Testing the mediating role of cognitive responses in the Elaboration Likelihood Model
Stephenson, Michael T
Cognitive response theories such as the Elaboration Likelihood Model (ELM) have been criticized for lacking appropriate and rigorous empirical testing. As a result, the ELM in particular has little support for the proposed causal mechanism, cognitive responses, as a mediating variable between messages and attitude change. Using argument quality and source credibility manipulations, this investigation examines several causal models to assess the role of cognitive responses in the ELM. In general, the data supported the ELM’s prediction that high quality arguments produce more favorable thoughts and subsequently more attitude change for high involvement receivers. Although a structural equation model supports the mediating role of cognitive responses in the ELM, the data are more consistent with a model derived from Information Processing Theory.
Petty and Cacioppo’s (1981, 1986a, 1986b) Elaboration Likelihood Model (ELM) is perhaps one of the most researched theories of attitude change. Nevertheless, the considerable research on this theory provides scant evidence for the mediating role of cognitive responses in producing the observed effects (see Mongeau & Stiff, 1993). This study addresses the shortage of empirical evidence on the ELM by testing the mediating role of cognitive responses in attitude change. First, however, a description of the ELM, its criticisms, and review the literature on these outcomes is provided.
THE ELABORATION LIKELIHOOD MODEL
The cognitive response model generally (Greenwald, 1968; Perloff & Brock, 1980) and the ELM specifically (Petty & Cacioppo, 1981, 1986a, 1986b; Petty, Haugtvedt, & Smith, 1995) portray receivers as active participants in the persuasion process (also see Chaiken, 1980). Receivers produce cognitions (thoughts, elaborations) in response to the stimulus of persuasive discourse.
Petty and Cacioppo (1986a, 1986b) posit two “routes” to persuasion: central and peripheral. The central route to persuasion consists of thoughtful consideration of the arguments (ideas, content) in the message, and occurs only when a receiver possesses both the motivation and ability to think about the message and topic. The peripheral route occurs when the receiver lacks ability and/or motivation to engage in much thought on the issue. Using the peripheral route, the listener decides whether to agree with the message based on other cues besides the strength of the arguments in the message, such as whether the source is credible or attractive, the number (but not the quality) of arguments in the message, or length of the message.
The ELM represents a powerful model of persuasion, as indicated by the considerable empirical testing it has amassed. Eagly and Chaiken (1993) explain that:
The assumption that systematic or central route processing requires motivation and ability has been documented in many studies, using a variety of motivational and ability variables: Persuasive argumentation is a more important determinant of persuasion when recipients are motivated and able to process attitude-relevant information than when they are not. There is also substantial empirical support for the hypothesis of these models that heuristic or peripheral cues exert a sizable persuasive impact when motivation or ability for argument processing is low, but little impact when motivation and ability are high. (p. 333)
However, the ELM has also provoked considerable controversy in the communication literature. See the critiques by both the “Michigan State” researchers (Allen & Reynolds, 1993; Hamilton, Hunter, & Boster, 1993; Mongeau & Stiff, 1993; Stiff, 1986; Stiff & Boster, 1987) and the subsequent replies by “Ohio State” researchers (Petty, Kasmer, Haugtvedt, & Cacioppo, 1987; Petty, Cacioppo, Kasmer, & Haugtvedt, 1987; Petty, Wegener, Fabrigar, Priester, & Cacioppo, 1993). The criticisms focus on multiple theoretical, conceptual, and methodological issues, including single versus dual processing, the imprecision of central and peripheral routes of processing, the interpretation of empirical tests, and the falsifiability of the theory, among others. The ELM has been questioned in other domains as well, largely regarding the influence of involvement on low quality messages Johnson & Eagly, 1989, 1990; Petty & Cacioppo, 1990). These exchanges focused on Johnson and Eagly’s (1989) meta-analysis which found “massive variation” in the influence of weak arguments on attitudes when comparing studies by non-Ohio State researchers against those conducted by Petty, Cacioppo, and their other Ohio State colleagues (Hamilton, Hunter, & Boster, 1993, p. 54). Hamilton et al. (1993) review the controversy and offer explanations for those divergent findings. All of these exchanges provide extensive insight into the conceptualization of the ELM and offer a considerable advancement in how this model is understood.
