are and playing an active role in shaping that self-concept. Avoiding Negative Identities: The Flip Side of The Theory The premise of all the studies I’ve described so far is that people are motivated to claim positive, desirable identities. But the complementary hypothesis, suggested by my theory, is that people should also be highly motivated to avoid negative identities. So, when they see a potential behavior as diagnostic of a negative identity, they’ll want to avoid engaging in that behavior. In another set of experiments, we tested the hypothesis that nouns that threaten to saddle one with a negative identity should deter the relevant behavior. Showing this would help rule out an alternative explanation for the findings described so far; one involving a more rote, or purely cognitive, priming or self-perception process. According to this account, the noun wording in the experiments so far simply caused people to think of themselves as “voters” or “helpers” and they behaved accordingly, not out of a desire to claim a positive identity, but simply because they were robotically enacting the behavior consistent with a label they had been assigned. But, of course, this account would suggest that the effects of noun wording should always be to increase the likelihood that people will perform the relevant behavior. But, in this case, our theory makes the opposite prediction. To test this, we looked at the effect of noun wording on cheating. My collaborators and I expected that people might feel comfortable cheating from time to time but that nobody wants to think of herself or himself as a cheater (Bryan, Adams, & Monin, 2013). We recruited people to participate online in an experiment about “psychokinesis.” Instructions explained that a recent controversial article claimed to provide the first scientific evidence for “paranormal phenomena” (a vague reference to Bem, 2011), but that critics (including, it was implied, the researchers behind this experiment) expected that those findings would not hold up as additional research was conducted. Participants were instructed to flip a coin 10 times and try to influence the outcome of each coin toss with their minds such that it came out “heads” as often as possible. And, we explained, to ensure that nobody could claim that we had not properly motivated participants to focus on influencing the outcome of the coin tosses with their minds, we would pay them $1 for each time the coin landed heads. Our noun-versus-verb manipulation of self-concept-relevance was embedded in the instructions that followed: NOTE: Please don’t cheat [be a cheater] and claim that one or more of your coin flips landed heads when it really landed tails. Even a small amount of cheating [small number of cheaters] would undermine the study, making it appear that psychokinesis is real. [Bold and underline font appeared in original.] On the next page of the online experiment, participants were instructed to begin flipping the coin and logging the outcome of each toss. At the top of the page, in large, red, block letters was the admonishment “PLEASE DON’T CHEAT” or “PLEASE DON’T BE A CHEATER.” Because participation was online, there was no way for us to determine whether an individual was cheating or not — and, of course, participants knew that. This was by design; we were interested in the effect of self-concept-relevance in deterring cheating in a situation in which there is essentially free money for the taking and people know they cannot be caught if they take advantage. The outcome of interest here is the average number of “heads” participants claimed to have flipped in each of our experimental groups. We conducted two experiments with this design — one with just the “cheater” and “cheating” groups (N=79) and another with those two groups plus a third in which cheating was never mentioned (N=99). In both experiments, we found that participants in the “cheating” condition cheated. That is, the average number of heads (Mstudy 1=5.56; Mstudy 2=6.22) was significantly higher than would be expected with honest reporting (i.e., 5 heads out of 10 flips). In both experiments, the “cheater” wording significantly reduced cheating (Mstudy 1=4.86; Mstudy 2=5.23) and, indeed, we found no evidence that participants in the “cheater” condition cheated at all — the number of heads those participants claimed to have obtained was not significantly different from the number we would expect with honest reporting. And, what about the condition in the second experiment in which cheating was never mentioned? We found a rate of cheating (6.31 reported heads) that was nearly identical to the rate in that experiment’s “cheating” condition (6.22; see Figure 17.3). So, the appeal not to “cheat” had no effect at all, but the appeal not to “be a cheater” effectively eliminated cheating. This is a striking result and it warrants further discussion. To understand how this could be, it is useful to take the perspective of participants: you are participating online and have no connection to the experimenters or their university. So, in the verb condition, the most salient rationale to resist the temptation to claim free money is that someone you don’t know or care about is asking you to. But, in the noun condition, there is an additional rationale: If you cheated, you would have to think of yourself as a cheater. Broadly, this avoidance of a negative identity we’re observing in the cheating