Forbid/Allow Asymmetry in Persuasion
The Forbid Frame Decreases Biased Elaboration and Increases Attitude Change
Abstract
Abstract. Previous research showed that responses to questions about forbidding something differed from those to the seemingly equivalent questions about allowing the same object (forbid/allow asymmetry). We postulate that the effect of the forbid vs. allow framing may be also consequential for the processing of attitude related information and attitude change. The forbid frame (compared with the allow frame) may increase the impact of negative (vs. positive) arguments and/or reduce the impact of initial attitudes on the elaboration the presented information. To test these predictions we conducted three experiments (one preregistered, total N = 655). Participants were reading both pro and con arguments, differing in consistency with their initial attitudes, and concerning three different attitude objects: genetically modified organisms (GMOs), euthanasia, and barbecuing in public places. The results show that the forbid (vs. allow) frame decreases the tendency for generating thoughts prevailingly consistent with participants, initial attitudes (Experiment 2). It also reduces bias in the evaluation and interpretation of the presented arguments and yields more similar assessments of arguments that are consistent and inconsistent with initial attitudes (Experiment 3). As a result, the attitudes are more susceptible to change within the forbid frame (they move more in the direction opposite to the initial attitude) than within the allow frame (Experiments 1-3). The results for the first time show the existence of forbid vs. allow asymmetry in persuasion. This effect has practical consequences, e.g., when designing referenda.
In September 2018, in the Swiss canton of St. Gallen, a referendum was held about covering one's face in public (Shields, 2018). The media called it the burka ban referendum (“Voters approve ‘burka ban’ in St Gallen,” 2018). Referendums are held fairly often in Switzerland, and at least some of them focus on whether something should be forbidden. For instance, in 2008, the Swiss voted on whether to forbid training flights of fighter aircraft over tourism areas in times of peace (Khetani, 2012), and in 2009, the Swiss voted on whether building minarets should be banned (Wyler, 2017). In each of these cases, the issue put to vote could have just as easily been presented within the allow frame. Instead of asking whether training flights or minarets should be banned (forbidden), the question could be whether these objects should be allowed.
The effect of wording questions in public opinion surveys is well known: Asking about forbidding something produces different responses than asking about allowing something (e.g., Chessa & Holleman, 2007; Hippler & Schwarz, 1986; Rugg, 1941). In a classic study, Rugg (1941) showed that if the question concerned allowing public speeches against democracy, 75% of the respondents chose an answer indicating that they were against such speeches (i.e., they answered the speeches should not be allowed). However, if the question concerned forbidding public speeches against democracy, only 54% of the respondents were against such speeches (i.e., they answered they supported the ban). Subsequent studies confirmed the existence of the forbid versus allow asymmetry (Hippler & Schwarz, 1986; Holleman, 2006; Schuman & Presser, 1981): Responses to questions about forbidding something differed from those to the seemingly equivalent questions about allowing the same object. Specifically, more people were willing not to allow something than to forbid it, and vice versa – more persons were willing to not to forbid something than to allow it.
We postulate that the effect of the forbid versus allow framing is not limited to differences in responses to surveys or referendums. This effect can also concern the processes that take place when the respondent comes into contact with arguments during a referendum campaign. Instead of trying to persuade the voters that training flights or minarets should be or should not be banned (forbidden), the prereferendums campaigns could be focused on whether these objects should be allowed. However, would the attempts at persuasion have progressed the same way in both cases (forbid vs. allow)? It is possible that framing can affect how different types of arguments and information (pro vs. con and attitude-consistent vs. attitude-inconsistent) are processed. Moreover, these potential differences in elaboration of different types of arguments can also affect the scope and direction of attitude change.
Factors Affecting Elaboration of Different Types of Arguments
During prereferendum campaign, citizens are confronted with positive and negative information (pro and con arguments) concerning the object of the referendum, but these different types of arguments or information are not evaluated and processed in the same way. As research on the negativity effect showed, people are more attentive to negative than positive information, and the negatives shape attitudes stronger than the positives (Baumeister et al., 2001). As a result, con arguments are more persuasive than pro ones (Cobb & Kuklinski, 1997).
Moreover, looking at arguments as positive or negative (pro or con) is only one way of encoding them. In addition to this descriptive encoding, these arguments can also be encoded evaluatively – as compatible or incompatible with the person's preexisting attitude (Wyer et al., 1991). As a result, influence of pro versus con arguments can be moderated by the compatibility or incompatibility of these arguments to the preexisting attitude. Arguments consistent with attitudes are accepted uncritically, while the inconsistent ones are prone to critical scrutiny and, as a result, are evaluated as weaker than consistent ones (disconfirmation bias; Edwards & Smith, 1996; Taber et al., 2009). This biased elaboration (biased assimilation; Lord et al. 1979; or biased evaluation; Cohen et al., 2000) leads to the perception of incoming evidence as supporting one's previous attitude, which in turn allows leaving this attitude unchanged or even to bolster it (Kuhn & Lao, 1996; Lord et al., 1979).
Both the tendency to concentrate on the negatives, as well as the tendency for a biased elaboration of evidence, may be moderated by individual-related factors, for example, by the strength of initial attitude (e.g., Pomerantz et al., 1995; Taber et al., 2009). What from our study perspective is more interesting – the aforementioned effects can be also moderated by situational or external factors (e.g., Cohen et al., 2000; Lord et al., 1984). Simple and seemingly unimportant variations in framing the question or task can change the way the object-related information is processed or weighted. This may also increase responders' propensity to correct their automatic or biased responses and affect the susceptibility of the attitudes to change (e.g., Bizer & Petty, 2005; Sokolova & Krishna, 2016).
Here, we focus on the attribute framing, where only a single attribute in the object's description is changed, which results in differences in the evaluations of the object (Levin et al., 1998). According to Levin and Gaeth (1988), attribute-framing effects occur due to information being encoded relative to its descriptive valence. Positive frames lead to an encoding of the information that evokes favorable associations in memory, and negative frames evoke unfavorable associations. However, this is not only a matter of associations. Research on the task-type effect (which is an example of the attribute framing) showed that framing affects salience of different features of the object. In the task-type paradigm, the person has to decide which of the available options to choose or which of them to reject. For example, a person can be asked which of the two dishes offered they choose (positively framed task), but the question could also be which of the two dishes they reject (negatively framed task). With a positively (vs. negatively) framed task, people focus more on the positive (vs. negative) attributes of the objects. As a result, a rejection task, when compared with a choice task, showed a greater impact of negative information on decision-making (Ganzach, 1995; Shafir, 1993; Shafir et al., 1993; Wedell, 1997). Ganzach and Schul (1995) state that this effect may be mediated by confirmation bias. If the decision is framed as an acceptance decision, people are likely to test hypotheses concerning the acceptance of alternatives, and therefore, they attend to, and rely on, positive information more than negative information. On the other hand, when the decision is framed as a rejection decision, the reverse is likely to occur.
Moreover, in the case of previously existing attitudes, framing may also affect encoding and reliance on attitude consistent versus inconsistent information. It may also reduce the tendency for preferential treatment of information or arguments consistent with an initial attitude. Laran and Wilcox (2011) propose that, in a choice task, people focus more on the options consistent with the currently activated preferences and goals. In a rejection task, they focus on the preference-inconsistent alternatives. As a result, a rejection task shifts preferences toward options that are inconsistent with one's baseline preference. Other research showed that simply switching the task from choice to rejection led to more deliberative processing and increase responders' propensity to correct their automatic or biased responses (Sokolova & Krishna, 2016).
To sum up, attribute framing can change associations evoked by particular arguments and switch the salience of these arguments. This may result in changing the impact of different arguments (pro vs. con and consistent vs. inconsistent with initial attitude) on final attitude, which in turn can affect the susceptibility of attitudes to change and determine the direction of this change.
Forbid Versus Allow Framing in Persuasion
To date, no studies have addressed susceptibility to persuasion in the forbid versus allow frame. However, for several reasons, we can expect that formulating the question within the forbid versus allow frame may lead to differences in the respondents' susceptibility to persuasion.
First, as Chessa and Holleman (2007) suggest, the forbid versus allow asymmetry in public opinion surveys is the consequence of an increased response threshold in forbid answers. If respondents are expected to take an extreme position (especially if the respondent is expected to agree to a ban), they may want “to retrieve extra arguments to be sure about their stance, or to defend it to an outsider” and, as a result, “an extra (negative) attribute of the same attitude or attitudinal network is being activated” (p. 222). This suggests the possibility that answering a question about forbidding something may be associated with a stronger susceptibility to the presented arguments than answering a question about allowing something. Moreover, it may lead to increased focus on some specific (i.e., negative) information.
Second, the forbid versus allow framing, like choosing versus rejecting framing, can be considered as an example of attribute framing. Moreover, both rejecting and forbidding, as well as choosing and allowing, have common characteristics.1 Both rejecting and forbidding involve a lack of possibility for further action related to the object that has been rejected or forbidden (Holleman, 2006). Similarly, both choosing and allowing involve creating a possibility for further action related to the object that has been chosen or allowed. Moreover, in Polish, that is, the language in which this study was conducted, the words “reject” and “forbid” (odrzucać and zakazywać) and the words “choose” and “allow” (wybierać and pozwalać) have similar evaluative and emotional connotations (Wierzba et al., 2015). The valence of both odrzucać (reject) and zakazywać (forbid) is negative, and the level of excitement evoked by these words is similar. The valence of both wybierać (choose) and pozwalać (allow) is slightly positive, with a slightly lower level of excitement evoked than with negative counterparts. Therefore, we can expect some analogies between the task-type effect described earlier and influence of forbid versus allow frame.
