In the first 25 generations of his classical mutation accumulation experiment, T. Mukai estimated a large rate of early linear decay for the relative viability of Drosophila melanogaster chromosome II (ΔMII = 0.004). Mukai forced through zero the regression of viability decline on generation number, but it has recently been shown (Fry, 2001) that a similar decline (ΔMII = 0.006) is obtained from unforced regression even if generation 32 instead of generation 25 (whose validity has been questioned) is included. We show that, from the perspective of the whole long-term experiment, it is hard to decide up to which generation viability can be considered to decline linearly. Depending on this decision, and on whether or not the regression is forced through the origin, very different estimates are obtained. Furthermore, the particular behaviour of the lines used as control suggests that they could have been different from the remaining lines at the beginning of the experiment, and casts doubts on the adequacy of a forced regression. Estimates from the linear unforced regression (ΔMII = 0.011) or from the linear term in a quadratic unforced regression (ΔMII = 0.001) are very different. The data fit both models very well, and the choice between them should be based on biological grounds.