Multiple Roles for Persuasion Variables
* * *
One important contribution of the ELM to the general understanding of persuasion is its emphasizing that a given variable might play different roles in persuasion under different conditions. From the perspective of the ELM, a variable might influence persuasion in three general ways. First, it might affect the degree of elaboration (and thus influence the degree to which central-route or peripheral-route processes are engaged). Second, it might serve as a peripheral cue (and so influence persuasive outcomes when peripheral-route persuasion is occurring). Third, it might influence the valence of elaboration (and so influence persuasive outcomes when central-route persuasion is occurring). (Some presentations of the ELM provide a more elaborated list of possible roles [e.g., Petty & Wegener, 1999, p. 51], but the three identified here will serve for present purposes [see O’Keefe, 2002, pp. 164–165].)
The question naturally arises, however, as to exactly when a given variable is likely to serve in one or another of these roles. The ELM offers a general rule of thumb for anticipating the likely function for a given variable, based on the overall likelihood of elaboration (Petty, Wegener, Fabrigar, Priester, & Cacioppo, 1993, p. 354). When elaboration likelihood is low, then if a variable affects attitude change, it most likely does so by serving as a peripheral cue. When elaboration likelihood is high, then any effects of a variable on attitude change are likely to come about through influencing elaboration valence. And when elaboration likelihood is moderate, then any effects of a variable on attitude change are likely to arise from affecting the degree of elaboration (as when some aspect of the persuasive situation suggests that closer scrutiny of the message will be worthwhile).
One might wonder about the degree to which this ELM rule of thumb is genuinely informative, because it amounts to little more than a restatement of the distinction between the two routes to persuasion. For instance, the proffered principle says in effect that “when elaboration is low, attitude change happens through peripheral processes and so anything that affects attitude change under such conditions does so by serving as a peripheral cue.” This might appear to verge on a tautology, in which by definition something that influences attitude change under conditions of low elaboration must be operating as a peripheral cue. The value of this rule of thumb thus turns on the degree to which one can independently assess whether peripheral or central processes are engaged, and such independent assessments are elusive (as acknowledged by Petty & Briñol, 2006, p. 217).
However, the ELM’s analysis does point to distinctive predictions about the different roles of a given variable, predictions derived from the operation of moderating variables. For example, if the physical attractiveness of a communicator in an advertisement is processed as a peripheral cue (and so activates a general liking heuristic), then the nature of the advertised product is unlikely to influence the cue’s effects. By contrast, if attractiveness influences elaboration valence because of being processed as an argument, then attractiveness’s effects should obtain for some products (namely, those for which attractiveness is a plausible argument, such as beauty products) but not for others (Petty & Briñol, 2006, p. 218). The implication is that by examining the observed effects of a moderator variable, one can distinguish whether a given property is activating a heuristic or influencing elaboration valence.
The larger point to be noticed is that the ELM draws attention to the mistake of thinking that a given variable can influence persuasive outcomes through only one pathway. For example, the credibility of the communicator might serve as a peripheral cue (and so activate a credibility-based heuristic)—but it could also influence the amount of elaboration, as when the communicator’s apparent expertise leads receivers to think that it will be worthwhile to pay closer attention to the message’s arguments. Recognizing this complexity of persuasion processes represents an especially important contribution of the ELM.
Future Research
* * *
As fruitful as the ELM has been as a framework for stimulating research, at least three areas of research deserve future attention: the nature of argument quality, the nature of involvement, and the relationship of central and peripheral processes.
The Nature of Argument Quality
In ELM research, the nature of argument quality (or argument strength) has not been a focus of explicit attention, because argument quality variations have been defined in terms of persuasive effects. That is, a high-quality argument is one that, in pretesting, is relatively more persuasive (compared to a low-quality argument) under conditions of high elaboration.
This way of defining argument quality reflects the role that argument quality has played in ELM research designs. In ELM research, argument quality variations have been used “primarily as a methodological tool to examine whether some other variable increases or decreases message scrutiny, not to examine the determinants of argument cogency per se” (Petty & Wegener, 1998, p. 352). The idea is that if message receivers are sensitive to argument quality variations (as displayed by their being more persuaded by high-quality arguments than by low-quality arguments), then those receivers must have been engaged in close message processing (relatively high elaboration). For example, in Petty, Cacioppo, and Goldman’s (1981) classic study discussed earlier, argument quality variations affected persuasive outcomes under conditions of high involvement but not under conditions of low involvement; the inference to be drawn is that under conditions of higher involvement, audiences were more closely processing the message and so were more attentive to argument quality variations.
But a thorough understanding of persuasion processes requires some analysis of the nature of these argument quality variations. As a way of seeing the importance of this matter, consider: What advice would the ELM offer to a persuader presenting a counterattitudinal message to an audience likely to engage in a great deal of elaboration? Presumably the advice would be “use high-quality arguments.” But because argument quality has been defined in terms of effects (a high-quality argument is one that persuades under conditions of high elaboration), this advice amounts to saying “to be persuasive under conditions of high elaboration, use arguments that will be persuasive”—which is obviously unhelpful (for some elaboration of this line of reasoning, see O’Keefe, 2002, 2003). And, unfortunately, the experimental messages used in ELM experiments appear to have confounded a great many different appeal variations, making it challenging to identify just which features might have been responsible for the observed effects.
