INTRODUCTION. 5 



Using the preceding terminology, the measurements made may be 

 listed as follows: 



(1) Length of the meripodite of leg i, from the edge of its proximal dorsal process 



for articulation with the ischiopodite to the distal edge of the joint on the 

 median line. (Fig. 1, a~b.) 



(2) Length of the meripodite of leg ii. Measurements from the points correspond- 



ing to those given in (1). 



(3) Length of the meripodite of leg iii. Measurements from the points corre- 



sponding to those given in (1) and (2). 



(4) Length of the carpopodite of leg i, from the proximal dorsal edge to the end of 



the distal dorsal process, where this joint articulates with the propodite. 

 (Fig. l,d-e.) 



(5) Length of the carpopodite of leg ii. Measurements from the points corre- 



sponding to those given in (4). 



(6) Length of the carpopodite of leg iii. Measurements from the points corre- 



sponding to those given in (4) and (5). 



(7) Length of the propodite of leg i, from the proximal dorsal process where it 



articulates with the carpopodite to the extreme distal end. (Fig. l,f-g.) 



(8) Length of the propodite of leg ii. Measurements from the points correspond- 



ing to those given in (7). 



(9) Length of the propodite of leg iii. Measurements from the points correspond- 



ing to those given in (7) and (8). 



(10) Length of the cephalothorax, from the tip of the rostrum to the posterior 



margin on the dorsal median line. 



(11) Breadth of the head, between symmetrical points on either side in the cervical 



groove just in front of the two lateral spines. 



The dimensions were taken by means of a pair of fine-pointed dividers. 

 They were then read off on a steel scale graduated to fifths of a millime- 

 ter. All readings were made with the aid of a hand lens. The records 

 were made to tenths of a millimeter, and may certainly be considered 

 accurate to fifths. 



In reducing the material we have determined the correlation between 

 every possible pair of characters. Since we are dealing with eleven 



11X10 



characters, the number of possible pairs is , or 55. From this 



material we are enabled to analyze in detail the phenomena of correla- 

 tion in the eleven characters enumerated. 



In calculating the constants of the frequency distributions we have 

 throughout used Sheppard's corrections for the moments. The work 

 of computation was much facilitated by the use of a large Brunsviga 

 arithmometer. 



