148 D.J. Kitchener, I. Maryanto 
Figure 1 Locality of specimens used in this study. Pteropus g. griseus (O); P. g. pallidus (•); P. t. temmincki (▼); P. /. 
lombocensis (■); P. 1. heudei (□) and P. 1. salottii (*). 
Adult condition was judged to be when both the 
basioccipital/basisphenoid and basisphenoid/ 
presphenoid sutures were fused. 
The statistical procedures were as described in 
Kitchener et al. (1995a). Both sexes were combined. 
The stepwise canonical variate (discriminant 
function) analysis was initially run for all 20 skull 
characters for islands separately and islands 
grouped. The DFA was then run for a subset of 5 
of these characters on the grouped populations. 
This subset was selected to minimise Wilk's 
lambda. This was because the full set of characters 
exceeded the sample size of the smallest island 
group examined. Because the DFA plots using the 
subset of characters were very similar to those 
when all characters were employed in the DFA, 
only the analysis carried out with the reduced set 
of characters are presented below. 
RESULTS 
Discriminant Function Analysis 
The preliminary analysis using all 20 characters 
clearly placed the single Komodo Island specimen 
with Pteropus l. heudei. 
All groups/taxa 
A DFA was run based on 6 taxon/groups (the 
P. L lombocensis 
-6-4-2 0 2 4 6 
Function 2 
Figure 2 Plot of Functions 1 and 2 from DFA based on 
six taxa [Pteropus I. salottii (Timor, T); P. 1. 
lombocensis (Lombok, O); P. I. heudei (Flores, 
F; Lembata, E; Pantar, P; Alor, A; and 
Komodo, K); P. g. griseus (Wetar, W; Timor, 
1; Semau, S; and Roti, R); P. g. pallidus (Banda 
Neira, C); and P. 1. temmincki (Ambon, B; 
Seram, S)] and five selected skull characters. 
