Assessing the Impact of Message Cues and Arguments in Persuasion: Conceptual and Methodological Issues

Charles S. Areni, Texas Tech University
K. Chris Cox, Nicholls State University
ABSTRACT - Postulates of the Elaboration Likelihood Model (ELM), the HeuristicBSystematic Model (HSM), and related models of persuasion in the marketing literature are compared and contrasted. Several recommendations are offered regarding directions for future research efforts. In general, these suggestions fall into the following categories: (1) propositions for future research efforts, (2) recommendations to enhance the quality of the tests of various aspects of the ELM and HSM, and (3) suggestions regarding more complete and practical conceptualizations of key constructs in each model.
[ to cite ]:
Charles S. Areni and K. Chris Cox (1995) ,"Assessing the Impact of Message Cues and Arguments in Persuasion: Conceptual and Methodological Issues", in NA - Advances in Consumer Research Volume 22, eds. Frank R. Kardes and Mita Sujan, Provo, UT : Association for Consumer Research, Pages: 198-202.

Advances in Consumer Research Volume 22, 1995      Pages 198-202

ASSESSING THE IMPACT OF MESSAGE CUES AND ARGUMENTS IN PERSUASION: CONCEPTUAL AND METHODOLOGICAL ISSUES

Charles S. Areni, Texas Tech University

K. Chris Cox, Nicholls State University

ABSTRACT -

Postulates of the Elaboration Likelihood Model (ELM), the HeuristicBSystematic Model (HSM), and related models of persuasion in the marketing literature are compared and contrasted. Several recommendations are offered regarding directions for future research efforts. In general, these suggestions fall into the following categories: (1) propositions for future research efforts, (2) recommendations to enhance the quality of the tests of various aspects of the ELM and HSM, and (3) suggestions regarding more complete and practical conceptualizations of key constructs in each model.

Petty and Cacioppo's (1981a, 1986a) elaboration likelihood model (ELM) and Chaiken's (1980, Eagly & Chaiken, 1993) heuristicBsystematic model (HSM) of persuasion have become quite influential in the marketing literature, due largely to their implications regarding the persuasive impact of various message elements commonly employed in advertising (cf. Petty, Cacioppo, & Schumann, 1983; MacKenzie, Lutz, & Belch, 1986; Munch & Swasy, 1988; Miniard et al. 1991).

Although the ELM and HSM initially posited relatively simple and distinct persuasive effects for the arguments and cues in a communication, as empirical research accumulated, the proposed relationships increased in complexity. Adding to this intricacy was the discovery that the impact of a given communication variable depended upon: (1) the mode in which it was presented (Chaiken & Eagly, 1983), (2) its salience within the communication (Pallak, 1983), (3) the symbolic form in which it was presented (Yalch & ElmoreBYalch, 1984), and most importantly, (4) the manner in which it was processed (Petty & Cacioppo, 1981b; Miniard et al. 1991). Finally, streams of research have emerged in the marketing literature that suggest even more complexity in the relationships among visually and verbally presented elements (MacKenzie & Spreng, 1992; Miniard, Bhatla, & Rose, 1990).

Despite these conceptual advances, however, the basic method of examining the ELM and HSM has evolved much more slowly. The diagnosticity of tests of the more intricate relationships in the model has been questioned (Stiff & Boster, 1987; Allen & Reynolds, 1993; Mongeau & Stiff, 1993). The following discussion addresses a number of conceptual and methodological issues relating to theories of persuasion in general, and the ELM and the HSM in particular. The issues discussed can be categorized as: (1) propositions for future research efforts,(2) recommendations to enhance the quality of the tests of various aspects of the ELM and HSM, and (3) suggestions regarding more complete and practical conceptualizations of key constructs in each model.

ARGUMENT-VERSUS CUE-BASED PERSUASION

Although numerous communication variables have been examined within the ELM, the HSM, and related streams of persuasion research, researchers have continually drawn the distinction between those that comprise the fundamental message and those that create a setting for the message (Chaiken, 1987; Petty & Cacioppo, 1986; Lutz, 1985). In the ELM, the former are termed central arguments and the latter peripheral cues, whereas in the HSM they are referred to as persuasive arguments and heuristic cues, respectively. For purposes of simplicity and clarity, we adopt the term argument to refer to a communication element that constitutes part of the fundamental message, and cues to refer to a communication element that creates a setting for the message. [It should be emphasized that the terms cue and argument do not refer to an element of the communication per se, but rather, to the audience member's interpretation of that communication element.]

