The SAGE Handbook of Persuasion

by James Price Dillard

  Based on the analysis of the over-time data, no evidence of nonmonotonicity was found, which supports the single-push-with friction model rather than the push-with-pullback model. However, cognitive responses did play a role in some of the study’s outcomes: There is evidence that in some conditions, the effects of source and of discrepancy increased over time, which may be due to cognitive responses.

  Although the study by Chung and colleagues is complex, it analyzed the questions about discrepancy we posed earlier, and it provided a tentative answer about the shape of the discrepancy-belief change relation: With high involvement, the curve is monotonic; with low involvement and a high-credibility source, the curve is monotonic; and with low involvement and a low-credibility source, the curve is nonmonotonic. Furthermore, over-time data were not consistent with the idea that strong counterarguments create a downturn in the over-time movement to equilibrium.

  Auxiliary Issues

  * * *

  Discrepancy or Disconfirmation?

  At this point we note something that should be obvious, at least in retrospect: In almost all the investigations concerning discrepancy, and in all the ones included here, messages that were greatly discrepant were also greatly surprising. Consider Bochner and Insko’s (1966) messages: Some of them argued for 4, 3, 2, 1, or even 0 hours of sleep per night as appropriate for the average young adult; these messages are clearly surprising. What’s more, the level of surprise correlates with the level of discrepancy. In other words, these two variables were confounded, and for all we know it may have been the surprise value, rather than the level of discrepancy, that accounted for the results of the studies that have been reviewed here.

  Kaplowitz and Fink (1991), discussed previously, labeled a message’s surprise value as its level of disconfirmation; they conducted two experiments in which manipulated discrepancy and manipulated disconfirmation were orthogonal. The topic of both experiments was criminal sentencing. One dependent variable was the number of years of imprisonment respondents recommended for a convicted armed robber (P1). The second dependent variable was the comparative evaluation of the robber: “How bad is Defendant X?” In Experiment 1, a third independent variable was focus of attention: Respondents were directed either to focus on the source (the judge) or the reasons given for the sentence. In Experiment 2, a third independent variable was the size of the sample of defendants (3 vs. 100) previously sentenced by the judge. These additional independent variables were associated with hypotheses designed to tease out the role of cognitive elaboration.

  In both experiments, discrepancy was found to directly and positively affect P1 (the person’s position after the message) and not comparative evaluation, whereas disconfirmation was found to directly and negatively affect comparative evaluation and not directly affect (Experiment 1) or weakly and negatively affect (Experiment 2) P1. The proportion of variance directly explained by discrepancy in predicting P1 (about 30%) was much greater than the proportion of variance that disconfirmation directly explained in predicting comparative evaluation (about 4%). Finally, the “effects of comparative evaluation on position [P1] appear to require substantial cognitive elaboration” (Kaplowitz & Fink, 1991, p. 191), although the effects due to discrepancy did not. This study clearly shows that it is discrepancy rather than disconfirmation that accounts for the effects on belief position. However, a second process also occurs: Disconfirmation affects the evaluation of the focal object. Furthermore, disconfirmation’s effect on comparative evaluation “appears to require thinking about one’s expectancy regarding the source and about the disconfirmation of that expectancy” (p. 205).

  Social and Psychological Factors Examined Over Time

  The research that has been presented to this point has dealt principally with psychological processes: the perceptions, thoughts, and other cognitive activities related to processing discrepant messages. (We note that emotions, which may play a role, have not been the focus of our discussion.) But implicit social processes are clearly entwined with the psychological ones. For one thing, every study cited involved humans interacting with humans, even if the experimenter merely gave out questionnaires in a classroom; that human-to-human interaction undoubtedly has some effect. Second, recall the two responses that Bochner and Insko (1966) described as “unavailable”: arguing with the message source against the position of the message and obtaining social support. In responding to messages outside of the lab, both of these behaviors are clearly social and generally available (see Smith & Fink, 2010).

  To incorporate potential social processes, Kaplowitz et al. (1986) conducted an experiment using panel data at two points in time. The study replicated Fink and colleagues (1983) research, which gathered data immediately after respondents read the messages: The 1986 study used the same topic (a tuition increase) at the same university, and five of the experimental conditions appear in both the 1983 and 1986 studies. Kaplowitz et al. (1986) asked participants the same question, both immediately and four to eight days after the initial response, about the tuition increase that they (the participants) would propose. The second, later data gathering was disguised in several ways and seemed to be part of a different study. In addition, after the initial (time 1) data were collected, the participants were debriefed about the deception involved in that part of the study.

  The time-1 results were essentially identical to Fink and colleagues’ (1983) results: The rank order correlation of the means of the five conditions that were in common across these two studies was 1.00, and the Pearson correlation was .89. This replication was successful.

