220 WILLIAM J. FOX. 



are inconstant characters is proven by our own species of Mutilla 

 (sens. lat.). Considering the round-eyed species only (representing 

 Sphcerophthalma), the number of submarginals is variable in the 

 different species. Generally the third subnuirginal, if not distinct, 

 is faintly marked ; it is rarely as distinct as the second. This 

 characteristic occurs in species with the marginal cell truncate, and 

 in some with it acuminate {propinqua, oceola) ; and we find species 

 with two submarginal cells and a truncate marginal (fulvohirta). 

 The same type of mandible seems to exist in Pseudomethoca Cres- 

 sonii and in Sphcerophthalma fenedrata, i.e., bidentate, or in other 

 words, with a tooth before the apex on inner margin. 



It is not the present writer's intention to assert that the form and 

 dentition of the mandibles are entirely valueless as characteristics ; 

 they can be used to some extent in separating the groups into which 

 I have arranged Mutilla, as in the groups hexagona and scrujiea, in 

 both of which the dentition of the $ mandible differs. 



In the present work an almost complete series of the species of 

 America, north of Mexico, has been available, including the types 

 of Blake, Cresson, Ashmead and others, as well as a number of 

 Mexican and South American species. 



The author acknowledges with thanks the loan of valuable mate- 

 rial from Wra. H. Ashmead. 



The Mutillidse is divisible into subfamilies as follows : 



FE3IALES. 



Thorax of 9 i^ot divided, at least not on its upper surface IVIutilliiiie. 



Tliorax of 9 divided into two or more parts Thyniiinse. 



3fALES. 



Tip of abdomen beneath supplied with two slender appendages- • • -Nutilliiise. 

 Tip of abdomen otherwise supplied Thyiininse. 



1, Subfamily Mutillin.e. 

 This subfamily consists of one vast genus Mutilla Linne, of which 

 twelve groups are at present known from the United States. 

 Several of these have been described and given generic rank, but a 

 study of them with a large series shows that such a course is not 

 tenable, as they intergrade to an extent which renders a sharp line 

 of demarkation impossible. Closely related groups will be found in 

 both the tri- and bidentate mandibled series ; thus the group a,ioput> 

 with tridentate mandibles is remarkably close to group oceidentalis, 



