Waiting for a Match: Mitigating Reactance in Prosocial Health Behavior Using Psychological Distance

ABSTRACT

This study examined how psychological distance, both social and temporal, can be leveraged in prosocial health behavior messages to mitigate perceived psychological reactance. Following the construal level and psychological reactance theories, we conducted a 2 × 2 between-subjects factorial design (N = 245), which manipulated naturalistic messages regarding a prosocial communications campaign. Structural equation modeling showed that far temporal distance combined with far social distance could significantly reduce threat to freedom and therefore positively affect attitudes and behavioral intentions toward prosocial health topics. The effect of social distance was found not significant, differing from past findings. Further, intertwined and parallel psychological reactance models were tested and discussed. We suggest the need for more psychological reactance research, particularly examining prosocial health behavior. Strategies for practical persuasion strategies in prosocial messages are proposed.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1. The experimental design did not include a control group because this study focused on testing whether far psychological distance will exert less reactance than close psychological distance, rather than comparing far or close psychological distance to the control condition in which the message does not indicate any psychological distance.

2. All stimuli messages are available from the corresponding author.

3. The 14-item HPRS did not fit well in our model, so the validity of the scale was checked before proceeding to the next step. We found that previous research had difficulty identifying the psychometric properties of this scale (Sinclair et al., 2015). For example, some studies found a four-factor model fit the HPRS scores the best (e.g., Brown et al., 2011; Hong & Faedda, 1996; Hong & Page, 1989), some studies supported a second-order model, in which four factors existed at the first-order and an overall trait reactance being treated as unidimensional at the second-order (e.g., Shen & Dillard, 2005), while some studies demonstrated a bifactor model resulted in greater model fit (e.g., Yost & Finney, 2018). It seems the modeling solution of the HPRS depends on the specific studies. We applied all the above-mentioned modeling solutions to the HPRS scores in the current study, but none of them produced an acceptable fit (for the four-factor model: RMSEA = .099 (.085 – .112); for the second-order model: RMSEA = .148 (.131 – .165); for the bifactor model: RMSEA = .126 (.112 – .140)). Therefore, we conducted our own factor analysis for the current study and adjusted the items to fit the measurement model for this study. In the current final measurement model, the items kept for analysis on the HPRS scale included: item 1, item 2, item 3, item 5, item 9, item 13, and item 14 (See Hong & Page, 1989 for specific items). In addition to relying on the modeling results, we finalized these items that were also based on conceptual reasoning. These items explained three out of four factors that concluded in the previous four-factor model (Hong & Faedda, 1996; Shen & Dillard, 2005): reactance to compliance (item 1, 2, and 3), resisting influence (item 13 and 14), reactance to advice (item 5 and 9). The items related to the last factor, emotional response (item 4, 6, 7, and 8), significantly hurt the model fit, so we considered deleting those items. Miron and Brehm (2006) also cast doubt on the items on the HPRS measuring an “affective state” (p. 7). They commented that “the explanatory power of these scales is low perhaps because of the different threat situations covered by these scales items;” therefore, “there is little to gain from the conceptualization of reactance as a personality trait” (p. 7). This may also explain why these items did not fit well in our model. Thus, we decided to remove items related to the emotional response from the final measurement model in this study.

4. To improve the model fit, we correlated two items of the freedom threat scale as the model inspection results indicated: Item 1 (The message threatened my freedom to choose) and Item 4 (The message tried to pressure me). The text meanings of these two items are similar, so we considered it appropriate to correlate them.

5. Model #1 and Model #2 had identical model fit indices because they were the same model just with different reference groups.