We suggest that a previously conjectured relation between structure functions (SF) for nuclei and nucleons also links distribution functions (df) for partons in a nucleus and in nucleons. The above proposal ensures in principle identical results for SF ${F}_{2}^{A}$, whether computed with effective hadronic or partonic degrees of freedom. In practice there are differences, because of different input for ${F}_{2}^{n}$. We show that the thus-defined nuclear parton distribution functions (pdf) respect standard sum rules. We observe close agreement between moments of nuclear SF, computed in the hadronic and partonic descriptions. Despite substantial differences in the participating SF, we nevertheless find approximately the same EMC ratios in the two representations, as well as reasonable agreement with data. The apparent correlation between the above deviations is ascribed to a sum rule for ${F}_{2}^{A}$. We conclude with a discussion of alternative approaches to nuclear pdf.