⎯ The major aim of the research is to analyze the variability of thermal seasons characteristics (onset and length), at 7 stations in Poland during the 40-year-long period of 1971-2010. The analysis comprised the selection of the optimal method for season onset identification and the statistical analysis of the season onset and length characteristics, which was augmented with the interdecadal comparison of statistics using one-way analysis of variance (ANOVA). Also, it was investigated if the season onset type (late/normal/early) induces the season length distribution characteristics. The results show that the variability of the onset of thermal seasons is most pronounced in the case of summer and early spring. Summer also shows the tendency to occur faster, whereas autumn tends to start later. The interdecadal change of the thermal seasons length is clearly visible in the case of summer which length systematically increases and winter where the trend is reversed at most of the stations. The research also confirmed that for winter, early spring, and summer, the onset type clearly determines the distribution of the season length. Other seasons either show little variability or the differences appear only at some stations.