  More important for our current discussion is what was found at time 2. There was a dramatic change in the relative effectiveness of the experimental conditions. The six time-1 conditions that had messages advocating a tuition increase of some amount (i.e., excluding a no-position control condition) were initially ordered (from most change to least change; E = extreme, M = moderate) E/E (i.e., two messages, the first and the second extremely discrepant), E/M, M/M, M/E, E, and M. At time 2, the messages formed two clusters: The most-change cluster consisted of messages in which the first or only message was M; the least-change cluster consisted of messages in which the first or only message was E. Note that the message that was most effective at time 1 became one of the least effective messages at time 2, and the message that was least effective at time 1 became one of the most effective at time 2.

  Although the processes involved were not directly assessed, the data analysis allowed the authors to make a reasonable interpretation of what took place over time. Summarizing the relevant results (Kaplowitz et al., 1986, pp. 525–526):

  Forgetting affects the long-term effectiveness of messages.

  Recipients of the moderately discrepant message received more messages in the days between time 1 and time 2. These messages could be external (from others) or internal (based on the recipient’s cognitive elaboration); furthermore, these messages supported a belief position that was greater than P0.

  Recipients of the extremely discrepant message received messages in the days between time 1 and time 2 that supported a belief position that was less than P0.

  The more discrepant message “was either remembered better or produced fewer delayed messages” (italics in original; pp. 525–526). There is reason to believe that the former explanation was more plausible.

  In other words, over time, forgetting, thinking, and social processes—such as arguing with the communicator’s position and seeking social support—changed the initial response and changed it dramatically. This study clearly shows that a complete understanding of discrepancy of beliefs requires data over longer times—at least several days—to understand the interplay of the cognitive and social processes that may be at work.

  Remaining Questions and Future Research

  * * *

  Discrepancy and Oscillation

  The first study that examined oscillation of beliefs and discrepancy has already been discussed: the study by Kaplowitz and c
olleagues (l983), which used a between-participants design. Since that study, oscillation studies have used within-participant designs, relying on participants making decisions between belief alternatives, such as whom to recommend for college admission (McGreevy, 1996; Wang, 1993; this research has been reviewed in Fink et al., 2002).

  The relevance of oscillation for modeling discrepancy is clear: If different discrepant messages induce oscillations of different amplitudes or phases, conclusions about their relative effectiveness have a good chance of being incorrect. If the process has not yet reached equilibrium, results reflect the belief that exists at the moment of measurement. For example, Chung and Fink (2008, based on McGreevy’s, 1996, data) examined the number of belief changes induced by univalent versus mixed-valence messages. Using a computer mouse, participants continuously reported their belief while reading a message (message-receipt phase; average time = 126.41 s), and after receipt of the message, they continuously reported their belief while making their decision (postmessage phase; average time = 59.22 s). During the postmessage phase, the mixed-valence message was found to cause more changes in belief than did the univalent message, and these temporary beliefs could be mistaken for equilibrium values.

  Future research needs to examine the trajectories and impacts of discrepant messages on oscillation. The current models of over-time effects have been only partially successful in capturing the processes at work.

  Cognitive Responses, Cognitive Dissonance, and Discrepancy

  Cognitive Responses

  Related to the analysis of discrepancy and oscillation is the role of cognitive responses. When a trajectory of beliefs indicates oscillation, are there accompanying thoughts that are associated with that change? Given the findings of Chung and Fink (2008), it seems likely that that thinking is associated with oscillation. To further examine this question, research needs to be conducted that interrupts a participant to find what, if any, thoughts are being considered while the participant is moving the computer mouse—indicating a change in belief position—in one direction or another. It may be that cognitive responses direct the movement toward a new belief position, but it is also possible that the position, arrived at by some dynamic cognitive algebra (Anderson, 1974), forms the cognitive response. Himmelfarb (1974), supporting the linear discrepancy model (referring to it as information integration), raised this same issue with regard to apparent resistance effects in persuasion: “Resistance effects cannot simply be inferred from differences in the overall attitudinal response” (p. 413). The relationship between belief trajectories and cognitive responses needs to be determined.

  The elaboration likelihood model (ELM; Petty & Cacioppo, 1986) suggests that “a given variable may play different roles in the persuasion process” (O’Keefe, 2002, p. 161). The roles that source credibility and discrepancy play in discrepancy models are not fully resolved. In the studies that have been reviewed, discrepancy appears to have induced central processing in some research and peripheral processing in other research. On one hand, it seems that discrepancy causes beliefs almost automatically, as if a response to a cognitive algebra mechanism: Note that Kaplowitz and Fink’s (1991) finding that focus of attention (source vs. reasons) and alleged size of the behavioral sample on which the expectations of the source’s position were based had little effect on the relation between discrepancy and P1. On the other hand, the psychological-discrepancy-discounting model has a role for psychological discrepancy, which may seem to suggest elaboration and resistance could also be associated with levels of attention or other factors (Fink et al., 1983). In addition, a relationship between discrepancy and counterargument production has been found (e.g., Brock, 1967; Kaplowitz & Fink, 1995), but this relationship does not seem integral to the relationship between discrepancy and belief position. Research to clarify the role of cognitive elaboration in discrepancy processes would be valuable for formulating a more complete model of discrepancy and belief change.

