RESPONSES TO LIGHT OF THE ROACH. 373 



Three kinds of glass discrimination boxes were used. Box 

 number one was 25 centimeters long, 8 centimeters wide and 8 

 centimeters deep. By means of transverse glass partitions, this 

 box was divided into three compartments. One partition was 

 stationary, shutting off an end compartment 15 centimeters long. 

 The other partition was adjustable, thus making it possible to 

 vary the lengths of the other two divisions. In these experiments 

 the middle compartment was usually a little less than 4 centi- 

 meters long, thus making the other end compartment a little more 

 than 15 centimeters in length. The middle compartment com- 

 municated with each of the others by doors which faced each 

 other. Each door was a square 2.5 centimeters long. By means 

 of a special hood and an opaque screen the shape of the transverse 

 partition, one of the end compartments was transformed into a 

 dark chamber. The animal forming the subject of the experi- 

 ment was placed in that end of the narrow middle compartment 

 which was most remote from the exits. When the animal reached 

 the opposite end of the narrow passage it had an opportunity to 

 select which of the doors it would enter. 



After reading Szymanski's paper, which appeared soon after 

 I had begun my experiments, I decided to use a box similar to 

 his. This decision was reached, not because the box described 

 in the above paragraph proved unsatisfactory, but because, on 

 account of the greater freedom, it took longer to perform an 

 experiment with the box I designed than with the one designed 

 by Szymanski. The fact that box number one permitted a roach 

 in one compartment to select which of two others it would enter 

 while Szymanski's box simply permitted it to enter the one 

 towards which it was moving or else remain in the one where it 

 was caused me, at first, to consider this box superior to Szyman- 

 ski's; but, after much thought, it was decided that box number 

 one was not sufficiently superior to Szymanski's to offset the 

 advantage of shortening the time necessary for conducting an 

 experiment. The shorter the time required for an experiment 

 the less fatigue interferes with the reactions. Then, too, there 

 are other reasons for desiring to shorten the time of intimate con- 

 tact with these stench-engendering creatures. Box number two 

 was 30 centimeters long, 3 centimeters wide and 8 centimeters 



