Finch- or rather Siskin-like appearance other characters which reveal these birds as 

 typical Drepanididse : — (1) The tongue, which, of the length of the bill, forms a typical 

 brush-tube ; it is as typically developed as in the most intensified Drepanididse. 

 (2) The arrangement of the bones of the palate conforms with that of the more slender- 

 billed forms, namely, the maxillo-palatines are rather long and slender, the inter- 

 palatine spurs form a pair of narrow vertical plates, the transpalatines are very little 

 broadened, but the right and left halves are fused in the middle into one plate, together 

 with the vomer. (3) The nostrils have a small but distinct soft operculum, which, 

 however, does not close the narial opening, which is long-oval; the general configu- 

 ration closely resembles that of Oreomyza, while the differences from Fringillidse, if 

 examined side by side, are obvious. 



We have to consider the following hypotheses : — 



I. All the Sandwich Island birds in question are Drepanididge. The most central, 



or least modified, are the small-sized genera Oreomyza, Loxops and Chryso- 

 mitridops, and Himatione. Thence have sprung in two divergent lines — 



1. The long- and slender-billed ultra-Drepanine forms with long tubular 



brush-tongues : Hemignathus, Viridonia, Vestiaria, Drepanis. 



2. The thick-billed Fringilloid genera with short, more fleshy, reduced 



tongues: Psittacirostra, Pseudonestor, Loxioides, Bhodacanthis, Chlo- 

 ridops. 



II. The majority of the birds are Drepanididse, while the thick-billed forms without 



tubular tongues are Fringillidge. The relationship of these thick-billed genera 

 does, however, not lie with the palsearctic Clitoris, e. g. kawarahiba, nor with 

 any of the Coccothraustinre, least so with the genus Geospiza from the Galapagos 

 Islands. The very thick-billed Fringillidse indicate a terminal, not an indifferent, 

 stage of development. Such forms as Geospiza, Loxigilla, Coccothraustes, and 

 Pyrrhula are undoubtedly instances of convergent analogies. 



III. All the Drepanididae have started from Fringillidee, some of which have 

 developed further in essentially Fringilloid lines, leading to Chloridops and 

 Bhodacanthis as extremes. We should have to assume that Loxops and 

 Chrysomitridops are least removed from the hypothetical starting-point; but it 

 so happens that these two genera are closely allied to Oreomyza, while by 

 their tongue, nostrils, and palatal arrangement they are far removed from 

 moderate, not exaggerated, true Fringillidse, e. g. Chrysomitris. We should 

 further have to assume that, by the development of a more slender bill — in fact, 

 by departing from typical Fringilline features, — some of the birds in the Sandwich 

 Islands have produced the tubular brush-tongue, have weakened and lengthened 

 the palatal arrangement (although retaining the fused palatines), have elongated 

 the trans- and interpalatine portions, and last, but not least, have developed 

 operculated nostrils. Anatomically, at least, the development of the nasal 