Mongeau and Stiff (1993) in particular focus on the need for a more rigorous test of cognitive response theories, including the ELM. The cognitive response model assumes that cognitive responses are a mediating variable in the persuasion process. Yet, in comparison to the vast amount of literature generated by the ELM, there have been relatively few rigorous tests of this assumption via path analysis or structural equation modeling. In fact, “studies rarely test the hypothesized relationship between cognitive responses and attitudes” (Mongeau & Stiff, 1993, p. 67). To address this critique, this study employs structural equation modeling to assess the role of cognitive responses as a mediating variable in the ELM. Two independent variables are employed as the basis for this study: argument quality and source expertise.
Research demonstrates that argument strength or quality is positively related to attitude change (Cacioppo, Petty, & Morris, 1983; Petty, Cacioppo, & Goldman, 1981), and it appears that cognitive responses mediate this attitude change. For example, Petty and Cacioppo (1984) reported that subjects produce more favorable cognitive responses to messages with strong than weak arguments, and more unfavorable thoughts to messages comprised of weak than strong arguments. Benoit (1987a) found that messages with strong arguments produced more favorable thoughts, fewer unfavorable thoughts, and more attitude change than messages with weak arguments.
The predictions of the ELM are largely contingent upon the impact of several moderator variables, including issue involvement (relevance). Under high involvement conditions, receivers are believed to have the motivation to evaluate the message. Consequently, in the ELM, the influence of argument quality on involving topics should increase the likelihood of central processing (Andrews & Shimp, 1990; Petty & Cacioppo, 1984). Under low involvement conditions, receivers are less motivated to evaluate the message, thereby decreasing the importance of argument quality. Petty and Cacioppo (1986a) explain that “when people are highly motivated and able to process message arguments, strong arguments are more effective than weak ones despite the presence of peripheral cues such as source credibility and attractiveness” (p. 205). Therefore, consistent with the assertions of the ELM regarding the interaction between involvement and argument quality, it is predicted:
H1: As issue involvement increases, the effect of argument quality on cognitive responses increases, such that strong arguments produce increasingly more favorable cognitive responses relative to weak arguments.
H2: As issue involvement increases, the effect of argument quality on cognitive responses increases, such that strong arguments produce increasingly more favorable attitudes relative to weak arguments.
Petty, Cacioppo, and their colleagues have consistently maintained that an influence such as argument strength on attitudes is contingent upon one’s cognitions. That is, cognitions play an important mediating role substantiating their function as the proposed causal mechanism of the ELM. Petty and Cacioppo (1986a) claim that “a message with strong arguments should tend to produce more agreement when it is scrutinized carefully [central processing] than when scrutiny is low, but a message with weak arguments should tend to produce less overall agreement when scrutiny is high rather than low” (p. 44). Ironically, there are few path analytic or covariance structure modeling tests of this relationship (Mongeau & Stiff, 1993). Those investigations that have been conducted are not tests of the ELM per se, but focus on related content such as fear-arousing messages (Hale, Lemieux, & Mongeau, 1995), mood (Petty, Schumann, Richman, & Strathman, 1993), and source credibility (Benoit & Kennedy, 1999). Therefore, in order to provide a direct test of the mediating role of cognitive responses in the ELM, it is predicted:
H3: The influence of argument quality on attitudes will be mediated via cognitive responses under high involvement conditions.
Because argument quality is less important in persuading low involved receivers, it is predicted that cognitive responses would mediate the argument quality-attitude relationship only for high involvement receivers.
Research shows that source credibility influences persuasion only if the source is identified before the message, which suggests that credibility influences persuasion by altering message processing or elaboration (O’Keefe, 1987). In the ELM, source factors can influence attitude change in three ways: “they can serve as arguments [an attractive model is evidence for a beauty product], they can serve as [peripheral] cues, and they can affect argument processing” (Petty & Cacioppo, 1986a, p. 205). For example, a belief that the source of a message is an expert can reduce motivation to scrutinize messages (develop counterarguments, elaborate) attributed to that source. Given that most messages attempt to change attitudes (i.e., are discrepant messages), the most likely response to persuasive messages is unfavorable thoughts. Reduced levels of motivation (because the expert source reassures receivers) means fewer unfavorable thoughts, which should result in more attitude change. Conversely, receivers exposed to messages attributed to non-expert sources are unlikely have reduced levels of motivation to think critically about messages; thus they ought to produce more counterarguments which reduces persuasion from messages attributed to such sources (see Benoit, 1991; Gillig & Greenwald, 1974; Hass, 1981; Perloff & Brock, 1980).
