Growth refers to an increase in dimensions with time and is implicit in the expected continual increase in tree dimensions. Tree diameters, however, could decrease during the growing season due to water depletion. Annual negative growth measurements are usually attributed to human error and not to other physiological or physical processes. Although seasonal and diurnal fluctuations of diameter have been well documented, perennial decrement of diameter has not been the focus of physiological research. The aim of this work was to analyse the potential causes of decrease in annual diameter growth related to tree mortality due to self-thinning in Nothofagus pumilio forests and to quantify the variations in water depletion of the tree trunks. Some trees did present negative annual diameter increments associated with a water content decrease in the trunks (77 per cent in live trees compared with 56 per cent in recently dead individuals), which produced a contraction (more than 8 per cent of the initial diameter) in the wood and the bark. Trees could survive during 2-5 growth seasons with continual decreases in their diameters (14 per cent, standard error 5 per cent of the trees in the studied stand) until the water content reached a limit where mortality resulted. Therefore, the occurrence of data showing a diameter decrease in successive forest inventories may be due to physiological and physical processes in the natural dynamics of the stand, and not exclusively be explained away as the results of human measurement errors.