Abstract
When people seek support for a cause, they typically present the strongest case they can muster. The present research suggests that under some conditions, the opposite strategy may be superior—in particular, presenting weak rather than strong arguments might stimulate greater advocacy and action. Across four studies, we show that when individuals already agree with a cause (i.e., it is pro-attitudinal), receiving weak arguments in its favor can prompt them into advocating more on its behalf. Perceived argumentation efficacy mediates this effect such that people exposed to weak arguments are more likely to think they have something valuable to contribute. Moreover, consistent with the notion that it is driven by feelings of increased efficacy, the effect is more likely to emerge when initial argumentation efficacy and attitude certainty are moderate or low. Individuals with high argumentation efficacy and high certainty generally advocate more, regardless of the strength of arguments received.
A great deal of research has tested the effect of argument quality—or message strength—on persuasion (for reviews see Petty, Rucker, Bizer, & Cacioppo, 2004; Petty & Wegener, 1998). This vast literature consistently shows that, when argument quality does exert an impact on persuasion (e.g., under conditions of high message processing), strong arguments produce more persuasion than do weak arguments. Indeed, it is virtually a truism in the persuasion literature that as long as message recipients are attending to and processing the arguments contained in a message, strong arguments will outperform weak ones. The general notion is that people will form more message-consistent attitudes following strong rather than weak arguments, and from a practical perspective the hope is that these attitudes will translate into behavioral intentions and, ultimately, actual behavior.
Weak > Strong?
The current research takes a different tack. We explore the possibility that under specifiable conditions, weak arguments can outperform strong ones. Of particular interest in the present research is the notion that weak arguments might sometimes prove more effective than strong arguments at stimulating advocacy among message recipients. Consider an example in which an individual supports a particular cause (e.g., marriage equality) and attends a town hall meeting at which the issue will be discussed. We submit that this individual might be more likely to speak up and offer her own arguments in support of gay marriage if she first hears another person at the meeting make the case poorly—for example, using weak arguments. In other words, the target individual might be more likely to advocate for and promote the cause herself after first observing another person do it with weak rather than strong arguments.
Why would weak arguments be more effective than strong ones at stimulating advocacy? Although there are a variety of possible mechanisms for such an effect, we examine one in the current work. Specifically, we explore the possibility that individuals’ own perceived ability to make the case, or their argumentation efficacy, can be elevated following exposure to other people’s weak arguments. For instance, when people observe others make a poor case for something they care about, they might think, “I could make this case more effectively . . . I better speak up.” Stated differently, seeing others make weak arguments for a cause might lead observers to feel as though they have something unique and valuable to contribute to help promote that cause. Seeing others make strong arguments, however, might foster the impression that the cause is already well-supported and that there is nothing unique or important to contribute that would substantially impact the outcome. If true, specious arguments by others on the same side could lead an individual to advocate more aggressively for the cause. Thus, ironically, weak rather than strong arguments might be more effective at creating advocates among like-minded message recipients.
If this is true, a variety of constraints on the effect seem to be likely. First, the effect should only emerge under pro-attitudinal message conditions; that is, when individuals hear or observe weak arguments promoting a position that they already endorse. After all, if one receives weak arguments promoting a counter-attitudinal position, those arguments might also increase one’s perceived argumentation efficacy, but there may be less motivation to actually advocate or argue a case when the opposing side is already undermining itself. When one receives weak arguments that are pro-attitudinal in nature, however, there could be an increase in perceived argumentation efficacy and an apparent need to intervene and help promote the cause. Therefore, we submit that the advantage of weak over strong arguments in stimulating advocacy will emerge under pro- but not counter-attitudinal message conditions.
Second, if indeed this advocacy effect is driven by perceived argumentation efficacy, as postulated, it should be most likely to manifest under conditions in which argumentation efficacy or attitude certainty is not already high. First consider argumentation efficacy. In general, perceived efficacy has been the subject of considerable research attention over the past few decades. One consistent finding to emerge from this literature is that people are more likely to take action when they believe that they have the necessary skills or ability to effectively execute that action (Bandura, 1982, 1993). Accordingly, we expect that individuals who feel greater argumentation efficacy will be more likely than individuals with lower argumentation efficacy to advocate for a position or take other action to promote a cause they support. If an individual is already high in argumentation efficacy, there would be little additional benefit to gain from the observation of others’ weak arguments. We therefore expect such individuals’ advocacy to be less affected by the arguments they receive. However, if an individual is initially lacking in argumentation efficacy, he should be unlikely to advocate unless something in the immediate context elevates his felt efficacy. We hypothesize that the boost in perceived efficacy derived from others’ weak arguments will provide this elevation, creating the predicted advocacy effect under low but not high initial efficacy conditions.
Following a similar logic, we submit that the advantage of weak arguments also will be moderated by attitude certainty. Attitude certainty refers to the sense of confidence or conviction one has about one’s attitude (Tormala & Rucker, 2007). This sense of confidence or conviction can emanate from feelings of attitude clarity or attitude correctness (Petrocelli, Tormala, & Rucker, 2007). Most germane to the present concerns, attitude certainty has been shown to be closely associated with advocacy-relevant action. For example, when people hold their attitudes with high rather than low certainty, they are more likely to behave in ways that are consistent with their attitudes (e.g., Fazio & Zanna, 1978; Tormala & Petty, 2002). Furthermore, high attitude certainty has been shown to directly increase willingness to vote and sign petitions (Barden & Petty, 2008), and to encourage people to express their opinions and seek to persuade others (Visser, Krosnick, & Simmons, 2003). Thus, in the current research we expect that high certainty individuals will tend to evince greater advocacy in general (i.e., unaffected by argument quality), whereas low certainty individuals will show greater advocacy under weak rather than strong argument conditions. Again, it is those individuals who initially lack certainty who have the most room to increase in advocacy. We hypothesize that if weak arguments can foster uncertain individuals’ perceived argumentation efficacy, they will stimulate advocacy as well.