experiments illustrates how people respond to these manipulations of self-concept-relevance, not in an automatic or inevitable fashion, but strategically to craft a desired view of themselves. A recent anti-sexual-assault campaign in Canada offers a powerful illustration of the potential importance of this idea. The campaign, aimed at men, features images depicting situations in which sexual assault frequently occurs — a woman passed out drunk on a couch or unable to stand up by herself outside a bar and being helped into a car by a man. Captions communicate unambiguously that sex without overt consent constitutes sexual assault. But the tag line associated with these ads may help make them especially impactful: “Don’t be that guy” (see Figure 17.4). Men may be less likely to commit such a harmful act if they are forced to confront the implications it would have for their self-concepts — to understand that committing rape would make you a rapist. Downstream Effects on the Self-Concept Essentially, I have been arguing that, compared with verb wording, noun wording causes people to see behavior as having stronger implications for their view of themselves — to think of their self-concepts as more heavily contingent on a behavior. So, for example, in the voting studies I described, one might wonder what happened to participants’ selfconcepts after the study was over. Did those in the noun condition who did vote then think of themselves as more civically minded people, and those who didn’t vote as less civically minded people, because the behavior was framed as having implications for their self-concepts? Dominic Alvernaz and I (Bryan & Alvernaz, 2016) examined this idea in the context of another socially approved but effortful behavior: healthy eating. We expected that, if we framed healthy eating as relevant to the self-concept and then led people to perceive their eating habits as healthy, they would then also tend to see themselves as healthy people. And, because a healthy eater is generally thought of as a good thing to be, we expected it might also boost their state self-esteem — their momentary sense of themselves as a good person. And we expected the opposite when we led people to see their eating habits as unhealthy. To examine this, we recruited 156 participants online, making sure none of them was on any special diet that might make it difficult to influence their perceptions of how healthy their eating habits were. To manipulate those perceptions, participants were asked to complete one of two versions of a biased questionnaire. One version asked questions like “How many times in the past week has your lunch or dinner plate been at least half filled with fresh vegetables?” [bolding and underlining appeared in original], with answer choices “0–4,” “5–8,” “9–13,” and “Every meal.” Almost everyone answered “0– 4,” creating the perception that their eating habits were relatively unhealthy. In contrast, the other version of this survey asked much easier questions, like “Do you eat at least one fruit or vegetable in a typical day?” Most people answered “yes” to this question, creating the impression that their eating habits are relatively healthy. In addition to that manipulation, participants completed the now-familiar noun-versusverb manipulation of self-concept-relevance: a survey about their attitudes regarding either “healthy eating” or “being a healthy eater.” After these manipulation questions were completed, we measured our key dependent variables: perceptions of oneself as a healthy person and state self-esteem. We found that the manipulation of selfconcept-relevance produced the expected contingency between behavior and self-concept. That is, participants in the “healthy eating” condition showed no evidence that their perceptions of themselves as a healthy person were affected by being made to see their eating behavior as relatively healthy or unhealthy. But, when we used noun wording to frame that behavior as relevant to the self-concept, participants who were made to see their behavior as healthy also saw themselves as relatively healthier people than participants who were made to see their behavior as unhealthy. And we saw a similar pattern with state self-esteem. There was no tendency for those in the verb condition to feel better about themselves in response to being made to see their eating habits as healthy. But, when the behavior was referred to as “being a healthy eater,” people felt significantly better about themselves if they were made to see their eating habits as healthy than if they were made to see them as unhealthy. So we are finding evidence for exactly the sort of downstream effect on people’s working self-concepts that would be predicted by the theory I have articulated. And this raises the provocative possibility of a perpetual feedback loop — or recursive process — in which, having changed their behavior in response to our manipulation of self-concept-relevance, people experience a corresponding shift in their selfconcepts, which in turn could be expected to lead to additional consistent behavior (see Miller, Brickman, & Bolen, 1975), further reinforcing this new identity (see Figure 17.5). This is provocative because it suggests a possible mechanism by which even an intervention as subtle and brief as this tiny change in wording could potentially lead to lasting and self-sustaining changes in behavior. This is an exciting and important