Ineffective Attribution Testing: An exploration of individual differences in cognition between Liberals and Conservatives

Stephanie Nail   (Stanford University)

Abstract: Previous literature has suggested that there are underlying differences in cognitive processes between liberals and conservatives. Conservatives are said to have more structured cognitive patterns compared to liberals in addition to being less able to inhibit their responses on behavioral inhibition tasks (Amodio et al. 2007). These differences may manifest in implicit data collected by using the Implicit Association Test (IAT) and may prove to have both methodological and substantive implications. This paper tests the conjecture that conservatives have less variability in their behavioral response times to IAT tasks, as well as longer response times and higher error rates compared to liberals. Using existing non-political IAT experiment data, this paper investigates the idea that conservatives exhibit these characteristics only after establishing a baseline behavioral pattern on the first block of the experiment. Because they exhibit more structured cognitive patterns, they should make more errors, have less variation, and slower response times in subsequent blocks, after establishing a behavioral pattern in the first block of the IAT. There are two ways that conservatives being slower could generate methodological and substantive problems for IAT studies. The first is a methodological problem: conservatives are not slower overall, but they are slower after they are forced to change the behavior pattern they established previously. This creates a problem when comparing block response times within respondents. More specifically, this could manifest as a methodological problem when scholars do not use political ideology as a factor when they analyze the results of an IAT. The second is a substantive problem: subtracting within respondents to calculate the IAT’s “D” score will be hazardous to the results. If conservatives are slower on the last three blocks than the first block, the subtraction of response times between blocks will lead to a high number – not because conservatives are biased against particular word pairs, but because they are slower in the later blocks. This poses validity issues – conservatives will look worse (i.e. more prejudiced or more biased) because of the methodology of the test and individual cognitive differences compared to liberals, not because they are biased against the substantive concept tested. The results explain if underlying individual differences in cognitive processing between conservatives and liberals have methodological and/or substantive implications. If true, without the inclusion of political ideology as a control, scholars may falsely attribute substantive findings to conservatives when the results are actually due to a methodological artifact.

View Poster in a New Tab