However, research has identified the active ingredient in ELM messages as a variation in the perceived desirability of the outcomes associated with the advocated view (Areni & Lutz, 1988; Hustinx, van Enschot, & Hoeken, 2007; van Enschot-van Dijk, Hustinx, & Hoeken, 2003; see also Johnson, Smith-McLallen, Killeya, & Levin, 2004). So (for example) when receiver involvement is low, the persuasiveness of a message is relatively unaffected by variation in the desirability of the outcomes, whereas when involvement is high, persuasive success is significantly influenced by whether the outcomes are thought to be highly desirable or only slightly desirable. That is, under conditions of high elaboration, receivers are led to have more positive thoughts about the advocated view when the message’s arguments indicate that the advocated view will have outcomes that the receivers think are relatively desirable than they do when the arguments point to outcomes that are not so desirable—but this difference is muted under conditions of low elaboration.
That outcome desirability should turn out to be a key determinant of the persuasiveness of arguments under conditions of high involvement—that is, direct personal relevance—is perhaps not entirely surprising. When the outcomes affect the message recipient directly, the desirability of the outcomes becomes especially important.
The open question is whether other message variations might function in a way similar to outcome desirability. That is, are there other quality-related features of persuasive appeals that function as
outcome desirability does—features whose variation makes relatively little difference to persuasive outcomes under conditions of low elaboration, but whose variation makes a more substantial difference under conditions of high elaboration?
One candidate that naturally comes to mind is outcome likelihood. A general expectancy-value conception of attitudes would suggest that attitudes are a joint function of evaluative judgments (how desirable the attitude object’s characteristics are seen to be) and likelihood judgments (the likelihood with which those characteristics are associated with the object). Correspondingly, one might expect that messages varying in the depicted likelihood of outcomes might have effects parallel to those of messages varying in the depicted desirability of outcomes: variation in outcome likelihood might make a greater difference to persuasiveness under conditions of high elaboration than under conditions of low elaboration.
There is not much direct evidence concerning whether the effects of outcome-likelihood variations are moderated in this way by involvement. However, the general research evidence concerning the persuasive effects of outcome-likelihood variation is not very encouraging. Some research finds that outcome-likelihood variations have persuasive effects akin to those of outcome-desirability variations (e.g., Witte & Allen, 2000), but other studies have found very different patterns of effects (e.g., Johnson et al., 2004; Smith-McLallen, 2005). Perhaps only under yet-to-be-discovered conditions do variations in outcome likelihood function in ways akin to outcome desirability.
In any case, the general question remains open: There may be additional quality-related message characteristics—beyond outcome desirability—that enhance message persuasiveness under conditions of high elaboration. Identification of such message properties would represent an important advance in the understanding of persuasion generally and argument quality specifically.
The Nature of Involvement
In persuasion research, the concept of “involvement” has been used by a variety of theoretical frameworks to describe variations in the relationship that receivers have to the message topic. The most notable historical example is social judgment theory’s use of the concept of “ego-involvement” (Sherif, Sherif, & Nebergall, 1965). The ELM has extended this tradition in its emphasis on the role of involvement as an influence on elaboration likelihood. But various commentators have suggested distinguishing different kinds of “involvement,” on the grounds that different varieties of involvement have different effects on persuasion processes.
For example, Johnson and Eagly (1989) distinguished value-relevant involvement (in which abstract values are engaged) and outcome- relevant involvement (in which concrete short-term outcomes or goals are involved). Their meta-analytic evidence suggested that value-relevant involvement leads receivers to defend their opinions when exposed to counterattitudinal messages, regardless of whether the message contains strong or weak arguments. By contrast, outcome-relevant involvement produces the pattern of effects expected by the ELM, in which variations in argument strength produce corresponding variations in persuasive effects. Johnson and Eagly’s argument thus is that the ELM describes the role that one kind of involvement plays in persuasion, namely, outcome-relevant involvement, but does not capture the effects of variations in value-relevant involvement. Petty and Cacioppo (1990), however, have argued that the same process might underlie these apparently divergent patterns of effect (for some further discussion, see Johnson & Eagly, 1990; Levin, Nichols, & Johnson, 2000; Petty & Cacioppo, 1990; Petty, Cacioppo, & Haugtvedt, 1992; see also Park, Levine, Westermann, Orfgen, & Foregger, 2007).
As another example, Slater (2002) has approached the task of clarifying involvement’s role in persuasion not by starting with different kinds of “involvement” but by starting with different kinds of message processing—and then working backward to consider how different kinds of involvement (and other factors) might influence the different sorts of processing. Slater’s analysis includes such processing varieties as “outcome-based processing” (motivated by the goal of self-interest assessment), “value-affirmative processing” (motivated by the goal of value reinforcement), and “hedonic processing” (motivated by the goal of entertainment)—with these influenced by, respectively, outcome relevance (akin to “outcome-relevant involvement”), value centrality (akin to “value-relevant involvement”), and narrative interest. Slater (2002, p. 179) thus argues that “simply distinguishing value-relevant involvement from the issue- or outcome-relevant involvement manipulated in ELM research does not go far enough.”
