The authors appreciate the time Mizael and de Almeida (2022) took to discuss the implications of our recent manuscript, Effects of Perspective Taking and Values Consistency in Reducing Implicit Racial Bias. Their input adds greatly to our discussion and provides future directions for the use of RFT- and ACT-based interventions in the field of behavior analysis. By identifying some of the limitations of our study, the research community can design additional interventions that provide answers concerning the use of these empirically based procedures to study and reduce implicit racial bias. Our study is an additional effort that other behavioral-oriented researchers have taken in the past (Matsuda et al., 2020).
We would like to specifically clarify four of the queries raised by our colleagues. First, it was suggested that there was only one difference between the groups in the intervention stage. To be clear, the intervention stage in our study was different in many ways for each of the groups. The experimental group completed both values consistency and perspective taking tasks. The values consistency tasks included a sorting exercise adapted from Lejeune and Luoma (2019) as well as writing an autobiography tasks, adapted from Stoddard and Afari (2014). The perspective taking tasks included the anthropologist metaphor, adapted from Hayes and Smith (2005), and completing a writing task that was intended to look at the photo of a Black man and write about a day in his life from his perspective, adapted from Todd et al. (2011). In contrast, the control group was asked to watch a TED talk on changing careers and write about the life of a Black man (without any direction as to perspective).
Second, a question was raised as to whether lower than average empathy levels might explain why some participants were less affected by the perspective taking tasks. To address this concern, it is important to briefly comment on the different measures of derived relational responding (DRR). In the relational coherence model (REC) there are two measures of bias. The first is elaborated and extended relational responses (EERRs) (Barnes-Holmes et al., 2010). These are evoked when participants have time to reflect on their answers while filling out a questionnaire/survey. Consequently, they reflect the social norm and are, therefore, less reliable. This measure is deemed as explicit. In contrast, brief and immediate relational responding (BIRRs) (Hughes & Barnes-Holmes, 2013) is emitted under the pressures of latency and accuracy. While many factors influence one’s perceived empathy, the IRAP is an objective measure of BIRRs. As participants do not have time to reflect on their answers, this measure is deemed as more reliable. BIRRs represent an implicit measure of bias. No correlation was found between explicit and implicit measures that could predict responding, making it difficult to draw any conclusions about how individual levels of empathy were affected by perspective taking tasks. Future research might answer this question by creating an IRAP to measure exactly the same stimuli included in the Toronto Empathy Questionnaire (TEQ) (Spreng et al., 2009).
Third, another concern was that this study employed some parameters not common to recent IRAP studies. The latency requirement of 3,000 ms was used for the acquaintance IRAP and practice blocks of the experimental IRAP only. The actual test trial block criteria were consistent with other IRAP literature (i.e., ≥ 80% correct and median response latency of ≤ 2,000 ms) (Vahey et al., 2015).
A final concern was that the study employed very few trials per block (only 12). In our paper we indicated that each of the six target words was presented once with each of the two stimulus labels to produce 12 trials in each block. In total four different trial blocks containing 12 trial types each. Furthermore, the results for trial-type 1 were said to be so robust for the control group as to indicate a single trial type dominance effect (Finn et al., 2018). A better explanation to these data is found by discussing the differential arbitrarily applicable relational responding model (DAARRE) (Barnes-Holmes et al., 2018. Briefly, the DAARRE model posits that single trial-type effects occur because the properties of the stimuli contained within specific blocks are more coherent within the natural language of the participant. For example, the word “black” describes many negative attributes in the English language (e.g., black mail, black death, etc.). In contrast, the word “white” does not. As a color, it is used primarily to describe positive attributes (e.g., purity, cleanliness, etc.,). In addition, the word “true” is more coherent with affirming responses than “false”. Considering that “white” describes mostly positive attributes, it is not surprising that the data would reveal that Trial-Type 1 in which participants were asked to affirm positive attributes (i.e., “true” + “white” + “careful” or “reliable” or “honest”) demonstrated a single trial-type dominance effect because it is the trial type most coherent with the preexperimental history of the participants.
An explanation to the unexpected results of trial types 3 and 4 is that as experimental participants were asked to sort their values and then act upon them shortly thereafter (i.e., perspective taking writing task), the ACT intervention package exposed the participants’ own prejudicial bias. Support for this hypothesis is found in research by Lillis and Hayes (2007): the probability of behaving in a values consistent manner is enhanced by experiential exercises. It seems plausible, therefore, that both the tasks and their order contributed to this outcome.
We thank again our colleagues for bringing out these important concerns. We hope we have clarified some of the issues raised in their response. As our colleagues mentioned more research derived from RFT technology will be essential to clarify possible interventions of prejudice. We hope to see more research in the future like the one we conducted in this paper (Matsuda et al., 2020; Mizael & de Almeida, 2019).