Morphological variation in Hipposideros bicolor 
Function 2 
185 
Function 3 
Figure 3 Canonical Variate Analysis among four population groupings of Hipposideros bicolor (Selaru, Timor, Roti/ 
Savu and Peninsula Malaysia) - with the few Sumba, Sumbawa and Java specimens unallocated - and 
based on a subset of five cranial , dentary and dental characters (see text), (a) plots of Functions 1 and 2 
and (b) plots of Functions 1 and 3. Population codes as for Figure 2. 
the two Java specimens and single Sumba and 
Sumbawa specimens unallocated, and again using 
the above five characters, extracted three 
significant functions which allocated 98.2% of the 
specimens to their correct group. Function 1, which 
explained 56.8% of the variance, separated the 
Selaru and Malay Peninsula Group from the Timor 
Group (Figure 3a). The characters with high 
coefficients (>0.5) on Function 1 were dentary 
length and least interorbital breadth (Table 3a). 
Function 2 (35.3% of variance) separated the Roti/ 
Savu Group from the Selaru, Malay Peninsula and 
Timor Groups (Figure 3a). The character with the 
highest coefficient (>0.5) on Function 2 was 
interlachrymal distance (Table 3a). Function 3 
(7.9% of variance) separated the Malay Peninsula 
Group from the other three groups (Figure 3b). The 
character with the highest coefficient on Function 3 
was least interorbital breadth (Table 3a). Of the 
unallocated specimens, one of the Javanese 
specimens allocated to the Peninsula Malaysia 
Group and the other to the Roti/Savu Group; the 
Sumba and Sumbawa specimens allocated to the 
Selaru Group. 
External body characters 
The DFA based on five characters (digit 3 
phalanx 1 length, digit 4 metacarpal length, 
forearm length, ear length and digit 2 metacarpal 
length), and using as groups each island 
population and the Peninsula Malaysia population, 
extracted four significant functions but with only 
61.8% of individuals classified to their correct 
locality. Bivariate plots of these functions did not 
reveal the descrete clusters apparent for the cranial, 
dentary and dental characters, although there was 
partial separation between the Timor and Selaru 
Groups (Figure 4). 
A DFA was run based on five characters (digit 3 
phalanx 1 length, digit 5 metacarpal length, digit 4 
metacarpal length, ear length and forearm length) 
and using as groups those identified in the earlier 
cranial, dentary and dental analysis. This DFA 
extracted three significant functions with 80.3% of 
individuals classified to their correct groups. There 
was considerable overlap between the Timor and 
Selaru Groups, with 11.8% of individuals from the 
Timor Group misclassified to the Selaru Group. A 
total of 37.5% of the Roti/Savu Group were 
misclassified to either the Selaru Group (12.5%) or 
the Timor Group (25.0%). Of the ungrouped 
specimens, the two from Sumbawa and the two 
from Java grouped with the Selaru Group; the 
single specimen from Sumba island grouped with 
the Selaru Group. Only the Peninsula Malaysia 
Group was discrete. Function 1, which explained 
60.2% of the variance, separated the Peninsula 
Malaysia Group from the other groups and from 
the unallocated Sumbawa, Java and Sumba 
specimens (Figure 5). The characters with high 
coefficients (> 0.5) on Function 1 were digit 5 
metacarpal length, digit 3 phalanx 1 length and 
forearm length (Table 3b). Function 2 (27.2% of 
variance) also separated the Peninsula Malaysia 
