The intervening task method: Implications for measuring mediation
This research challenges the assumption that the effects of an independent variable can be assessed with an unrelated filler task preceding the focal task, a method often used to study the mechanisms underlying stereotype threat. For example, poor performance on a Stroop task following a stereotype threat manipulation has been offered to suggest that stereotype threat reduces self-regulatory abilities (e.g., Carr & Steele, 2010; Inzlicht, McKay, & Aronson, 2006). Instead, it is argued that performance on unrelated filler tasks might vary as a function of their perceived relevance to the primary task in the experiment. Experiment 1 involved female undergraduates (N = 64) who were told they would answer math problems from the GRE. Participants were told either that gender differences in performance typically occur (stereotype threat) or do not (control) on this test. Participants were then informed that the experimenter ran out of test copies and were asked to complete another task while copies are being made. Participants were told that the new task was either unrelated to math ability (Stroop performance stereotype irrelevant) or was a measure of processing efficiency and predictive of mathematical ability (Stroop performance stereotype relevant). When told that the Stroop task was described as irrelevant to the stereotype, participants in the stereotype threat condition performed more poorly than participants in the control condition, as has been found in previous studies. Specifically, these participants were slowest to name aloud color words printed in incongruent colors (e.g., reading aloud red when the word is printed in blue ink). However, when the task was described as stereotype relevant, the effect of stereotype threat disappeared. Also, performance on compatible Stroop trials (e.g., simply naming the ink color of a series of XXXXs) was faster in the conditions where participants were told that the task was stereotype relevant. This might suggest that participants who were told that the Stroop performance was stereotype relevant were more motivated to perform well on the task compared with participants who believed the task was stereotype unrelated. Experiment 2 involved female undergraduates (N = 60) who completed a procedure similar to Experiment 1 except the Stroop was replaced with a prosaccade task. In this task, participants were asked to stay visually fixated on a central crosshair on a monitor but to look at cues that randomly appear on the right or left side of the screen. An arrow then appeared in the same location as the cue, and participants identified the orientation of the arrow (up, right, down, left) as quickly and as accurately as possible. As in Experiment 1, performance on the prosaccade task was described as either relevant or irrelevant to math performance. Results showed that participants were slowest to indicate the orientation of the arrow if they were under stereotype threat and believed that performance on the prosaccade task were irrelevant to math performance. These results show that performance on a presumably unrelated test are not necessarily pure measures of intervening processes. Participants expectations about the relation between performance on such tasks and the primary task can affect motivation on the task. With regard to research on stereotype threat, these findings suggest that poorer performance on tasks following stereotype threat manipulations (such as the Stroop task) do not necessarily demonstrate depleted capacity or reduced self-regulatory ability. Instead, it may simply show that individuals operating under stereotype threat disengage from tasks they they view as irrelevant to the stereotype with which they have been threatened.