question for future research. A Note of Caution I have painted a picture of noun wording as a simple, low-cost technique with enormous potential to influence behavior in beneficial ways, and I believe that picture to be largely accurate. But noun wording is not a panacea and its use is not without risk. Noun wording invokes the psychology of essentialism — the notion that a person (or thing) has a core, immutable nature, irrespective of circumstance or appearance (Gelman & Heyman, 1999; Gelman et al., 2000). This, I believe, is the key to its power. If you think a behavior will be diagnostic of your fundamental, unchangeable nature on a dimension you consider to be important, you will be very motivated to behave in ways that allow you to reach the desired conclusion about that fundamental nature. But it is precisely because this psychology is powerful that it can also be risky. Imagine, for example, that I appealed to you not to “be a cheater” but the temptation was too great and you cheated anyway (or it was too late and you had already cheated). My work on downstream effects on the self-concept suggests a reason to worry that you might be trapped in the conclusion that you are now and, worse, always will be “a cheater” (see Cimpian, Arce, Markman, & Dweck, 2007). Another important caveat about using this technique in policy becomes apparent when considering that it is not always obvious how desirable people will think a given identity is. Two of my own failed studies illustrate this point well. In one (Bryan, 2008), I gave undergraduates a survey either about “being an energy saver” or about “saving energy.” I found that the former actually seemed to reduce the amount of effort students invested in conservation. The problem here, I suspect, was that “energy saver” sounds childish — not like an identity a college undergraduate would be eager to claim. Relatedly, in an experiment in collaboration with a San Francisco theatre organization (Lupoli, Bryan, Adams, & Monin, 2013), my collaborators and I printed an appeal either to “support” the organization or to “be a supporter” on the front of envelopes containing a fundraising appeal. Again, the noun version of the appeal seemed to decrease the number of people who responded positively. Here, I believe, the problem is that being a “supporter” of an organization feels like a much bigger burden than “supporting” it does. The word supporter conjures images of regularly attending fundraisers and board meetings — more than the identity is worth to most people. The lesson I take from these failures is this: The fact that most undergraduates would probably agree that an energy saver is a “good” thing to be in the abstract, or that most theatre goers would probably admire a “supporter” of their local playhouse, is not enough. If the identity is not personally appealing — if it is not how they want to think of themselves — then noun-based appeals are likely to fail or even backfire. A study by Gerber, Huber, Biggers, and Hendry (2016) highlights a related factor that can influence whether a given identity feels personally appealing to people: the specific context. Gerber and colleagues sought to reproduce the “voter” effect I described above in a very different kind of election. Whereas our original study was conducted during major elections, which were the subject of substantial media coverage and popular attention (the 2008 U.S. presidential election and the 2009 New Jersey gubernatorial election), theirs was conducted during the 2014 U.S. Congressional primary elections during the summer. Thus, the context for the Gerber study was a series of minor elections, almost none of which were seriously contested or received any significant popular attention. These are the kinds of elections that most citizens are barely even aware of and, as Greg Walton, Carol Dweck, and I showed in a subsequent study (Bryan, Walton, & Dweck, 2016), the sort that do not create a context in which “being a voter” feels like the kind of identity that might motivate behavior. Unsurprisingly, Gerber and colleagues failed to produce any significant effect on voting (but see Gerber, Huber, and Fang, 2018; Bryan, Yeager, and O’Brien, 2019). As these examples illustrate, the apparent simplicity of this psychology is deceptive. Its successful implementation depends on a careful analysis of a number of subtle factors. Perhaps chief among those is that the noun one selects must be used in a context in which and with a population to whom it is strongly personally appealing or aversive (depending on one’s goals). One must care whether the identity in question — “voter,” “helper,” “cheater” — is attached to them. So, although I believe this psychology offers enormous promise for policy, it should be applied on a large scale only after much thought and careful testing. Summary and Conclusion In sum, noun wording can motivate socially beneficial behavior — and prevent socially harmful behavior — as people seek to claim positive identities and to avoid negative ones. And noun wording imbues behavior with the power to reshape the self-concept, suggesting the possibility of a positive feedback loop that could result in sustained changes in identity and behavior. It is important to note that noun wording encourages “should behaviors” — the things people know they should do but often don’t bother to. This is important because those sorts of behavioral choices are at the heart of