Overview
Synthesizing the above logic, we predict that weak (vs. strong) arguments will lead to greater advocacy among pro- but not counter-attitudinal participants. Moreover, we posit that perceived argumentation efficacy will mediate this effect, such that seeing others make weak arguments will lead people to feel that they have something valuable to contribute to the cause, and this perception will motivate them to advocate more. Following directly from this logic, we also predict that the effect of weak arguments on advocacy will be most likely to manifest when people are not already high in argumentation efficacy or attitude certainty, because it is precisely these individuals for whom there is room to increase perceived efficacy and, ultimately, advocacy. Among high efficacy and high certainty individuals, we expect to observe greater levels of advocacy in general, independent of the strength of arguments they have received. We present 4 studies exploring these issues.
Study 1
Study 1 aimed to provide an initial test of the possibility that weak arguments can lead to greater advocacy than strong arguments. All participants were presented with persuasive arguments on a topic (Barack Obama’s reelection bid) and subsequently reported their intentions to advocate for their position, or side, on that topic. We predicted that among pro- but not counter-attitudinal participants, those who were exposed to weak rather than strong arguments would report greater intentions to advocate. We also predicted that perceived argumentation efficacy would mediate this effect, such that those who saw weak arguments would feel greater argumentation efficacy, which would in turn boost their advocacy intentions. Of importance, we also measured attitude change and voting intentions in this study, but we did not necessarily expect effects on these indices. Indeed, we classified participants as pro- or counter-attitudinal according to their initial attitudes toward Barack Obama, meaning they entered our experiment with preexisting opinions of the president. We did not expect to change those opinions or shape voting outcomes with our manipulation. Rather, our focus was on whether we could influence participants’ intention to advocate.
Participants and Design
One hundred sixty-five participants, U.S. citizens from a national pool maintained by Survey Sampling International, took part in an online survey in exchange for being entered in a raffle for a US$25 Amazon.com gift certificate. Participants were randomly assigned to one of two argument quality conditions: strong or weak.
Procedure
Before the experiment began, participants were told that the researchers conducting the study were part of a real national initiative titled Voter’s Voice. The initiative was ostensibly designed to inform Americans about why other citizens support or oppose particular political candidates. All participants were told that the researchers were investigating the ways in which voters view and make sense of their fellow citizens’ reasons for supporting and opposing those candidates. Finally, to underscore our cover story that the Voter’s Voice initiative was real and that their responses would have actual consequences, all participants were told that the information they provided would be used in future get-out-the-vote campaigns and political advertisements.
Initial attitude
Following this introduction to the study (which was conducted several months prior to the 2012 presidential election), participants were asked to report their attitudes toward Barack Obama’s bid for reelection, using a scale ranging from 1 (strongly oppose) to 9 (strongly support). Overall, the sample hovered near the midpoint of this scale (M = 5.39, SD = 2.84). We defined participants as “pro-attitudinal” if they scored higher than 5 on the initial attitude question (n = 93) and as “counter-attitudinal” if they scored lower than 5 (n = 58). We removed 14 participants who chose the midpoint of the scale (i.e., were neutral), leaving 151 participants for analysis.
Argument quality manipulation
Next, participants were asked to read and rate “a representative sample” of their fellow citizens’ arguments on this topic, which ostensibly had been collected by the Voters’ Voice organization prior to the study. Participants read that they would see arguments either in favor of or against President Obama’s reelection bid (in reality, all participants received pro-Obama arguments), and that when the study was complete, the top-rated arguments would be used in the initiative’s get-out-the-vote promotional materials. Participants were then randomly assigned to receive either three weak or three strong arguments in support of Obama (see the appendix), which they rated one at a time.
Argument quality rating
After reading each argument, participants rated the quality of that argument on a scale ranging from 1 (not persuasive at all) to 9 (very persuasive). Ratings were averaged to form a composite argument quality index for strong (α = .82) and weak (α = .89) arguments.
Attitude change and vote preference
Next, participants again reported their attitudes using the exact same scale as in the beginning of the study. Attitude scores before the manipulation were subtracted from attitude scores after the manipulation to create an attitude change index; higher numbers indicated more attitude change in the pro-Obama direction. In addition, participants were asked to report their voting intention on a scale ranging from 1 (definitely voting against Obama) to 9 (definitely voting for Obama).
Advocacy intentions
Following the voting measure, we assessed advocacy intentions. The first item was directed to supporters and opponents of Obama. In particular, participants answered a question on a 1 to 9 scale assessing the extent to which they would try to persuade others to their view: “How likely would you be to try to persuade others to your position on Barack Obama’s reelection?” (very unlikely–very likely). The next two advocacy items were directed specifically to Obama supporters, for whom the persuasive arguments were pro-attitudinal: “How willing are you to make phone calls for Barack Obama’s 2012 presidential campaign?” (1 = not at all to 9 = very much). “How willing are you to donate money to Barack Obama’s 2012 presidential campaign?” (1 = not at all to 9 = very much). For all subsequent analyses specific to pro-attitudinal participants (i.e., Obama supporters), these two items along with the first persuasion intention item were averaged together to form a composite index of supportive advocacy (α = .87).
Argumentation efficacy
Finally, participants were asked two questions about their perceived argumentation efficacy: “Do you think you know a lot more or a lot less about this issue compared with other people?” (1 = know a lot less to 9 = know a lot more). “How effective do you think you are at arguing for your side of this issue?” (1 = not effective at all to 9 = very effective). Responses were averaged to form a composite index (α = .84).
