BIRDS OF NORTH AND MIDDLE AMERICA. 25 



sive and therefore may require still further restriction. As commonly 

 understood and accepted, the two supposed families are clearly purely 

 artificial, and the arbitrary- line that has usually been drawn between 

 them is manifestly far out of place, the Tanag-rid{\^ having- been made 

 to include forms (those mentioned above' ) which are unquestionably 

 fringilline in their relationships. 



In the absence of any knowledge concerning the internal structure 

 of a large proportion of the genera comprised in the two groups any 

 dividing line must necessarily be more or less arbitrary; but I feel 

 sure that by shifting its position as here indicated the two groups 

 become much more natural, since they are now susceptible of fairly 

 definite characterization, whereas until this was done their intelligi- 

 ble diagnosis was simply impossible. I am not at all sure Imt that 

 still further subdivision, at least of the Tanagridse, would better 

 express the facts of relationship, since even now, with their respective 

 limits certainly more correctly drawn, each of the two g-roups contains 

 forms extremel}^ different in their general appearance, structural 

 details, and habits.' However, this question as to whether the Frin- 

 gillida? and Tanagridaj are really distinct family groups or not, and if 

 they are, where the line between them should >)e drawn, is one which 

 can not now be exactly determined. 



While, as above stated, the reference of the genera Buafrremon^ 

 Arrennon., Pitylus^^ and Saltator to the Fringillida? renders possible an 

 intelligible diagnosis of the two supposed families, it does not in the 

 least lessen the difficulty of defining the genera or of arranging them 

 into definite subordinate groups. This is indeed a matter so extremely 

 difficult that after repeated, patient, determined, and prolonged 

 attempts I must confess my inability to solve the problem. It is 

 very evident, according to my judgment, that Dr. Sharpe's so-called 

 subfamilies, Coccothraustinaj, Fringillinje, and Emberizinse, are 

 unnatural groups, especially the first; certainly Geosjyiza, Guiraca, 

 Spermophila, Cardinalis^ etc., are not at all closely related to Cocco- 

 thraustes^ He^periphona^ Eophona, PycnorhahiphHS^ and Mycerohas^ 

 which together form a very distinct group, though evidently closely 

 related to, if not directly connected with, the group which Dr. Sharpe 

 designates as his '' Subfamily Fringillina3." The latter is another very 



^The only reasonable doubt pertains to the genera "PiVy/us" and Saltator. 



'^Compare the heavily built, crass-billed, short-legged, and arboreal true grosbeaks 

 (Coccothrausteae) with the slender, small-billed, long-legged, and terrestrial grass 

 buntings (genera Ammodrnmus, Passerculus, etc. ) on the one hand, and the broad- 

 billed, short-tailed, and long-winged frugivorous Euphonite (genera Eapltonia and 

 Chlorophonia) with the slender-billed, long-tailed, insectivorous genera Tachyphonus, 

 Nemosia, etc., on the other. 



^In the wider sense, as these genera are given by Dr. Sclater. 