The larger point is that involvement (simpliciter) is a concept that is insufficiently well-articulated to do the work asked of it. Although a broad distinction between value-relevant and outcome-relevant involvement has merit, further conceptual and empirical work is surely to be welcomed.
The Relationship of Central and Peripheral Processes
One prominent alternative to the ELM has been Kruglanski and Thompson’s (1999b) “unimodel” of persuasion, which suggests that the ELM’s two routes to persuasion are in fact not fundamentally different: In each route, receivers try to reach conclusions about what to believe, using whatever evidence is available. In the two persuasion routes, different kinds of evidence are employed (peripheral cues in the peripheral route, message arguments in the central route), but from the point of view of the unimodel, these are not actually two fundamentally different processes. In each route there is a process of reasoning to conclusions based on evidence, and thus a unitary picture—a “unimodel”—will suffice.
This underlying similarity, it is argued, has been obscured in ELM research by virtue of a confounding of the contrast between cue and arguments and the contrast between simple and complex inputs. From the unimodel point of view, both peripheral cues and message arguments can vary in their complexity (ease of processing, brevity, etc.). The argument is that in ELM research, peripheral cues have typically been quite simple and arguments have typically been quite complex. This produces differences in how these two inputs are processed, but (the suggestion is) if cues and arguments are made equally complex, then they will be seen to be processed identically and produce identical effects. This unity of underlying processing thus is taken to undermine the ELM’s distinction between the two persuasion processes.
As an illustration of research supporting the unimodel’s view, Kruglanski and Thompson (1999b, Study 1) found that when communicator expertise information was relatively lengthy, expertise influenced the attitudes of receivers for whom the topic was personally relevant but not the attitudes of receivers for whom the topic was not relevant. That is, topic relevance and expertise interacted in exactly the same way as topic relevance and argument quality did in earlier ELM studies. The apparent implication is that peripheral cues (such as expertise) and message arguments function identically in persuasion, once the level of complexity of each is equalized. (For some presentations of the unimodel and related research, see Erb & Kruglanski, 2005; Erb, Pierro, Mannetti, Spiegel, & Kruglanski, 2007; Kruglanski, Chen, Pierro, Mannetti, Erb, & Spiegel, 2006; Kruglanski, Erb, Pierro, Mannetti, & Chun, 2006; Kruglanski & Thompson, 1999a, 1999b.)
The unimodel raises both empirical and conceptual issues concerning the ELM, and these issues are sufficiently complicated that it will take some time to sort them out. (For some discussion of these and related issues, see, e.g., Chaiken, Duckworth, & Darke, 1999; Petty & Briñol, 2006; Petty, Wheeler, & Bizer, 1999; Wegener & Claypool, 1999.) Empirically, it is not yet clear exactly when (or, indeed, whether) the ELM and the unimodel make genuinely different predictions. That is, there is at present some uncertainty about just what sort of empirical findings will represent confirmation of one view and disconfirmation of the other. Consider, for instance, the just-mentioned finding indicating that complex information about source expertise had more influence on persuasive outcomes when the topic was personally relevant to receivers than when it was not (Kruglanski & Thompson, 1999b, Study 1). From a unimodel perspective, this is tak
en to be inconsistent with the ELM, because the ELM is assumed to expect that source cues will have a smaller influence on persuasion as topic relevance increases. But—bearing in mind that a given variable might affect persuasion through various pathways—the ELM might explain this result in several ways, including the possibility that expertise information was processed as an argument or provoked elaboration of self-generated (as opposed to message) arguments (Petty, Wheeler, & Bizer, 1999, pp. 159–160). The general point is that it is not yet clear whether (or exactly how) the ELM and the unimodel can be made to offer contrasting empirical predictions.
Conceptually, the unimodel points to some aspects of the ELM that are not clear. Consider, for example, the question of whether it is true by definition that peripheral cues are easy to process. If part of the very concept of a peripheral cue is that it is easy to process, then it does not make sense to speak of there being any “confounding” of cues and simplicity—and so the unimodel’s suggestion that there might be complex cues is conceptually malformed. On the other hand, if peripheral cues are not by definition easy to process, then it becomes more plausible to explore, as the unimodel suggests, the effects of hard-to-process peripheral cues.
In sum, the unimodel has raised valuable issues concerning the ELM. Continuing attention to these issues offers the promise of better-articulated conceptual frameworks and more finely tuned empirical predictions.
The ELM: A Model of Attitude Change, Not Persuasion
* * *
The ELM can be placed in a broader context by noticing that it is better described as a theory of attitude change than as a theory of persuasion. To be sure, influencing attitudes is often an important aspect of persuasion. Attitudes influence what products people buy, what policies they prefer, what candidates they favor—and so persuaders often have the goal of ensuring that people have the desired attitudes.
The SAGE Handbook of Persuasion Page 29