  Cognitive Dissonance

  If, as Aronson et al. (1963) and Bochner and Insko (1966) proposed, dissonance is caused by receipt of discrepant messages, the stress or tension associated with dissonance should be present after receipt of such messages, and greater discrepancy should cause greater dissonance. Furthermore, misattribution of stress should eliminate associated belief change (Drachman & Worchel, 1976; Fazio, Zanna, & Cooper, 1977; Pittman, 1975; Zanna & Cooper, 1974) as well as the oscillations that could indicate dissonance and regret (Walster, 1964). Research to clarify the role that dissonance plays in discrepancy is long overdue.

  The analysis of social processes needs to be more carefully investigated. Long-term effects due to messages that differ in discrepancy need to incorporate the two “laboratory unavailable” responses to dissonance mentioned by Bochner and Insko (1966).

  Involvement

  Laroche’s (1977) model included involvement as a key factor. Chung et al. (2008) used two topics that differed in level of involvement, and some important differences in model parameters between these topics were found; however, the topics differed in many unspecified ways, so that conclusions concerning the differences due to involvement must be tentative. Given the extensive research on and theory regarding involvement and belief change (e.g., Freedman, 1964; Johnson & Eagly, 1989) and given the intriguing findings in Chung et al. (2008) experimentally manipulating involvement seems to be a necessary next step to clarify its role in the discrepancy-belief change process.

  Methodological Wish List

  Measurement

  The next steps in theory construction regarding discrepancy will benefit from significant improvements in methodology. The discrepancy models in the sciences, some of which were briefly mentioned earlier, rely on conventional, agreed-on measurement rules, which are lacking in belief-change research. Furthermore, the scales that form the basis for the International System of Units (ISU; meter, kilogram, second, ampere, kelvin, candela, and mole) all have a lower bound of zero and have, in principle, no upper bound (although in practice there may be an upper bound); other scientific quantities are defined in terms of these fundamental units. The need addressed here is not just to create more reliable measures but measures that have greater precision and that can be used to derive other measures within a specified theoretical framework (see Torgerson, 1958).

  Following this logic of scientific measurement, Woelfel and Fink (1980) examined cultural and cognitive processes using distance (in their case, psychological and cultural distance), time, and related concepts to formulate theory. The study of discrepancy of beliefs, with discrepancy considered as a distance, can readily be studied using equations that are tied to fundamental measurements, such as those of distance and time. The recommendation here is to create and utilize a system of measures, rather than separate scales (typically measurement by fiat; Torgerson, 1958), that is tied to theory.

  Dynamic Models and Longitudinal Designs

  Dynamic models are best for explicating processes, which are typically written as mathematical equations. Longitudinal research designs (e.g., panel studies, time-series designs, pooled cross-sectional time-series designs) used to estimate dynamic models are generally not applied to the study of belief change, but they can be and should be; Chung and colleagues’ (2008) work is an exception. To understand process, we must see it unfold over time. Static models can only get us so far.

  Multidimensional Models

  Finally, we note that a message can induce change in concepts that are unmentioned in the message as well as change along dimensions other than the belief-position dimension. A multidimensional framework can examine both of these kinds of changes (see Dinauer & Fink, 2005; Woelfel & Fink, 1980). By focusing almost exclusively on belief position, we have not seen the whole picture, which a multidimensional analysis can provide.

  Conclusion

  * * *

  Studying the effect of discrepant messages on belief change would have seemed, at the onset, to be an easy and straightforward task. After yea
rs of considering this issue, and after different researchers, theories, and models have been brought to bear on it, there have been advances with regard to the shape of the relationship, the factors that do and do not play a role, as well as the temporal parameters of the process. There are significant questions that remain, and, alas (or hooray!), more research is needed.

  References

  * * *

  Anderson, N. H. (1974). Cognitive algebra: Integration theory applied to social attribution. In L. Berkowitz (Ed.), Advances in experimental social psychology (Vol. 7, pp. 1–101). New York, NY: Academic Press.

  Anderson, N. H., & Hovland, C. (1957). The representation of order effects in communication research. In C. Hovland, W. Mandell, E. H. Campbell, T. Brock, A. S. Luchins, A. R. Cohen, et al. (Eds.), The order of presentation in persuasion (pp. 158–169). New Haven, CT: Yale University Press.

  Aronson, E., Turner, J. A., & Carlsmith, J. M. (1963). Communicator credibility and communication discrepancy as determinants of opinion change. Journal of Abnormal and Social Psychology, 67, 31–36.

  Bar-Meir, G. (2011). Basics of fluid mechanics. Chicago, IL: Bar-Meir. Retrieved from www.potto.org/FM/fluidMechanics.pdf

  Berlo, D. K. (1960). The process of communication. New York, NY: Holt, Rinehart, & Winston.

  Bochner, S., & Insko, C. A. (1966). Communicator discrepancy, source credibility, and opinion change. Journal of Personality and Social Psychology, 4, 614–621.

  Brock, T. C. (1967). Communication discrepancy and intent to persuade as determinants of counterargument production. Journal of Experimental Social Psychology, 3, 296–309.