Third, the word “forbid” seems to have some specific status (especially when compared with the word “allow”; cf. Chessa & Holleman, 2007; Schuman & Presser, 1981). This word is connected with different signs informing that some actions cannot be taken in the given time or situation. Moreover, in many cases, the word “forbid” (or sign representing it graphically) implies that an ongoing activity cannot be continued or at least that the person should decide whether their activity is breaking the communicated rule. In the course of life, people are learning to associate the word “forbid” with stopping the action, an interruption of ongoing activity, an obstacle to fulfilling their goals, a disagreement with some activity, the consideration of potential fines, etc. It is possible, then, for the word “forbid” to have the ability to induce some specific state, where caution is evoked and previous actions or goals are reassessed; the state when people are motivated to attend to all relevant information more carefully and objectively (Kunda, 1990). In the context of persuasion, this can result in lowering the reliance on previously existing attitudes. As Houston and Fazio (1989) noted, when a person is found in a situation with potential serious consequences, they may be motivated to evaluate the incoming information with less attitudinally driven processes. The consistency or inconsistency of information with previous attitudes can become less important than an accurate evaluation of the situation. Persuasive consequences of this state can be similar to the effect of fear on the processing of information. Fear increases cautiousness and motivates a search for information, decreases the salience of prior beliefs and reliance on previous preferences, and encourages reconsideration of choices based on contemporary evaluations, which results in increasing the susceptibility to persuasion (Brader, 2005; Griskevicius et al., 2009; Koniak & Cwalina, 2020).
On the other side, the word “allow” is connected rather with the possibility of doing, the continuation of ongoing activity, an absence of obstacles to fulfilling actual goals, agreement, etc. In the course of life, people learn that when something is allowed, there is no need for constraining and re-evaluating their existing goals, that obstacles should not be expected, and that ongoing actions can be continued. There is no need to consider information other than one's own preferences. In the context of attitude change, this can result in lowering reliance on the presented arguments and increasing reliance on previously existing attitudes. Moreover, the impact of preexisting attitudes on the evaluation and interpretation of presented information or arguments should not be interrupted. As a result, preexisting attitudes can be kept without significant change.
Therefore, it can be expected that forbid versus allow framing will have consequences for the processes of changing attitudes toward the framed object. Specifically, we can expect that forbid versus allow frame can result in increased susceptibility of attitude to change. This change can result from how the negative versus positive (or attitude-consistent vs. attitude-inconsistent) information about the object is weighted and processed and even what of this information is seen as more persuasive.
Pro Versus Con Information
When people are considering whether an object should be forbidden, the arguments or information concerning negative aspects of the object (con arguments) should be more influential than those concerning positive ones (pro arguments). When they are considering allowing this object, the reverse is likely to occur. The differential influence of these arguments can be reflected in a change in their evaluation, in the ease of processing or memory for positive versus negative information, as well as in the pattern of eventual attitude change. From this perspective, which can be called a descriptive valence account, we can expect that when both pro and con arguments are presented, attitudes toward the object should change more in the negative direction when the forbid frame is used and/or more in the positive direction when the allow frame is used. Moreover, because of the word “forbid” arousing more excitement than the word “allow” (Wierzba et al., 2015), as well as for the negativity effect, we can expect the forbid frame to have a stronger ability to focus people on negatives than the allow frame would have to focus them on positives. As a result, we can expect that the absolute amount of attitude change will be higher in the case of the forbid frame than that of the allow frame.
Attitude-Consistent Versus Attitude-Inconsistent Information
From another perspective, which can be called an evaluative encoding account, it can be expected that forbid (vs. allow) framing would mitigate the overall tendency to rely mostly on arguments and information consistent with the previous attitude (regardless of their descriptive valence). Moreover, it can also mitigate the tendency to discredit inconsistent arguments and/or the preferential treatment of consistent ones, that is, it can reduce the biased processing of information or arguments. This can result in weakening of tendency to perceive or interpret presented evidence as consistent or supportive of one's previous attitude. As a result, we can expect that when both arguments consistent and inconsistent with previous attitude are presented, and the forbid frame is used, attitudes toward the object should change more in a direction opposite to the initial attitude. When the allow frame is used, we can expect the attitude to remain unchanged or even to slightly polarize in a more extreme direction consistent with initial attitude (Kuhn & Lao, 1996; Lord et al., 1979).
However, it is also possible that forbid versus allow framing affects attitude change in a more direct way, where the mediating role of biased interpretation and/or evaluation of arguments is unnecessary. If the forbid frame is inducing some caution, the previous attitude can be simply withdrawn or at least mitigated because of this caution regardless of how this framing is affecting the perception and evaluation of arguments.
Study Overview
In this study, we are reporting the results of the three experiments (one preregistered) dealing with the consequences of forbid versus allow framing on changing preexisting attitudes (toward different objects: GMOs, euthanasia, and barbecuing in public places). First, in Experiments 1–3, we have tested whether the forbid versus allow framing leads to a decrease in the biased elaboration, evaluation, and interpretation of presented arguments. In Experiments 2 and 3, we have also verified whether the forbid versus allow frames simply affects processing of con versus pro arguments (descriptive valence account) or whether the relationship of these arguments to the preexisting attitude is crucial (evaluative encoding account). Next, we have tested whether forbid versus allow framing results in a different pattern of attitude change and whether this pattern is consistent with predictions of the descriptive valence account or evaluative encoding account (Experiments 1–3). We have also tested whether this attitude change is mediated by a decrease in the biased processing, evaluation, and interpretation of presented arguments (Experiments 2 and 3). Moreover, in Experiments 1 and 3, we have also tested potential alternative explanations for the proposed forbid/allow asymmetry in persuasion.
For all experiments, we report how we determined our sample size, all data exclusions (if any), all manipulations, and all measures. The experimental procedures were approved by the Ethics Commission of SWPS University of Social Sciences and Humanities.
Experiment 1
The primary aim of Experiment 1 was to test whether the forbid versus allow framing affects susceptibility to the persuasion of attitudes toward forbidding or allowing barbecuing in public parks. Moreover, the arrangement of this study also helped to determine whether an account alternative to the ones mentioned above should be taken into consideration. According to Sokolova and Krishna (2016), the task-type effect can be explained by the deliberative processing account – they conclude that the rejection frame led to more deliberative processing. However, more deliberative processing (or elaboration, to use the persuasion studies terminology) can be decomposed as two separate processes (Petty & Cacioppo, 1986). First, the direction of processing may be affected, where more deliberative elaboration results in a decrease in biased processing (this is, in fact, equal to what we called the evaluative encoding account). Second, it can also be a matter of increased extent of elaboration, where the amount of elaboration is affected regardless of the direction of this elaboration. In this experiment, we have checked whether the forbid versus allow framing also affects the extent of the processing.
Method
Participants and Design
A total sample of 151 participants (Mage = 23.05 years; SD = 3.46; 97 women, 59.2%) were recruited to participate in the experiment via social media (invitations to participate were sent by email and posted on Facebook pages and other social media). A larger sample size was desired for this study; however, because our participants were recruited from the student sample, data were collected for as many participants as possible before the end of the semester. All participants completed the survey online and were not compensated. The participants were randomly assigned to a 2 (attitude question framing: forbid vs. allow) × 2 (strength of arguments: strong vs. weak) between-participant factorial design. Participants evaluated the presented arguments, and the change of their attitudes was assessed by the repeated measures (attitudes toward the issue before vs. after the message). A sensitivity analysis conducted with G*Power (Faul et al., 2007) showed that our sample sufficed to detect the effects of f = 0.23 with a power of 0.80 for a between-subject ANOVA and f = 0.14 for a mixed-design ANOVA.
Material and Procedure
The participants were first asked about their attitudes toward various issues: forbidding cycling on the sidewalk, allowing GMO crops, requiring every dog to be walked on a leash and aggressive breeds to wear also a muzzle, barbecuing in public parks, requiring drivers to use winter tires between November and March, forbidding the use of mobile phones in schools, and allowing the sale of homemade alcohol. The only topic of interest in this study was the attitude toward barbecuing in public parks; all the other topics were buffer questions. The participants were randomly shown one of two versions of the question about barbecuing in public parks: One group declared their attitude toward forbidding it and the other declared their attitude toward allowing it. They declared their attitude toward each issue on three 11-point items ranging from −5 (very strongly against this proposal; this is a very bad idea; this should definitely not be introduced) to 5 (very strongly in favor of this proposal; this is a very good idea; this should definitely be introduced). The reliability of this scale regarding the attitudes toward barbecuing in public parks was α = .98.
A pilot study (N = 26) indicated that the population from which we planned to recruit the participants of Experiment 1 was dominated by the proponents of barbecuing in public parks. For this reason, only those participants who were in favor of barbecuing in public parks were recruited to Experiment 1 (i.e., those who answered higher than 0 to the question about allowing it and those who answered lower than 0 to the question about forbidding it).
After the initial measurement of their attitudes, the participants read a message containing only arguments that were inconsistent with their attitude, that is, arguments against barbecuing in public parks. One group read a message containing strong arguments (e.g., “Barbecuing in public places will result in these spaces getting more litter”), and another group read a message containing weak arguments (e.g., “Vegetarians, vegans, and other persons who do not eat meat may feel excluded from the public space”). Each version of the message contained three arguments, and the strength of these arguments had been tested in a pilot study (N = 25). The author of the message, described as Professor Andrzej Goździcki, President of the Polish City Planners Forum, explicitly declared at the beginning and end of the message that he was against barbecuing in parks owned by the city or village and that in his opinion, public parks were not the place for such entertainment as barbecuing.
