The paper tests whether technological change has been neutral in Canadian manufacturing industries, using a system of translog cost share equations for 1962 through 1986. The model features two classes of labor treated as distinct inputs. Tests rejected homotheticity in all industries. Hicks-neutrality was also rejected in 16 of 18 industries. The most common pattern of non-neutral technical change was a bias away from blue-collar workers. Formal tests for skill-neutral innovation rejected the hypothesis in 10 industries in favor of skill-using technical change. The results suggest that in studies of Canadian manufacturing aggregation across labor inputs is inappropriate.