of treatment effects increased with treatment duration, and because one of the variables which originally had been intended for use as a concomitant control was affected by treatments. No satisfactory form of analysis for dealing with these complications was reached, and the problems raised seem worth re-examining. The weights of hind-leg muscles of rats, one intact (v), one denervated (u), were observed at 4, 8, and 12 days after operation followed by treatment twice per day with one of four drugs (of which D was merely control saline). There were four randomly chosen rats in each of the twelve groups. Initial (x) and final body weights (y) were also observed. The objective of the experiment was to determine whether the drugs delay atrophy of denervated muscles. Apparently it had been hoped that drugs would not affect intact muscles, so that they could be used as concomitant control. But inspection of the data revealed that drugs seriously affected both intact muscles (v) and final body weights (y). The question was then asked whether a drug might delay the rate of decay of denervated muscle relative to the shrinkage to be expected if there is no drug X denervation interaction. The common empirical definition of null interaction for additive effects is, however, inappropriate and irrelevant: shrinkages caused by drug and denervation cannot possibly be additive when measured in grammes or as proportions, and there is no hypothesis to define how the two effects should superimpose and thence to indicate a transformation by which null interaction would be additively expressed.