Next, the participants used scales ranging from −5 to 5 to assess the strength of the arguments in the message (very unconvincing–very convincing, makes no sense–makes total sense, very weak–very strong, completely unimportant–very important; α = 0.96) and the direction of the arguments (strongly against barbecuing in public parks–strongly in favor of barbecuing in public parks). Last, the participants again declared their attitudes toward forbidding or allowing barbecuing in public parks using the same scale as at the beginning of the experiment (α = .98).
Results
Evaluation of Arguments
We submitted both the perceived location of arguments on the against-for scale and the assessment of their strength to a two-way ANOVA. The assessment of the direction of arguments was not affected by the frame, argument strength, or the interaction between these factors (Fs < 1).
As intended, the weak arguments were perceived as weaker than the strong arguments (M = −0.52, SD = 2.89 and M = 0.67, SD = 2.87, respectively; F(1, 147) = 6.44, p = .012, ηp2 = .042). The assessments of the strength of the arguments were not in any way affected by the frame (Fs < 1).
Attitude Change
The participants' attitudes were recoded from attitudes toward forbidding or allowing barbecuing in public parks into an index of attitudes toward barbecuing in public parks. We reversed the responses given by the participants who had been asked about forbidding barbecuing in public parks such that high values in both the forbid and allow groups indicate a more positive attitude toward barbecuing in public parks. Next, we conducted a 2 (attitude question framing: forbid vs. allow) × 2 (strength of arguments: strong vs. weak) × 2 (attitude measurement: before vs. after the message) ANOVA (where the last factor was within-participants). Here, as well as in the further experiments, we report only the results directly related to our predictions and to the influence of the forbid versus allow frames. Due to the space limitations, we report the results which are of secondary importance (i.e., the effects not interacting with the frame) in the Electronic Supplementary Material, ESM 1.
We found a significant interaction between the frame and the time of attitude measurement, F(1, 147) = 5.38, p = .022, ηp2 = .035. As shown in Figure 1, before the presentation of the message, participants in the forbid (M = 3.31, SD = 1.46) and allow (M = 3.54, SD = 1.43) groups had similar attitudes toward barbecuing in public parks (F(1, 147) = 1.01, p = .317). After the presentation of the counter-attitudinal message, attitudes in both the allow (M = 2.92, SD = 2.48; F(1, 147) = 5.20, p = .024, ηp2 = .034) and the forbid groups shifted significantly in a more negative direction (M = 1.78, SD = 2.54; F(1, 147) = 29.66, p < .001, ηp2 = .168). However, as a result of this shift, the final attitudes were less positive in the forbid than allow group (F(1, 147) = 7.83, p = .006, ηp2 = .051), which indicates that the amount of change was higher within the forbid than allow group. Finally, neither of the interactions with the strength of arguments was significant (Fs < 1).
Discussion
Attitudes toward forbidding barbecuing in public parks were more susceptible to persuasion than attitudes toward allowing barbecuing in public parks. Moreover, we did not find any difference in the initial attitudes between the forbid and allow groups. In other words, the forbid/allow asymmetry in persuasion worked even when the forbid/allow effect known from the survey research did not occur.
Although manipulating the strength of the arguments proved to be effective, attitude change was independent from the quality of the presented arguments. Thus, in this experiment, we did not find any evidence that the forbid versus allow framing leads to a different extent of elaboration (Petty & Cacioppo, 1986). As a result, the difference in the scope of elaboration does not seem to be a plausible explanation for the observed difference in attitude change. It seems that it is the nature of this elaboration that is crucial here. The pattern of results of Experiment 1 is aligned with the overall prediction that the forbid frame increases the influence of negatives. However, because participants' preexisting attitudes were unipolar and positive, the arguments we have presented them were, at the same time, the con and the attitude-inconsistent. For this reason, the results do not allow us to indicate which of the accounts proposed (descriptive valence or evaluative encoding) is a more valid explanation of how forbid versus allow framing works. We will explore this issue in the next two experiments.
Finally, we would also expect that framing should influence the perception or evaluation of the arguments presented – an effect we did not find in this experiment. It is possible that this effect is not detectable with such direct measures as an evaluation of arguments. These processes can manifest, for example, in the fluency of processing different types of arguments, etc. It can also be detectable by tracking thoughts generated by the participant when they are thinking about forbidding versus allowing. Moreover, here only one type of argument was presented (con or inconsistent). It is possible that differences in evaluations are easier to detect when both pro and con (consistent and inconsistent) arguments are presented (see Edwards & Smith, 1996; Lord et al., 1979). It should also be stressed that framing effect can manifest itself in differential treatment of negative (con or inconsistent) or positive (pro or consistent) information, or both. In other word, it can affect the perception or evaluation of one type of information without affecting the perception or evaluation of the other one. Moreover, the perception of which type of information will be more strongly affected can be the result of the initial views held by the participants. For example, we can expect that for proponents, as the group holding weaker attitudes and thus inclined to bolster these attitudes, framing will primarily affect the perception of arguments consistent with an attitude (see Bizer et al., 2013, 2011; Clark & Wegener, 2013). It is possible, then, that by measuring the evaluation of only con (inconsistent) arguments, we simply missed the opportunity to catch changes in the evaluation of arguments that are positive. All the above considerations have been taken into account in the following two experiments.
Experiment 2
In this experiment, we have investigated how the forbid versus allow framing affects attitude change toward different objects – GMOs and euthanasia. These issues were chosen due to the fact that they allow for a distinction between two groups of participants: the opponents and the proponents. The participants were presented with messages containing a set of arguments for and against GMOs or euthanasia. This allows us to manipulate both encoding of arguments (pro-con and consistent-inconsistent) orthogonally. Moreover, to capture eventual shifts in the focus on negative versus positive arguments, we have used the thought-listing technique.
Method
Participants and Design
A total sample of 175 participants (Mage = 35.03 years; SD = 8.86; 96 women, 54.9%) were recruited to participate in the experiment via social media. A larger sample size was desired for this study; however, because our participants were recruited from the student sample, data were collected for as many participants as possible before the end of the semester. All participants completed the survey online and were not compensated. They were randomly assigned to a 2 (issue: GMOs vs. euthanasia) × 2 (attitude question framing: forbid vs. allow) between-participant factorial design. After the presentation of the message, participants listed thoughts that came to their minds when reading the message, and the change of their attitudes was assessed by the repeated measurements (attitudes toward the issue before vs. after the message). For an analytical reason, based on the initial attitude scale, we divided participants into two groups: proponents and opponents of the discussed issue (details described later). Eleven participants were excluded from further analysis due to being identified as outliers with an extremely high attitude change index (equal to a full range on the attitude scale). Their exclusion does not affect the results substantially.2 A sensitivity analysis conducted with G*Power (Faul et al., 2007) showed that our final sample was sufficient for detecting the effects of f = 0.15 with a power of 0.80 for a mixed-design ANOVA.
Material and Procedure
The participants were randomly assigned one of two issues: Half of the participants indicated their attitudes toward GMO crops and read a message about GMOs, and the other half indicated their attitudes toward euthanasia and read a message about euthanasia. The initial measurement of the participants' attitudes toward GMOs or euthanasia was placed between questions about other attitudes: legalizing same-sex marriage, introducing the death penalty, penalizing possession of marijuana, removing religion classes from the school curriculum, and building a nuclear power plant. The issue of interest in this study was the question about GMOs or euthanasia; all other topics were buffer questions.
The participants in the GMOs group were randomly assigned one of two versions of the question about GMOs: Half of the participants declared their attitudes toward forbidding GMO crops and the other half declared their attitudes toward allowing GMO crops. Similarly, the participants in the euthanasia group were randomly assigned one of two versions of the question about euthanasia: They declared their attitudes toward either forbidding or allowing euthanasia. The participants indicated their attitudes toward each issue using an 11-point scale ranging from −5 (strongly against) to 5 (strongly in favor). Because the aim of the experiment was to measure the susceptibility of attitudes to change, only those participants who declared having an attitude toward the issue (responses between −5 and −1 and between 1 and 5) proceeded with the experiment.3
Next, the participants were informed that they will see a message about one of the aforementioned issues, in their case, the GMO crops (in one group) or euthanasia (in the second group). Participants assigned to the GMO version read a message with three pro arguments (e.g., “GMO farming provides high yields of crops even with bad soil”) and three con arguments about GMOs (e.g., “The long-term effect of GMOs on human health is difficult to predict”). Similarly, participants in the euthanasia group read a message with arguments for (e.g., “A small number of medical errors or inexplicable recuperations cannot justify the inordinate amount of pain and suffering experienced by many other, terminally ill patients”) and against euthanasia (e.g., “The wish to undergo assisted suicide can be analogous to the wish to die caused by depression, and can also be eliminated using the same means as with other groups of patients with suicidal thoughts”). All arguments had been tested in pilot studies (N = 15 for GMOs and N = 37 for euthanasia). We only used the arguments that the participants of the pilot studies viewed as unambiguously for or against and did not differ in terms of understandability and strength.
In both versions of the message, the author (described as an expert in the respective field) only provided arguments for and against and did not state any opinion on the matter. Arguments for and against were provided in two separate paragraphs, preceded by the expressions “on the one hand” and “on the other hand.” To control the effect of the order of arguments, the paragraphs were randomly presented in the for-against or the against-for order. Preliminary analyses showed that the effect of the forbid versus allow framing was not affected by the order of argument presentation; therefore, this factor was omitted in the subsequent analyses.
Next, the participants described the thoughts that came to their minds when reading the message. They were instructed to spend up to two minutes listing a maximum of 12 thoughts in separate boxes. Two judges, blind to the conditions in which the thoughts were generated, coded the thoughts as positive, negative, neutral, or irrelevant to GMOs or euthanasia (depending on the group). All disagreements were resolved through discussion.
