The logic for the effect of source credibility is straightforward: More credible sources should widen the latitude of acceptance and, as a result, induce more belief change. Greater ego involvement should result in “larger latitudes of rejection” because with high involvement “a person’s own attitude acts as a stronger anchor” (Petty & Cacioppo, 1981, p. 107).
The evidence for the social judgment effects is not strong. For example, Eagly and Telaak (1972) found that it is the width of the latitude of acceptance, rather than the discrepancy level of the message, that determined the amount of change induced by a message: the greater the width, the greater the change. A study by Miller (1965) showed that the latitude of rejection did not increase for those highly involved with the relevant message issue as compared to control group members who were highly involved with an unrelated issue. Both these studies’ findings are inconsistent with social judgment theory (see also Petty & Cacioppo, 1981). Thus, the social judgment approach lacks sufficient evidence of its validity. Eagly and Chaiken (1993) summarized this view by stating that “existing research provides little, if any, convincing evidence that the perceptual processes of assimilation and contrast covary with attitude change, let alone precede attitude change as the theory maintains” (p. 380, emphasis in original).
Cognitive Dissonance Theory
Both Aronson et al. (1963) and Bochner and Insko (1966) proposed that the theory of cognitive dissonance (Festinger, 1957) was relevant to understanding the effect of discrepancy on attitude change. Aronson et al. proposed that a discrepant message, assumed to be counterattitudinal, is a cause of dissonance. Furthermore, they suggested that there should be more dissonance when the message source is credible.
Bochner and Insko stated that discrepancy causes dissonance that can be reduced in any of four ways: conformity to the communicator’s [i.e., the source’s] point of view, disparagement of the communicator, persuasion of the communicator that he is incorrect, and obtained social support. (p. 614)
Because laboratory studies are not amenable to the third or fourth of these ways of reducing dissonance, Bochner and Insko (1966) proposed that an individual must respond to a discrepant message by being persuaded, by disparaging the source, or by some combination of the two. (Notice that this conclusion is not a requirement of theory but rather of the research design.) They then suggested that at low levels of discrepancy, belief change occurs, and that at high levels of discrepancy, disparagement occurs. Thus, they predicted the same curvilinear relationship that was predicted by social judgment theory, with disparagement and belief change treated as functional alternatives in response to a discrepant message: They predicted (1) the relationship between discrepancy and belief change is curvilinear, and (2) the extremum of the curve occurs at higher values of discrepancy the more credible the source; see Figure 6.4.
Cognitive Elaboration: Counterarguing
It seems reasonable that messages that are more discrepant induce more counterarguments, both as thoughts and as vocal disparagement of the communicator and the communication. Brock (1967) found empirical support for this relationship when considering subvocal counterarguments. This finding suggests that discrepancy causes processing through the central rather than the peripheral route (Petty & Cacioppo, 1986). The central route involves greater elaboration, indicated by a greater number of thoughts, which may be pro-attitudinal, counterattitudinal, or both. Because discrepancy is counter to one’s beliefs, when an individual has the ability and motivation to think about a message, greater discrepancy → greater elaboration → more counterarguments → reduced belief change. The reduced effectiveness of the more discrepant messages may also be associated with the greater scrutiny that these messages receive as well as their perceived weakness, because arguments in messages that are incompatible with prior beliefs are judged to be weaker than arguments in messages compatible with one’s beliefs (Edwards & Smith, 1996).
