Studies examining job knowledge often combine job knowledge tests with work-sample tests and situational judgement tests (Roth, Bobko, McFarland & Buster, 2008). Thus, it is often difficult to isolate differences on job knowledge. For example, Schmitt and colleagues (1996) reported a Black-White d of 0.38 in their meta-analysis (k = 37, N = 15,738), based on a combination of work-sample, job knowledge and situational judgement tests. One meta-analysis that did (in part) isolate job knowledge to estimate an effect size was that conducted by Roth and colleagues (2008). Although the study focused on work samples, these authors performed analyses in which they isolated constructs based on their saturation in work samples. They were able to estimate a d of 0.80 (corrected to 0.82 for measurement error in the work-sample tests/exercises; k = 13, N = 785) for work samples with a high saturation of cognitive and job knowledge. It should be noted that the authors combined these constructs because ‘the two sets of constructs were so tightly related within work sample exercises’ (p. 652). Thus, African-American and White group mean score differences on job knowledge appear to be high, but it is difficult to determine the degree to which the cognitive ability aspect biases this estimate.