The aim of this paper is to analyse Solow's model, introducing the consideration that steady state labour productivity (in efficiency units) and its determinans are not constant values, for a country over a given period of time. They are, in fact, time series with unit roots. First, the paper shows that Solow's model can be interpreted as an error correction model and it could be consistent with the stochastic nature of the variables. Secondly, implications about integration and coinstegration of the relevant series are tested. We use data coming from four countries over the period 1960-88. The error correction mechanism implied by Solow's model never appers to have been operative: convergence of current productivity towards its (stochastic) steady state path does not emerge in any considered case.