Finally, the participants declared their attitudes toward forbidding or allowing GMOs (or euthanasia) again using the same scale as at the beginning of the experiment.
Results
To distinguish between effects supporting the descriptive valance account or evaluative encoding account, we need to include the valence of participants' preexisting attitudes in the analyses. For this reason, as well as for simplicity of presentation, we recoded the participants' attitudes from attitudes toward forbidding or allowing GMOs or euthanasia into a GMOs or euthanasia attitude index. The responses given by the participants who had been asked about forbidding GMOs or euthanasia were reversed such that high values in both the forbid and allow groups mean a more positive attitude toward GMOs or euthanasia. Based on this, we divided participants into two groups: the proponents of the GMOs or euthanasia (N = 52) and the opponents of the GMOs or euthanasia (N = 112).
Valence of Thoughts
Our object of interest here was the number of pro versus con thoughts reported by participants related to the issue. To control for individual differences in the total number of thoughts reported, for each subject, we calculated ratios of each category over the total number of issue-related thoughts (pro, con, and neutral; Blankstein et al., 1989). Next, we conducted a 2 (issue: GMOs vs. euthanasia) × 2 (attitude question framing: forbid vs. allow) × 2 (initial attitude: pro vs. con) × 2 (thoughts reported: pro vs. con) ANOVA (where the last factor was within-participants).
We found a significant interaction between the forbid versus allow framing and the numbers of pro versus con thoughts reported (F(1, 156) = 5.33, p = .022, ηp2 = .033). For the forbid frame, the con thoughts prevailed over the pro thoughts (M = 0.36, SD = 0.41 and M = 0.11, SD = 0.27, respectively, F(1, 156) = 18.51, p < .001, ηp2 = .106). In the allow frame condition, numbers of con and pro thoughts were similar (M = 0.27, SD = 0.48 and M = 0.22, SD = 0.30, respectively, F < 1). From the other side, pro thoughts were more numerous for the allow frame than the forbid frame (F(1, 156) = 6.40, p = .012, ηp2 = .039), but the quantity of the con thoughts in the forbid frame was not significantly higher than in the allow frame (F(1, 156) = 1.62, p = .205). Overall, these results may suggest that processes predicted by the descriptive valence account were working here to some degree and that the participants were just more prone to generate more negative thoughts for the forbid frame and/or more positive thoughts for the allow frame. However, even if the frame itself skewed the valence of reported thoughts, this effect was modified by the processes predicted by the evaluative encoding account. Specifically, the frame and thoughts interaction was qualified by a significant three-factor interaction between the forbid versus allow framing, the number of pro versus con thoughts, and the participants' initial attitudes (F(1, 156) = 21.70, p < .001, ηp2 = .122).
As shown in Figure 2, in the allow frame condition, thoughts consistent with initial attitudes prevailed over inconsistent ones, for both proponents and opponents. Specifically, when the allow frame was applied, the opponents reported more con than pro thoughts (M = 0.45, SD = 0.38 and M = 0.07, SD = 0.24, respectively, F(1, 156) = 45.35, p < .001, ηp2 = .225), while the proponents, conversely, reported more pro than con thoughts (M = 0.37, SD = 0.24 and M = 0.08, SD = 0.39, respectively, F(1, 156) = 6.49, p = .012, ηp2 = .040). However, this tendency was mitigated (for the opponents) or even reversed (for the proponents) when the forbid frame was applied. In the forbid frame condition, the proponents reported more inconsistent (con) than consistent (pro) thoughts (M = 0.42, SD = 0.41 and M = 0.12, SD = 0.26, respectively, F(1, 156) = 12.40, p < .001, ηp2 = .074). The opponents still reported more consistent (con) than inconsistent (pro) thoughts (M = 0.29, SD = 0.39 and M = 0.10, SD = 0.25, respectively, F(1, 156) = 6.26, p = .013, ηp2 = .039). However, in the forbid condition, they reported less con (consistent) thoughts than in the allow condition (F(1, 156) = 4.76, p = .031, ηp2 = .030). The amount of pro (inconsistent) thoughts reported by the opponents was similar for the forbid and allow frames (F < 1). Thus, the forbid frame decreased the generation of thoughts consistent with the preexisting attitude and, to some extent, also increased the generation of thoughts inconsistent with this attitude. Overall, participants in the forbid group showed a less-biased pattern of thoughts.
Attitude Change
To explore whether forbid versus allow framing affects attitude change, we conducted a 2 (issue: GMOs vs. euthanasia) × 2 (attitude question framing: forbid vs. allow) × 2 (initial attitude: pro vs. con) × 2 (attitude measurement: before vs. after the message) ANOVA (where the last factor was within-participants). We found a significant interaction between preexisting attitudes, the time of measurement of attitudes, and the forbid versus allow framing (see Figure 3; F(1, 156) = 5.14, p = .025, ηp2 = .032). The opponents' initial attitudes were similar in the forbid and allow groups (M = −4.23, SD = 1.11 and M = −4.50, SD = 1.09, respectively, F(1, 156) = 1.62, p = .206). However, the proponents' reported more positive initial attitudes in the forbid than allow group (M = 3.84, SD = 1.16 and M = 2.76, SD = 1.09, respectively, F(1, 156) = 10.91, p = .001, ηp2 = .065). In other words, in the last case (and only in this case), the forbid/allow asymmetry known from the research on the survey questions was found (see Hippler & Schwarz, 1986). What is more important, attitudes changed significantly only in the forbid frame groups. The opponents' attitudes shifted after the message in a positive direction when the forbid frame was applied (F(1, 156) = 15.53, p < .001, ηp2 = .091) but remained unchanged in the allow frame condition (F(1, 156) = 2.53, p = .114). Similarly, the proponents' attitudes moved in a negative direction within the forbid frame group (F(1, 156) = 9.17, p = .003, ηp2 = .056) but remained at the initial level within the allow frame group (F < 1). As a result of these shifts, the opponents' final attitudes were less negative in the forbid than allow group (M = −3.06, SD = 2.30 and M = −4.14, SD = 2.25, respectively, F(1, 156) = 5.87, p = .017, ηp2 = .036). The proponents' final attitudes were similar in the forbid and allow groups (M = 2.78, SD = 2.40 and M = 2.45, SD = 2.26, respectively, F < 1), but note that this was a consequence of the significant differences in the initial attitudes of this group of participants.
Mediational Analysis
To check whether the effect of forbid versus allow framing on attitude change is mediated by the decrease in biased elaboration, we used PROCESS model 4 (Hayes, 2018; with allow frame coded as 0, forbid frame coded as 1, continuous variables that define products mean centered, and CIs estimated using 5,000 bootstrapping samples). The overall consistency of each participant's thoughts with their initial attitude was indexed by subtracting the number of thoughts inconsistent from the number of the consistent ones and dividing this difference by the total number of issue-related thoughts. Thus, a high score indicated a dominance of thoughts that were consistent with a participant's initial attitude, or – to speak differently – thoughts more biased by the preexisting attitude. An attitude change index was created by subtracting the attitudes at Time 1 from the attitudes at Time 2 for participants with initially unfavorable attitudes and subtracting the attitudes at Time 2 from the attitudes at Time 1 for participants with initially favorable attitudes. As a result, positive scores indicated a change in the opposite direction from a participant's initial attitude (cf. Taber et al., 2009).
The forbid frame led to less biased thoughts, b = −0.37, SE = 0.08, t(162) = −4.69, p < .001. Decreasing the bias of thoughts was associated with higher attitude change, b = −0.73, SE = 0.29, t(161) = −2.55, p = .012. The direct effect of a frame on attitude change was not significant, b = 0.41, SE = 0.31, p = .182. More importantly, the effect of the forbid versus allow frame on attitude change was fully mediated by the thought consistency index, effect = 0.27, boot SE = 0.12, 95% CI [0.06, 0.55].
Discussion
The results of Experiment 2 confirm that attitudes are more susceptible to change when the forbid frame is used than when the allow frame is used. Specifically, the forbid frame led to attitude change in the direction opposite to the initial attitude (which confirms the predictions of the evaluative encoding account), not in the direction of a more negative attitude (as the descriptive valence account predicts). Moreover, the change was induced solely by presenting a set of arguments for and against GMOs or euthanasia to the participants, with no clear favoring of either viewpoint. We can assume, then, that this change reflects the focus of participants on different types of arguments. Analyses of the thoughts generated by participants seem to confirm this assumption – the forbid frame decreased the generation of thoughts consistent with the preexisting attitude and to some extent also increased the generation of thoughts inconsistent with this attitude. Thus, the forbid frame may decrease the biased elaboration of the presented information (Edwards & Smith, 1996; Taber et al., 2009) and the tendency to confirm preexisting attitudes, or even to lead to a greater acceptance of information that is inconsistent with the initial attitudes.
Experiment 3
The aim of Experiment 3 was to replicate the results concerning the impact of the forbid versus allow frame on attitude change and biased processing of information. Moreover, this time we have checked whether framing can affect not only focus on the specific arguments but also how they are evaluated and interpreted. We have also administered some auxiliary measures, which can help to specify the exact nature of the forbid/allow asymmetry in persuasion. Primarily, we have measured associations evoked by the words “forbid” and “allow” (Wierzba et al., 2015). Furthermore, a series of other measures was administered to check for possible alternative explanations and antecedents of the forbid/allow asymmetry in persuasion (e.g., attitude certainty, perceived risk, psychological distance). Because none of these variables proved to serve as a reliable alternative explanation of the results obtained, and due to space limitations, we present the rationale behind all of them and their description and results in ESM 1. This study was preregistered at https://aspredicted.org/2c56v.pdf.
