Reducing Implicit Stigmatizing Beliefs and Attitudes Toward Depression by Promoting Counterstereotypic Exemplars

ABSTRACT

Stigma prevents people with depression from seeking professional help, and effective strategies to reduce implicit stigmatizing beliefs and attitudes toward depression are needed. We developed stepwise strategies to promote counterstereotypic exemplars of people with depression and conducted an experiment (N = 105) with a pretest–posttest control group design to examine effects of these strategies. Our results showed that implicit stigmatizing beliefs and attitudes were reduced more in the experimental group, in which the counterstereotypic strategies were implemented in addition to the provision of education regarding basic knowledge of depression, than in the control group, which received only education.

Notes

Greenwald et al. (2009) reported that the average effect size regarding the predictive validity of implicit beliefs and attitudes was r̅ = .24 and that the predictive validity was not greatly impaired in the studies of socially sensitive topics. Although Oswald, Mitchell, Blanton, Jaccard, and Tetlock (2013) conducted another meta-analysis using broader inclusion criteria than those used by Greenwald et al. (2009) and reported a smaller effect size (r̅ = .15), both effect sizes were large enough to predict complex intergroup behavior in realistic situations (Greenwald, Banaji, & Nosek, 2015). Considering these findings and discussions, we assumed that implicit beliefs and attitudes are predictive of actual intergroup behavior.

We conducted the BIATs and presented the slideshows (mentioned later in the Procedure subsection) on Hewlett-Packard (Intel Core i5 processor) laptop computers equipped with 15.6-in. screens and Windows 8 operating systems.

In designing the stepwise counterstereotypic strategies, we referred to previous studies that conducted multiple tasks to reduce implicit intergroup biases (e.g., Asgari et al., 2012; Devine et al., 2012; Interventions 4, 6, and 7 in Lai et al., 2014). We intended to affirm the counterstereotypic associations by conducting the mental imagery tasks so that participants in the experimental group could vividly imagine the two counterstereotypic characters introduced in the second slideshow.

Previous studies that examined the effects of counterstereotypic strategies frequently instructed the participants to keep counterstereotypic images in mind while conducting IATs (e.g., Blair et al., 2001; Lai et al., 2014). Following these previous studies, we used similar instructions for the BIATs at the posttest.

As summarized in Table 4, we reported estimated true scores (i.e., scores without random error components) of standard deviations in addition to observed standard deviations when these true scores were applicable. We estimated a true score for each measure by multiplying the square root of a reliability coefficient and an observed standard deviation using Equation 2 described in Trafimow (2014). Furthermore, we calculated Cohen’s d scores by using these estimated true scores of standard deviations, to handle issues of low reliabilities and to examine strengths of stigma and effectiveness of interventions rigorously.

Following previous studies using IATs (e.g., Lai et al., 2014; Monteith & Pettit, 2011; Sriram & Greenwald, 2009), we assumed that the 0 points of D-scores were equivalent to lack of biases. However, a recent study by Blanton et al. (2015) raised a skeptical view of this assumption. The issue regarding the interpretation of D-scores is further discussed in the Limitations and Future Directions subsection of the Discussion section.

We did not calculate effect sizes for interaction of group and time point here. Although such effect sizes are conventionally calculated in frameworks of analyses of variance, such effect sizes only tell us how large the interaction effects are and cannot provide detailed information regarding the effectiveness of manipulations (e.g., how large stigma-reduction effects were within each group, which manipulations had larger stigma-reduction effects). Instead, with the editorial of the Basic and Applied Social Psychology (Trafimow & Marks, 2015) in mind, we intensively reported descriptive statistics and effect sizes regarding the effectiveness of manipulations in a stepwise manner.

Although we stated that reliability coefficients in the present were low according to the usual cutoff criteria (.70), it is uncertain if these coefficients could be regarded as lower than that of previous studies administering IATs in longitudinal experiments (e.g., Devine et al., 2012). These previous studies certainly reported grand reliability coefficients at pretests, but they did not report reliability coefficients in each condition at follow-up assessments. Most previous studies and meta-analyses concluding that IATs are more reliable than other implicit measures (e.g., Bosson, Swann, & Pennebaker, 2000; Hofmann, Gawronski, Gschwendner, Le, & Schmitt, 2005; Nosek et al., 2007) also did not focus on reliability coefficients after administering kinds of interventions. It seems to be a challenge for future research using IATs to accumulate and integrate findings regarding reliability coefficients in longitudinal experiments.

Additional information

Funding

This work was supported by JSPS KAKENHI Grants 15J10159 and 16H03741. Portions of the present study were presented as a poster at the Second International Convention of Psychological Science in Vienna, Austria (March 2017). We thank our colleagues at Nihon University who supported the present study: Prof. Takashi Oka who gave us helpful comments regarding the procedure of the experiment and Asami Kudo, Mizuto Yoshii, Sawa Maruyama, and Takahiro Tomita who conducted the experiment. We also thank Editage (www.editage.jp) for English language editing.