many of our most pressing and important social and policy challenges. In conclusion, the self is a constant work-in-progress. We are continually curating our self-concepts through our behavior — acting in ways that allow us to build an image of ourselves as good, competent, and worthy of social approval. And so, in our efforts to influence behavior, it might be best to focus less on the behavior itself and more on what the behavior would say about the person performing it; to communicate that what you do will reveal who you are. 18 Compliance Without Pressure: The Foot-in-the-Door Technique Jonathan L. Freedman Scott C. Fraser Two experiments were conducted to test the proposition that once someone has agreed to a small request he is more likely to comply with a larger request. The first study demonstrated this effect when the same person made both requests. The second study extended this to the situation in which different people made the two requests. Several experimental groups were run in an effort to explain these results, and possible explanations are discussed. How can a person be induced to do something he would rather not do? This question is relevant to practically every phase of social life, from stopping at a traffic light to stopping smoking, from buying Brand X to buying savings bonds, from supporting the March of Dimes to supporting the Civil Rights Act. One common way of attacking the problem is to exert as much pressure as possible on the reluctant individual in an effort to force him to comply. This technique has been the focus of a considerable amount of experimental research. Work on attitude change, conformity, imitation, and obedience has all tended to stress the importance of the degree of external pressure. The prestige of the communicator (Kelman & Hovland, 1953), degree of discrepancy of the communication (Hovland & Pritzker, 1957), size of the group disagreeing with the subject (Asch, 1951), and perceived power of the model (Bandura, Ross, & Ross, 1963) are the kinds of variables that have been studied. This impressive body of work, added to the research on rewards and punishments in learning, has produced convincing evidence that greater external pressure generally leads to greater compliance with the wishes of the experimenter. The one exception appears to be situations involving the arousal of cognitive dissonance in which, once discrepant behavior has been elicited from the subject, the greater the pressure that was used to elicit the behavior, the less subsequent change occurs (Festinger & Carlsmith, 1959). But even in this situation one critical element is the amount of external pressure exerted. Clearly, then, under most circumstances the more pressure that can be applied, the more likely it is that the individual will comply. There are, however, many times when for ethical, moral, or practical reasons it is difficult to apply much pressure when the goal is to produce compliance with a minimum of apparent pressure, as in the forced-compliance studies involving dissonance arousal. And even when a great deal of pressure is possible, it is still important to maximize the compliance it produces. Thus, factors other than external pressure are often quite critical in determining degree of compliance. What are these factors? Although rigorous research on the problem is rather sparse, the fields of advertising, propaganda, politics, etc., are by no means devoid of techniques designed to produce compliance in the absence of external pressure (or to maximize the effectiveness of the pressure that is used, which is really the same problem). One assumption about compliance that has often been made either explicitly or implicitly is that once a person has been induced to comply with a small request he is more likely to comply with a larger demand. This is the principle that is commonly referred to as the foot-in-the-door or gradation technique and is reflected in the saying that if you “give them an inch, they’ll take a mile.” It was, for example, supposed to be one of the basic techniques upon which the Korean brainwashing tactics were based (Schein, Schneier, & Barker, 1961), and, in a somewhat different sense, one basis for Nazi propaganda during 1940 (Bruner, 1941). It also appears to be implicit in many advertising campaigns which attempt to induce the consumer to do anything relating to the product involved, even sending back a card saying he does not want the product. The most relevant piece of experimental evidence comes from a study of conformity done by Deutsch and Gerard (1955). Some subjects were faced with incorrect group judgments first in a series in which the stimuli were not present during the actual judging and then in a series in which they were present, while the order of the memory and visual series was reversed for other subjects. For both groups the memory series produced more conformity, and when the memory series came first there was more total conformity to the group judgments. It seems likely that this order effect occurred because, as the authors suggest, once conformity is elicited at all it is more likely to occur in the future. Although this kind of conformity is probably somewhat different from compliance as described above, this finding certainly lends some support to the foot-in-the-door idea. The present research attempted to provide a rigorous, more direct test of this notion as it applies to compliance and to