Method
Participants and Design
A sample of 329 participants (Mage = 23.01 years; SD = 5.08; 266 women, 78.9%) were recruited to participate in the experiment via four Polish university mailing lists and social media. All participants completed the survey online and were not compensated. Participants were randomly assigned to one of two experimental groups using the forbid or the allow frames. Next, pro and con arguments were presented and participants were asked to evaluate each argument using the scales provided. Moreover, initial and final attitudes were measured to assess the attitude change. Basing on the initial attitude scale, participants were divided into two groups: proponents and opponents of the discussed issue (details described later). A sensitivity analysis conducted with G*Power (Faul et al., 2007) showed that the final sample was sufficient for detecting the effects of f = 0.09 with a power of 0.80 for a mixed-design ANOVA.
Material and Procedure
At the beginning of the study, the participants declared their attitudes toward forbidding or allowing GMOs using an 11-point scale ranging from −5 (strongly against) to 5 (strongly in favor). Because the aim of the experiment was to measure the susceptibility of attitudes to change, only those participants who declared having an attitude toward the issue (responses between −5 and −1 and between 1 and 5) proceeded with the experiment. Based on this question, similar to Experiment 2, we divided participants into two groups: the proponents of GMOs (N = 188) and the opponents of GMOs (N = 141). Next, a battery of auxiliary measures was administered (see ESM 1).
After finishing answering the auxiliary measures, participants were informed that their task would be to indicate whether GMO crops should be allowed (in one group) or whether they should be forbidden (in the second group), and that to make their task easier, they would be shown several arguments about GMOs. Next, three pro and three con GMOs arguments were presented in random order (the arguments were the same as in Experiment 2). The participants answered questions concerning each argument's interpretation (against GMOs–for GMOs) and evaluation (unpersuasive–persuasive), scales ranging from −5 to 5. Overall indices of interpretation and evaluation of the pro and con arguments were created. Specifically, the assessments of all three pro arguments were averaged for interpretation (α = .86) and evaluation (α = .76) scales. A similar procedure was followed for the assessments of the three con arguments (interpretation index α = .83 and evaluation index α = .64).
Next, associations with the words “allow” (in the first group) and “forbid” (in the second group) were assessed. Participants indicated their associations using a series of bipolar adjective scales (ranging from −5 to 5). Associations with blocking the action and goals were indexed by averaging answers on the following scales: difficulty–facilitation, obstacle–incentive, stopping action–continuing action, limitation–increase of available options, discord–consent (α = .94). To index associations with cautiousness and consideration of costs, answers on six scales were averaged: caution–confidence, danger–safety, prevention–encouragement, receding–approaching, loss–gain, penalty–reward (α = .87).
Finally, the participants again declared their attitudes toward forbidding or allowing GMOs using the same scale as at the beginning of the experiment.
Results
Forbid Versus Allow Associations
Our analyses confirm that the word “forbid” brings up associations related to blocking an action and goals (M = −3.24, SD = 1.64), while the word “allow” is instead connected with the smooth continuation of an action and the realization of a goal (M = 2.52, SD = 1.62, t(327) = 31.98, p < .001, d = 3.535). Moreover, while the word “forbid” is associated with caution and consideration of costs (M = −2.65, SD = 1.47), the word “allow” is related to safety and orientation about gains (M = 1.75, SD = 1.50, t(327) = 26.78, p < .001, d = 2.961).
Interpretation of Arguments
One consequence of biased elaboration can be the tendency to perceive the direction of presented arguments or information as skewed in the direction of a preexisting attitude (cf. Houston & Fazio, 1989; Kunda, 1990). For example, pro arguments can be perceived as more pro by proponents, but in the eyes of the opponents, the same arguments may not support the pro stance so much. To check whether the forbid versus allow framing affects the perceived direction of an argument, we conducted a 2 (attitude question framing: forbid vs. allow) × 2 (initial attitude: pro vs. con) × 2 (arguments: pro vs. con) ANOVA, where the last factor was within-participants and the dependent variable was the assessment of arguments on the against GMOs–for GMOs scales.
None of the interactions with the framing and the type of arguments were significant. However, rather than differences in the perception of the pro and con arguments, we are interested in whether framing mitigates the overall tendency to interpret arguments or information as confirming a preexisting point of view. Overall, the proponents' perception of all presented arguments was skewed in the direction for GMOs (M = 0.20, SD = 0.99), and the opponents perceived the same arguments as speaking rather against GMOs (M = −0.29, SD = 0.98, F(1, 325) = 20.68, p < .001, ηp2 = .060). This tendency to perceive the arguments in accordance with one's point of view was lowered in the forbid frame condition (see Figure 4; F(1, 325) = 9.50, p = .002, ηp2 = .028). Specifically, the opponents assessed the presented arguments as less anti-GMOs in the forbid condition than they did in the allow condition (M = −0.13, SD = 0.98 and M = −0.46, SD = 0.98, respectively, F(1, 325) = 4.10, p = .044, ηp2 = .012). The proponents assessed the presented arguments as less for GMOs when the forbid rather than allow frame was applied (M = 0.03, SD = 0.97 and M = 0.38, SD = 0.98, respectively, F(1, 325) = 5.60, p = .019, ηp2 = .017).
Evaluation of Arguments
To check whether forbid versus allow framing affects the evaluation of an argument, we conducted a 2 (attitude question framing: forbid vs. allow) × 2 (initial attitude: pro vs. con) × 2 (arguments: pro vs. con) ANOVA, where the last factor was within-participants and the dependent variable was the assessment of arguments on the unpersuasive–persuasive scales. We found a significant interaction between the type of arguments, the initial attitude, and the framing (see Figure 5; F(1, 325) = 5.01, p = .026, ηp2 = .015). Framing did not affect the evaluation of the con arguments. The proponents of GMOs evaluated these arguments as equally unpersuasive in the forbid and allow conditions (M = 1.24, SD = 2.05 and M = 1.36, SD = 2.05, respectively, F < 1). The opponents of the GMOs assessed the persuasiveness of these arguments as similarly high when the forbid versus allow frames were used (M = 2.46, SD = 2.05 and M = 2.54, SD = 2.05, respectively, F < 1). However, the forbid versus allow framing affected the evaluation of pro GMOs arguments. Specifically, the proponents evaluated the pro (attitude consistent) arguments as less persuasive in the forbid than allow condition (M = 2.87, SD = 1.74 and M = 3.38, SD = 1.74, respectively, F(1, 325) = 3.99, p = .047, ηp2 = .012). The reverse was true for the opponents – they assessed the pro (attitude inconsistent) arguments as more persuasive when the forbid frame was used rather than allow (M = 2.10, SD = 1.74 and M = 1.32, SD = 1.74, respectively, F(1, 325) = 7.15, p = .008, ηp2 = .022).
As a result, the overall tendency toward a preferential evaluation of arguments consistent versus inconsistent with preexisting attitudes (measured as the difference between the evaluation of consistent and inconsistent arguments) was lower in the forbid than allow frame (M = 0.99, SD = 2.50 and M = 1.62, SD = 2.56, respectively, F(1, 325) = 5.01, p = .026, ηp2 = .015).
Attitude Change
Similar to Experiment 2, we recoded the participants' attitudes from attitudes toward forbidding or allowing GMOs into a GMOs attitude index. The responses given by the participants who had been asked about forbidding GMOs were reversed such that high values in both the forbid and allow groups mean a more positive attitude toward GMOs. To explore whether the forbid versus allow framing affects attitude change, we conducted a 2 (attitude question framing: forbid vs. allow) × 2 (initial attitude: pro vs. con) × 2 (attitude measurement: before vs. after the message) ANOVA (where the last factor was within-participants).
We found a significant interaction between preexisting attitudes, the time of measurement, and the forbid versus allow framing (see Figure 6; F(1, 325) = 14.76, p < .001, ηp2 = .043). The opponents' initial attitudes were similar in the forbid and allow groups (M = −3.31, SD = 1.22 and M = −3.30, SD = 1.22, respectively, F < 1). The same was true for the proponents – their initial attitudes were similar in the forbid and allow groups (M = 3.38, SD = 1.22 and M = 3.61, SD = 1.22, respectively, F(1, 325) = 1.71, p = .192). The opponents' attitudes shifted after the presentation of arguments in a positive direction, both when the forbid frame (F(1, 325) = 56.79, p < .001, ηp2 = .149) and the allow frame were applied (F(1, 325) = 10.93, p = .001, ηp2 = .033). Similarly, the proponents' attitudes moved in a negative direction both within the forbid frame group (F(1, 325) = 35.49, p < .001, ηp2 = .098) and the allow frame group (F(1, 325) = 9.46, p = .002, ηp2 = .028). However, as a result of these shifts, the opponents' final attitudes were less negative in the forbid than allow group (M = −1.14, SD = 2.61 and M = −2.33, SD = 2.61, respectively, F(1, 325) = 7.41, p = .007, ηp2 = .022). Analogically, the proponents' final attitudes were less positive in the forbid than allow group (M = 1.75, SD = 2.61 and M = 2.89, SD = 2.61, respectively, F(1, 325) = 8.78, p = .003, ηp2 = .026). Thus, although attitudes changed significantly in both the allow and forbid groups, the change was higher when the forbid frame was applied.
Mediational Analysis
To check whether the effect of the forbid versus allow framing on attitude change is mediated by the decrease in biased interpretation and/or biased evaluation of arguments, we ran a mediation analysis. Attitude change was indexed by subtracting the attitudes at Time 1 from the attitudes at Time 2 for participants with initially unfavorable attitudes and subtracting the attitudes at Time 2 from the attitudes at Time 1 for participants with initially favorable attitudes. As a result, positive scores indicated a change in the opposite direction from a participant's initial attitude. To create an index of the biased perception of arguments, we reversed the opponents' assessments of the direction of the arguments such that high results mean the arguments are perceived as consistent with the initial attitude. The biased evaluation of arguments was indexed by subtracting the evaluation of arguments inconsistent with the participant's initial attitude from the evaluation of consistent ones. As a result, positive scores indicated that arguments consistent with initial attitudes were evaluated as more persuasive than inconsistent ones.