Second, source credibility may facilitate a “biased scanning” effect. There, high credibility sources would tend to elicit more favorable and fewer unfavorable thoughts than less credible sources, who in turn would encourage more unfavorable and fewer favorable thoughts (Benoit & Kennedy, 1999). In this study, the effects of expertise on cognitive responses and attitude change are evaluated. Therefore, to test the ELM assertion that issue involvement is a moderating variable with source credibility, persuading via the peripheral route, it is predicted:
H4: Source expertise will be more influential on cognitive responses with a low involvement topic rather than a high involvement topic.
H5: Source expertise will be more influential on attitudes with a low involvement topic rather than a high involvement topic.
Message type was manipulated in a 2 (high involvement, low involvement) X 2 (strong arguments, weak arguments) X 2 (high source expertise, low source expertise) factorial design. From the 364 participants (students in a communication course at a large Midwestern University), about 45 were randomly assigned to each one of the eight message conditions. All received extra credit for their participation. Participants were 55% female and the average age was 20 years.
After signing the consent form, participants read a written message advocating that seniors be required to pass a comprehensive examination to graduate. When all participants finished reading the message, they were given three minutes to generate cognitive responses in a thought-listing task. After thought-listing was complete, subjects rated their own thoughts as favorable, unfavorable, or neutral toward the message. This avoids difficulties in interpretation from irony, sarcasm, ellipses, or poor handwriting. Participants then completed measures for attitude, source credibility, argument quality, message relevance (involvement), and need for cognition.
Eight messages were written to vary involvement, strength of argument, and source credibility. All messages were equated for length (919 to 941 words), number of arguments (six), and number of times the University pertinent to the message was mentioned (15). Strong and weak argument messages were written using the arguments of Petty, Harkins, and Williams (1980) as guides. On the cover page of the message, the fictional high expertise source was credited as a Ph.D. and a Professor of Education while the fictional low expertise source was a M.A. in Botany. The source’s name was the same for both conditions. Relevance (involvement) was manipulated by advocating this change occur either at the university at which data were collected or at a distant university.
Nine-point Likert-type response formats were employed. Participants responded on the general purpose NCS Scan Form from which data were subsequently scanned into a digital file for analyses.
Attitudes. Attitudes were assessed with four items: (a) It is a good idea to require that seniors at [insert School name] pass a mandatory assessment exam, (b) We should not require seniors at [insert School name] to pass a mandatory assessment exam, (c) I support the idea of requiring that seniors pass a mandatory assessment exam at [insert School name], and (d) I am opposed to requiring that seniors pass a mandatory assessment exam at [insert School name]. These measures generated good reliability (alpha = .96).
Involvement. Personal involvement was assessed with four items: (a) This topic is not important to me, (b) This topic is not relevant to me, (c) This topic is one that really matters to me, and (d) This topic affects me personally. These measures were reliable (alpha = .85).
Perceived argument quality. Perceived argument quality was measured with four items: (a) Regardless of whether I agreed with the message, the arguments in it were strong, (b) Regardless of whether I agreed with the message, the ideas in it were poorly thought-out, (c) Regardless of whether I agreed with the message, it used weak reasons, and (d) Regardless of whether I agreed with the message, the points it made were powerful. These measures were reliable (alpha = .90).
Perceived source credibility. Perceived source credibility was assessed with four items: (a) The source was not competent to discuss these ideas, (b) The source was poorly qualified to speak on this issue, (c) The source was an expert on this question, and (d) The source was an authority on this topic. These measures were reliable (a = .81).
Cognitive responses. As noted, participants were given three minutes to write down their thoughts (both message-related and source-related) to the message. Then each subject was asked to return to his or her own listed thoughts and code each one as favorable (in agreement with the message), unfavorable (not in agreement with the message), or neutral (neither in agreement or disagreement with the message).
Manipulation checks, assessed with t-tests, are reported first. Hypotheses were tested on the omnibus model depicted in Figure 1. We employed structural equation modeling in EQS for Windows 5.7 and used normal theory maximum likelihood) to estimate the free parameters (Loehlin, 1992).
