ABSTRACT
People often continue to rely on misinformation in their reasoning after they have acknowledged a retraction; this phenomenon is known as the continued-influence effect. Retractions can be particularly ineffective when the retracted misinformation is consistent with a pre-existing worldview. We investigated this effect in the context of depressive rumination. Given the prevalence of depressotypic worldviews in depressive rumination, we hypothesised that depressive rumination may affect the processing of retractions of valenced misinformation; specifically, we predicted that the retraction of negative misinformation might be less effective in depressive ruminators. In two experiments, we found evidence against this hypothesis: in depressive ruminators, retractions of negative misinformation were at least as effective as they were in control participants, and more effective than retractions of positive misinformation. Findings are interpreted in terms of an attentional bias that may enhance the salience of negative misinformation and may thus facilitate its updating in depressive rumination.
Acknowledgements
We thank Charles Hanich and Blake Cavve for research assistance. This work was supported by an Australian Government Research Training Program Scholarship to the first author, and an Australian Research Council grant (DP160103596) to the second author. The funding agencies had no involvement in the study design, collection, analysis, and interpretation of data, writing of the report, or the decision to submit the article for publication. The authors declare no conflicts of interest. All procedures performed were approved by the Human Ethics Office of the University of Western Australia.
Disclosure statement
No potential conflict of interest was reported by the authors.
ORCID
Ee Pin Chang http://orcid.org/0000-0002-9932-6267
Ullrich K. H. Ecker http://orcid.org/0000-0003-4743-313X
Andrew C. Page http://orcid.org/0000-0003-3133-2844
Notes
1 We selected participants on both rumination and depression measures simultaneously because we have argued that rumination could be either (1) a primary symptom arising from a generic updating deficit, which contributes to depressive symptomatology (Vanderhasselt & De Raedt, Citation2012), or (2) a secondary symptom arising from a valence-specific updating deficit associated with depression (Joormann & Gotlib, Citation2008, Citation2010). Given this uncertainty and previous findings that rumination, but not depression, correlated with updating impairments in depression (Joormann & Gotlib, Citation2008, Citation2010), it was decided to screen on both constructs simultaneously (see Chang, Ecker, & Page [Citation2017] for more detailed justification for sample selection criteria).
2 We selected participants based on a combination of rumination and depression scores because of the exploratory nature of this work. While selecting participants on both rumination and depression scores does not allow us to disentangle the relative contributions of these two constructs, it provides a preliminary insight into whether, and how, these constructs collectively contribute to the CIE with valenced misinformation in depressive rumination. Future research can explore the specific relative contributions of rumination and depression. Due to a clerical error in the screening exercise, four participants were recruited based only on their RRS score.
3 For sake of transparency, we note that we additionally administered the Dysfunctional Attitude Scale (DAS-A; Weissman & Beck, Citation1978) for potential use as a continuous predictor in lieu of the DASS-Dep/RRS grouping. The DAS-A is a 40-item self-report questionnaire that assesses the intensity of dysfunctional attitudes. Analyses using the DAS-A closely mirrored the analyses reported in the following, and are detailed in the Supplemental Online Material 1.
4 The mean inference score is based on the combination of open-ended and rating scale questions, which was the a-priori analysis plan. There was no difference in the results between using the combined inference scores compared to using only the open-ended questions. Furthermore, there were 11 open-ended questions but only 2 rating scale questions, so it is not ideal to base any analysis on rating scale questions only.
5 The effects of group and scenario × group were non-significant after exclusion of participants with recall accuracy < 2; ps > .103.
6 Excluding the 15 participants with retraction-awareness scores of zero, the main effect of group was non-significant, F(1,174) = 3.34, p = .069, MSE = .08, = .02.
7 The task involved participants encoding a set of three valenced (positive or negative) words; individual words were then repeatedly replaced with other valenced words before a final recall test.
8 As this is the first study to investigate the effects of emotional valence on the CIE in depressive rumination, we compared negative scenarios with positive scenarios to increase test sensitivity. Future studies could include a neutral scenario to better ascertain the role of cognitive biases towards negative information, or cognitive biases away from positive information.
9 However, in Experiment 2, two participants, one with a test-day DASS-Dep score of 11 and one with an RRS score of 36, were inadvertently included in the low-DR group. Exclusion of these two participants did not change any of the results, hence they were retained for the analyses.
10 Excluding the seven participants with no retraction-awareness, this effect was non-significant, F(1,55) = 3.30, p = .075.
11 Excluding the seven participants with no retraction-awareness, this effect was non-significant, F(1,55) = 2.37, p = .129.
12 Regarding the question of whether rumination or depression scores were the better predictor of inference scores, we can only speculate: Exploratory regression and commonality analyses reported in the Online Supplementary Materials suggested that in the high-DR group, depression but not rumination scores (negatively) predicted reliance on negative (mis)information. However, we cannot know whether those who scored high on depression but low on rumination measures might attenuate this effect, since this group was not included in our sample.
13 We use the term “effective” here purely as a means of comparison, that is, in terms of relative (as opposed to absolute) effectiveness of a retraction to reduce mean inference scores.