We used PROCESS model 4 (Hayes, 2018), with biased interpretation and the biased evaluation of the arguments serving as two mediators (allow frame coded as 0, forbid frame coded as 1, continuous variables that define products mean centered, and CIs estimated using 5,000 bootstrapping samples). The forbid frame led to a less-biased interpretation of the arguments, b = −0.33, SE = 0.11, t(327) = −3.06, p = .002, as well as a less-biased evaluation of them, b = −0.68, SE = 0.28, t(327) = −2.42, p = .016. The decreased bias in the interpretation of the arguments was associated with higher attitude change, b = −0.49, SE = 0.13, t(325) = −3.67, p < .001. The same was true for the decreased bias in evaluation, b = −0.27, SE = 0.05, t(325) = −5.34, p < .001. The direct effect of the frame on attitude change was significant, b = 0.71, SE = 0.25, t(325) = 2.81, p = .005. However, more importantly, the effect of the forbid versus allow frame on attitude change was mediated by both the biased interpretation index, effect = 0.16, boot SE = 0.07, 95% CI [0.05, 0.33], and the biased evaluation index, effect = 0.19, boot SE = 0.08, 95% CI [0.04, 0.37].
Discussion
Overall, the pattern of results in this study confirms the predictions of the evaluative encoding account. First, the interpretation of arguments, regardless of whether they were pro or con, was less biased in the forbid than allow frame – results not predicted by the descriptive valence account. Second, the pattern of results for the assessment of arguments' persuasiveness also seems to indicate that the consistency versus inconsistency of arguments with initial attitude was crucial here. For the proponents of GMOs, the forbid frame mitigated the tendency toward a higher evaluation of arguments consistent with attitudes. For the opponents, the forbid frame reduced the tendency toward a low evaluation of arguments inconsistent with attitudes (we will come back to this asymmetry between proponents and opponents in the general discussion). Despite these shifts in evaluation not concerning both types of arguments, they still led to an overall decrease in the biased evaluation of the information presented. Finally, decreased bias in the interpretation and evaluation of arguments seems at least a partial reason for the greater susceptibility to change of attitudes within the forbid versus allow frame.
General Discussion
In three experiments (one preregistered), we showed that the forbid/allow asymmetry affects the effectiveness of persuasion and may impact public discourse. Asking citizens about forbidding GMOs, rather than allowing GMOs, may not only yield different survey results (e.g., Chessa & Holleman, 2007; Hippler & Schwarz, 1986), but it may also increase citizens' susceptibility to persuasion. As the results of our experiments show, the forbid frame (compared with the allow frame) reduced the impact of initial attitudes on the direction of processing and evaluating the presented information. As a result, the forbid frame caused people to defend the arguments or thoughts consistent with preexisting attitudes less strongly and/or less strongly discredit the inconsistent ones. Specifically, the forbid frame decreases the tendency for generating thoughts prevailingly consistent with participants' initial attitudes (Experiment 2). Moreover, when compared to the allow frame, the forbid frame also reduces bias in the evaluation and interpretation of the presented arguments and yields more similar assessments of arguments that are consistent and inconsistent with initial attitudes (Experiment 3). This contributes to a relative increase in the influence of inconsistent arguments on final attitude. This, in turn, led to a change of attitude in the direction opposite to the initial attitude. Overall, attitudes are more susceptible to change within the forbid frame than within the allow frame (Experiments 1–3).
These results indicate the existence of new, previously undescribed effects. The forbid versus allow framing seems to be more than simply choosing between two alternative and compatible formulations of an issue. It can be very impactful in the case of prereferendum campaigns or in the case of campaigns aimed at changing social attitudes. Inducing people to think about forbidding (instead of allowing) some object, for example, GMO or euthanasia, leads them to process the information more objectively and can result in attitude change.
Limitations and Further Research
Our results can also shed some light on the potential limitations of the effect described. Further research could help to specify the conditions in which the forbid versus allow framing works as well as to describe the exact mechanism of its influence.
Issue Importance and Scope of Elaboration
First, it should be noted that the issues used in Experiments 2 and 3 were fairly relevant to the participants (GMOs and euthanasia). Our two other supplementary studies asked the participants about how relevant these topics were to them (the participants assessed the topics on scales ranging from −5 to 5, i.e., completely irrelevant–very relevant). The results show that both GMOs (M = 2.00, SD = 2.62, N = 251) and euthanasia (M = 2.79, SD = 2.29, N = 104) were assessed as fairly important. This suggests the possibility that to initiate the process described in the evaluative encoding account, not only is the preexisting attitude necessary, but this attitude should also be important or relevant enough to motivate the biased processing of arguments (the processing, perception, and evaluation of information inconsistent vs. consistent with initial attitudes). In the case of less important attitudes, or when the object is novel and previous attitudes do not exist, the predictions of the descriptive valence account could be more applicable and forbid versus allow framing may simply change the influence of positive (pro) versus negative (con) information. Moreover, in this case, attitude change can also be the simple result of vigilance – people can withdraw their initial assessments without considering the available information at all.
Furthermore, research on valence framing shows it is limited to situations of high cognitive elaboration: Differences in the strength of positively and negatively framed attitudes disappear when attitudes are of low relevance or when the participants are under cognitive load (Bizer et al., 2013). Similarly, the focus on preference-consistent information in a choice task and on preference-inconsistent information in a rejection task is eliminated or even reversed when the information is processed under cognitive load (Laran & Wilcox, 2011). Our study involved no factors that could have led to cognitive load. The initial level of motivation for elaboration was also not manipulated. However, the aforementioned results of the supplementary studies suggest that the issues used in Experiment 2 and Experiment 3 were relevant enough to motivate participants to process the information. With the issue used in Experiment 1, it should be noted that participants were informed at the beginning of the study that the issues they would be presented with could concern their neighborhoods, which may have increased the initial scope of processing (Petty & Cacioppo, 1986). Therefore, it is possible that, similar to the valence framing effect and the task-type effect, the precise mechanism behind the forbid versus allow effect or even the existence of this effect is related to participants' motivation or ability to process.4
Attitude Strength
Second, it is also worth noting that we found some asymmetry between the evaluations by proponents and opponents. In Experiment 2, the forbid frame affected both the consistent and inconsistent thoughts for the proponents only. Specifically, the forbid frame led these participants to generate less consistent and more inconsistent thoughts. For the opponents, we found only a forbid frame-related decrease in the generation of thoughts consistent with the initial attitude. This asymmetry may be explained with the help of the previous research, which showed that the attitudes of opponents are more resistant and stronger than the attitudes of proponents (Bizer & Petty, 2005; Bizer et al., 2011, 2013). Switching the focus of less resistant participants from the arguments consistent to inconsistent may be relatively easy. However, when participants are more resistant, undermining what they accept may be easier than making them think the way that they initially opposed (see Clark & Wegener, 2013). Furthermore, in Experiment 3, the forbid frame affected (decreased) the evaluation of arguments consistent with the attitudes of the proponents, while for the opponents the evaluation of inconsistent ones was changed (increased). This asymmetry could be explained by the different motives prevailing within the proponents and opponents. The proponents may be more inclined to bolster their attitudes, that is, to look for information confirming them. The opponents may be oriented toward defending their attitude, which can be achieved by attacking information inconsistent with this attitude (Clark & Wegener, 2013). Thus, it seems that for both the proponents and the opponents, the forbid frame decreased the predominant motivation. As motives involving bolstering are more probable for relatively weak attitudes, and defense motives for relatively strong ones, these results suggest that the attitudes of proponents were weaker than the attitudes of opponents, which once again seems to be in accord with Bizer and Petty's results (2005; see also Bizer et al., 2011, 2013). Overall, this asymmetry indicates that the frame is not the only factor affecting the influence of different types of arguments. Factors such as strength of initial attitude may be important here – frame and the attitude strength may conjointly shape the processing of arguments and scope and direction of attitude change.
Moreover, forbid versus allow framing can also be consequential for attitude strength. In two of our studies (see footnote 3 and ESM 1), we did not find evidence that the forbid versus allow asymmetry (we found for other measures) is mediated by the frame-related shift in the certainty of initial attitudes. In fact, participants answered questions about forbidding versus allowing an object with a similar level of certainty. However, it does not mean that forbid versus allow framing has no consequences for the attitude certainty at all. We measured certainty of attitudes at the beginning of the study, just after the initial question about attitude toward forbidding versus allowing, when no other attitude object-related information was presented. Thus, subtle difference in the initial framing of the situation (asking about forbidding vs. allowing) can be not enough to change the certainty of attitudes. In other studies, where impact of different frames on attitude certainty was found, the certainty measure was applied after participants have an occasion to process some attitude object-related information (Bizer et al., 2013, 2011; Rucker et al., 2008). It is possible, then, that forbid versus allow framing can affect the certainty of attitudes at later stages of persuasion process when people encounter some argumentation and when they start thinking about the object of persuasion.
Furthermore, basing on the other research findings, we can even expect that frame can affect attitude change via two separate routes: affecting consistency of the thoughts with initial attitude and affecting the confidence of these thoughts (Clarkson et al., 2011). Since we have found that forbid versus allow framing differently affects processing and evaluation of thoughts consistent versus inconsistent with initial attitude, especially interesting would be checking how these frames affect the certainty of these two types of thoughts. As Clarkson et al. (2013) showed, attitude-consistent thought confidence and attitude-inconsistent thought confidence can play an independent role in mediating attitude change (see also Koniak & Cwalina, 2020).