Results
Argument quality rating
We began by submitting participants’ argument quality ratings to analysis. To allow for any effect of initial attitudes on these ratings, we conducted a hierarchical linear regression analysis, in which the main effect terms for argument quality (dummy coded: weak = 0, strong = 1) and initial attitude (continuous) were included in the first step, and their interaction was included in the second step. As intended, argument quality ratings were significantly higher in the strong rather than weak argument condition (β = .46, SE = .05), t(145) = 8.57, p < .001. 1 Not surprisingly, the main effect for initial attitude was also significant (β = 1.38, SE = .32), t(145) = 4.32, p < .001, such that the more participants supported Obama, the higher they rated the quality of the arguments made in favor of him. Finally, the interaction was also significant (β = .24, SE = .11), t(144) = 2.28, p = .024. To interpret the interaction, we examined pro-attitudinal and counter-attitudinal participants—defined as being above or below the midpoint of the initial attitude scale, respectively—separately and tested for argument quality effects. That is, rather than test the simple effects of argument quality at 1 SD above and below the mean on initial attitudes, we dichotomized initial attitude to ensure that pro-attitudinal participants scored 6 or higher on that index, and counter-attitudinal participants scored 4 or lower. This analysis revealed that the argument quality effect was stronger among pro-attitudinal participants (Mweak = 4.47, SDweak = 2.50; Mstrong = 6.19, SDstrong = 1.22), t(90) = 4.10, p < .001, than it was among counter-attitudinal participants (Mweak = 2.11, SDweak = 1.80; Mstrong = 3.00, SDstrong = 1.48), t(54) = 2.04, p = .04. Of importance, though, the effect was significant and in the expected direction (i.e., strong > weak) in both cases.
Advocacy intentions
We next submitted the advocacy intention data to analysis. First, we examined the persuasive intent item, which applied to pro- and counter-attitudinal participants. Specifically, we conducted a hierarchical regression analysis with argument quality (weak vs. strong), initial attitude, and their interaction predicting intent to persuade others. The main effect was significant neither for argument quality (β = −.44, SE = .40), t(146) = −1.10, p = .27, nor for initial attitude (β = .01, SE = .07), t(146) = .20, p = .84. As illustrated in Figure 1, however, we did obtain the predicted interaction (β = −.25, SE = .13), t(145) = −1.90, p = .05. Following the same approach we used with argument quality ratings, we further explored the interaction by examining pro- and counter-attitudinal participants separately. As hypothesized, pro-attitudinal participants reported greater intention to persuade others following weak (M = 6.49, SD = 2.05) rather than strong (M = 5.51, SD = 2.34) arguments, t(90) = −2.13, p = .036. Among counter-attitudinal participants, argument quality had no effect on persuasive intent (Mweak = 5.62, SDweak = 2.89; Mstrong = 6.06, SDstrong = 2.52), t(55) = .63, p = .53.
Intention to persuade others as a function of initial attitude and argument quality in Study 1.
Having established that the effect of weak over strong arguments on our initial advocacy item occurred only among pro-attitudinal participants, we focused our attention on pro-attitudinal participants and supportive advocacy in our subsequent analyses. Again, the latter two advocacy items—pertaining to phone calls and donations—were specifically geared toward supporters of Obama (i.e., pro-attitudinal participants). Using the composite index including these two items and the initial persuasive intent item, we found the predicted effect among pro-attitudinal participants: Weak arguments (M = 5.51, SD = 2.19) led to greater supportive advocacy than did strong arguments (M = 4.43, SD = 2.29), t(88) = −2.28, p = .025.
Argumentation efficacy
Conducting the same hierarchical regression analysis on argumentation efficacy, we found only a main effect for argument quality; weak arguments produced higher argumentation efficacy scores than did strong arguments (β = −.65, SE = .32), t(145) = −2.02, p = .04. Neither the main effect for initial attitude (β = .09, SE = .08), t(145) = 1.14, p = .25, nor the interaction (β = −.08, SE = .11), t(144) = −.72, p = .47, was significant.
Mediation
Because argument quality influenced advocacy only among pro-attitudinal participants, we restricted our mediation analysis to pro-attitudinal participants to determine if perceived efficacy mediated the effect of argument quality on advocacy. As noted previously, among pro-attitudinal participants, argument quality influenced advocacy. Argument quality also had a tendency to influence pro-attitudinal participants’ argumentation efficacy (Mweak = 6.12, SDweak = 1.99; Mstrong = 5.18, SDstrong = 1.91), t(89) = −1.72, p = .089. When we entered argument quality and argumentation efficacy into a regression analysis predicting advocacy, the relationship between argumentation efficacy and advocacy was strong and significant (β = .81, SE = .09), t(85) = 9.16, p < .001, whereas the relationship between argument quality and advocacy was reduced (β = −.63, SE = .34), t(85) = −1.82, p = .07. A bootstrapped mediation analysis (Preacher & Hayes, 2004) confirmed that the mediated effect of argument quality on advocacy through perceived efficacy was significant, 95% confidence interval (CI) = [0.03, 1.30]. Thus, among pro-attitudinal participants, weak arguments led to greater perceived argumentation efficacy, which in turn led to greater advocacy.
Attitude change and vote preference
Finally, we examined whether argument quality influenced attitude change and vote preference. First, a regression analysis with an argument quality (weak vs. strong) × initial attitude interaction predicting attitude change showed no significant main effects (ps > .77). There was a marginally significant interaction (β = .08, SE = .05), t(146) = 1.84, p = .07. It appears that within the strong argument condition, Obama supporters (but not opponents) showed some degree of positive attitude change; however, none of the simple effects were reliable (ps > .18). Of greatest import, regardless of initial attitudes, the argument quality manipulation had no impact on attitude change. Not surprisingly, though, the same regression analysis predicting vote preference did reveal a significant main effect for initial attitude (β = 1.03, SE = .03), t(146) = 34.4, p < .001, such that Obama supporters reported a stronger preference for voting for him. Neither the main effect for argument quality, p = .26, nor the interaction term, p = .55, affected vote preference. Thus, although weak arguments did produce greater advocacy among pro-Obama participants, they did not appreciably alter attitudes or voting intentions.
