Fluctuating asymmetry (FA), small random differences between left and right, has been extensively used as a measure of individual quality, though its usefulness in that respect is controversial. Whether FA is heritable has implications for sexual selection theory and for its usefulness as an indicator of stress. Heritability (h2) of FA is, however, difficult to estimate precisely and reliably. Here we report h2s of FA for two pupal traits in the speckled wood butterfly (Pararge aegeria). We used a restriction error maximum likelihood (REML) analysis in combination with a jackknife procedure to analyse a large mixed offspring–parent/half-sib/full sib data set. A five-generation selection experiment provided a second set of narrow sense h2s. Narrow sense h2s were not significant and on average - 0.029 (REML-analysis) and 0.031 (selection experiment) for the pupal segment covering the fore leg (LEG) and 0.057 and 0.004 for a SPOT on that segment. Estimated percentage dominance variances were 0.057 (LEG) and 0.027 (SPOT) and not significantly different from 0. The h2 estimates had been slightly increased by cage effects. Average FA for LEG after five generations of selection were higher in the high lines than in the low lines, and the control lines were in between. No difference in FA between lines was found for SPOT. Although differences between lines were not significant, a slight h2 (<3% ) for LEG could not be excluded. The genetic effect was, however, small compared with the effect of foodplant quality. Larvae grown on foodplants that were not watered enough for good growth showed significantly higher FA for LEG, but not for SPOT, compared with larvae grown on good foodplants.