160 HAROLD H. PLOUGH 



the Fi females were allowed to lay for ten days. No. 6 was run 

 in conjunction with another experiment. The characters used 

 were star-black-curved. The data given in No. 2 represents the 

 summaries of the series of counts from flies raised in tubes, the 

 parents being shifted every two days instead of remaining in 

 one bottle for ten. For most of the temperatures tested, a care- 

 fully regulated refrigerator and a large and small electric incu- 

 bator sufficed. The test summarized in No. 4, involving a 

 temperature of 17.5°C. was made by placing the bottles in Dr. 

 A. M. Banta's artificial cave at the Carnegie Institution, Cold 

 Spring Harbor, Long Island. Here the temperature remains 

 nearly constant for long periods, and was observed to vary less 

 than 1°C. during the period of the test. I am indebted to Dr. 

 Banta for the opportunity to use the cold-room during the 

 summer of 1916. No. 5 represents the average of all the controls 

 with the dextrose total added, and is introduced to make the 

 table complete. The temperature was subject to a variation 

 of about 2°C. in either direction. The summaries for successive 

 temperatures are shown in table 8. 



The results given in the column at the left are brought out 

 with great clearness when plotted in the form of a curve. For 

 reasons already discussed, the percentages of the shortest re- 

 gion — black-purple — are probably more accurate, and there- 

 fore these alone are used for the curve. The actual percentage 

 given in No. 6 cannot be used, for obvious reasons. Because 

 of the fact that the smaller region shows no significant increase 

 over the control, we are justified in continuing the temperature 

 curve for 22°C. through the 27°C. point. 



In constructing the curve, the value of the corresponding 

 control must be taken into consideration. It has already been 

 noted that sibs from the same Pi parents show very little varia- 

 tion in percentage of crossingover, while cultures made up at 

 different times may vary as much as three units for a cross- 

 over percentage of this size. The curve will probably be more 

 accurate, therefore, if the temperature value is increased or 

 decreased in the same proportion that its corresponding control 

 shows an increase or decrease over the mean value for the con- 