Discussion
Study 1 provided an initial demonstration of the hypothesized advocacy effect. When participants received pro-attitudinal arguments, they ironically showed greater intention to advocate for the cause (President Obama)—specifically, to try to persuade others, to make phone calls, and to donate money—when those arguments were weak rather than strong. This effect appeared to stem from perceived argumentation efficacy. When participants received weak arguments, they believed they had more to contribute, and they displayed more supportive advocacy. In essence, when participants came to perceive themselves as more efficacious than their like-minded peers, they displayed greater willingness to promote, endorse, and advocate for their cause. The novel finding is that this perceived efficacy was derived from weak arguments.
As expected, this pattern did not hold for counter-attitudinal participants, who had similar levels of advocacy regardless of whether they saw weak or strong arguments. Exposure to weak arguments led to greater perceived argumentation efficacy regardless of whether participants were pro- or counter-attitudinal, but the boost in efficacy did not translate into greater advocacy for counter-attitudinal participants. Although our primary focus is on understanding pro-attitudinal advocacy, it is worth noting that there are a variety of possible reasons why we did not find an advantage of weak over strong arguments among counter-attitudinal participants. For example, even if counter-attitudinal participants felt more efficacious after seeing weak arguments, they might not advocate more because it is not their side that is making the weak arguments and, thus, there is no need to intervene or help the cause. Alternatively, for these participants, the argument quality manipulation might have had conflicting effects. Perhaps some intended to advocate more when they saw strong arguments because they perceived a greater need to rise to the challenge, whereas others intended to advocate more when they saw weak arguments because they thought their own arguments would be more likely to dominate those of the opposition. These two effects could have canceled each other out. In any case, we found no overall effect of argument quality on advocacy in the counter-attitudinal message case.
Finally, it is important to reiterate that the argument quality manipulation (even among pro-attitudinal participants) did not change participants’ attitudes or affect their voting preference. The effect was constrained to measures of advocacy. Again, we do not propose that weak arguments are more persuasive than strong ones, nor that they will boost voting intentions, but rather that under some conditions they can stimulate advocacy by elevating the perceived argumentation efficacy of the recipient. In the next study, we seek to further delineate these conditions.
Study 2
The primary objectives of Study 2 were to replicate the advocacy effect among pro-attitudinal participants and to identify an important boundary condition. Specifically, we explored the moderating role of argumentation efficacy. Recall that in Study 1, we found that the advantage of weak over strong was mediated by increases in perceived argumentation efficacy. As outlined earlier, if the effect is driven by feelings of increased efficacy, it should be especially likely to emerge under conditions in which people do not already feel highly efficacious. To test this possibility, we manipulated perceived argumentation efficacy such that some participants would feel like they were highly effective at argumentation, whereas others would feel like they were only moderately effective at argumentation. We hypothesized that weak arguments would foster greater advocacy under moderate but not high efficacy conditions. When participants felt just moderately efficacious, we expected weak arguments from others to increase feelings of argumentation efficacy as in Study 1, resulting in greater advocacy. In the high argumentation efficacy condition, we expected that participants would be more willing to advocate in general and be less affected by the quality of arguments they received. This is not to say that high argumentation efficacy individuals would always advocate regardless of contextual factors, but that in the current context they might generally be more likely to do so and presumably would receive less boost from weak arguments.
Participants and Design
Because our focus was on pro-attitudinal participants, we pre-selected participants from a national online pool (maintained by Survey Sampling International) who were “pro-attitudinal” on the message topic (President Obama’s reelection), using the same criteria as in Study 1. Ninety-five pro-attitudinal participants then took part in exchange for being entered in a raffle for a US$25 Amazon.com gift certificate. They were randomly assigned to one of two argument quality conditions (weak vs. strong) and one of two argumentation efficacy conditions (moderate vs. high).
Procedure
The procedure for this study was virtually identical to the procedure in Study 1, with two exceptions. First, we manipulated perceived argumentation efficacy. Before participants received any arguments for Obama’s reelection, they were told that the researchers conducting the study wanted to collect some information about participants’ argumentation skills before they proceeded with the Voters’ Voice initiative. Participants were asked to read an unrelated policy proposal, regarding comprehensive exams as a university graduation requirement (see Petty & Cacioppo, 1986), and then write their own arguments for or against the proposal. After writing their own arguments, participants were led to believe that a reputable software program would analyze the content of their writing and determine their argumentation skill level. That skill level would be reported to the participant as a percentile score relative to a national sample.
After reading about the exam issue and writing their arguments, participants were randomly assigned to receive feedback that their argumentation efficacy was high or moderate. Specifically, they were told that their argumentation skills were either excellent and in the 91st percentile, or average and in the 53rd percentile. The feedback read “91st percentile (Excellent)” or “53rd percentile (Average)” followed by the statement “Based on an analysis of your arguments, you scored in the [91st/53rd] percentile of argumentation skill. That means that your argumentation skills are better than [91%/53%] of our sample.”
After participants received their scores, they returned to the Voters’ Voice initiative, evaluated either weak or strong arguments, and then completed the study in the same manner as in Study 1. The only difference was that we simplified the measure of argumentation efficacy to one item to focus it squarely on argumentation: “How effective do you think you are at arguing for your side of this issue [i.e., Obama’s reelection]?” (1 = not effective at all to 9 = very effective).
Results
Argument quality ratings
As in Study 1, we began by submitting the argument quality ratings to analysis. In this case, we conducted a 2 (argumentation efficacy: moderate vs. high) × 2 (argument quality: weak vs. strong) ANOVA predicting perceived argument quality. First, the main effect of argument quality was significant, such that participants rated the arguments as weaker in the weak (M = 3.47, SD = 2.08) rather than strong (M = 6.55, SD = 1.22) argument condition, F(1, 91) = 74.90, p < .001. Neither the main effect for argumentation efficacy, p = .66, nor the interaction, p = .58, was significant, suggesting that argument quality ratings did not differ as a function of efficacy condition.
