A highly inbred line of Drosophila melanogaster was subdivided into replicate sublines that were subsequently maintained independently with 10 pairs of parents per generation. The parents were randomly sampled for 19 'unselected' sublines and artificially selected for high or low abdominal or sternopleural bristle number for 12 'selected' sublines (with 3 replicate selection lines/trait/direction of selection). Divergence in mean bristle number among the unselected sublines, and response of the selected sublines to selection, are attributable to the accumulation of new mutations affecting bristle number. The input of mutational variance per generation, VM, can be estimated from the magnitude of response or divergence, assuming neutrality of mutations affecting the bristle traits. We reared unselected lines at generations 222 and 224, and selected lines at generations 182-184 of mutation accumulation at each of three temperatures (18 degrees C, 25 degrees C, 28 degrees C), and estimated the mutational variance common to all environments and the mutational variance from genotype x environment interaction. For sternopleural bristle number, the mutational interaction variance was 26% of the mutational variance common to all temperatures, and the interaction variance was due to temperature x line interaction. For abdominal bristle number, the mutational interaction variance was 142% of the mutational variance common to all temperatures, and the interaction variance was due to interactions of temperature x line, sex x line, and temperature x sex x line. It is possible that segregating variation for bristle number is maintained partly by genotype x environment interaction, but information on the fitness profiles of mutations affecting bristle number in each environment will be necessary to evaluate this hypothesis quantitatively.