Manipulation checks were performed on argument strength, source credibility, and involvement. The argument strength manipulation was significantly different for the two conditions as individuals in the strong argument condition (M = 6.28, SD = 1.97) perceived that the message they read was significantly stronger than individuals in the weak argument condition (M = 5.65, SD = 2.14), t (362) = 2.95, p
Beyond the manipulations, the effect size of the argument strength manipulation on perceived source credibility and perceived involvement was nonsignificant (r = .07 and .05 respectively). Additionally, the effect size of the source credibility manipulation on perceived argument strength and perceived involvement was nonsignificant (r = .07 and .01 respectively). Further, the effect size of the involvement manipulation on perceived argument strength and perceived source credibility was nonsignificant (r = .05 and .05 respectively). Finally, two and three-way interactions on each of the three dependent variables were nonsignificant, suggesting no evidence of confounded factors.
While there was a difference between high and low involvement conditions, the involvement manipulation failed to produce a true “low” involvement condition (both involvement means were above the midpoint of the involvement measure). Rather, the two conditions appear to represent high and moderate levels of involvement. Involvement is the critical moderating variable in the ELM. Without a true low involvement condition, it was inappropriate to test hypotheses that predict an involvement interaction.
The weak argument quality and low source credibility manipulations also did not fall below the midpoint of the scale (although low source credibility sits on the midpoint). Clearly this is not ideal for this type of research. However, we considered these issues to be slightly less troublesome than the involvement manipulation, given the centrality of issue involvement to the tenets of the ELM. Therefore, to compensate for the failed involvement manipulation, we collapsed all subjects into one involvement condition (M = 6.02) for subsequent analyses and report direct effects where appropriate. Interestingly, however, we find others that have experienced similar issues with scale midpoints. For example, Petty, Schumann, Richman, and Strathman’s (1993) manipulation check on argument strength revealed that their weak message was above the midpoint, as was the “low” status source manipulation in Tasaki, Kim, and Miller (1999).
Structural Equation Models
A measurement model establishing the content validity of the measures is reported first, followed by the structural model that hypothesizes the relations among variables depicted in Figure 1.
Measurement model. A total of 12 indicator variables measured the three latent constructs (perceived argument quality, perceived source credibility, and attitudes) in the hypothesized measurement model. The initial run indicated a good fit of the model. However, modification indices suggested one pair of error terms for the perceived source credibility scale (items 2 and 4 in Table 1) covaried significantly, indicating that responses to the items covaried for reasons beyond the common influence of the perceived source credibility latent variable. When the constraint is relaxed, the re-specified model fits, X^sup 2^ (77, N = 364) = 87.20, p =.20; CFI = .997; RMSEA = .019 (90% Confidence Interval = .000-.037). Because the correlated errors are theoretically not expected and may be a function of the latent variable, the model was tested using Ordinary Least Squares (OLS) path analysis where the assumption about latency is not required. This solution fits the data, X^sup 2^ (3, N = 364) = 1.5, p > .05.
Structural model. The structural model depicted in Figure 1 proposes relations among three theoretical latent factors (perceived argument quality, perceived source credibility, and attitude toward topic) and two observed factors (argument quality induction and source credibility induction) as proposed by Petty, Cacioppo, and Goldman (1981). Table 2 provides the correlation matrix and standard deviations for the study’s variables.
The test of the moderator hypotheses predicted in H1, H2, H4, and H5 was not conducted because of the failed involvement manipulation. However, the data are still meaningful in assessing the expected relations among the variables in the collapsed high to moderately high involvement condition. That is, instead of comparing path coefficients from two separate models, with one model representing high involvement and the other low involvement, we examined only the former. While comparisons between high and low involvement conditions could not be made as planned, the ELM is relatively clear regarding the influence of argument quality and source credibility on cognitions and attitudes under high involvement conditions. For high involved receivers, argument quality should exert a substantial influence on cognitions and attitudes whereas source credibility should not.
The Influence of Argument Quality and Source Credibility on Cognitive Responses
The model in Figure 2 provides evidence of central processing for highly involved participants, X^sup 2^ (86, N = 364) = 106.89, p = .06; CFI = .994; RMSEA = .026 (90% Confidence Interval = .000-.041). The amount of variance explained in the attitude variable was considerable, R^sup 2^ = .37. Specifically, the model indicates a direct effect of perceived argument quality on cognitive responses, indicating that perceptions of a strong argument were associated with increasingly positive cognitive responses (o = .42, z = 2.64). Additionally, the model indicates that perceived source credibility had no measurable influence on cognitive responses. Together, these results are consistent with the ELM’s expectations for highly involved receivers.
As predicted in Hypothesis 3, cognitive responses mediated the perceived argument quality-attitude relationship, indicating the data are consistent with the ELM. Perceived stronger arguments were associated with more positive cognitions (beta = .42, z = 5.70), which in turn facilitated more positive attitudes (beta = .60, z = 12.04). Further evidence of mediation is indicated by the almost negligent residual for the argument quality-attitude covariance (.02).
