A study was made of the possible basis for naturally occurring variations in the number of segments in individual dendritic trees. Distributions of the number of terminal segments have been studied in dendrites from rat, cat, and frog motoneurons, basal dendrites from rat visual cortex pyramidal and non-pyramidal neurons, in rat cerebellar Purkinje cell dendritic trees, and in human hippocampal dentate granule cells. By means of a mathematical model for dendritic branching, it was shown that the variation in the number of dendritic segments can be accounted for by assuming that new branches during neurite outgrowth are formed randomly at terminal segments. The observed terminal segment number distributions could be closely approximated by additionally assuming that branching probabilities decline with increasing number of terminal segments in growing dendrites. The pyramidal neuron group differed significantly from the other neuron groups in such a way as to suggest that this decline is stronger than in the dendrites of other types of neurons. By using literature data on the mean number of terminal segments in rat cerebellar Purkinje cells, measured at different times during early development, an estimate could be obtained of the time-course of the branching probabilities. The branching probability of a terminal segment was found to be in the order of 0.002 per hour in the first 4 weeks postnatal with a 5-fold transient increase in the second week.