Both the ELM and the HSM focus on the amount of thought the audience devotes to a communication as a primary moderator of the extent to which cues or arguments produce the strongest direct attitudinal effects. Specifically, they imply that when the amount of thought the audience devotes to the communication is high, consideration of arguments drives persuasion, via the central route and systematic processing, respectively. On the other hand, attitude shifts are directed by the consideration of cues when the amount of thought devoted to the communication is lower. In a typical test of this basic proposition, the valence of various communication elements and the amount of thought the audience devotes to the communication are manipulated. Those components that produce stronger attitudinal effects when an audience devotes little thought to the communication are termed peripheral or heuristic cues, whereas those that produce more pronounced effects when the audience devotes a great deal of thought to the communication are referred to as central or persuasive arguments.

HIERARCHICAL VERSUS ADDITIVE PERSUASION PROCESSES

The results of such tests imply a hierarchical relationship wherein the persuasive impact of cues are eclipsed by argumentBbased persuasion when the audience devotes more thought to the communication. Importantly, however, there are some differences in the status of the hierarchical persuasion postulate in the ELM, the HSM, and other theories of persuasion. For purposes of clarity, we will discuss these differences in terms of an equation introduced by MacKenzie and Spreng (1992). Their equation, however, is modified to reflect the terminology used in this discussion. As discussed above, the ELM and HSM imply:

(EQ 1)Attitude = b0 + b1ARG + b2CUE + E

where ARG is argumentBrelated thought and CUE is cueBrelated thought. The hierarchical persuasion postulate implies that when the audience devotes a great amount of thought to a persuasive communication, the product b2CUE approaches zero. However, some researchers have suggested that the ELM is unclear as to whether the coefficient, b2, the variable, CUE, or both converge to zero (Stiff, 1986; Stiff & Boster, 1987). In other words, when an audience devotes a great deal of thought to a persuasive communication, does the processing of cues diminish (CUE _ 0), are cues discounted when the attitude is formed or changed (b2 _ 0), or, do both occur?

Due to this ostensible ambiguity, Stiff (1986) advanced Kahnemann's (1973) elastic capacity model (ECM) as an alternative to the ELM. He suggested that, unlike the ELM, the ECM suggests that CUE approaches zero when an audience allocates a great deal of thought, or cognitive capacity, to a persuasive communication. Though this interpretation of the ECM offers some interesting possibilities, it is important to note that the model was intended to account for attention and encoding processes, not persuasion processes. There is some question as to whether it can be adapted to the latter context (Petty et al. 1987).

The ELM does not preclude that the encoding of message cues and arguments can coincide (Petty et al. 1987), despite Stiff's claims that it takes a singleBchannel, "eitherBor" approach to information processing. Rather, the ELM posits that when an audience allocates a great deal of thought to a persuasive communication, attitude change tends to be determined by the consideration of arguments rather than cues BB b2 approaches zero (Petty et al. 1993). MacKenzie and Spreng (1992) tested this postulate against the rival explanation that cueBrelated persuasion wanes at higher levels of elaboration likelihood because cueB related thought decreases (CUE approaches zero). In general, MacKenzie and Spreng's results were consistent with the ELM's strength of relationship predictions, but not the mean level predictions.

The HSM not only allows the encoding of message cues and arguments to coincide, but also posits that if the most salient cues are consistent with message arguments in terms of their evaluative implications, then the consideration of both will drive persuasion when an audience commits a great deal of thought to the communication. However, when cues and arguments are in conflict with respect to evaluative implications, the hierarchical persuasion postulate holds, and argumentBbased persuasion attenuates the impact of cueBbased persuasion. Thus, rather than strictly adopting the hierarchical persuasion postulate, the HSM posits that argumentB and cueBbased persuasion can produce additive effects under certain conditions (i.e., when they are evaluatively congruent). If we let VALcue represent the evaluative implications of cueBrelated thought and VALarg represent the evaluative implications of argumentB related thought, then the HSM posits that when an audience devotes a substantial amount of thought to a persuasive communication: [It is important to note that the HSM conceptualizes heuristic and systematic processing as the endpoints of a persuasion continuum rather than as a strict dichotomy. The equations presented above, thus, represent extreme conditions where heuristic (i.e., cue-based) persuasion is either completely additve or completely attenuated.]

