We ask whether the role of employer learning in the wage-setting process depends on skill type and skill importance to productivity. Combining data from the NLSY79 with O*NET data, we use Armed Services Vocational Aptitude Battery scores to measure seven distinct types of pre-market skills that employers cannot readily observe, and O*NET importance scores to measure the importance of each skill for the worker's current three-digit occupation. Before bringing importance measures into the analysis, we find evidence of employer learning for each skill type, for college and high school graduates, and for blue and white collar workers. Moreover, we find that the extent of employer learning – which we demonstrate to be directly identified by magnitudes of parameter estimates after simple manipulation of the data – does not vary significantly across skill type or worker type. Once we allow parameters identifying employer learning and screening to vary by skill importance, we find evidence of distinct tradeoffs between learning and screening, and considerable heterogeneity across skill type and skill importance. For some skills, increased importance leads to more screening and less learning; for others, the opposite is true. Our evidence points to heterogeneity in the degree of employer learning that is masked by disaggregation based on schooling attainment or broad occupational categories.