The primary purpose of this study was to conduct a more rigorous test of the predictions made by the Elaboration Likelihood Model in order to extend the conceptual and theoretical understanding of cognitive responses in message processing. Overall, the models tested in this study support the idea that cognitive responses mediate the argument strength-attitude relationship for high to moderately high involvement participants. Similarly, the models also show the limited role of source credibility in processing at this level of involvement. These data are not only consistent with the predictions made by the ELM, but also with those of other processing persuasion models including the Heuristic-Systematic Model (Chaiken, 1980, 1987) and the Elastic Capacity Model (Stiff, 1986). The involvement manipulation in this investigation did not produce what we felt to be a true low involvement condition, consequently these results apply only to the expectations for message processing under high to moderately high involvement conditions.
As depicted in Figure 2, messages perceived to have strong arguments generated more favorable cognitive responses than those messages perceived to have weak arguments. These results are consistent with prior research on the effects of high quality arguments (e.g., Benoit, 1987a; Cacioppo, Petty, & Morris, 1983; Petty & Cacioppo, 1984; Petty, Cacioppo, & Goldman, 1981). Additionally, strong argument messages generated more favorable attitudes toward the topic than weak argument messages.
But perhaps theoretically most informative is evidence for mediation specified in Hypothesis 3. Specifically, the argument strength-cognitive responses and cognitive responses-attitude relationships were significant. Further, the error in the covariance between argument strength and attitudes was negligible when cognitive responses mediated the argument strength-attitude relationship. These findings indicate complete mediation as the covariance between argument strength and attitudes dropped to zero and was explained by the intervening variable. As a result, this study adds to the current meager support for the causal mechanism proposed by the ELM under high involvement conditions.
It should be noted that the correlation between the argument quality induction and perceived argument quality was low. While the two perceived argument quality conditions were statistically different, the argument manipulation was clearly weak.
Unlike argument quality, source credibility is expected to exert more influence on those with less motivation to attend to the message (i.e., who are engaged in peripheral processing). Consequently, considering previous tests of source credibility in the ELM (e.g., Petty & Cacioppo, 1979; Petty, Cacioppo, & Schumann, 1983), one would not expect source credibility to be a strong source of influence for high involvement receivers. The opposite would be true for low involvement receivers, although we were unable to test this prediction. In this study, the effect of perceived source credibility was trivial. Source credibility exerted neither a direct effect nor an indirect effect on attitude toward the issue for high or moderately-high involvement receivers.
Limitations and Suggestions for Future Research
Perhaps the greatest limitation to this research was the inability to obtain a true low involvement condition. Although participants’ perceptions of involvement were significantly different in the two conditions, the mean score for the condition initially designated as low involvement fell about the midpoint of the nine-point scale. While statistically we could have argued for two different levels of involvement, the theoretical contribution would have been minimal as the Elaboration Likelihood Model’s conceptualization of low involvement is much different than that which we obtained. In all likelihood, the participants in the low involvement condition did not notice that the message arguments were directed at a distant university. Rather, they treated the arguments as if they were applicable to their own university. This means that this study must be viewed as providing support for the causal mechanism of the ELM for high to moderately high involved participants. The data do not permit us to test the ELM’s mechanism for uninvolved subjects.
There is some available evidence on the relationship between argument quality and cognitive responses. Both messages were rated above the mid-point on perceived argument quality (strong = 6.23; weak = 5.65, on a 9-point scale). The message with higher quality arguments elicited 61% unfavorable thoughts. The message with weaker quality arguments elicited more unfavorable thoughts, 74%. Presumably this occurred because participants tended to be highly opposed to the position advocated by the messages. This finding, while not conclusive, calls into question Petty and Cacioppo’s decision to operationally define a “strong argument” as one that elicits predominantly favorable responses. Both messages were rated as above the mid-point in quality, yet both elicited predominantly unfavorable thoughts. This indicates that the quality of argument is related to, but should not be defined by, the profile of the cognitive responses it elicits.