Advocacy intentions
Next, we conducted the same 2 × 2 ANOVA on our advocacy intentions index, which used the same three items as in Study 1 (α = .77). This analysis revealed a main effect for argument quality, F(1, 91) = 5.70, p = .019, such that weak arguments (M = 5.07, SD = 2.05) led to greater advocacy intentions than did strong arguments (M = 4.72, SD = 1.97). The main effect for argumentation efficacy was not significant, F(1, 91) = 1.53, p = .22. Most importantly, as depicted in Figure 2, the interaction was significant, F(1, 91) = 6.74, p = .011. As hypothesized, we found that weak arguments (M = 5.43, SD = 2.17) led to greater advocacy than strong arguments (M = 4.08, SD = 1.50) in the moderate efficacy condition, t(46) = −2.50, p = .016, but not in the high efficacy condition (Mweak = 4.75, SDweak = 1.91; Mstrong = 5.50, SDstrong = 2.20), t(45) = 1.24, p = .22.
Advocacy intentions as a function of argumentation efficacy and argument quality in Study 2.
Perceived argumentation efficacy
On the perceived efficacy item, there was no main effect for argument quality, p = .96, but we did find a marginal tendency for perceived argumentation efficacy to be higher in the high (M = 6.47, SD = 1.78) rather than moderate (M = 5.83, SD = 1.93) efficacy condition, F(1, 91) = 2.88, p = .09. This tendency was qualified by a significant interaction between argument quality and argumentation efficacy, F(1, 91) = 5.47, p = .022. As predicted, in the moderate efficacy condition, weak arguments (M = 6.28, SD = 2.05) tended to produce greater perceived argumentation efficacy than did strong arguments (M = 5.43, SD = 1.74), t(46) = −1.64, p = .11, though this simple effect was not significant. Interestingly, in the high argumentation efficacy condition, we found an unexpected reversal of this tendency, such that strong arguments (M = 6.96, SD = 1.59) produced marginally greater perceived efficacy than did weak arguments (M = 6.04, SD = 1.84), t(45) = 1.72, p = .09. Although we did not anticipate this reversal, the overall interaction suggesting that weak arguments tended to foster greater perceived efficacy than strong arguments under moderate efficacy conditions was consistent with our predictions. Also noteworthy, the means for perceived efficacy were above the scale midpoint (5) in each condition, consistent with our intention to create “moderate” and “high” efficacy conditions.
Mediation
To examine whether perceived argumentation efficacy mediated the Argument Quality × Argumentation Efficacy interaction on advocacy, a mediated moderation analysis was conducted (Muller, Judd, & Yzerbyt, 2005). As noted above, the interaction between argument quality and argumentation efficacy was significant on perceived argumentation efficacy and advocacy intentions. When we submitted the advocacy index to a regression analysis with argument quality, argumentation efficacy, and their interaction as predictors, controlling for perceived argumentation efficacy, the relationship between perceived efficacy and advocacy remained strong and significant (β = .49, SE = .10), t(90) = 4.89, p < .001, but the direct effect of the interaction was reduced (β = 1.23, SE = .74), t(90) = 1.66, p = .10. Most important, a bootstrapped mediation analysis (Preacher & Hayes, 2004) indicated that the mediating pathway was significant, 95% CI = [0.18, 1.75]. Thus, when participants were initially induced to have moderate levels of argumentation efficacy, seeing weak rather than strong arguments led to greater perceived efficacy, which in turn fostered advocacy.
Attitude change and vote preference
Finally, using the same measures of attitude change and vote preference as in Study 1, we found no significant effects on either index (all ps > .19).
Discussion
Study 2 replicated the key advocacy effect from Study 1 while also highlighting a boundary condition. Participants advocated for a cause more when they believed they would be effective at doing so, whether that belief came from our efficacy manipulation or the fact that their peers had made weak arguments. In fact, in comparing the perceived efficacy ratings from the moderate efficacy/strong arguments condition with the ratings in the other three conditions (in which high perceived efficacy was induced through our efficacy manipulation or through weak arguments), we found that perceived efficacy was significantly lower in the former condition (M = 5.42, SD = 1.74) than in the latter conditions (M = 6.39, SD = 1.86), F(1, 93) = 5.10, p = .026, which did not differ from each other, F(2, 68) = 1.55, p = .22. The same pattern of results held for advocacy. Advocacy intention was lower in the moderate efficacy/strong arguments condition (M = 4.08, SD = 1.50) than in the other three conditions (M = 5.19, SD = 2.08), F(1, 93) = 5.76, p = .018, which did not differ from each other, F(2, 68) = .97, p = .38. It could be that when participants were told that they were excellent at arguing, they felt no need to observe the quality of others’ arguments to make an inference about their own argumentation skills. However, when participants initially were made to feel less confident about their argumentation skills, seeing others make specious arguments made them feel more effective at arguing, which in turn made them more willing to advocate.
Notably, the advocacy effect among those in the moderate efficacy condition of Study 2 paralleled the effect among all participants in Study 1. Indeed, the mean and median argumentation efficacy ratings among all participants in Study 1 were 5.78 and 6.00, respectively, which were essentially equivalent to the argumentation efficacy scores among participants in the moderate efficacy condition of Study 2 (M = 5.83, median = 6.00). Participants in the high argumentation efficacy condition in Study 2 had mean and median argumentation efficacy scores of 6.47 and 7.00, respectively. Collectively, then, Study 1 and Study 2 suggest that when one is only moderately confident about one’s argumentation skills, seeing others make weak rather than strong pro-attitudinal arguments can lead to greater advocacy.
