Differences from _richardsonii_ are indicated in the formal description just given. Additional to differences therein noted, _celenda_ differs from _initis_ in larger interorbital and preorbital parts of the skull although dimensions of other parts of the skull and the teeth are about the same or even less. From _salva_, _celenda_ differs in larger average size in every measurement taken, except for the inner moiety of M1 which is about the same. The skull of _celenda_ is 35 per cent heavier. In relation to the basilar length the skull of _celenda_ is wider, especially in the interorbital and preorbital regions. In comparison with _alascensis_ the tympanic bullae are of approximately the same length; otherwise essentially the same differences obtain as are noted in comparison with _salva_ and the zygomatic breadth is relatively more in _celenda_. From _seclusa_, in which the teeth are of comparable size, _celenda_ differs in that every cranial measurement is more and the skull is 28 per cent heavier. Because the skull of _celenda_ is so much longer, its dimensions in other planes are less in relation to the length than in _seclusa_. _M. e. celenda_ is larger in every part measured than _haidarum_, 21 per cent heavier, and in relation to the basilar length the interorbital, and preorbital, parts of the skull are smaller, the braincase is shallower, and the skull is relatively wider across the zygomata and mastoid processes. In comparison with _kadiacensis_, differences are: 26 per cent lighter, skull shorter; in relation to the basilar length, braincase shallower as measured at the anterior end of the basioccipital, tooth-rows shorter but orbitonasal length more. In comparison with _arctica_ all parts measured of the teeth and skull of _celenda_ are smaller and its skull is 34 per cent lighter. In relation to the basilar length, the interorbital breadth of _celenda_ is only slightly less but its skull is narrower across the rostrum and zygomata, the tooth-rows are shorter, and the braincase is shallower. _Remarks._--The late George Willett in the course of his work in Alaska collected most of the known specimens of this strongly differentiated subspecies. In both coloration and cranial characters the distinguishing features are so well marked that the zoölogist could with reason accord full specific rank to _celenda_. Nevertheless it obviously is an ermine. Also, races from other islands of southeastern Alaska tend to bridge the gap, as regards cranial features, between _celenda_ and the mainland ermine. The specimen from Dall Island agrees in all respects with topotypes. The specimen from Howkan on Long Island is in white winter pelage and the skull has suffered shrinkage from some chemical solution; the reference of this specimen to _celenda_ is tentative. _Specimens examined._--Total number, 25, as follows: Arranged by localities from north to south. Unless otherwise indicated, in U. S. National Museum.