The question of which features of discourse (e.g., violations of tests of reasoning, presence of fallacies, use and quality of evidence) influence perceptions that an argument is strong or high quality has been considered (Benoit, 198 7b; Hample, 1981; McKerrow, 1977; O’Keefe, 1998; Rowland, 1985) but has not been thoroughly investigated. There has been a recent spate of studies comparing narrative and statistical evidence (e.g., Allen & Preiss, 1997; Baesler, 1997; Boster et al., 2000). Nevertheless, more research into the effects of persuasive message features on perceptions of argument quality would enhance understanding of the nature and meaning of argument quality.
Finally, problems with the empirical properties of the source credibility measure should be acknowledged. Error terms for two items were correlated in the measurement model with latent variables, an issue that is problematic for some researchers. The use of these measures of source credibility, without additional research on the structure of these items, is not recommended.
Alternative Models of Message Processing
While it was demonstrated that cognitions mediated the argument quality-attitude relationship, a secondary interest emerged regarding the consistency of the data with other related processing models. Mongeau and Stiff (1993) have suggested that “the predictions of the ELM are sufficiently imprecise and ambiguous as to prevent an adequate test of the entire model” (pp. 67-68). While Figure 1 represents the ELM, the testing of a different model that was more consistent with the methodology of this study and the related Information Processing Theory was warranted (cf. Hamilton, 1997, 1998; Stiff, 1994).
First, it is not certain that all receivers actually form an impression of argument quality during message processing. Following Information Processing Theory, Hamilton (1997, 1998) argues that cognitions precede inference-making processes such as assessing the quality of message arguments. Methodologically, the questionnaire measured cognitive responses before perceived argument quality because cognitive responses were believed to be a more important dependent variable than perceived message quality in the ELM. Because participants reported cognitive responses before perceived message quality, it is possible that their cognitive responses influenced perceptions of message quality. It is conceivable, then, that this judgment was made only when they responded to the dependent measure (i.e., well after message processing was finished).
Second, source credibility was manipulated before participants read the message. In fact, early persuasion theory by Hovland, Janis, and Kelly (1953) contends that source credibility may facilitate the attention and comprehension of a message’s arguments (cf. Hamilton, Hunter, & Boster, 1993; Hunter, Danes, & Cohen, 1984). However, judgments about the source may not occur independently of judgments about the quality of the message. Furthermore, the perceived source credibility measures followed the assessment of cognitive responses.
In light of these issues, we present the model in Figure 3 which tests both of these possibilities. Statistically, the model in Figure 3 provides a good fit to the data, X^sup 2^ (86, N = 364) = 89.89, p =.37; CFI = .999; RMSEA = .012 (90% Confidence Interval = .000 -.031). In fact, the fit of this model is significantly better than the fit of the model derived from the ELM in Figure 1, Delta^sub X^^sup 2^ (2, N = 364) = 17.0, p
This study provided support for the ELM’s prediction that, for moderately to highly involved receivers, high quality arguments elicited more favorable cognitions and attitudes than low quality arguments. The data also show that the influence of source credibility was insignificant for such readers. Perhaps most importantly, this study offers a more rigorous test of mediation that resulted in support for the causal mechanism proposed by the ELM. Comparing models of persuasive processing was not initially a goal of this study. However, the data are more consistent with Information Processing Theory. In particular, perceived argument quality and perceived source credibility were significantly correlated. This relationship is not predicted by the ELM. Nevertheless, cognitive responses still mediated the argument quality– attitude relationship in these data. More research of this type, particularly with low involvement participants, is warranted.
Maximum likelihood estimation assumes normally distributed data.
Allen, M., & Preiss, R. W. (1997). Comparing the persuasiveness of narrative and statistical evidence using meta-analysis. Communication Research Reports, 14, 125-131.
Andrews, J. C., & Shimp, T. A. (1990). Effects of involvement, argument strength, and source characteristics on central and peripheral processing in advertising. Psychology &Marketing, 7, 195-214.
Baesler, E. J. (1997). Persuasive effects of story and statistical evidence. Argumentation and Advocacy, 33, 170-175.
Benoit, W. L. (1987a). Argument and credibility appeals in persuasion. Southern Speech Communication,journal, 52, 181-197.
Benoit, W. L. (19876) Argument evaluation. In F. H. van Eemeren, R. Grootendorst,J. A. Blair, & C. A. Willard (Eds.), Argumentation: Across the lines of discipline (pp. 189-297). Dordrecht, Holland: Foris.
Benoit, W. L. (1991). A cognitive response analysis of source credibility. In B. Dervin & Mj Voigt (Eds.), Progress in communication sciences (vol. X, pp. 1-19). Norwood, NJ: Ablex.