Two additional studies examined the relation of discrepancy to cognitive elaboration (see Kaplowitz & Fink, 1997). Kaplowitz and Fink (1991) manipulated discrepancy, measured the individual’s belief, measured other evaluations (including manipulation checks), and finally assessed cognitive elaboration; they found that discrepancy and its effects were not related to elaboration. In a later study (Kaplowitz & Fink, 1995), in which participants indicated their belief continuously and then reported their thoughts, discrepancy was significantly associated with elaboration in terms of the number of counterarguments but not in terms of the number of pro-belief thoughts. In this 1995 study, participants considered their position repeatedly before reporting their cognitive responses, whereas in the 1991 study other measures intervened between the belief measure and the elaboration measure. Thus, it appears that when participants are directed to consider their belief as it is being formed, discrepancy increases cognitive responses at least with regard to counterarguments. However, based on a comparison of the 1991 and 1995 studies, it appears that this elaboration reduces the effect of discrepancy on belief change, suggesting that discrepancy is more effective if it serves as a peripheral cue. Given the procedural differences between these two studies, more research on this issue is needed.
Assuming that discrepancy does increase elaboration, the elaboration likelihood model predicts the same outcome as social judgment theory and the theory of cognitive dissonance: The relation between discrepancy and belief change should be an inverted-U shaped curve, as shown in Figure 6.4. On the other hand, if discrepancy or belief position serves as a peripheral cue, a discrepant message may be rejected without any need for elaboration. Kaplowitz and Fink’s (1991) study found a large effect of discrepancy on belief change, but that study found no evidence of a downturn in belief position or of cognitive elaboration associated with discrepancy.
Is the Relation Between Discrepancy and Belief Change Nonmonotonic?
* * *
Bochner and Insko Redux
Bochner and Insko (1966) conducted an experiment that tested the functional form of the relationship of discrepancy to belief change using daily hours of sleep as the topic. Through a pilot study, P0, the participants’ initial position, was found to be 7.89 hours. Bochner and Insko’s dependent variables were belief position (“For maximum health and well being, how many hours of sleep per night do you think the average young adult should get?”) and two measures of disparagement: disparagement of the source and of the message. They manipulated three variables: message discrepancy, with nine levels (messages advocating 8, 7, 6, 5, 4, 3, 2, 1, and 0 hours of sleep); source credibility, with two levels (high vs. moderate); and the order in which the dependent variables were measured (belief then disparagement vs. disparagement then belief).
Figure 6.5. shows Bochner and Insko’s (1966) results, combining data for the two different orders, which did not differ in their effects. The figure shows that messages attributed to the source with greater credibility were not more persuasive over all levels of discrepancy: For intermediate levels of discrepancy, the messages from the moderately credible source were more persuasive.
More important, the figure and its analysis shed light on the functional form of the relationship between discrepancy and belief change. These results, however, have often been misstated because the statistical analysis of the data appears inconsistent with what is shown in the figure: The statistical analysis reveals that the relationship is linear for discrepancy induced by the high credibility source, whereas the relationship is curvilinear and nonmonotonic for the discrepancy induced by the moderately credible source. Although there were some significant findings regarding disparagement, it did not appear that disparagement acted simply as a functional alternative to belief change.
Figure 6.5 Bochner and Insko’s (1966) Findings Relating Discrepancy to Belief Change
A Mathematical Integration: Laroche (1977)
Laroche (1977) created a mathematical model that incorporated source credibility, message discrepancy, and ego involvement in predicting belief change.
Laroche’s model requires that all variables are transformed to be between 0 and 1. Laroche’s key equation is
where Dp is message discrepancy, P0 = 0, and yEq = belief position at equilibrium. The parameter γ is a function of source credibility and noninvolvement.
The critical feature of Laroche’s (1977) model is that, depending on the levels of credibility and noninvolvement, three different relations between discrepancy and belief position are possible: When γ > 1.00, there is a downturn in the graph (i.e., the relation is nonmonotonic); when 0 < γ < 1.00, the curve is monotonic, decelerating with a positive slope; when γ = 0, the relation is linear. In other words, we can think of γ as a dial that changes the shape of the curve relating discrepancy to belief change. Greater credibility and greater noninvolvement reduce γ, thereby making the relation approach linearity. Using the results of prior studies, Laroche found that “γ was generally higher for low-credibility sources than for high-credibility sources and higher for conditions with high involvement than for those with low involvement” (Chung, Fink, & Kaplowitz, 2008, pp. 161–162).
