![Sample our Communication Studies journals, sign in here to start your access, 2013 & 2014 volumes FREE to you for 14 days]({"type":"image" , "src" : "/sda/5927/TOC-Subject-Sample-2021-Information-Science2.jpg"})
Abstract
This investigation introduced the concept of associative networks to the resistance domain. A four-phase experiment was conducted involving 298 participants. The results confirmed the role of known elements in the process of resistance: the core elements of threat, issue involvement, and counterarguing output facilitated resistance, as did the more recently introduced element of attitude certainty. Also, the results indicated that inoculation treatments modified the structure of associative networks, planting additional nodes and increasing the linkages between nodes. Subsequently, changes in the structure of associative networks contributed to resistance to counterattitudinal attacks.
The authors thank Tim Roberts, Rachel Savant, Brad Thomason, and Erik O'Kelley for their assistance with data collection and thank instructors in the Department of Communication for their cooperation in providing subjects for the study.
Notes
1. The issues of restricting T.V. violence and legalizing marijuana scored 6.69 and 6.28, respectively, on a three-item, 10-interval involvement measure, which was based on a tool originally developed by Traylor (Citation1981). Scores on involvement across the 16 issues ranged from 5.33 to 7.63. Seven issues score lower in involvement than the two issues selected for the study, whereas seven scored higher in involvement. The distribution of opinion—for, against, and neutral/no opinion—was, for restricting T.V. violence: 45%, 53%, and 12%, and, for legalizing marijuana, 31%, 49%, and 19%.
2. Researchers were interested in participants’ thoughts as they processed an experimental message. However, the wording, “attitude measure,” was used instead of “message” in order to insure uniformity in instructions across participants. This was important because about one-fourth of respondents at Phase 2 and three-quarters of respondents at Phase 4 did not receive a message.
3. Preliminary intercoder reliabilities for coder pairs ranged from .91 to .98 for number of nodes and links and from .73 to .91 for node strength.
4. Experimental condition was not a measured variable. Those participants who served as controls were assigned a value of 1; those who were assigned to inoculation conditions were assigned a value of 2. Since participants were assigned to the experimental conditions randomly, and because we are confident, but not absolutely certain, that this assignment was recorded correctly, the error term for inoculation was set at .04. Other variables required special treatment in estimating error. For example, as indicated in the Methods section, Phase 3 attitude certainty was a single-item measure. Although some recommend that error terms of single-item measures be set at zero, assuming the item is measured without error, we reasoned that the most accurate estimate of its reliability would be this variable's performance across time. Hence, its reliability rating was computed by combining Phase 3 and 4 attitude certainty measures. This reliability coefficient (.81) was plugged into the formula for estimating measurement error, producing error estimates of 31.365. Finally, the reliabilities for Phase 2 counterarguing output and for Phase 3 nodes/links could not be computed in the standard fashion since, as explained in Methods, these measures consisted of: for Phase 2 counterarguing output, the number of arguments opposing initial attitudes and number of responses to those arguments (counterarguments); and, for Phase 3 nodes/links, the number of nodes and links divided by two. In neither case, was there an ideal way to compute reliability, but it was unreasonable to assume that these items were measured without error. Thus, the following approach was used. For Phase 2 counterarguing output, reliability was assessed by combining ratings of arguments opposing initial attitudes and counterarguments. This produced a reliability rating of .78, which was inserted in the formula for estimating measurement error, producing an error estimate of 1.21. For Phase 3 nodes/links, reliability was computed by combining the ratings of the number of nodes and the number of links. This produced an alpha reliability rating of .99, which was plugged into the formula for estimating measurement error, producing an error estimate of .08.
5. Standard errors for the parameters of the final model were reasonable. The normalized residuals (skew: 0.72 to 1.18; kurtosis: 0.74 to 3.75) were within what West, Finch, and Curran (Citation1995, p. 74) view as acceptable ranges. There were a limited number of missing observations, all random. Missing observations were handled by imputation of missing observations using one of two approaches (see Kline, Citation1998, p. 75). For multiple-item scales in which a single value was missing, the average of the remaining items for that participant was inserted. For missing single observations, such as attitude certainty, it was determined whether the participant was assigned to the treatment or control condition and whether the mean for that condition on the item in question was inserted.
6. The significant factor-to-factor paths included: experimental condition to Phase 2 counterarguing output (z=4.28, p<.05), Phase 3 network nodes/links (z=5.82, p<.05), Phase 3 attitude certainty (z=4.22, p<.05), and Phase 4 attitude toward the counterattitudinal attack (z=–2.76, p<.05); Phase 2 counterarguing output to Phase 4 attitude toward the persuasive attack (z=–2.74, p<.05); Phase 3 network nodes/links to Phase 4 attitude toward the counterattitudinal attack (z=–1.93, p<.05, one-tailed); and Phase 3 attitude certainty to Phase 4 attitude toward the counterattitudinal attack (z=–5.36, p<.05).