The question-behavior effect from an action control perspective

The question-behavior effect is discussed by taking an action control perspective. With respect to the determinants of the effect, it is asked whether its strength is dependent on the features of the behaviors at issue (e.g., difficulty), how the question targets the critical behaviors (e.g., in terms of attitudes or intentions, frequency or intensity), and in which context it is asked. With respect to the psychological processes, it is questioned whether the activation of positive implicit attitudes underlies the effect. Finally, to counter unwanted question-behavior effects on risky behaviors, an intervention treatment geared at blocking such effects is suggested that is based on the self-regulation of goal setting (by mental contrasting; Oettingen, 2000) and goal implementation (by forming if-then plans; Gollwitzer, 1999). The psychology of action (Gollwitzer & Bargh, 1996; Oettingen & Gollwitzer, 2001) has developed conceptual tools that allow for the analysis of questions such as: When is action control easy or difficult, failure-ridden or successprone, conscious and effortful or automatic and efficient? Research using these conceptual tools sheds light on how action control is guided by people's motives, their wishes and attitudes, their various kinds of expectations, goals, and plans as well as their evaluative thoughts with respect to achieved outcomes. In the present commentary, we will use the research perspective of the psychology of action to reflect on the question-behavior effect as discussed by Fitzsimons and Moore (2008). Fitzsimons and Moore caution to ask adolescents about risky behaviors (e.g., drugs, alcohol, dangerous sports, unsafe sex, smoking), because this might increase the frequency of engaging in exactly these risky behaviors. Shall parents therefore stay away from asking their children about risky behaviors? And shall researchers and educators stop conducting individual and population screening? On a first glance, it always seems to be beneficial for parents, educators, health professionals and people who design public policies to learn about the extent of adolescents’ engagement in various kinds of risky behaviors, as such information yields important benefits for the individual, the care taker, and the society at large (e.g., individuals can avoid tempting situations, caretakers can encourage alternative engagements, researchers and policy makers can corroborate in developing targeted intervention programs for prevention and cure). But Fitzsimons and Moore raise an intriguing counter-argument stating that these benefits may be offset by the question-behavior effect: People who are asked about a particular behavior show an increased probability of carrying out this very behavior. In our opinion, the question of costs and benefits of asking people about their risky behaviors is a question that the relevant decision makers (e.g., parents, politicians, health professionals) have to answer for themselves. The scientists, on the other hand, have to deliver the necessary factual information.More specifically, scientists are to provide to decision makers factual information on the strength of the question-behavior effect, its determinants and processes, and how it can be prevented and attenuated. In the following, we will address these issues one by one. Strength of the question-behavior effect With respect to the first question (how strong is the effect?), a sophisticated meta-analysis is needed that considers all of the individual studies on the question-behavior effect that have been conducted so far. With a sophisticated analysis we mean that differential ds are computed for different types of samples (e.g., children, adolescents, adults; students, professionals; men, women), different types of behaviors ⁎ Corresponding authors. E-mail addresses: [email protected] (P.M. Gollwitzer), [email protected] (G. Oettingen). (e.g., consumer, risky, communication), and behavior in different life domains (e.g., interpersonal, achievement, health). Regarding the Fitzsimons and Moore paper, it would be important to know whether risky behaviors show significantly different ds for the question-behavior effect than non-risky behaviors, and whether the behavior-question effects for these risky behaviors are of a small, medium, or large magnitude. It is vital to know for computing reliable costbenefit analyses (of whether one should ask adolescents about certain risky behaviors) how strong the question-behavior effects actually are for the risky behaviors at issue. Determinants of the question-behavior effect Let us now turn to the determinants of the question-behavior effect. First, an important determinant of the strength of the effect may be features of the risky behavior itself. For example, it is possible that only easy to implement behaviors are vulnerable to the effect (e.g., accepting an offered drug) while more difficult to implement behaviors are protected from it (e.g., trying to buy drugs). The strength of the question-behavior effect might also depend on whether the critical behavior is linked to norms embraced by the majority of