Additional Evidence From Static Models
Aggregating Prior Studies
To investigate the functional form of the relationship between discrepancy and belief position, we can look at the average slope for different levels of discrepancy. Kaplowitz and Fink (1997) did just that, using their own prior research. In the works of Fink et al. (1983), Kaplowitz, Fink, Armstrong, and Bauer (1986), and Kaplowitz and Fink (1991), the relationship between discrepancy and belief position was a curve with a positive slope that decelerated; in other words, as discrepancy went up belief change went up, but the rate of change decreased. However, these differences in slope were relatively small. Furthermore, after reviewing other investigations, Kaplowitz and Fink (1997) concluded that there was little evidence for nonmonotonicity or boomerang effects (see Kaplowitz & Fink, 1997, pp. 83–85; cf. McGuire, 1985, p. 276); they speculated that “strong supportive arguments may increase the effectiveness of an extremely discrepant message from a mildly credible source and thus inhibit nonmonotonicity” (p. 84).
Positional Versus Psychological Discrepancy
Laroche’s key equation (Equation 3, presented earlier) has discrepancy in two places, as a coefficient and as an exponentiated value: Fink and colleagues (1983) created a variant of this model that incorporates two different aspects of discrepancy: positional discrepancy, which is the same as what was previously referred to as discrepancy, and psychological discrepancy, which is the “level of discrepancy between two positions as experienced by an individual” (Fink et al., 1983, p. 415). Using our notation, the model’s key equation for the receipt of one message is:
P1 = [w0P0 + wA Δ (Ψ)PA]/[w0 + wA Δ (Ψ)],
where w0 is the weight of the initial position (P0), wA is the weight of the message position (PA), and Δ(Ψ) is a discounting function that reduces the weight of the message as psychological discrepancy increases (compare with Equation 5). More specifically,
Δ (Ψ) = e-τΨ,
where γ > 0 and Ψ is a positive monotonic transformation of measured psychological discrepancy. As psychological discrepancy increases, Δ (Ψ) decreases, which reduces the impact of the message.
Psychological discrepancy reflects positional discrepancy as well as context effects. For example, holding positional discrepancy constant, the message environment can change a message’s psychological discrepancy. Compare the following scenarios: (1) you are asked to give $100 to your alma mater; (2) you are asked to give $1,000 to your alma mater, and then, after a moment, you are asked to give $100. The request for $100 that follows the $1,000 request should seem less psychologically discrepant than the single $100 request, and therefore it should be more effective.
Fink et al. (1983), using what was termed the psychological-discrepancy-discounting model, showed that the greater the psychological discrepancy, the less the effectiveness of messages with the same positional discrepancy, and that psychological discrepancy can be manipulated by the combination of messages that are presented to the receiver. Not all the model’s predictions were supported, but many were. For example, it was found that the “psychological discrepancy of the moderate message is less when the extreme message precedes the moderate message than vice versa,” resulting in “greater attitude change in the former condition” (p. 429).
More critical to the present discussion, the psychological-discrepancy-discounting model allows for nonmonotonic effects, and this model was found to be statistically superior to the linear discrepancy model.
Evidence From Dynamic Models
The theories and studies previously reviewed have been used to assess questions of statics, which is the study of forces in equilibrium rather than the movement toward equilibrium. Another way to examine the process activated by discrepant messages is to examine the changes over time or the movement from one equilibrium to another, the dynamics (see Eiser, 1994, for a general statement on dynamics, and for material related to the dynamics discussed next, see Fink & Kaplowitz, 1993; Fink, Kaplowitz, & Hubbard, 2002; Kaplowitz & Fink, 1996).
