One of the reasons for the weak and inconclusive effect of human capital on growth in existing cross-country studies may be the use of inappropriate specifications that do not account for different channels through which human capital affects economic growth. It has been suggested that both the initial stocks and changes in the human capital stocks have positive growth effects when considered together, while each channel often appears insignificant in isolation. This implies that the effect of human capital is likely to be underestimated in restricted (single-channel) specifications. However, this paper shows that extended (two-channel) specifications might also fail to provide a reasonable explanation for the inconsistent human capital effects in existing growth literature. It seems that the omitted variable bias problem resulting from the ‘incomplete’ specification of human capital in growth regression may not be as pronounced as expected. In addition, we consider some aspects of measurement error and low signal in the changes in human capital.