198 
D.J. Kitchener, N. Cooper, I. Maryanto 
Table 3 Canonical Variate Function coefficients for the 
three locality groups: Lesser Sunda; Western 
Australian; and Victorian (see text for 
clarification). Standarised values followed by 
(in brackets) unstandardised values for (a) 10 
skull and dental characters and (b) seven 
external characters. For explanation of 
character codes see Materials and Methods 
section. 
Table 3a 
Character 
Function 1 
Function 2 
M 3 M 3 W 
0.5469 
(3.3069) 
0.1786 (1.0802) 
BB 
0.0790 
(3.1840) 
0.4706 (3.1285) 
LIB 
0.1569 
(1.4119) 
-0.1350 (-1.2154) 
PPL 
0.3856 
(2.8807) 
0.0581 (0.4341) 
C’C’W 
-0.5574 
(-4.2458) 
0.0018 (0.0134) 
GSL 
1.2016 
(3.5616) 
-1.8370 (-5.4449) 
CBL 
-1.9381 
(-5.5309) 
2.1724 (6.1994) 
RL 
0.1723 
(0.9570) 
-0.9471 (-5.2606) 
I'M 3 L 
0.7461 
(4.3581) 
0.5446 (3.1809) 
M 2 B 
-0.0812 
(-1.1189) 
-0.4262 (-5.8712) 
Constant 
-58.8462 
-22.3024 
Variation 
Explained (%) 72.2 
27.8 
Table 3b 
Character 
Function 1 
Function 2 
D3P1 
0.5373 (0.7249) 
-0.7477 (-1.0087) 
FA 
0.0894 (-0.0752) 
0.8681 (0.7301) 
D5P2 
0.4108 (-0.9332) 
0.6058 (1.3762) 
TIB 
0.3731 (0.6049) 
-0.3616 (-0.5862) 
D3P2 
0.3885 (0.5117) 
-0.1650 (-0.2174) 
D3M 
-0.0544 (-0.0414) 
0.4549 (0.3458) 
EL 
0.3825 (0.5846) 
0.0720 (0.1100) 
Constant 
-35.8578 
-23.6352 
Variance 
Explained (%) 
94.6 
5.4 
loading most heavily (>0.5) on this Function were: 
condylobasal length; greatest skull length; I'M 3 
length; CC 1 width; and M 3 M 3 breadth (Table 3a). 
Function 2 separated the Victorian Group from 
both the Lesser Sunda Group and the Western 
Australian Group. The characters loading most 
heavily (>0.5) on this Function were: condylobasal 
length; greater skull length; rostrum length; and 
I'M 3 length (Table 3a). It appears that the Western 
Australian Group will be distinguished from the 
other two groups on overall size, particularly skull 
length, palatal breadth and tooth row length. The 
Victoria Group separated from the other two 
groups on a shape difference involving the 
relationship between condylobasal and greatest 
skull length, and tooth row length. The Lesser 
Sunda Group separated from the other two groups 
on a combination of the above characters. 
All locations - external characters 
DFA was run on a reduced set of seven 
characters (listed in Table 3b) using the three above 
location groupings determined for the skull, 
dentary and dental characters. This DFA extracted 
two very significant Functions (Figure 2b). 
Function 1 explained 94.6% of the variance and 
Function 2, 5.4%. A total of 95.5% of individuals 
were classified to their correct group; two 
specimens from the Victorian Group were 
misclassified to the Western Australian Group and 
three specimens from this latter group were 
misclassified to the Victorian Group. 
Function 1 separated the Lesser Sunda Group 
from both the Western Australian and Victorian 
Groups. The character loading most heavily (>0.5) 
on Function 1 was digit 3 phalanx 1 length (Table 
3b). Function 2 partially separated the Victorian 
and Western Australian Groups. The characters 
loading most heavily (>0.5) on Function 2 were: 
forearm length; digit 3 phalanx 1 length; and digit 
5 phalanx 2 length. (Table 3b). This suggested that 
these three groups can be separated by aspects of 
the wing structure, particularly digit 3 phalanx 1 
length. 
The Lesser Sunda Group - skull and dental characters 
DFA was run on a reduced set of five skull and 
dental characters (listed in Table 4a) for all island 
locations in the Javan Group. Three groupings of 
islands were apparent. These were: 
(i) The Alor Group (Java, Kangean, Nusa Penida, 
Sumbawa, Moyo, Flores, Lembata, Pantar, 
Alor, Timor and Savu); 
(ii) Wetar; and 
(iii) Yamdena. 
The single New South Wales specimen was left 
unallocated. 
A DFA, using both the above five characters and 
the three above a priori island groupings, extracted 
two very significant Functions. Function 1 
explained 65.1% of the variance and Function 2, 
34.9%. A total of 88.9% of specimens were allocated 
to their correct group. Misclassifications were as 
follows: Two of the Alor Group were misclassified 
to Yamdena; three to Wetar. Two from Wetar were 
misclassified to the Alor Group and one to 
Yamdena. 
The single specimen from New South Wales was 
classified to the Yamdena population. 
Function 1 almost completely separated the Alor 
Group from the Yamdena population (Figure 3a). 
The characters loading heavily (>0.5) on Function 1 
were M 3 M 3 width; anteorbital foramen breadth; M 2 
breadth; and mesopterygoid fossa breadth (Table 
4a). Function 2 partially separates the Wetar 
population from both the Alor Group and the 
Yamdena population (Figure 3a). The character 
loading heavily on Function 2 was M 2 breadth. 
