19
data yielded nonsignificant paths for the interaction term onto reactance and for attitude onto
intention. The interaction term was removed, first, and the remaining paths re-estimated. When
the modification indices suggested the need for elimination of the attitude-intention link and the
addition of a direct path from reactance to intention, we altered the model accordingly then re-
estimated the parameters. Figure 3 presents the path coefficients for the final Intertwined Process
Model.
TABLE 7 & FIGURE 3 ABOUT HERE
Research Questions 1 and 2: The Nature of Reactance
The first two research questions inquired as to the nature of reactance. RQ1 asked: Is
reactance best conceptualized as cognition, affect, or both? In fact, both data sets showed that
cognition and affect mediated the effects of threat and proneness on attitude. Thus, the two single
process models were rejected.
RQ2 posed the question: If reactance has both cognitive and affective components, how
are the two components combined? Here too, the results were consistent across the two data sets.
The Intertwined Process Model was superior to the Dual Process Model, in terms of fit to the
data, on both absolute (i.e., AGFI, RMSEA, BIC) and relative (BIC difference) indices.
Hypotheses 1 and 2: Antecedents of Reactance
H1 predicted a positive association between threat to freedom and reactance. The path
from threat to reactance was .31, p< .05, in the flossing data and .51, p< .05, in the drinking data.
Therefore, Hypothesis 1 received consistent support across the two data sets.
H2 anticipated a positive correlation between reactance proneness and reactance. The
path from proneness to reactance was .41, p< .05, in the flossing data and .29, p< .05, in the
drinking data. Therefore, Hypothesis 2 was supported in both studies.
Convention