In this paper we revisit the work of Jones (1995a) which provided strong evidence against the empirical validity of AK-type models that predict increasing output growth rates as a result of permanent movements in the physical investment rate. For that purpose, we employ recently developed unit root tests with good size and power which are also able to accommodate the existence of a structural break in the data for 26 OECD countries over the period 1950-1992. Overall, the analysis of deterministic and stochastic trends in output growth and investment rates (in total physical investment, and in producer durables and total structures) do not render broad support for the empirical validity of AK models. This result appears to be confirmed by the estimation of autoregressive distributed lag growth models which consistently render insignificant long-run coefficients on the investment rates.