This paper analyses the convergence of regional products in Finland using two different data sets. Firstly, β- and σ-convergence was estimated for the 12 Finnish provinces during 1960"“94. Convergence was found to be strong in 1960"“80, but after 1980 regional disparities started growing again. Secondly, a similar study was conducted for the 88 small-scale subregions in 1988"“94. As with the provinces, the subregions' relative growth performance and cross-sectional convergence dynamics were evaluated using Markov chain transition matrices. No clear evidence for σ- or β-convergence was found here, but the dynamic analysis revealed a rapidly evolving distribution of gross regional products. Thus the type of regional classification and method used can markedly affect the results obtained in a convergence study. (JEL O4, R1)