The aim of this paper is to test for the presence of dualism in a standard wage regression. The disparity in wages between primary and secondary workers, according to labour market segmentation theory, is not provided by worker characteristics, but rather by job characteristics. A standard way to assess this situation is by looking at the estimated coefficients in a standard regression for comparable workers across different labour market segments. In an attempt to avoid arbitrary modelling choices, we deploy mixture regression methods which allow for endogenous determination of the number of existing labour market segments. Using Italian data, our modelling strategy outlines stark differences in returns to human capital between homogeneous workers in different markets. Thus, future policies should consider these findings when implementing labour market measures.