Study 3
Our goals in the third study were to replicate the advocacy effect in a completely different paradigm and determine its relationship with a well-established, conceptually related moderator. As noted previously, individuals high (compared with low) in attitude certainty generally are more likely to behave in ways that are consistent with their attitudes (Fazio & Zanna, 1978), express willingness to vote and sign petitions (Barden & Petty, 2008), and try to persuade others to their views (Visser et al., 2003). In this respect, attitude certainty bears some relation to argumentation efficacy and advocacy as we have described them. In Study 3, we directly explored the role of attitude certainty in moderating the effect of weak (pro-attitudinal) arguments on advocacy. Based on past research, we generally expected individuals high in attitude certainty to express a willingness to advocate and to show little influence of argument quality (Tormala & Rucker, 2007). By contrast, individuals lower in attitude certainty were expected to advocate to a greater extent when they received weak rather than strong arguments. Again, the logic was that for individuals lacking maximal certainty, receiving weak arguments in support of their own views might boost their perceived argumentation efficacy, producing greater advocacy under weak rather than strong argument conditions. In addition to investigating the role of attitude certainty in this study, we changed the attitude issue to establish the robustness of the advocacy effect. Specifically, whereas the previous studies involved pro-attitudinal arguments for a political candidate, Study 3 presented pro-attitudinal advocacy against a social policy (i.e., arguments that took the same side as participants, but were against a policy).
Participants and Design
An initial sample of 141 participants from a national online pool (maintained by Survey Sampling International) took part in exchange for being entered in a raffle for a US$25 Amazon.com gift certificate. Participants were randomly assigned to one of two argument quality conditions: strong or weak.
Procedure
At the beginning of the experiment, participants were asked to imagine that officials in their town recently had proposed a policy that would make all school cafeterias vegetarian by law. According to this policy, all existing K-12 (i.e., primary and secondary) schools would have to change their menus to provide only vegetarian options, and all future schools within the same districts would only be allowed to provide vegetarian options for students.
Attitude
After reading the introduction and set up, participants were asked to indicate their attitudes toward the vegetarian-only school policy on a 9-point scale anchored at strongly oppose (1) and strongly support (9). Because we sought to obtain a sample of pro-attitudinal participants to test our focal hypotheses, we pretested the issue to confirm that most participants’ attitudes toward the policy would be in the same direction as the arguments to which we would later expose them. In this case, most participants opposed the policy, and all participants were subsequently exposed to arguments against it later in the experiment. Thus, in this study, pro-attitudinal participants were defined as those who scored lower than 5 on the attitude measure—that is, they opposed the policy and later received arguments that also opposed the policy. Our sample included 118 pro-attitudinal participants, and all subsequent analyses included only those participants.
Attitude certainty
Immediately following the initial attitude item, participants were asked five questions assessing their attitude certainty on 9-point scales (adapted from Petrocelli et al., 2007): “How certain are you about your attitude toward this policy?” (1 = not certain at all to 9 = very certain). “How certain are you that you know what your true attitude toward this policy really is?” (1 = not certain at all to 9 = very certain). “To what extent is your true attitude toward this policy clear in your mind?” (1 = not clear at all to 9 = very clear). “How certain are you that your attitude toward this policy is the correct attitude to have?” (1 = not certain at all to 9 very certain). “To what extent do you think other people should have the same attitude as you about this policy?” (1 = not certain at all to 9 = very certain). These five items were averaged together to form a composite index (α = .86).
Argument quality manipulation
Next, participants were told to imagine that their local government was going to hold a town hall meeting in which members of the community would be able to express support for or opposition to the vegetarian-only policy, and that the government would take these opinions into consideration when making a decision on whether or not to enact it. All participants were told to imagine that they were at this town hall meeting and that some community members were expressing their views about the policy. Participants were told that on the next screen of the experiment they would read the views expressed by these community members. When participants continued, they were exposed to either three weak or three strong arguments opposing the vegetarian-only schools policy (see the appendix).
Argument quality rating
After reading all three arguments presented in their condition, participants gave those arguments one overall rating on a scale ranging from 1 (very low quality) to 9 (very high quality).
Advocacy intentions
Finally, participants were asked three questions assessing their advocacy intentions: “How likely would you be to try to persuade others to your position on this policy?” (1 = very unlikely to 9 = very likely). “How much effort would you be willing to put into supporting or opposing this policy in the future?” (1 = none at all to 9 = a lot). “To what extent would you share your views on this policy with your friends?” (1 = not at all to 9 = very much). Responses to these items were averaged to form a composite advocacy index (α = .85).
Results
Argument quality ratings
Again, we started by submitting argument quality ratings to analysis. As in Study 1, we conducted a hierarchical linear regression analysis. In this case, however, the main effect terms for argument quality (dummy coded: weak = 0, strong = 1) and attitude certainty (continuous, mean centered) were included in the first step, and their interaction was included in the second step. As expected, the main effect for argument quality was significant (β = 3.45, SE = .42), t(115) = 8.15, p < .001; strong arguments were rated as significantly stronger than weak arguments. There was no main effect for attitude certainty, p = .40. The interaction between argument quality and attitude certainty was significant (β = .73, SE = .31), t(114) = 2.38, p = .019. To interpret this interaction, we examined the effects of argument quality among high (1 SD above the mean) and low (1 SD below the mean) attitude certainty participants (Aiken & West, 1991). 2 This analysis revealed that the argument quality effect was greater among high (β = 4.50, SE = .60), t(114) = 7.43, p < .001, rather than low (β = 2.49, SE = .58), t(114) = 4.29, p < .001, certainty participants, though it was significant and in the predicted direction in each case.
Advocacy intentions
Having established that our argument quality manipulation was successful, we examined the advocacy effect. Here too, we conducted a hierarchical regression analysis in which argument quality, attitude certainty, and their interaction predicted advocacy intentions. This analysis revealed a main effect for attitude certainty (β = .54, SE = .10), t(115) = 5.21, p < .001, such that higher attitude certainty generally led to greater advocacy. Although participants also tended to advocate more in the weak rather than strong argument condition, the main effect for argument quality was not significant, p = .12. Most germane to our primary concerns, the interaction was significant (β =.61, SE = .21), t(114) = 2.94, p = .004. As illustrated in Figure 3, among low certainty participants (1 SD below the mean), weak arguments led to greater intent to advocate than did strong arguments (β = −1.25, SE = .39), t(114) = −3.20, p = .002. Among higher certainty participants (1 SD above the mean), there was no effect of argument quality on advocacy intentions (β = .43, SE = .41), t(114) = 1.05, p = .30, which were generally high irrespective of argument quality.