Kaplowitz, Fink, and Bauer (1983)
Kaplowitz et al. (1983), using a between-participants design, manipulated discrepancy and the amount of time respondents had to consider a message. The topic was the health service fee at the respondents’ university, and the message advocated an increase in the fee. Participants, who were not in favor of a fee increase, were presented with a message that took no position on the fee, proposed a moderate increase in the fee, or proposed an extremely large increase in the fee.
Based on a dynamic model, the estimated equilibrium message effectiveness (i.e., change achieved divided by change advocated) was about 0 for the no position advocated message, whereas the moderately discrepant message was about 15% effective, and the extremely discrepant message was about 18% effective. Furthermore, it was estimated that the time to achieve 90% of the movement toward the equilibrium position was about 2 ¼ minutes. The trajectories were found to exhibit oscillation with a period of oscillation of about 13.5 seconds (s), and the amplitude of oscillation was greatest for the extremely discrepant message and much less and not that different for the no-position message and the moderately discrepant message.
The key findings of this study are that (1) greater discrepancy caused a greater change at equilibrium; (2) integrating a message takes time, and for these particular messages, it took well over two minutes; and (3) because of oscillations, the trajectories of messages with different discrepancies will likely cross prior to the achievement of equilibrium, which means that “if messages of different discrepancies cause oscillations which have different frequencies … one’s conclusion as to which message was most effective may be determined by the time interval from message to measurement” (Kaplowitz et al., 1983, p. 247). If we do not control the time between a discrepant message and the belief measurement, we may misidentify the message that caused the greatest change at equilibrium.
Chung and Colleagues (2008)
Chung et al. (2008) gathered data by having participants, who were university students, respond via a computer mouse, indicating their view approximately every 77 milliseconds (ms). Each participant responded to two topics: criminal sentencing, a topic relatively low in ego-involvement for university students, and a tuition increase, a topic relatively high in ego-involvement for them. For both topics, message discrepancy (with 3 levels) and source credibility (with 2 levels) were manipulated, and eleven time points for each individual, spread out over each individual’s trajectory, were analyzed. Participants were not limited in the time that they took to consider the message that they received. For the criminal-sentencing issue, participants took from 8.55 s to 146.92 s, with an average of 48.48 s; for the tuition issue, participants took from 3.93 s to 151.23 s, with an average of 48.46 s. The relationship of discrepancy and belief position was found to be monotonic and positive in three of the four combinations of topic by credibility; nonmon
otonicity was found only in the condition with the low-credibility source and the low-involvement message.
These data allowed the simultaneous testing of both Laroche’s model, which is static (i.e., it relates discrepancy to the equilibrium belief position), as well as the dynamic model from Kaplowitz et al. (1983; referred to in Chung et al., 2008, as the single-push with friction model). The dynamic single-push with friction model predicts that over-time (i.e., the x-axis is time, not discrepancy), belief trajectories are positive and decelerating, and that trajectories associated with different levels of source credibility and discrepancy do not cross each other. (The possibility of oscillation was not included in this model.)
An alternative model, a push-with-pullback model, was proposed to represent two different possible ways for counterarguing to have an effect: First, discrepancy is associated with subvocal counterarguments (Brock, 1967; see also Cook, 1969; Toy, 1982), which occur after the discrepant message has been considered. Second, the Spinozan procedure (Gilbert, Krull, & Malone, 1990) may suggest that the recipient of a discrepant message may first entertain it as true before rejecting it. Both these processes—the generation of counterarguments and the rejection of discrepant information—should take time. Therefore, if counterarguing or the Spinozan procedure apply, we should find that after being presented with a discrepant message, (1) the early part of the over-time belief trajectory should move in the direction advocated by the message, and (2) then, when counterarguments have been generated sufficiently, the movement of one’s belief should be in the opposite direction. Therefore, the push-with-pullback model proposes a nonmonotonic relationship between time and belief change, making the push-with-pullback model an alternative to the single-push with friction model.
The SAGE Handbook of Persuasion Page 19