(EQ 2) Attitude = b0 + b1ARG + b2CUE +

when cor(VALcue, VALarg) is positive

(EQ 3) Attitude = b0 + b1ARG +

when cor(VALcue, VALarg) is negative

The evidence for the additivity hypothesis is still emerging. However, it is difficult to assess in experiments that cross two levels of a cue with two levels of argument quality, because there is no "baseline" from which to assess each effect. If we were to find that when subjects devote a great deal of thought to a persuasive communication, the credible source/strong argument version of the communication results in more positive attitudes than the version in which subjects are exposed to the unreliable source/strong argument, we could be observing attenuation in the unreliable source/strong argument treatment and additivity in the credible source/strong argument treatment, as the HSM predicts. But, the difference might also be due to attenuation in the credible/strong condition and additivity in the unreliable/strong condition, or to additive effects in each condition.

In fact, most of the studies employing the aforementioned two level, fully crossed experimental design have found little or no persuasive influence of cues when an audience allocates a great deal of thought to a persuasive communication, suggesting that complete attenuation occurred. Maheswaran and Chaiken (1991) observed that when consistent in terms of valence, both cueB and argumentBrelated thought are correlated with attitudes, but when they are inconsistent, only the argumentBrelated thought is correlated with attitudes. However, they point to the need for a "message only" condition to provide a baseline for identifying additivity effects.

At least two variables would appear to moderate the likelihood of observing additive cueBand argumentBbased persuasion effects. Chaiken, Axsom, Liberman, & Wilson (1992) have discussed individual predispositions to use certain heuristics (i.e., experts are always right, length equals strength, etc.) as individual difference variables and as the results of learning via repeated use. "Chronically" accessible heuristics are likely to be associated with high levels of perceived reliability. Additivity would, therefore, appear to be more likely to occur when the communication contains a salient cue corresponding to a chronically accessible heuristic in memory.

The second moderating variable reflects an underresearched topic in the ELM and HSM programs of research, the temporal sequence of the presentation of cues and arguments. In tests of the ELM and HSM in advertising settings, the presentation of cues and arguments generally coincide; that is, both are presented as elements in a single communication. In some of the studies reporting additive cueBbased persuasion, however, cues and arguments were presented in two distinct communications with the former preceding the latter (Maheswaran & Chaiken 1991). In the absence of more diagnostic information, the presentation of a simple cue may induce audience's to form an initial evaluation, even if motivation and ability to think about the focal topic are relatively high (Srull & Wyer 1989). The HSM suggests exactly this possibility. Chaiken et al. (1989) note that if highly motivated audience members cannot engage in systematic processing, then they will focus on a salient cue in the persuasion context and settle for an "insufficient" level of confidence in the judgment. An initial impression, once formed, might be expected to persist, even in the presence of subsequently presented communications containing conflicting information (Srull & Brand, 1983). Thus, prior presentation of a simple cue would appear to increase the likelihood of observing additive persuasion effects.

INDEPENDENCE VERSUS INTERDEPENDENCE

A second assumption implicit in the hierarchical persuasion postulate is that argumentB and cueBbased persuasion processes operate relatively independently in influencing attitude change. However, both the ELM and the HSM accommodate the notion that the processing of message cues can bias the processing of message arguments so as to make the latter consistent with the former in terms of valence.

Petty and Cacioppo (1986a) have described a number of variables that induce biased elaboration on a persuasive communication. They have found that: (1) preBexisting knowledge structures, (2) forewarnings of message content, (3) forewarnings regarding the persuasive intent of a message, (4) bogus personality feedback, (5) high levels of message repetition, and (6) audience expressions of approval or disapproval serve to bias the valence of argumentBrelated thought and self generated argumentation.

