VARIATIONS OF GARTER-SNAKES. 17 



body there are 19 rows of scales. If these rows be traced backward 

 somewhat beyond the middle of the body, however, it will be noticed 

 that a row is dropped on either side at approximately the same place, 

 leaving 17. The decrease posteriorly in the number of scale rows, 

 thus described for sirtalis, occurs in all of the garter-snakes, and the 

 number usually given in diagnoses as typical of the different forms 

 (23, 21, 19, or 17) is in every case the maximum number of rows, 

 which, as we have just indicated, always occurs on the anterior half 

 of the body. Furthermore, an examination of large series of these 

 snakes has established beyond question that the decrease in the number 

 of scale rows posteriorly is brought about in all of the forms of garter- 

 snakes by the loss of certain definite rows. On all of the specimens 

 examined with a maximum number of 23 scale rows anteriorly 

 (megalops), which is the largest number that occurs in the genus, the 

 fifth is first discontinued posteriorly, leaving 21, and shortly after- 

 wards the sixth (now the fifth) leaving 19, and finally the fourth, 

 leaving 17. In the specimens with 21 rows (elegans, radix, and 

 hammondi, for example) it is the fifth row that is dropped first to 

 leave 19, and then the fourth to leave 17. In those with 19 rows 

 {sirtalis, eques, sauritus, etc.) the fourth is dropped posteriorly to 

 leave 17, while when the maximum number of rows tends to be less 

 than 19 (butleri, ordinoides, and scalaris) the fourth row drops out 

 first to leave 17 and then the fifth (now the fourth) to leave 15. 



It will be seen at once that the order in which the rows are lost poste- 

 riorly in the different forms is the same as in the form having the maxi- 

 mum number of rows for the genus. In fig. 1 a diagram has been con- 

 structed to show this law of reduction as it would be represented on 

 .a snake having a maximum number of 23 rows anteriorly and a 

 minimum of 15 posteriorly — the extreme range of variation in the 

 genus. An examination of this diagram shows that although the 

 order in which the rows are lost is, when counted from the ventral 

 series each time, as is usually done, 5, 5, 4, 4, the real sequence when 

 expressed in terms of the maximum number of rows for the genus is 

 5, 6, 4, 7. The diagram also illustrates the manner in which the rows 

 are lost, for in the great majority of cases the scales of the row which 

 is dropped become smaller and finally cease, the adjacent rows con- 

 verging to occupy the space. At the point where the row is discon- 

 tinued there is often, however, a rather large scale bearing two keels, 

 indicating that the last scale of the lost row has fused with its 

 neighbor. In case such a fusion occurs it may be either with the 

 row above or below, but the fact that the scales of the row to be 

 dropped usually become smaller toward the termination of the 

 series, and that the keel of this row is usually only on the edge of the 

 large scale, generally denotes clearly which row is realh^ lost. 



