Parameter assessment in site index modelling has proved useful prior to model specification. A base-age invariant model that supports polymorphism and variable asymptotes was fitted for Mediterranean Maritime pine in the inland part of Iberian Peninsula. A generalized algebraic formulation of Hossfeld's model was found to be the most appropriate model in terms of biological consistency and fit statistics throughout the range of the data. Models were fitted to data from stem analysis and permanent sample plot measurement using a dummy variable approach and a continuous-time autoregressive structure to correct for autocorrelation. Model performance was evaluated using Jack-knife techniques applied to each growth series. The selected model was used in a regional-based comparison of growth patterns. The parameter identified as 'global' in the first part of the study was expanded with dummy variables to assess the differences among regions. The stands located in the Northern Plateau, the Central Mountain Range, and South Eastern Mountains, exhibited different growth patterns from one another. However, these regions show a similar growth pattern to stands located in their origin of dispersion (Iberian Mountain Range). Consideration was also given to other factors which might lead to variation, such as environmental conditions. © 2007 Elsevier B.V. All rights reserved.