The HSM has focused more on the biasing effects of cues that are elements of the persuasive communication itself. Chaiken and Maheswaran (1990) have examined the biasing impact of message cues on an audience's processing of message arguments. They manipulated the credibility of the source of a communication regarding a telephone answering machine, the ambiguity of the product claims in the communication, and subjects' motivation to process the communication. They observed that even when subjects' motivation was high, and product claims were ambiguous, the credibility of the source influenced their attitudes. Since they measured subjects' perceptions of source credibility and claim validity, they were able to demonstrate that the former influenced brand attitude indirectly via its biasing effect on the latter.

Several variables appear to moderate the extent to which the consideration of message cues biases the processing of message arguments. First, a cueBargument biasing effect is more likely when the arguments in the message are ambiguous with respect to the merits of the advocated position. Second, Lutz (1985) has noted that in an advertising pretest setting, subjects typically view an ad for a novel product, and are then asked to evaluate the product. Having little information on which to evaluate the unfamiliar product, an individual's thoughts about the product (argumentBbased thinking) are more likely to be biased by his or her assessment of the ad itself (cueBbased thinking). MacKenzie et al. (1986) have provided evidence in support of Lutz's contention. Finally, it is important to note that the ambiguous arguments in Maheswaran and Chaiken's experiment were attributed to a (credible or unreliable) source that served as the most salient cue. This may be an important condition for observing a biasing effect. When the arguments in a message are not likely to be attributed to the most salient source, such as when celebrity endorsers appear in advertisements, or when there is no inherent relationship between the arguments and cues in a message, the biasing effect may be less likely to occur.

INFLUENCING THE ALLOCATION OF THOUGHT

The amount of thought devoted to a communication was initially conceptualized as an audience predisposition that was independent of the specific elements of the communication itself. However, recent evidence suggests that certain cues influence the amount of thought given to a persuasive communication by influencing an audience's motivation, ability, or opportunity process it (cf. Heesacker et al., 1983; Yalch & ElmoreBYalch, 1984; Swasy & Munch, 1985; Pechmann & Stewart, 1990). Petty and Cacioppo (1986a) have suggested that these effects tend to occur when the personal relevance of a communication is ambiguous and the audience is predisposed to allocate a moderate level of thought to it.

The HSM provides an interesting theoretical account regarding the conditions under which a communication element will enhance or decrease an audience's motivation to think about a persuasive communication. It posits that an audience member has a sufficiency threshold, or desired level of confidence, for every judgment task s/he faces. Thus, consideration of information relevant to a given judgment occurs until the audience's perceived level of confidence equals or exceeds their desired level. As an audience's motivation and ability to think about the focal topic increases, so too does its desired level of confidence.

According to the sufficiency principle, a communication variable or variables can enhance the amount of thought given to a persuasive communication by either increasing the audience's desired level of confidence, or by undermining the audience's perceived confidence. A message presented by a prestigious, expert source increases the amount of thought devoted to a persuasive communication because it increases the audience's desired level of confidence. In other words, the appearance of the prestigious source suggests to the audience that the communication must be worthy of a great deal of thought. Conversely, a communication containing evaluatively inconsistent information undermines the audience's perceived confidence, making further consideration of relevant information necessary to attain a sufficient level of certainty (Maheswaran & Chaiken, 1991).

MacInnis, Moorman, and Jaworski (1991) have recently concluded that much more research is needed in the marketing literature regarding factors that enhance the amount of thought devoted to persuasive communications. Their framework identifies a number of communication elements that may influence an audience's ability, motivation, or opportunity to think about the information presented in an advertisement. The research propositions they offer represent potentially fruitful directions for future research efforts.

