Diversity in the Scutes of Chelonia. 73 



Finally, in tlie abnormalities we have to do not merely with 

 ■ supernumerary scutes but with many instances of turtles with less 

 than the typical number of elements, which cases it would be diificult 

 to expilain primarily by reversion. If turtles may have less than the 

 typical nimiber (even to only three scutes in each dorsal series, in 

 'No. 35), otherwise than by reversion, why may not they have moi^e 

 in the same way ? 



Summary of Part III, 



1. The experiments in incubation are suggestive of the possible 

 significance of external conditions in the causing of abnormalities 

 of scutes. 



2. I can atribute to supernumerary scutes no value for deduc- 

 tions as to the phylogeny of turtles. 



3. AMiether or not atavism may have a remote or indirect connec- 

 tion with the abnormalities (for which I see no evidence) I do not 

 regard individual scutes as atavisms. 



4. Many of the abnormalities are clearly not atavisms. 



Note. 

 As my conclusions in regard to the significance of abnormality of 

 scutes is in opposition to that with which IsTe^vmann starts out, and 

 may, therefore, seem inimical to his hypothesis, I think it proper 

 to state that the value of ISTewmaun's view of the evolutionaiy history 

 of the carapace appears to me to l)e largely independent of the ques- 

 tion of atavism since the more essential features Avere really based on 

 comparative anatomy, not systematic atavism. He supposes fourteen 

 primitive series of scutes; of the five not present in most turtles, 

 only one was found in his ahnornudities (interplastrals), while three 

 were represented by normal series of certain species (interpl astral 

 and supramarginals, two) and another only by normal scutes of the 

 tail of Chelydra. My obsen^ations might, it is true, show in abnor- 

 mality representatives of the supramarginal series, but, just as well, 

 they would give reason to infer primitive division of neurals and 

 the primitive presence of a pair of siihnarginal series, neither of 

 which inferences would be in hamionv with his view. Hence the 