Benoit, W. L., & Kennedy, K. A. (1999). On reluctant testimony. Communication Quarterly, 47, 376-387.
Boster, F. J., Cameron, K. A., Campo, S., Liu, W., Lillie, J. K., Baker, E. M., & Yun, K. A. (2000). The persuasive effects of statistical evidence in the presence of exemplars. Communication Studies, 51, 296-306.
Cacioppo,J. T., Petty, R. E., & Morris, K. (1983). Effects of need for cognition on message evaluation, recall, and persuasion. Journal of Personality and Social Psychology, 45, 805-818.
Chaiken, S. (1980). Heuristic versus systematic information processing and the use of source versus message cues in persuasion. Journal of Personality and Social Psychology, 39, 752-756.
Chaiken, S. (1987). The heuristic model of persuasion. In M. P. Zanna,J. M. Olson, & C. P. Herman (Eds.), Social influence: The Ontario Symposium (Vol. 5, pp. 3-39). Hillsdale, NJ: Lawrence Erlbaum.
Eagly, A. H., & Chaiken, S. (1993). The psychology of attitudes. Fort Worth: Harcourt Brace Jovanovich. Gillig, P. M., & Greenwald, A. G. (1974). Is it time to lay the sleeper effect to rest?Journal of Personality and Social Psychology, 29, 132-139.
Greenwald, A. G. (1968). Cognitive learning, cognitive response to persuasion, and attitude change. In A. G. Greenwald, T. C. Brock, & T. M. Ostrom (Eds.), Psychological foundations of attitudes (pp. 147-170). New York: Academic Press.
Hale, J. L., Lemieux, R, & Mongeau, P. A. (1995). Cognitive processing of fear-arousing message content. Communication Research, 22, 459-474.
Hamilton, M. A. (1997). The phase interfaced omnistructure underlying the processing of persuasive messages. In G. A. Barnett & F. J. Boster (Eds.), Advances in communication science (pp. 1-42). Greenwich, CT: Ablex.
Hamilton, M. A. (1998). Message variables that mediate and moderate the effect of equivocal language on source credibility. Journal of Language and Social Psychology, 17, 109-143.
Hamilton, M. A., Hunter,J. E., & Boster, F.J. (1993). The elaboration likelihood model as a theory of attitude formation: A mathematical analysis. Communication Theory, 3, 50-65.
Hample, D. (1981). What is a good argument? In G. Ziegelmueller &J. Rhodes (Eds.), Dimensions of argument (pp. 875-893). Annandale, VA: Speech Communication Association.
Hass, R. G. (1981). Effects of source characteristics on cognitive responses and persuasion. In R E. Petty, T. M. Ostrom, & T. C. Brock (Eds.), Cognitive responses in persuasion (pp. 44-72). Hillsdale: Erlbaum.
Hovland, C. I.,Janis, I. L., & Kelley, H. H. (1953). Communication and persuasion: Psychological studies of opinion change. New Haven, CT: Yale University Press.
Hunter,J. E., Danes,J., E., & Cohen, S. (1984). Mathematical models of attitude change, 1: Change in single attitudes and cognitive structures. New York: Academic Press.
Johnson, B. T., & Eagly, A. H. (1989). Effects of involvement on persuasion: A meta-analysis. Psychological Bulletin, 106, 290-314.
Johnson, B. T., & Eagly, A. H. (1990). Involvement and persuasion: Types, traditions, and the evidence. Psychological Bulletin, 107, 375-384.
Loehlin,J. C. (1992). Latent variable models. Hillsdale, NJ: Erlbaum.
McKerrow, R. E. (1977). Rhetorical validity: An analysis of three perspectives on the justification of rhetorical argument. Journal of the American Forensic Association, 13, 133-141.
Mongeau, P. A., & Stiff, J. B. (1993). Specifying causal relationships in the Elaboration Likelihood Model. Communication Theory, 3, 65-72.
O’Keefe, D. J. (1987). The persuasive effects of delaying identification of high- and low-credibility communicators: A meta-analytic review. Central States Speech Journal, 38, 63-72.
O’Keefe, D.J. (1998). Justification explicitness and persuasive effect: A meta-analytic review of the effects of varying support articulation in persuasive speeches. Argumentation and Advocacy, 35, 61-75.
