This paper examines how primitive skills associated with occupations are formed and rewarded in the labor market over the careers of men. The objective task complexity measurement from the Dictionary of Occupational Titles enables a more direct look into the primitive skills of workers. I show that the optimal choice of task complexity is a linear function of unobserved skills, worker characteristics, and preference shocks, which implies that the observed task complexity is a noisy signal of underlying skills. Using career histories from the NLSY79, the growth of cognitive and motor skills as well as structural parameters are estimated by the Kalman filter. The results indicate that both cognitive and motor skills account for a considerable amount of cross-sectional wage variation. I also find that cognitive skills grow over careers and are the main source of wage growth; this pattern is particularly pronounced for the highly educated. In contrast, motor skills grow and contribute to wage growth substantially only for high school dropouts.