64 
Figure 2 Plot of Functions 1 and 2 from canonical 
variate analysis of five island groups 
(Lombok, O; Flores, F; Lembata, E; Pantar, P; 
and Alor, A) based on 27 characters (see text), 
with males and females combined. 
one Alor specimen was misclassified (to the 
Lembata-Flores-Pantar group). Because the number 
of characters in this latter analysis (N = 27) greatly 
exceeds the sample size of the smallest group 
(Alor, N = 5) the DFA was repeated for these three 
groups using a reduced set of five characters. This 
reduced set was selected from this latter analysis 
to minimise Wilk's lambda. These five characters 
were: C,M 3 length, M 1 width, C'M 2 length, rostrum 
length and digit 4 metacarpal length. The DFA 
using this reduced set of characters produced 
similar DFA plots to those based on the three 
island groups using the 27 characters. 
Consequently, only results based on these five 
characters are detailed below. This DFA extracted 
two significant functions that explained all the 
variance (Figure 3). Function 1, which explained 
87.9% of the variance, separated the Lombok Island 
group from the other two groups. The character 
with the canonical discriminant function coefficient 
that loaded heavily (>0.5) on Function 1 was C M 
length (Table 3). This suggested that this character 
was an important discriminant between the 
Lombok Island population and the other 
populations. Function 2, which explained 12.1% of 
the variance, partially separated the Alor Island 
population from the Flores-Lembata-Pantar 
population. The characters that loaded heavily 
(>0.5) on Function 2 were C'M 2 length (1.02), 
rostrum length (0.94), digit 4 metacarpal length 
(0.73) and M 1 breadth (Table 3). This suggested that 
D.J. Kitchener, W.C. Packer, Maharadatunkafisi 
Function 2 
Figure 3 Plot of Functions 1 and 2 from canonical 
variate analysis of three island group 5 
(Lombok, Alor and Flores-Lembata-Pantar) 
and a reduced set of 5 characters (see text), 
with males and females Combined- 
Abbreviations for island populations as f° r 
Figure 2. 
a number of dental, skull and external characters 
were involved in the partial separation of the Alor 
population from the other eastern Nusa Tenggara 
populations. A total of 89.7% of individuals was 
correctly classified to their appropriate island 
group. The Lombok Island population was most 
distinct, with 100% of its individuals correctly 
classified. All misclassifications occurred between 
the Alor population and those of the other eastern 
islands. One of the five Alor Island individuals was 
misclassified to the Flores-Lembata-Pantar group, 
while two of the 13 individuals in this latter group 
were misclassified to the Alor population. 
Summary of multivariate analysis 
The above analysis indicates that two broad 
morphological forms occurred among Pterop 
lombocensis in Nusa Tenggara. These were the 
Lombok Island form (Pteropus lombocensis 
lombocensis ) and a form involving the remaining 
eastern populations (Flores, Lembata, Pantar and 
Alor), in which the Alor population is slight!' 
differentiated from the others. This differentiation 
of the Alor form from the other eastern 
populations is slight and is not considered to have 
taxonomic significance. Of the two named forms in 
these eastern Nusa Tenggara islands (P. solitarius - 
Alor and P. heudei - Flores), P. heudei has priori!) 
and becomes the subspecific name (P. lombocensis 
heudei) for this eastern form. 