Perloff, R M., & Brock, T. C. (1980). `And thinking makes it so’: Cognitive responses to persuasion. In M. E. Roloff & G. R. Miller (Eds.), Persuasion: New directions in theory and research (pp. 67-99). Beverly Hills: Sage. Petty, R E., & Cacioppo,J. T. (1979). Issue involvement can increase or decrease persuasion by enhancing
message-relevant cognitive processes. Journal of Personality and Social Psychology, 37, 1915–1926.
Petty, R. E., & Cacioppo,J. T. (1981). Attitudes and persuasion: Classic and contemporary approaches. Dubuque, IA: William C. Brown.
Petty, R. E., & Cacioppo, J. T. (1984). The effects of involvement on responses to argument quantity and quality: Central and peripheral routes to persuasion. Journal of Personality and Social Psychology, 46, 69 –81. Petty, R. E., & Cacioppo, J. T. (1986a). Communication and persuasion: Central and peripheral routes to attitude change. New York: Springer-Verlag.
Petty, R. E., & Cacioppo, J. T. (1986b). The elaboration likelihood model of persuasion. In L. Berkowitz (Ed.), Advances in experimental social psychology (vol. 19, pp. 123-205). San Diego, CA: Academic Press.
Petty, R. E., & Cacioppo,J. T. (1990). Involvement and persuasion: Tradition versus integration. Psychological Bulletin, 107, 367-374.
Petty, R. E., Cacioppo,J. T., & Goldman, R (1981). Personal involvement as a determinant of argumentbased persuasion. Journal of Personality and Social Psychology, 41, 847-855.
Petty, R. E., Cacioppo, J. T., & Schumann, D. (1983). Central and peripheral routes to advertising effectiveness: The moderating role of involvement. Journal of Consumer Research, 10, 135-146.
Petty, R. E., Haugtvedt, C. P., & Smith, S. M. (1995). Elaboration as a determinant of attitude strength: Creating attitudes that are persistent, resistant, and predictive of behavior. In R.E. Petty &J. A. Krosnick (Eds.), Attitude strength: Antecedents and consequences (pp. 93-130). Mahwah, NJ: Lawrence Erlbaum.
Petty, R. W., Harkins, S. G., & Williams, K. D. (1980). The effects of group diffusion of cognitive effort on attitudes: An information processing view. Journal of Personality and Social Psychology, 38, 81-92.
Petty, RL E., Kasmer, J. A., Haugtvedt, C. P., & Cacioppo, J. T. (1987). Source and message factors in persuasion: A reply to Stiffs critique of the elaboration likelihood model. Communication Monographs, 54,233-249. Petty, R E., Cacioppo, J. T., Kasmer, J. A., & Haugtvedt, C. P. (1987). A reply to Stiff and Boster. Communication Monographs, 54, 257-262.
Petty, R E., Schumann, D. W., Richman, S. A., & Strathman, A. J. (1993). Positive mood and persuasion: Different roles for affect under high- and low-elaboration conditions. Journal ofPersonality and Social Psychology, 64, 5-20.
Petty, R. E., Wegener, D. T., Fabrigar, L. R., Priester, J. R., & Cacioppo, J. T. (1993), Conceptual and methodological issues in the elaboration likelihood model of persuasion: A reply to the Michigan State critics. Communication Theory, 3, 336-362.
Rowland, R. C. (1985). On argument evaluation. Journal of the American Forensic Association, 21, 123-132. Stiff,J. B. (1986). Cognitive processing of persuasive message cues: A meta-analytic review of the effects of supporting information on attitudes. Communication Monographs, 53, 75-89.
Stiff, J. B. (1994). Persuasive communication. New York: Guilford.
Stiff, J.B., & Boster, F. J. (1987). Cognitive processing: Additional thoughts and a reply to Petty, Kasmer, Haugtvedt, and Cacioppo. Communication Monographs, 54, 250-256.
Tasaki, K, Kim, M-S., Miller, M. D. (1999). The effects of social status on cognitive elaboration and post-message attitude: Focusing on self-construals. Communication Quarterly, 47 196-212. y, 47, 196-212.
Mike Stephenson (Ph.D., Kentucky, 1999) is an assistant professor in the Department of Communication at the University of Missouri-Columbia where William L. Benoit (Ph.D., Wayne State University,1979) is a Professor and David A. Tshida is a doctoral candidate. Address correspondence to the first author, Department of Communication, 115 Switzler Hall, Columbia, MO, 65211-2310. E-mail: firstname.lastname@example.org.
Copyright Central States Speech Association Winter 2001
Provided by ProQuest Information and Learning Company. All rights Reserved