1. _Signaticollis_ [F] × _diversa_ [M] bred at an average temperature of 80º F. by day and 75° F. by night, gave two groups in about equal numbers. The first (49) was pure _signaticollis_ and bred true. The second (53) was of an intermediate type, which on being bred together gave the typical Mendelian result--1 _sig._: 2 _intermediate_: 1 _div_. 2. Next, as the account originally stood in the published paper, we are told that _sig_ [F] × _div_ [M] bred together at a day-temp. average 75° F. and night average 50° F. gave an _intermediate_ only, which subsequently produced a normal 1:2:1 ratio. The two crosses were repeated eleven times with identical results. In a further experiment (3) _signaticollis_ [F] × _diversa_ [M] were bred under the same conditions as those used in expt. (1). They again gave _sig._ and intermediates as before in fairly equal numbers. The _sig._ as before bred true, and the intermediate gave 1:2:1, all exactly as in expt. (1). In expt. (4) _the same parents used_ in (3) were again mated under conditions of expt. (2) at the lower temperature, and this time gave _signaticollis_ exclusively, which bred true for four generations. This experiment was repeated seven times with uniform results. Diagrams are given representing all these histories in graphic fashion. From these observations, Tower concludes that the determination of dominance, and the ensuing type of behaviour, is clearly a function of the conditions incident upon the combining germ plasms. It will be observed that expts. (1) and (3) gave identical results but (2) and (4), though much the same conditions were applied, are at variance, for (2) gave all intermediates, while (4) gave all _signaticollis_. In _Amer. Nat._, XLIV, 1910, p. 747, Professor T. D. A. Cockerell commented on this paper of Tower's and pointed out that there must be an error somewhere, for when he discusses these experiments Tower speaks of (2) and (4) as confirming each other. To this Tower replied[12] that there had been a mistake. He states that in preparing the paper "certain minor experiments were taken from a larger series and combined to illustrate a general point in the behaviour of alternative characters in inheritance," and that expt. (2) was introduced inadvertently in place of another which he desires to substitute. In this, which I number (5), _signaticollis_ [F] × _diversa_ [M] from exactly the same stocks as those used in (1), were mated at the lower temperatures specified for (2), day average 75° F., night average 50° F. These gave all of the _signaticollis_ type with a narrow range of variability, which bred true, in some cases to F_{6}. Tower says he has repeated this experiment six times with identical results.