Modeling the Effects of Processing Effort and Ability in Response to Persuasive Message Arguments

Abstract

This study proposed a path model of the effects of message comprehensibility on attitude change with the measured cognitive mediators suggested by dual-process models of persuasion. An experiment was conducted using a 2 (strong arguments vs. weak arguments) × 2 (easy vs. difficult to understand arguments) fully crossed independent-groups design with 120 participants. The data were consistent with the traditional impact of message comprehensibility on difficulty of processing the message such that the effect of argument strength was diminished when the arguments were difficult, rather than easy, to understand. The path models were not consistent with the data, and the model was revised to produce adequate fit.

Keywords:

• Causal Modeling
• Cognitive Processing
• Social Influence

Notes

Note. Correlations below the diagonal are raw correlations, and those above the diagonal have been corrected for measurement error.

The data examined here were collected as part of a larger study such that other groups whose data were not examined here were given additional processing instructions.

The messages used can be obtained by contacting Christopher J. Carpenter.

A pilot test established that comprehension difficulty and argument strength were successfully induced using the same messages and the same measures as the focal study (for details, see Carpenter & Boster, 2011).

The full scale is available from Christopher J. Carpenter.

For all correlations reported used to test the model, scatterplots were examined and no evidence of nonlinearity was found. Restriction in range in one of the variables for one group and not the other or differential measurement reliability may explain why two correlations are different. For all of the analyses reported, differential variance was unlikely because the differences in variance were trivial and Levene's test was not statistically significant for any of them. Differential measurement reliability was also unlikely to be a problem as the reliability of the scales also did not substantially differ by condition.

The revised models for strong and weak arguments were tested together using multiple-groups analysis with Amos (Amos Development Corporation, Crawfordville, FL). The fit indexes indicated excellent model fit, χ²(8, N = 93) = 5.22, p = .73 (normed fit index = 0.97, comparative fit index = 1.00, root mean square error of approximation = 0.00).

Additional information

Notes on contributors

Christopher J. Carpenter

Christopher J. Carpenter (Ph.D., Michigan State University, 2010) is an assistant professor in the Department of Communication at Western Illinois University.

Franklin J. Boster

Franklin J. Boster (Ph.D., Michigan State University, 1978) is a professor in the Department of Communication at Michigan State University.