Object of Attitude
Finally, it is also worth taking into account the other characteristics of the object of an attitude. This study tested how the forbid versus allow frame affected the susceptibility of attitudes toward GMOs, euthanasia, and barbecuing in public parks to persuasion. It seems that, at least in Poland, where this study took place, the population is mostly unfamiliar with the legal status of these issues; neither GMOs nor euthanasia is persistently associated with the status of a forbidden or allowed object (CBOS, 2013; Szadowska-Szlachetka et al., 2019). Familiarity with the legal regulations concerning barbecuing in public places is also uncommon; moreover, the issue itself is not specifically regulated. With such indeterminate objects, both the forbid and the allow frames can easily be used. Asking about forbidding GMOs or euthanasia comes as naturally as asking about allowing them. However, the forbid versus allow asymmetry in persuasion may be reduced if the status of an object is commonly known. For instance, in Poland, asking respondents about allowing marijuana (which is currently an illegal substance) may seem more natural than asking about forbidding it (as the ban is currently in effect). Moreover, the actual status of the object, whether it is forbidden or allowed, or rather how this status is perceived by people, can be an important mediator of the effect. For the objects already allowed, when people know that they are free to use these objects, asking about forbidding them can rather rise a reactance than create an opportunity for the less-biased processing of object-related information (cf. Koniak & Cwalina, 2020). This would suggest that for such objects, the forbid frame could even lead to a polarization of attitudes. However, this effect could probably be modified depending on whether the proposed ban or restriction would be seen as an absolute (i.e., complete, certain, or permanent) or nonabsolute (i.e., incomplete, uncertain, or temporally or spatially limited; see Laurin et al., 2012).
Another possibly important factor is how an issue is discussed in the public space. For example, most respondents have likely an established attitude toward GMOs, an attitude that will translate into their attitude toward forbidding or allowing GMOs (Aerni et al., 2011; Rzymski & Królczyk, 2016). However, certain issues are immediately discussed with either frame, as with the ban on trading on Sundays in Poland (Sierpińska, 2007). The object of attitudes in this case was more likely forbidding trading on Sundays, rather than trading on Sundays. It is possible that an increased susceptibility of attitudes to change did not occur in response to this ban.
Relation to Other Research
Finally, some notes should be made about the relationship of our results to other studies. First, our results could be seen as contradicting those obtained by Bizer and Petty (2005), where negative framing resulted in attitudes more resistant to change. We believe that this is only an apparent inconsistency. Note that in the case of the valence framing described by Bizer and Petty (2005), the frame changed the valence of attitude, that is, it led people to think about themselves as opponents versus proponents. If a person is to choose between two options, one they like and one they do not (e.g., between liked candidate A and disliked candidate B), the attitude is framed positively by asking the person about their attitude toward the option they like. For example, a person asked what she thought of liked candidate A being elected would indicate that she supports this result. To frame the attitude negatively, the person needs to be asked about the attitude toward the option she dislikes (e.g., a person asked what she thought of disliked candidate B being elected would indicate that she opposes this result). In other words, in the case of the valence framing, frame concerns the attitude of the person and affects how people “conceptualize their own preferences” (Bizer et al., 2011, p. 64). However, in case of the forbid versus allow framing, as well as in case of the task-type effect and the attribute framing generally, framing does not determine the valence of attitudes. The forbid frame does not necessarily lead to negative attitudes, and the allow frame does not necessarily lead to positive attitudes. These frames concern the attitude object and change the way the object is perceived and evaluated.
It is possible, then, that framing primarily impacts processing, evaluation, or weighting a framed subject-related information. In the case of an attitude framing, frame changes the way this attitude is weighted, how it affects the persuasion process, etc. In the case of framing of an object of attitude, frame changes how the characteristics of the object or object-related information are used, etc. As a result, in the case of an attitude framing, the attitudes seen as negative are more impactful on later evaluation, cognition, and behavior than positive attitudes are. In the case of framing of an attitude object, the negative frame increase impact of negative information concerned the object (as the arguments inconsistent with one's attitude can be seen from the point of view of the evaluative encoding account). It could be said, then, that negative frame in the case of both the valence framing and the forbid versus allow framing results in an increase in the influence of negative information. What is changing is whether these are primarily an attitude itself or an attitude object-related information.
Moreover, the effects connected with these two cases of framing can work simultaneously toward shaping final attitudes. We could expect that even with the forbid frame, the opponents should still be less susceptible to attitude change than the proponents. We have not found such asymmetry in our research, but the valence of initial attitude was not systematically manipulated here. However, as we mentioned earlier, some of our results confirm that the attitudes held by the opponents were stronger than the proponents' attitudes, which is in accord with Bizer and Petty's findings (2005; see also Bizer et al., 2011, 2013).
Second, our results can shed some light on the potential mediators of the forbid/allow asymmetry known from the studies concerning the effect of forbid versus allow questions on answers in surveys (e.g., Chessa & Holleman, 2007; Hippler & Schwarz, 1986). In these survey research, participants usually answer the question relying only on their own attitudes. No additional information about the surveyed issue is provided for them. Our results suggest that providing participants with additional information about the surveyed issue should lower the extremity of the answers when the question about forbidding an object is answered. However, to our knowledge, this possibility has not been studied so far. It is thus a potential area for future research. It should also be noted that in our research, we found only one case of the forbid/allow asymmetry for the initial measurement of attitudes. The initial measurement of attitudes is a case similar to how this asymmetry was studied in the survey research. This result confirms that the forbid/allow asymmetry in survey answers is a quite elusive phenomenon (Holleman, 1999).
Third, our studies, especially Experiment 2 and Experiment 3, can be treated as an extension of research on attitude polarization. Some of the previous research found that when people are confronted with mixed evidence on some issue, their attitudes toward this issue tend to polarize (Lord et al., 1979). However, other studies found that this effect is not a very frequent phenomenon (Kuhn & Lao, 1996), and in some cases, attitudes can even depolarize (Miller et al., 1993). Stimuli materials presented to our participants in Experiment 2 and Experiment 3 were in fact a case of mixed evidence – a set of pro and con arguments without clear support for any of them. What we have found is in fact an attitude depolarization. Thus, our findings can add something to the understanding of the determinants of the phenomenon of attitude polarization versus depolarization. Specifically, the results showed in what conditions attitudes are prone to depolarization. It is possible that the forbid frame is a relatively easy way to make people consider the opposite, the strategy that was proven to be effective in reducing the biased evaluation of evidence and in preventing attitude polarization (Lord et al., 1984). This could be an important finding nowadays, when the growing polarization of attitudes in society is a serious problem. It looks like one way of depolarizing attitudes is to make people think about what should be or should not be forbidden, rather than making them think about what should be or should not be allowed.
Finally, we believe that the results obtained here are a consequence of the ability of the word “forbid” to induce some specific state where caution is evoked and previous actions or goals are reassessed. Some similarities between this state and the prevention focus can be observed (Higgins, 1998). This suggests the possibility that forbid versus allow frames are able to induce the prevention versus promotion focus. This, in turn, leads to the hypothesis that persuasive consequences of the prevention versus promotion focus are similar to those induced by the forbid versus allow framing. It is possible that the prevention (vs. promotion) focus leads people to process information more objectively and can result in attitude change. Similarly, Clark and Wegener (2013) suggested that a promotion orientation can lead to enhanced processing of proattitudinal information and prevention orientation can lead to enhanced processing of counterattitudinal ones. To our knowledge, this aspect of regulatory focus has not yet been studied empirically. It seems to be a promising area for future research. The practical importance of this new possibility is much larger than the present one stemming from our research.
Electronic Supplementary Materials
The electronic supplementary material is available with the online version of the article at https://doi.org/10.1027/1864-9335/a000469
The authors thank M. Drzewiecka, A. Błachnio, A. Przepiórka, P. Próchniak, P. Długoński, M. Płaczkowski, A. Milanowska, S. Niedźwiedź, A. Pałac, K. Piotrowska, K. Rosół, J. Siejka, I. Skoczek, W. Strzałkowska, and J. Tomczyk for help with recruiting participants.
