Four Perspectives on . . .
15
The response scale ranged from (0) “Certain that I will not” to (100) “Certain that I will” with
numeric anchors at 10-point intervals. Persuasion was assessed by computing the difference
between the post-measure and the pre-measure such that larger values indicated more persuasion.
Because these were single-item measures, no estimate of reliability was possible.
Results
Effects of Threat and Measurement
To test the effects of the experimental manipulations, we used a multivariate analysis of
variance that treated threat (high vs. low) and measurement condition (interrupted vs. non-
interrupted) as predictors. All of the remaining constructs were dependent variables (see Table
1). The results revealed a significant effect for level of threat [
7
= .85, F (4, 354) = 15.23, p <
.001], a nonsignificant effect for measurement [
7
= .99, F (4, 354) = .25, p = .90], and a
nonsignificant interaction [
7
= .99, F (4, 354) = .25, p = .90]. Inspection of the univariate results
indicated that the message effect was limited to fear acceleration, velocity, and deceleration and
the likelihood difference. Thus, the results suggested that (a) the threat manipulation did produce
variance in the fear measures, but (b) did not provide any indication that the measurement
procedures were reactive (RQ1). Consequently, we collapsed across measurement conditions in
all subsequent analyses.
A closer analysis of the effect of threat on fear appears in Figure 1. To supplement the
graphic, we ran a series of tests that contrasted the high and low threat groups on fear at each of
the three time points. The pre-message contrast yielded t (359) = .36, p = .71, r = .02 (M
hi
= .30,
SD = .59 vs. M
lo
= .28, SD = .56), which confirmed the absence of a demonstrable difference
between the high and low threat groups prior to the message. The results for the post-threat
Convention