126 D.J. Kitchener, W.C. Packer, A. Suyanto 
keasti keasti, □; N. k. tozeri subsp. nov.. A; N. cephalotes cephalotes, •; N. c. aplini subsp. nov., X. 
greatest skull length; CBL, condylobasal length; 
PIF, minimum length from posterior margin of 
incisive foramen to palate posterior margin; RL, 
rostrum length, from anteriomost internal margin 
of orbit to nares; RH, rostrum height, from upper 
canine alveoli to level of dorsal surface of nasals; 
BB, braincase breadth above zygoma; ZW, 
zygomatic width; M'M 1 and C'C 1 , width across 
M‘M‘ and CC 1 respectively, from the labial side; 
C'M 1 and C,M 2 upper and lower canine to last 
molar length; IOB, minimum interorbital breadth; 
POB, minimum postorbital breadth; MFB, 
mesopterygoid fossa breadth, at the widest point 
of the palatal flange; P J P, palatal width between 
the lingual aspect of PP; ML, mandible length, 
from condyle to anteriomost point of dentary; CH, 
dentary coronoid height; TV, tail to anus length; 
EAR, basal notch to apex length; FA, forearm 
length; and MC3-5, metacarpal 3,4 and 5 length. 
The skull, dentary and dental characters were 
measured to an accuracy of 0.01 mm, while the 
external body characters were measured to 0.1 
mm. Terminology used in the description of skull, 
dentary, dental and external body characters 
follows Hill and Smith (1984). Pelage descriptions 
follow the colour terminology of Smithe (1975). 
Adults were diagnosed as those specimens with 
the following sutures fused: basioccipital - 
basisphenoid, basisphenoid - presphenoid and 
palatine - maxillary (these sutures are illustrated in 
Heaney and Peterson 1984: Figure 4). 
The effect of sex and island on all characters was 
examined by multiple regressions for the islands: 
Yamdena, Selaru, Kai Kecil, Kai Besar, Aru, 
Ambon, Seram, Sulawesi and Papua New Guinea. 
Examination of the residuals from regression 
analyses gave no indication of heteroscedasticity. 
Stepwise canonical variate (discriminant 
function) analyses (DFA) were run for skull, 
dentary and dental characters and external 
characters using all 23 characters for males and 
females combined, after first testing for sexual 
dimorphism. A reduced set of five of these 
characters was used in all presented DFA because 
in all instances they provided similar DF plots to 
the full set of 23 characters. This reduced set of 
characters was selected in all these analyses 
because the sample size of the smallest a priori 
group selected was always less than the total 
number of characters. This reduced set of 
characters was selected because they provided 
values that minimise Wilk's Lambda. 
