110 THE AMERICAN NATURALIST [Vol. LIV 



studied by following the inheritance of three factors in 

 that region simultaneously. By a series of extensive 

 counts of this sort the nature of the linkage in each indi- 

 vidual section of the first chromosome has been studied, 

 and found to be linear. The data in Table II are derived 

 from an experiment which involves less extensive num- 

 bers than these, but illustrates to better advantage the 

 linear behavior of all parts of the chromosome at once. 

 These data are taken from Muller's cross of flies hetero- 

 zygous for twelve mutant sex-linked factors. The re- 

 sults for six of these factors— those scattered most 

 evenly along the chromosome— are shown in the table, 

 which gives the number and per cent, of separation be- 

 tween every one of these factors and each of the other 

 five. It will be seen that it happened that in this partic- 

 ular experiment, for all per cents of separation below 23, 

 the per cent, of separation between any two factors was 

 exactly equal to the sum of the per cents of separation of 

 each from a third factor lying between them, whereas for 

 factors less closely linked, the larger per cent, was less 

 than the sum of the other two by an amount varying 

 closely with the size of the large frequency itself. The 

 data obtained in this same experiment for the three 

 factors y, w and bi are given separately in Table III, in 

 order that they may be compared to better advantage 

 with the non-linear relation for these factors which 

 Castle claims, as a result of his combination into one map 

 of the results of separate experiments. Whereas Castle 

 obtained a triangular figure to represent the three fre- 

 quencies (y w 1.1, w bi 5.3, and y bi 5.5) it is seen that in 

 this experiment, where all three were followed at the 

 same time, an exactly linear relationship was obtained 

 (ywl.7,wbi3.8,ybi5.5). An experiment of Sturtevant's 

 involving just these three factors is shown in the same 

 table (III) ; here too the relations are entirely linear. 

 In like manner the values obtained in the 12-factor ex- 

 periment for the loci of y w and A are given in Table 

 IV (y w 1.7, w A 1.4, y A 3.1) to be compared by the 