Advocacy intentions as predicted by attitude certainty and argument quality in Study 3.
Discussion
As in the previous studies, Study 3 showed that weak pro-attitudinal arguments can under specifiable conditions increase people’s intent to advocate for a cause they support. In particular, we found that participants who were low in attitude certainty expressed greater intent to advocate when they saw others’ weak rather than strong arguments. By contrast, participants who were high in attitude certainty expressed generally greater advocacy intention and were insensitive to the quality of arguments others made. Thus, we replicated the advocacy effect in a new context and with a different moderator. Although we had no direct process evidence in Study 3, we suspect that the mechanism was the same as observed in Studies 1 and 2. That is, uncertain individuals presumably felt themselves to be less-efficacious arguers for their position but could be made to feel as though they had something unique and useful to contribute when exposed to others’ specious arguments. Highly certain individuals, by contrast, presumably did feel as though they could argue for their position, and thus expressed greater intent to do so regardless of the arguments others made.
Study 4
Studies 1 to 3 tested the proposed advocacy effect by measuring advocacy intentions. In Study 4, we assessed participants’ actual attempts at generating advocacy. Specifically, we replicated Study 3 but replaced the measure of advocacy intentions with a measure of actual argumentation, recording participants’ self-generated arguments in support of their position. In accord with past research (e.g., Gal & Rucker, 2010), we gauged advocacy by assessing how much participants wrote in support of their views when they were given an opportunity to do so. In accord with Study 3, we hypothesized that individuals low in attitude certainty would argue more for the cause when they received weak rather than strong (pro-attitudinal) arguments, whereas individuals high in attitude certainty would be unaffected by argument strength.
Participants and Design
An initial sample of 388 participants, recruited using Amazon’s Mechanical Turk, took part in exchange for credit toward Amazon.com purchases. Participants were randomly assigned to one of two argument quality conditions: strong or weak.
Procedure
The procedure for Study 4 was identical to that of Study 3, except that in place of the advocacy intention items, we asked participants to imagine that they had an opportunity at the town hall meeting to express their own arguments. We then provided them with a text box in which they could enter any arguments that they would want to make at the meeting (“In the space below, please write down everything that you would say at this point in the town hall meeting.”). Following past research (Gal & Rucker, 2010), we created an advocacy index by tallying the number of words each participant typed into the text box (range = 7-244 words, M = 57.40, median = 49, SD = 34.90). Attitudes, attitude certainty (α = .88), and argument quality ratings were assessed in the exact same manner, and at the exact same point in the study, as in Study 3. Of importance, using the same criteria from Study 3, we identified 315 pro-attitudinal participants in our sample. All subsequent analyses focused on these participants.
Results
Argument quality ratings
Following the same procedure as in Study 3, we found a main effect for argument quality on argument quality ratings (β = 4.98, SE = .18), t(310) = 27.4, p < .001, showing that strong arguments were rated as stronger than weak arguments. There also was a main effect for attitude certainty (β = .23, SE = .07), t(310) = 3.53, p < .001, such that participants who were more certain of their attitudes perceived the arguments as stronger than did participants who were not as certain. In addition, as in Study 3, the interaction between argument quality and attitude certainty was significant (β = .39, SE = .14), t(309) = 2.87, p = .004. The argument quality effect was greater among high (β = 5.50, SE = .26), t(309) = 21.5, p < .001, rather than low (β = 4.42, SE = .47), t(309) = 16.6, p < .001, certainty participants, though it was significant and in the predicted direction in each case.
Advocacy effort
Most germane to our central concerns, we conducted another hierarchical regression analysis in which argument quality, attitude certainty, and their interaction predicted actual advocacy effort, operationalized in terms of word count. Although participants tended to write more in the weak rather than strong argument condition, the main effect for argument quality was not significant, p = .13. The main effect for attitude certainty was not significant either, p = .35. Most important, the interaction was significant (β = 6.22, SE = 3.01), t(309) = 2.07, p = .039. As expected, among low certainty participants, weak arguments led to greater advocacy effort than did strong arguments (β = −14.91, SE = 5.87), t(309) = −2.54, p = .012. Among high certainty participants, there was no effect of argument quality on advocacy effort (β = 2.32, SE = 5.63), t(309) = .41, p = .68 (see Figure 4).
Actual advocacy effort (number of words typed) as predicted by attitude certainty and argument quality in Study 4.
Discussion
Study 4 demonstrated that the advocacy intention effect observed in Study 3 can also manifest in terms of actual advocacy behavior. As hypothesized, uncertain individuals actually argued more after receiving weak rather than strong pro-attitudinal arguments. This finding extended the earlier results by moving beyond mere intentions to advocate for a cause and studying the actual arguments people generate—or the persuasion effort they mount—after receiving weak arguments from someone else. Of importance, replicating Study 3, this effect did not obtain among high certainty individuals. Interestingly, it did not appear to be the case that high certainty individuals actually advocated at an exceptionally high level compared with other individuals, as observed in Study 3 and with argumentation efficacy in Study 2. Nevertheless, the overall pattern was consistent with our primary hypothesis that the benefit of weak over strong arguments in fostering advocacy would be limited to relatively uncertain individuals.
General Discussion
Across four studies, we found evidence for the proposed advocacy effect. Specifically, we found that weak arguments could sometimes outperform strong arguments in attitude-relevant contexts. As hypothesized, we observed this effect on measures of advocacy intentions (e.g., individuals’ willingness to try to persuade others, to share their views with others, and to donate to their preferred cause) and actual advocacy (i.e., the amount people wrote while arguing for their cause). In contrast, we did not observe any advantage for weak arguments on measures of voting intentions or attitude change. Thus, we do not mean to suggest that weak arguments are more persuasive than strong ones. Rather, we submit that they can be more effective at generating advocacy. Importantly, this effect appears to be restricted to pro-attitudinal persuasion contexts—that is, settings in which people are exposed to weak arguments supporting their own positions. There was no apparent benefit of weak arguments over strong ones when people received counter-attitudinal messages.