COMMUNICATION VARIABLES AND PERSUASION PROCESSES

Multiple Roles for Communication Variables

As empirical investigations of the HSM, and the ELM in particular, accumulated, researchers discovered that a given communication variable could take on multiple roles in the persuasion process depending on the audience's level of processing intensity. Investigators found, for example, that some characteristics of the source of a communication influence persuasion under conditions of high processing intensity as arguments rather than cues (Petty & Cacioppo, 1980; Kahle & Homer, 1985). Researchers also discovered that under conditions of low processing intensity, certain message characteristics persuade audiences as simple cues rather than arguments (Alba & Marmorstein, 1987). Further, several studies indicated that both source (Heesacker et al. 1983) and message (Swasy & Munch, 1985; Munch & Swasy, 1988) elements influenced the amount of thought devoted to persuasive communications; and, finally, source variables were found serve as biasing influences on an audience's processing of relevant arguments (Chaiken & Maheswaran, 1990).

The complexity that such findings add to theories of persuasion is obvious. A single variable, source credibility, has been shown to operate in three of the four roles described above: as a simple cue (Cacioppo et al., 1983), as a biasing influence (Chaiken & Maheswaran, 1990), or to enhance the amount of thought devoted to a persuasive communication (Heesacker et al., 1983). The discovery that specific communication variables can operate as cues in some situations and arguments in others suggests at least three directions for future research efforts. The first concerns the specification of the conditions under which a given variable will operate in a particular role. The second concerns the development of more rigorous conceptual definitions of some of the core constructs of the ELM and HSM. The third stems from methodological refinements of the standard ELM experiment. Each of these issues is discussed below.

Specifying Conditions Under Which a Variable Operates in a Given Role

Petty and Cacioppo (1986a) have suggested that communication variables operate: (1) as simple cues when an audience's elaboration is low, (2) as enhancers/inhibitors of issueBrelevant thinking when an audience's elaboration is moderate, and (3) as arguments or biasing influences when an audience's elaboration is high. Although a great deal of evidence has accumulated establishing the first and last of these relationships, there is still some question as to whether the enhancer/inhibitor role requires moderate levels of elaboration. Many of the investigations reporting such effects have suggested that they occur under conditions of "low" elaboration (Swasy & Munch, 1985; Maheswaran & Chaiken, 1991). Of course, such a determination is difficult to make because examinations of the ELM typically manipulate elaboration at two levels for purposes of creating "high" and "low" conditions; it is difficult to determine where each condition falls along the elaboration continuum across studies. Perhaps many of the "low" elaboration conditions reported above actually fell closer to the midpoint of the continuum. Regardless, a more rigorous test of the points above would entail a manipulation that more clearly establishes distinct low, moderate, and high levels of elaboration.

Defining the Key Constructs

Argument Quality: Within the ELM, Petty and Cacioppo have adopted an empirical definition of argument quality. "Strong" arguments are those that elicit predominantly favorable cognitive responses, and "weak" arguments generate mostly unfavorable cognitive responses (Petty & Cacioppo, 1986a,b). Other researchers adopting empirically driven definitions of argument quality have focused on subjective perceptions of the arguments in a communication (Swasy & Munch, 1985), and changes in belief structures relevant to the principle assertions of the communication (Fishbein & Ajzen, 1975; Areni & Lutz, 1988; Eagly, 1991).

Defining strong and weak in terms of the cognitive responses they generate in pretest experiments constrains the utility of the ELM for at least two reasons. First, if strong (weak) arguments are "defined" as those generating primarily supporting (counter) arguments in a pretest, then the pretest has shown the arguments to operate via the central route by definition. Mongeau and Stiff (1993) have even suggested that this aspect of argument quality manipulations within the ELM render the model impossible to falsify. That is not the position adopted here. There is no reason why the empirical definition of argument quality necessitates the interaction between argument quality and elaboration demonstrated in so many tests of the ELM. However, the empirical definition does seem to guarantee that if elaboration likelihood is sufficiently high, then argument quality must persuade the audience via the central route.

This reasoning suggests a second major limitation of the empirical definition of argument quality. Tests of the ELM do not offer insights as to why a given argument is persuasive to an audience (Areni & Lutz, 1988). Petty and Cacioppo (1986b) have discussed need for a stronger conceptualization of the argument quality construct for exactly this reason. Research from a number of apparently unrelated disciplines suggests directions for the development of such a conceptual definition.