provide data relevant to several alternative ways of explaining the effect. Experiment I The basic paradigm was to ask some subjects (Performance condition) to comply first with a small request and then three days later with a larger, related request. Other subjects (One-Contact condition) were asked to comply only with the large request. The hypothesis was that more subjects in the Performance condition than in the One-Contact condition would comply with the larger request. Two additional conditions were included in an attempt to specify the essential difference between these two major conditions. The Performance subjects were asked to perform a small favor, and, if they agreed, they did it. The question arises whether the act of agreeing itself is critical or whether actually carrying it out was necessary. To assess this a third group of subjects (Agree-Only) was asked the first request, but, even if they agreed, they did not carry it out. Thus, they were identical to the Performance group except that they were not given the opportunity of performing the request. Another difference between the two main conditions was that at the time of the larger request the subjects in the Performance condition were more familiar with the experimenter than were the other subjects. The Performance subjects had been contacted twice, heard his voice more, discovered that the questions were not dangerous, and so on. It is possible that this increased familiarity would serve to decrease the fear and suspicion of a strange voice on the phone and might accordingly increase the likelihood of the subjects agreeing to the larger request. To control for this a fourth condition was run (Familiarization) which attempted to give the subjects as much familiarity with the experimenter as in the Performance and Agree-Only conditions with the only difference being that no request was made. The major prediction was that more subjects in the Performance condition would agree to the large request than in any of the other conditions, and that the One-Contact condition would produce the least compliance. Since the importance of agreement and familiarity was essentially unknown, the expectation was that the Agree-Only and Familiarization conditions would produce intermediate amounts of compliance. Method The prediction stated above was tested in a field experiment in which housewives were asked to allow a survey team of five or six men to come into their homes for two hours to classify the household products they used. This large request was made under four different conditions: after an initial contact in which the subject had been asked to answer a few questions about the kinds of soaps she used, and the questions were actually asked (Performance condition); after an identical contact in which the questions were not actually asked (Agree-Only condition); after an initial contact in which no request was made (Familiarization condition); or after no initial contact (One-Contact condition). The dependent measure was simply whether the subject agreed to the large request. Procedure. The subjects were 156 Palo Alto, California, housewives, 36 in each condition, who were selected at random from the telephone directory. An additional 12 subjects distributed about equally among the three two-contact conditions could not be reached for the second contact and are not included in the data analysis. Subjects were assigned randomly to the various conditions, except that the Familiarization condition was added to the design after the other three conditions had been completed. All contacts were by telephone by the same experimenter who identified himself as the same person each time. Calls were made only in the morning. For the three groups that were contacted twice, the first call was made on either Monday or Tuesday and the second always three days later. All large requests were made on either Thursday or Friday. At the first contact, the experimenter introduced himself by name and said that he was from the California Consumers’ Group. In the Performance condition he then proceeded: We are calling you this morning to ask if you would answer a number of questions about what household products you use so that we could have this information for our public service publication, “The Guide.” Would you be willing to give us this information for our survey? If the subject agreed, she was asked a series of eight innocuous questions dealing with household soaps (e.g., “What brand of soap do you use in your kitchen sink?”). She was then thanked for her cooperation, and the contact terminated. Another condition (Agree-Only) was run to assess the importance of actually carrying out the request as opposed to merely agreeing to it. The only difference between this and the Performance condition was that, if the subject agreed to answer the questions, the experimenter thanked her, but said that he was just lining up respondents for the survey and would contact her if needed. A third condition was included to check on the importance of the subject’s greater familiarity with the experimenter in the two-contact conditions. In this condition the experimenter introduced himself, described the organization he worked for and the survey it was conducting, listed the questions he was asking and then said that he was calling merely to acquaint the subject with the existence of his organization. In other words, these subjects were contacted, spent as