The vaccine rift effect provides evidence that the source’s vaccination status determines rejection of vaccination calls

Our pattern (M= 49.09, SD= 13.32, range [15; 83]) was representative of the average age of the adult population (M= 49.74), t(1169) = − 1.68p= 0.094, i.e= − 0.05. However, we oversampled female respondents (57% vs. 52% in the population, e.g=3.43, pe.g= − 3.43, pe.g0.01, p > 0.999), but we undersampled partially vaccinated or recovered participants (29% vs. 34% in the population, e.g= − 3.63, p e.g= 4.65, p

confirmation analyses

Overall, we observed main effects consistent with the proposed rift effect of vaccination (correlations in Fig. 1). Participants attributed less constructive motives t(1167.50) = 4.80, pi.e= 0.28, 95% CI [0.17; 0.40]and less positive personality traits, t(1157.08) = 2.92, p= 0.004, i.e= 0.17, 95% CI[0.06; 0.29], to vaccinated commenters than to unvaccinated commenters, despite identical message content. Participants also reported feeling more threatened by calls to get vaccinated from the vaccinated source than from the unvaccinated source. t(1161.92) = − 5.17, pi.e= −0.30, 95% CI[− 0.42; − 0.19] (Table 1).

illustration 1

Correlation plot of dependent variable and condition.

Table 1 Mean comment ratings and by comment source (folded over participants’ vaccination status).

As behavioral indicators, we analyzed the behavioral planning and counterarguments of the participants in free text. Only 15.6% of participants requested additional information, and the source of the comment did not have a significant direct impact on this measure, χ2(1, N=1170) p> 0.999. We expected that criticism from the seeded source would generate more counter-arguments, as evidenced by a larger word count of the free-text response (assessed in LIWC201524). Previous research has observed that response length is strongly correlated with negative content when countering group criticism25, and the word count thus provides a good behavioral indicator of the vaccination gap. Participants who responded to the vaccinated commenter actually used more words to express their opinion about the vaccine (mouth= 22.50) as participants who responded to the unvaccinated commenter (mouth= 19.00), Wilcoxon signed rank tests W= −3.00, 95% CI [− 5.00; − 1.00], p= 0.009, right= 0.08 (Table 2).

Table 2 Measures of behavior by comment source (folded over participants’ vaccination status).

Exploratory analyses

vaccination status of the participants

Vaccination status may mitigate the Rift effect, as those who have already adopted the behavior are less likely to be suspicious of the message, although previous research suggests that even members of the commentator’s group are suspicious of the intergroup criticism21. Indeed, participants’ own vaccination status moderated the observed impact on news threat (Fig. 2b) and commenter rating (Fig. 2c), as indicated by significant interactions between news source and participants’ vaccination status. f(2, 1164) = 3.94, p= 0.008, ηp2= 0.01 and f(2, 1164) = 5.40, p= 0.001, ηp2= 0.01. This interaction was not significant for the comment motif (Fig. 2a), f(2, 1164) = 1.46, p= 0.225, ηp2ps = 0.002 to ps = 0.014 to p= 0.049, but no threat, p= 0.444 or commenter ratings, p= 0.162. Partially vaccinated participants showed no vaccination crack to the subject, p= 0.181 or commenter ratings, p= 0.479, but for threat, p= 0.032. In summary, we observed a consistent vaccination gap among the unvaccinated or recovered, but not among the (partially) vaccinated.

figure 2
figure 2figure 2

Box plots with violin plots (a) Rating of message motive by message source (unvaccinated vs. vaccinated) and participants’ vaccination status. (b) News threat by news source (unvaccinated source vs. vaccinated) and participants’ vaccination status. (c) News source rating by news source (unvaccinated source vs. vaccinated) and participants’ vaccination status. (i.e) Word count of free-text response by message source (unvaccinated source vs. vaccinated) and participant’s vaccination status. Eight extreme data points > 300 words were omitted for presentation. Note.Boxplot notches indicate 95% CIs.

Table 3 Mean comment ratings by comment source for each vaccination status participant group.

The impact on requests for additional information or written responses (Fig. 2d) was not moderated by participants’ vaccination status. ps > 0.454 (group tests and effect sizes in Tables 4 and 5). Ironically, however, participants who were already fully vaccinated (19%) requested additional information significantly more often than participants who were not vaccinated (08%), χ2(1, N= 836) = 20.66, p2(1, N= 632) = 11.98, p= 0.001. Partially vaccinated participants also requested additional information significantly more frequently than non-vaccinated participants (08%), χ2(1, N= 538) = 8.60, p= 0.003 or recovered (07%), χ2(1, N= 334) = 6.19, p= 0.013. No significant difference between partially and fully vaccinated participants, χ2(1, N= 634) = 0.34, p= 0.562, or recovered and unvaccinated participants arose, χ2(1, N= 536) p= 0.945. Fully vaccinated participants showed shorter responses than recovered ones, ppp= 0.018 or unvaccinated participants, p= 0.034. The duration of response did not differ between fully and partially vaccinated participants, p= 0.200, still recovered and unvaccinated participants, p= 0.888. Overall, the critical unvaccinated group showed the strongest vaccine rift effect (detailed results in Appendix 1).

Table 4 Behavior planning by comment source for each group of participants with vaccination status.
Table 5 Counter-arguments by comment source for each group of participants with vaccination status.

Structural Equation Modeling

Next, we investigated possible mechanisms underlying the vaccination rift effect (Fig. 3; see Appendix 1 for detailed results). The comment source predicted an ascribed constructiveness of the comment, which predicted a higher likelihood of participants engaging in behavior planning morecounterargument. The indirect effects of news source on planning behavior and counterargument about motive were significant. Sub-sample analysis (Figs. 4 and 5) showed that none of the mediators predicted planning or counterarguments among fully vaccinated participants. Partially vaccinated participants showed both a serial indirect effect and a simple serial effect via message constructivity and threat to planning but not counterargument. In recovered participants, there was an indirect effect on behavioral planning about message motive, as well as counterargument about motive and threat. Finally, a serial indirect effect of message constructivity via commentator ratings on behavioral planning (but not counterarguments) occurred in unvaccinated participants. Apparently, for the unvaccinated, trust in the source is a particularly important determinant of vaccine failure on behavioral planning.

figure 3
figure 3

Results of structural equation model analysis. Dashed lines are insignificant paths. Estimates are standardized regression coefficients and SEs in parentheses.

figure 4
figure 4

Results of structural equation model analysis for behavioral planning by vaccination status. Dashed lines represent non-significant paths. Estimates are standardized regression coefficients and SEs in parentheses; only significant paths are reported.

Figure 5
Figure 5

Results of structural equation model analysis for counterargument by vaccination status. Dashed lines represent non-significant paths. Estimates are standardized regression coefficients and SEs in parentheses; only significant paths are reported.

Comments are closed.