References
(2011). How would Swiss consumers decide if they had freedom of choice? Evidence from a field study with organic, conventional and GM corn bread. Food Policy, 36(6), 830–838. 10.1016/j.foodpol.2011.08.002
(2001). Bad is stronger than good. Review of General Psychology, 5(4), 323–370. 10.1037/1089-2680.5.4.323
(2011). Exploring the valence-framing effect: Negative framing enhances attitude strength. Political Psychology, 32(1), 59–80. 10.1111/j.1467-9221.2010.00795.x
(2005). How we conceptualize our attitudes matters: The effects of valence framing on the resistance of political attitudes. Political Psychology, 26(4), 553–568. 10.1111/j.1467-9221.2005.00431.x
(2013). When are support and opposition not opposites? Depth of processing as a moderator of the valence-framing effect. British Journal of Social Psychology, 52(1), 191–201. 10.1111/bjso.12014
(1989). Test anxiety and the contents of consciousness: Thought-listing and endorsement measures. Journal of Research in Personality, 23(3), 269–286. 10.1016/0092-6566(89)90001-9
(2005). Striking a responsive chord: How political ads motivate and persuade voters by appealing to emotions. American Journal of Political Science, 49(2), 388–405. 10.1111/j.0092-5853.2005.00130.x
. (2013). Polacy o bezpieczeństwie żywności i GMO [Poles on food safety and GMOs; Survey report]. https://www.cbos.pl/SPISKOM.POL/2013/K_002_13.PDF
(2007). Answering attitudinal questions: Modelling the response process underlying contrastive questions. Applied Cognitive Psychology, 21(2), 203–225. 10.1002/acp.1337
(2013). Message position, information processing, and persuasion: The discrepancy motives model. Advances in Experimental Social Psychology, 47, 189–232. 10.1016/B978-0-12-407236-7.00004-8
(2011). A self-validation perspective on the mere thought effect. Journal of Experimental Social Psychology, 47(2), 449–454. 10.1016/j.jesp.2010.12.003
(2013). Motivated reflection on attitude-inconsistent information: An exploration of the role of fear of invalidity in self-persuasion. Personality and Social Psychology Bulletin, 39(12), 1559–1570. 10.1177/0146167213497983
(1997). Changing minds: Political arguments and political persuasion. American Journal of Political Science, 41(1), 88–121. 10.2307/2111710
(2000). When beliefs yield to evidence: Reducing biased evaluation by affirming the self. Personality and Social Psychology Bulletin, 26(9), 1151–1164. 10.1177/01461672002611011
(1996). A disconfirmation bias in the evaluation of arguments. Journal of Personality and Social Psychology, 71(1), 5–24. 10.1037/0022-3514.71.1.5
(2007). G*Power 3: A flexible statistical power analysis program for the social, behavioral, and biomedical sciences. Behavior Research Methods, 39(2), 175–191. 10.3758/bf03193146
(1995). Attribute scatter and decision outcome: Judgment versus choice. Organizational Behavior and Human Decision Processes, 62(1), 113–122. 10.1006/obhd.1995.1036
(1995). The influence of quantity of information and goal framing on decision. Acta Psychologica, 89(1), 23–36. 10.1016/0001-6918(94)00004-Z
(2009). Fear and loving in Las Vegas: Evolution, emotion, and persuasion. Journal of Marketing Research, 46(3), 384–395. 10.1509/jmkr.46.3.384
(2018). Introduction to mediation, moderation, and conditional process analysis: A regression-based approach. Guilford Press.
(1998). Promotion and prevention: Regulatory focus as A motivational principle. Advances in Experimental Social Psychology,Advances in Experimental Social Psychology Volume, 30, 1–46. 10.1016/S0065-2601(08)60381-0
(1986). Not forbidding isn't allowing: The cognitive basis of the forbid-allow asymmetry. Public Opinion Quarterly, 50(1), 87–96. 10.1086/268961
(1999). Wording effects in survey research using meta-analysis to explain the forbid/allow asymmetry. Journal of Quantitative Linguistics, 6(1), 29–40. 10.1076/jqul.6.1.29.4145
(2006). The meanings of 'yes' and 'no'. An explanation for the forbid/allow asymmetry. Quality and Quantity, 40(1), 1–38. 10.1007/s11135-005-4479-6
(1989). Biased processing as a function of attitude accessibility: Making objective judgments subjectively. Social Cognition, 7(1), 51–66. 10.1521/soco.1989.7.1.51
(2012, February 21). Democracy in action? 10 weird examples of the Swiss referendum system in action. https://www.businessinsider.com/democracy-in-action-10-weird-examples-of-the-swiss-referendum-system-in-action-2012-2?IR=T
(2020). Fear of coronavirus and forbid/allow asymmetry as determinants of acceptance of COVID-19 pandemic related restrictions and persistence of attitudes towards these regulations. Social Psychological Bulletin, 15(4), 1–13. 10.32872/spb.4421
(2022). Data sheet file for the paper “Forbid/allow asymmetry in persuasion: The forbid frame decreases biased elaboration and increases attitude change”. 10.5281/zenodo.6091442
(1996). Effects of evidence on attitudes: Is polarization the norm? Psychological Science, 7(2), 115–120. 10.1111/j.1467-9280.1996.tb00340.x
(1990). The case for motivated reasoning. Psychological Bulletin, 108(3), 480–498. 10.1037/0033-2909.108.3.480
(2011). Choice, rejection, and elaboration on preference-inconsistent alternatives. Journal of Consumer Research, 38(2), 229–241. 10.1086/659040
(2012). Reactance versus rationalization: Divergent responses to policies that constrain freedom. Psychological Science, 23(2), 205–209. 10.1177/0956797611429468
(1988). How consumers are affected by the framing of attribute information before and after consuming the product. Journal of Consumer Research, 15(3), 374–378. 10.1086/209174
(1998). All frames are not created equal: A typology and critical analysis of framing effects. Organizational Behavior and Human Decision Processes, 76(2), 149–188. 10.1006/obhd.1998.2804
(1984). Considering the opposite: A corrective strategy for social judgment. Journal of Personality and Social Psychology, 47(6), 1231–1243. 10.1037/0022-3514.47.6.1231
(1979). Biased assimilation and attitude polarization: The effects of prior theories on subsequently considered evidence. Journal of Personality and Social Psychology, 37(11), 2098–2109. 10.1037/0022-3514.37.11.2098
(2003). Framing effects: An analytic–holistic perspective. Journal of Experimental Social Psychology, 39(6), 610–617. 10.1016/S0022-1031(03)00036-2
(1993). The attitude polarization phenomenon: Role of response measure, attitude extremity, and behavioral consequences of reported attitude change. Journal of Personality and Social Psychology, 64(4), 561–574. 10.1037/0022-3514.64.4.561
(1986). Communication and persuasion: Central and peripheral routes to attitude change. Springer.
(1995). Attitude strength and resistance processes. Journal of Personality and Social Psychology, 69(3), 408–419. 10.1037/0022-3514.69.3.408
(2008). What's in a frame anyway? A meta-cognitive analysis of the impact of one versus two sided message framing on attitude certainty. Journal of Consumer Psychology, 18(2), 137–149. 10.1016/j.jcps.2008.01.008
(1941). Experiments in wording questions: II. Public Opinion Quarterly, 5(1), 91–92. 10.1086/265467
(2016). Attitudes toward genetically modified organisms in Poland: To GMO or not to GMO? Food Security, 8(3), 689–697. 10.1007/s12571-016-0572-z
(1981). Questions and answers in attitude surveys: Experiments on question form, wording and context. Sage Publications.
(1993). Choosing versus rejecting: Why some options are both better and worse than others. Memory & Cognition, 21(4), 546–556. 10.3758/bf03197186
(1993). Reason-based choice. Cognition, 49(1–2), 11–36. 10.1016/0010-0277(93)90034-S
(2018, September 23). Swiss canton becomes second to ban burqas in public. https://www.reuters.com/article/us-swiss-burqas/swiss-canton-becomes-second-to-ban-burqas-in-public-idUSKCN1M30J2
(2007). Zakaz handlu po raz pierwszy 1 listopada 2007 r [Trading ban for the first time November 1, 2007]. http://www.prawo.egospodarka.pl/25608,Zakaz-handlu-po-raz-pierwszy-1-listopada-2007-r,1,34,3.html
(2016). Take it or leave it: How choosing versus rejecting alternatives affects information processing. Journal of Consumer Research, 43(4), 614–635. 10.1093/jcr/ucw049
(2019). Knowledge and attitudes of Medical University students with regard to euthanasia. Palliative Medicine, 11(2), 73–80. 10.5114/pm.2019.86531
(2009). The motivated processing of political arguments. Political Behavior, 31(2), 137–155. 10.1007/s11109-008-9075-8
. (2018, September 23). https://www.swissinfo.ch/eng/directdemocracy/face-look_voters-approve--burka-ban--in-st-gallen/44419680
(1997). Another look at reasons for choosing and rejecting. Memory & Cognition, 25(6), 873–887. 10.3758/BF03211332
(2015). Basic emotions in the nencki affective word list (NAWL BE): New method of classifying emotional stimuli. PLoS One, 10(7), e0132305–16. 10.1371/journal.pone.0132305
(1991). Image, issues, and ideology: The processing of information about political candidates. Journal of Personality and Social Psychology, 61(4), 533–545. 10.1037/0022-3514.61.4.533
(2017). The Swiss minaret ban referendum and Switzerland's international reputation: A vote with an impact. Journal of Muslim Minority Affairs, 37(4), 413–425. 10.1080/13602004.2017.1405506
1It should be noted that most reject versus choose framing studies analyze its effect on preference choice (Ganzach & Schul, 1995), where the respondent has to choose between different objects. Research on forbid versus allow framing is mostly focused on its effect on preference judgment, where only one attitude object is being considered. Then, expressing opinion about forbidding an object is not the same as deciding which object from a larger set to reject. Similarly, expressing opinion about allowing an object is not the same as deciding which object from a larger set to choose. Thus, the forbid versus allow framing and reject versus choose framing, although both are examples of the attribute framing, are not identical.
2For the full sample-based analyses, our crucial three-way interactions remained significant for both the valence of thoughts (F(1, 167) = 21.69, p < .001, ηp2 = .115) and the change of attitudes (F(1, 167) = 15.64, p < .001, ηp2 = .086).
3Moreover, this experiment also tested the role of another variable that could be related to the effect of the frame on the susceptibility of attitudes to persuasion, that is, attitude certainty (Bizer et al., 2011). Unfortunately, due to a programming error, the initial attitude certainty scale was only administered in the euthanasia group. Since we have not found a significant influence of the forbid versus allow frame on this measure, we skip the more detailed description of it.
4Although it should be noted that research on risky choice framing found exactly opposite effect, the way in which a decision was framed had a weaker influence among people for whom the issue was more relevant (McElroy & Seta, 2003). This discrepancy between the role of the relevance in the attribute framing and the risky choice framing research may confirm that these are distinctive types of framing and that the effects found for the one type of framing cannot be easily extrapolated on the second type of framing (Levin et al., 1998).
