The main determinants of soil respiration were investigated in 11 forest types distributed along an altitudinal and thermal gradient in the southern Italian Alps (altitudinal range 1520 m, range in mean annual temperature 7.8°C). Soil respiration, soil carbon content and principal stand characteristics were measured with standardized methods. Soil CO2 fluxes were measured at each site every 15–20 days with a closed dynamic system (LI-COR 6400) using soil collars from spring 2000 to spring 2002. At the same time, soil temperature at a depth of 10 cm and soil water content (m3 m−3) were measured at each collar. Soil samples were collected to a depth of 30 cm and stones, root content and bulk density were determined in order to obtain reliable estimates of carbon content per unit area (kg C m−2). Soil respiration and temperature data were fitted with a simple logistic model separately for each site, so that base respiration rates and mean annual soil respiration were estimated. Then the same regression model was applied to all sites simultaneously, with each model parameter being expressed as a linear function of site variables. The general model explained about 86% of the intersite variability of soil respiration. In particular, soil mean annual temperature explained the most of the variance of the model (0.41), followed by soil temperature interquartlile range (0.24), soil carbon content (0.16) and soil water content (0.05).