In crop species, the impact of temperature on plant development is classically modelled using thermal time. We examined whether this method could be used in a non-crop species, Arabidopsis thaliana, to analyse the response to temperature of leaf initiation rate and of the development of two leaves of the rosette. The results confirmed the large plant-to-plant variability in the studied isogenic line of the Columbia ecotype: 100-fold differences in leaf area among plants sown on the same date were commonly observed at a given date. These differences disappeared in mature leaves, suggesting that they were due to a variability in plant developmental stage. The whole population could therefore be represented by any group of synchronous plants labelled at the two-leaf stage and followed during their development. Leaf initiation rate, duration of leaf expansion and maximal relative leaf expansion rate varied considerably among experiments performed at different temperatures (from 6 to 26 degrees C) but they were linearly related to temperature in the range 6-26 degrees C, with a common x-intercept of 3 degrees C. Expressing time in thermal time with a threshold temperature of 3 degrees C unified the time courses of leaf initiation and of individual leaf development for plants grown at different temperatures and experimental conditions. The two leaves studied (leaf 2 and leaf 6) had a two-phase development, with an exponential phase followed by a phase with decreasing relative elongation rate. Both phases had constant durations for a given leaf position if expressed in thermal time. Changes in temperature caused changes in both the rate of development and in the expansion rate which mutually compensated such that they had no consequence on leaf area at a given thermal time. The resulting model of leaf development was applied to ten experiments carried out in a glasshouse or in a growth chamber, with plants grown in soil or hydroponically. Because it predicts accurately the stage of development and the relative expansion rate of any leaf of the rosette, this model facilitates precise planning of sampling procedures and the comparison of treatments in growth analyses.