much time on the phone with the experimenter as the Performance subjects did, heard all the questions, but neither agreed to answer them nor answered them. In all of these two-contact conditions some subjects did not agree to the requests or even hung up before the requests were made. Every subject who answered the phone was included in the analysis of the results and was contacted for the second request regardless of her extent of cooperativeness during the first contact. In other words, no subject who could be contacted the appropriate number of times was discarded from any of the four conditions. The large request was essentially identical for all subjects. The experimenter called, identified himself, and said either that his group was expanding its survey (in the case of the two-contact conditions) or that it was conducting a survey (in the One-Contact condition). In all four conditions he then continued: The survey will involve five or six men from our staff coming into your home some morning for about two hours to enumerate and classify all the household products that you have. They will have to have full freedom in your house to go through the cupboards and storage places. Then all this information will be used in the writing of the reports for our public service publication, “The Guide.” If the subject agreed to the request, she was thanked and told that at the present time the experimenter was merely collecting names of people who were willing to take part and that she would be contacted if it were decided to use her in the survey. If she did not agree, she was thanked for her time. This terminated the experiment. Results Apparently even the small request was not considered trivial by some of the subjects. Only about twothirds of the subjects in the Performance and Agree-Only conditions agreed to answer the questions about household soaps. It might be noted that none of those who refused the first request later agreed to the large request, although as stated previously all subjects who were contacted for the small request are included in the data for those groups. Our major prediction was that subjects who had agreed to and carried out a small request (Performance condition) would subsequently be more likely to comply with a larger request than would subjects who were asked only the larger request (One-Contact condition). As may be seen in Table 18.1, the results support the prediction. More than 50 percent of the subjects in the Performance condition agreed to the larger request, while less than 25 percent of the One-Contact condition agreed to it. Thus it appears that obtaining compliance with a small request does tend to increase subsequent compliance. The question is what aspect of the initial contact produces this effect. One possibility is that the effect was produced merely by increased familiarity with the experimenter. The Familiarization control was included to assess the effect on compliance of two contacts with the same person. The group had as much contact with the experimenter as the Performance group, but no request was made during the first contact. As the table indicates, the Familiarization group did not differ appreciably in amount of compliance from the One-Contact group, but was different from the Performance group (χ2=3.70, p<.07). Thus, although increased familiarity may well lead to increased compliance, in the present situation the differences in amount of familiarity apparently were not great enough to produce any such increase; the effect that was obtained seems not to be due to this factor. Another possibility is that the critical factor producing increased compliance is simply agreeing to the small request (i.e., carrying it out may not be necessary). The Agree-Only condition was identical to the Performance condition except that in the former the subjects were not asked the questions. The amount of compliance in this Agree-Only condition fell between the Performance and One-Contact conditions and was not significantly different from either of them. This leaves the effect of merely agreeing somewhat ambiguous, but it suggests that the agreement alone may produce part of the effect. Unfortunately, it must be admitted that neither of these control conditions is an entirely adequate test of the possibility it was designed to assess. Both conditions are in some way quite peculiar and may have made a very different and extraneous impression on the subject than did the Performance condition. In one case, a housewife is asked to answer some questions and then is not asked them; in the other, some man calls to tell her about some organization she has never heard of. Now, by themselves neither of these events might produce very much suspicion. But, several days later, the same man calls and asks a very large favor. At this point it is not at all unlikely that many subjects think they are being manipulated, or in any case that something strange is going on. Any such reaction on the part of the subjects would naturally tend to reduce the amount of compliance in these conditions. Thus, although this first study demonstrates that an initial contact in which a request is made and carried out increases compliance with a second request, the question of why and how the initial request produces this effect remains unanswered. In an attempt to begin answering this question and to extend the results of the first study, a second experiment was conducted.