Perhaps the most obvious basis for conceptualizing argument quality is the area of formal logic. Indeed, Petty and Cacioppo described many of their strong arguments as being "logical." Much of the research of in this area has focused on the mental processes used in assessing the validity of the conclusions in logical syllogisms. Among the findings that have emerged are that individuals are more likely to accept conclusions associated with desirable outcomes and reject conclusions associated with undesirable outcomes (McGuire & McGuire, 1991), and that individuals weigh objective correctness more heavily than logical validity when evaluating the truth value of conclusions (Braine & Rumain, 1983).

Unfortunately, the "formality" of formal logic makes it difficult to apply to the "natural language" of persuasive communications (Braine & Rumain, 1983). An interesting stream of research has emerged which attempts to link the rules of formal logic to arguments presented in natural language (Braine, 1980; Braine & Rumaine, 1983). While some research examining the logic of natural language arguments has emerged in psychology (Harris & Monaco, 1978) and marketing (Gardner, 1975; Burke et al., 1988), a more systematic incorporation of Braine's work into the literature on the ELM, HSM and other persuasion research is needed.

Research regarding the structural features of argument quality has focused on the strength of the supporting evidence for the principle assertions of a communication. Many models have proposed that persuasive arguments consistent solely of principle assertions and supporting arguments (Fishbein & Ajzen, 1975; Jaccard, 1981). However, more sophisticated models include additional elements. Boller, Swasy, and Munch (1990) have developed an interesting perspective based on Toulmin's (1958) model of argument structure. For purposes of understanding persuasive communications in an advertising setting, Boller et al. condense Toulmin's six argument elements into four dimensions: (a) claim assertions, (b) evidence (grounds), (c) authority (warrants and backing), and (d) probability (qualifiers and rebuttals). Boller et al. (1990) view persuasive arguments as consisting of various combinations of these four classes of statements.

A number of researchers have examined the impact of each of these components in persuasion, and have found that explicit statements of claim assertions are more effective when an audience is likely to devote little thought to a persuasive communication, whereas implicit statements are more effective for an audience willing to think more extensively (Kardes, 1988; Sawyer & Howard, 1991). The impact of supporting evidence and probability in terms of the ELM and HSM is less clear, although it has been suggested that the former operates via central or systematic processing (Areni & Lutz, 1988). In addition, the linguistics literature suggests that females may be more sensitive to the use of probability statements (Lakoff, 1975), and prefer supporting evidence based on the personal experiences of others rather than aggregate data (Tannen, 1990).

Elaboration Likelihood: Petty and Cacioppo's reference to persuasion's central route as entailing the diligent consideration of issue or object relevant information suggests that elaboration has two basic aspects: a level of intensity, which refers to the amount of cognitive capacity devoted to an information processing task, and direction, which refers to the nature of the information needed to attain specific processing objectives (Kahnemann, 1973).

The distinction between the intensity and directional components is central to the evaluation of empirical research because actual manipulations of elaboration have varied considerably regarding their impact on each dimension, and, as Wright (1980) notes, researchers have been slow to acknowledge the ramifications of these differences in the interpretation of persuasion outcomes. For example, a manipulation that requires one group of subjects to evaluate an advertisement for a fictitious new product (i.e., low brand elaboration) while a second group is asked to evaluate the featured product in the ad (i.e., high brand elaboration) would seem to be fundamentally different from a manipulation in which all subjects are given the latter task and the experimenters simply alter the economic incentive for completing the task successfully. In terms of the discussion above, the first manipulation primarily alters the direction component, whereas the second focuses predominantly on the intensity dimension. This distinction may be critical for evaluating the mean difference versus the strength of relationship hypotheses proposed by MacKenzie and Spreng (1992). A manipulation focusing on the direction component would almost certainly produce mean differences in the amount of thought given to arguments versus cues. However, the strength of relationship hypotheses may be more applicable for studies manipulating the intensity dimension of elaboration likelihood without influencing subjects' processing goals.

Explicitly Modelling the Constructs

Many of the investigations of the ELM, HSM, and related streams of research have relied on inferring the process by which various cues and arguments influence persuasion by observing the effect of experimental manipulations on subjects' attitudes. However, the review presented above suggests numerous processes by which a given communication variable can operate in the persuasion process. When two or more of these processes produce similar outcomes, the approach of inferring process from outcome may not allow a researcher to discriminate among them.