In addition to documenting a conditional advantage for weak over strong arguments, we also provided evidence for its mechanism. In Studies 1 and 2, we found that weak arguments stimulated advocacy via their effect on perceived argumentation efficacy. In Study 1, for instance, when participants received weak rather than strong arguments supporting their own opinions, they perceived their own argumentation efficacy to be higher. This feeling of argumentation efficacy, in turn, fostered advocacy intentions. Moreover, as predicted, given this mechanism, the advocacy effect occurred primarily when participants’ initial feelings of argumentation efficacy (Study 2) or attitude certainty (Studies 3-4) were not already high. Indeed, when message recipients start off feeling uncertain about their attitudes or their argumentation skills, there is more room for growth on these dimensions. The boost from weak arguments thus applies primarily when individuals initially feel inefficacious or uncertain. When initial feelings of efficacy or certainty are high, people tend to advocate more in general (Studies 2-3) and are less sensitive to the strength of arguments they receive from their like-minded peers (Studies 2-4).
Taken together, these findings reveal that weak arguments can play an important role in stimulating advocacy. They suggest, counterintuitively, that it might sometimes behoove advocacy groups to expose their supporters to weak arguments from others—especially if those supporters are initially uncertain about their attitudes or about their ability to make the case for them. Weak arguments appear to be capable of bolstering such supporters’ appraisals of their own abilities to advocate effectively, and perhaps also of the need to advocate in the first place. Ironically, then, prompting supporters to advocate on behalf of a cause might be facilitated by the strategic use of weak arguments. This suggestion might be somewhat radical, but the current results indicate that it is at least worth considering the potential benefits of weak arguments in advocacy contexts.
Although our studies focused on perceived argumentation efficacy as a mechanism by which weak arguments can increase advocacy, other possible processes could contribute to this effect. As one example, hearing others make weak arguments in favor of a cause could lead fellow supporters to think that the cause will fail without their intervention. Other recent research has shown that when people’s own view or preferred position is being promoted by a poorly organized source or by a nonexpert, they pay more attention to that message out of a concern that the source may fail to adequately support their view and/or persuade others (Clark & Wegener, 2009; Clark, Wegener, Habashi, & Evans, 2012). It is possible that seeing one’s own position argued weakly, as in the current studies, creates the same kind of concern that those weak arguments will fail to persuade others. In response, supporters might sometimes step in and advocate more themselves if they believe that it is up to them to intervene and make a more compelling case.
In fact, one of our advocacy items in Studies 1 and 2 tapped participants’ willingness to donate money to Barack Obama’s campaign, which is perhaps less about argumentation efficacy and more about the perception that help is needed or that something must be done. That we obtained the predicted effect on this measure as well as on more argumentation-type measures (i.e., trying to persuade others and making phone calls on the campaign’s behalf) is consistent with the possibility that weak arguments might have worked through more than one channel in our studies. Our mediation and moderation findings suggest that perceived argumentation efficacy played an important role in making weak arguments more effective than strong ones, but the door remains open to exploring whether other mechanisms might also be involved.
In addition to providing new insights and raising new questions for research in attitudes and persuasion, the current work has implications for understanding self-efficacy and attitude certainty. In particular, our studies uncover ways that people who lack these states can be motivated to act. For instance, although previous research has shown numerous ways to increase perceived self-efficacy (Bandura, 1977), our studies demonstrate for the first time that persuasive messages—particularly weak ones—can play a role in this regard. Indeed, simply by observing others make weak arguments, people can come to perceive that they themselves are more effective at arguing. Thus, whereas past research has shown that observing similar others perform tasks successfully can boost self-efficacy (Bandura & Barab, 1973), observing them perform tasks poorly might in some instances do so as well.
Finally, our findings point to a number of interesting and potentially important questions for future research. As but one example, in each of the present studies, the ostensible source of the arguments participants received was a peer, or at least there was no mention of the source having any relevant expertise. Perhaps the presumed similarity of the source to the recipient made it relatively easy for participants to feel as though they could draw inferences about themselves by observing the source’s arguments. Indeed, the perceived argumentation efficacy effect in our studies likely revolved around a social comparison process (Corcoran, Crusius, & Mussweiler, 2011; Festinger, 1954), whereby participants inferred greater efficacy after observing the relatively poor efficacy (i.e., weak arguments) of others. It remains to be seen how participants would respond if the source of the arguments were an established expert on the subject. Weak arguments from an unambiguous expert, for instance, might elicit the inference that the best arguments on the topic are truly unconvincing, which seems unlikely to motivate advocacy. Consistent with this general notion, there is evidence from past research that high (versus low) expertise can backfire when the source presents weak arguments (e.g., Bohner, Ruder, & Erb, 2002; Tormala, Briñol, & Petty, 2006), though that effect has been observed on persuasion rather than advocacy. We can only speculate for now, but it would be useful to explore the effects of source variations in the current context as well.
In closing, we submit that in addition to uncovering a novel and counterintuitive effect, the present studies make a broader contribution to the literature. In particular, they suggest that in at least some cases it may be important for persuasion research to shift the focus away from attitude change or persuasion per se, and onto other outcomes such as attitudinal advocacy. Over the past several years, persuasion researchers increasingly have moved away from emphasizing attitude change as the sole or primary outcome of interest following a persuasive appeal. With this shift, we have seen a greater focus on metacognitive processes and outcomes such as increased or decreased attitude certainty, among other things (see Petty, Briñol, Tormala, & Wegener, 2007; Tormala & Rucker, 2007). The current work underscores this departure. We contend that considering other outcomes beyond attitudes and attitude change can deepen our insight into persuasion and expand our understanding of the many important outcomes that can vary following exposure to a persuasive message. We hope that our findings will generate interest in studying advocacy among attitude and persuasion researchers, as understanding how to motivate action and change behavior remains a core goal of our discipline.
Footnotes
Appendix
Editor’s Note
Dr. Greg Maio served as guest action editor for this article.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.