a society (sun bathing, alcohol consumption) or to norms embraced only by a sub-culture (drug use, tattooing). Second, the question-behavior effect may also be determined by whether the question targets the critical behavior directly or indirectly. Questions targeting the behavior indirectly may focus on the psychological concepts related to this behavior such as relevant attitudes, expectations, and goals (intentions), as well as plans to carry out the behavior. If attitudes are targeted, what dimension of attitude is inquired about (e.g., ambivalence)? If expectations are targeted, how are they conceptualized (in terms of efficacy, outcome expectations, or general expectations)? Are the certainties of holding these expectancies asked for as well? If goals (intentions) are assessed, is it the strength of the commitment or the quantity and quality of the standard? And if the set standard is asked for, does it pertain to a specific performance outcome or to merely doing one's best? If a plan is asked about, is it an outcome versus process simulation, or an ifthen statement (i.e., an implementation intention that specifies what goal-directed behavior one wants to perform if a certain critical situation will be encountered)? Third, already the mere framing of the question may weaken or strengthen the question-behavior effect. Fitzsimons and Moore mention temporal focus (future versus past), valence focus (positively versus negatively), and means focus (approach versus avoidance) as well as the anchoring of the answer scale (7 times a week versus 8 times a week). Whether the question asks for frequency versus intensity of the behavior might also play a role. Fourth, the context as well as the physical and temporal location in which the question is asked might be a determinant of the question-behavior effect. For example, does the question appear in the context of other questions also addressing risky behavior or among questions addressing comparatively neutral behaviors? What reference points do the instructions use? For example, do the instructions implicitly convey the assumption that similar others are engaging in risky behavior too, that engaging in these risky behaviors is known or even accepted by the adult world, or that the risky behavior is a means to assert oneself against authority? Further, in what location do the adolescents answer the question? Are they providing their answer outside the critical context long before they are faced with the opportunity to act in a risky way (e.g., in school) or while they are already facing the decision to act on the risky behavior (e.g., on the internet or in public places such as parties or cafes)? Psychological processes underlying the question-behavior effect Fitzsimons and Moore propose that asking adolescents about their risky behaviors activates their implicit positive attitudes toward these behaviors which in turn facilitate carrying out these behaviors. Such a process hypothesis assumes a) that adolescents have implicit positive attitudes with respect to risky behaviors, and b) that it is the activation of these implicit positive attitudes what leads to risky behavior. Are these assumptions justified? What is the empirical basis of the assumption that adolescents hold implicit positive attitudes towards risky behaviors? Are there studies systematically assessing adolescents’ implicit attitudes towards risky behaviors by using various implicit measures of attitudes (e.g., the Implicit Association Test, IAT; the auto-evaluation test)? Do studies using these procedures indeed detect implicit positive attitudes towards risky behavior in adolescents? How strong are these positive attitudes, and how stable are they over time? With respect to the second assumption (the activation of implicit positive attitudes leads to risky behavior), we wondered whether a positive implicit attitude activated in the questioning situation stays in a state of activation until the opportunity to act is finally coming along. Such an enduring heightened state of activation seems rather unlikely as the activation of mental constructs is commonly known to decay over time. Thus the activation of an implicit positive attitude may produce questionbehavior effects only when the adolescents answer the question in the same context in which they also have an opportunity to perform the risky behavior (see the moderating role of context as pointed out above). Therefore, we suggest that questioning adolescents about risky behaviors may not only affect the state of activation of the implicit attitude but also increase the level of positivity of the implicit attitude. In order to test these competing (or complementing) process hypotheses, studies that allow computing mediation analyses would need to be conducted. In such studies, adolescents are first asked about risky behaviors, and then classic procedures of assessing the activation status versus the positivity of their implicit attitudes towards these behaviors are employed. Finally, participants are given the chance to enact the risky behaviors, and this opportunity would be provided either immediately or delayed. Statistical analyses that support 108