Consider an apparently straightforward experiment in which a researcher manipulates the physical attractiveness of the source of a communication, the valence of its arguments, and the audience's motivation to process its content. Suppose, as expected, only the source attractiveness manipulation influences subjects' attitudes when motivation is low. But, similar to Petty and Cacioppo (1981b), both the source attractiveness and the argument valence manipulation influence subjects' attitudes when motivation is high. Have we observed, in this instance, that the source manipulation operated as an argument rather than a cue in the high motivation condition as Petty and Cacioppo (1986a) have suggested? Or, have we produced evidence for Maheswaran and Chaiken's (1991) notion of additivity between argumentB and cueBbased persuasion? Even the inclusion of a "message only" control condition in the design would not eliminate this ambiguity (cf. Miniard et al. (1991).

Petty, Unnava, and Strathman (1991, p. 247) note that: "in most communication settings, a confluence of factors determines the nature of information processing rather than one variable acting in isolation. When multiple variables are involved, interaction effects are possible." It is obvious that communications in the "real world," particularly advertisements, are far more complex than those employed in previous ELM and HSM investigations. Researchers attempting to assess the impact of more realistic communications containing numerous elements may be faced with distinguishing among a multitude potential processes.

Now consider a magazine ad for a brand of toothpaste that contains the following components: (1) a photograph of an attractive model with a bright smile, (2) an endorsement by the American Dental Association (ADA), and (3) smaller copy regarding specific attributes of the toothpaste. The photograph of the model could influence persuasion as a cue, as a persuasive argument, or indirectly via its influence on the amount of thought given to the ad. Likewise, the endorsement of the ADA could operate as a cue, an influence on the amount of thought given to the ad, or a biasing influence on the audience's assessment of quality of the arguments in the copy. Further, the copy at the bottom of the page could serve as message arguments, or as a cue. Finally, if either the photograph or the ADA endorsement were perceived as incongruent with the copy of the ad, this evaluative inconsistency could further affect the amount of thought the audience devotes to the ad.

Suppose researchers were interested in examining the impact of this ad. The previous paragraph implies thirtyBsix [3 potential roles for the photograph X 3 potential roles for the ADA endorsement X 2 potential roles for the copy claims X 2 (presence versus absence of incongruity effects)] possibilities regarding its impact on an audience. Even if the audience's predisposition to think about the advertisement is taken into account, the possibilities are numerous. If our understanding of the persuasive impact of more complex, realistic advertisements is to progress, researchers must begin to accommodate and test the simultaneous effects of multiple communication variables. That is, we must begin assessing the relative importance of a communication element given the presence of many others. Hence, what is needed is a technique that explicitly models the causal relationships among the various latent constructs associated with the effects of source and message factors in persuasion (see also Mongeau & Stiff, 1993).

MacKenzie and Lutz (1989; MacKenzie et al., 1986) demonstrated the usefulness of applying structural equations modelling for assessing multiple potential advertising effects simultaneously representing an important deviation from the standard ELM and HSM experimental approach. As discussed above, ELM and HSM researchers have traditionally placed emphasis on assessing the impact of experimental manipulations on subjects' attitudes. But, the causal modelling approach places emphasis on predicted relationships among measured latent constructs rather than predicted stimulus effects. When causal models are used in conjunction with experimental design, the primary role of experimental manipulations is to create variance in the latent constructs hypothesized as antecedents to persuasion. Once experimental manipulations have influenced the appropriate latent antecedents, as assessed by an ANOVA, subjects can be pooled across treatments to create a single sample. Causal modelling may, in this manner, allow for tests of the persuasive effects of a communication containing a confluence of distinct elements.

CONCLUSION

The ELM and the HSM have yielded tremendous insights over the last fifteen years regarding the impact of various communication elements in persuasion. If these, and related theories of persuasion are to continue offering utility for understanding persuasion processes in more natural settings, several conceptual and methodological modifications will be facilitating. The present research has attempted to identify and discuss some of the most important of these issues in hopes of helping to initiate the next generation of research on persuasion processes.

REFERENCES

Available upon request from the first author.

----------------------------------------