Myotis adversus comple: 
197 
analysis a reduced set of 5-10 characters was 
selected. It is the DFA based on this reduced set of 
characters that is discussed in the text, because in 
all instances they provided similar discriminant 
function plots to those of the complete set of 
characters. These reduced set of characters were 
selected in all these analyses because the sample 
size of the smallest a priori group selected 
approximated, or was less than, the number of 
characters in the analysis. This reduced set of 
characters was chosen because they provided 
values that minimise Wilk's lambda. The statistical 
software used throughout was SPSS PC+. 
RESULTS AND DISCUSSION 
Univariate statistics 
Multiple regressions were run separately for 
skull, dentary and dental characters, excluding 
zygomatic width which was missing values from 
many specimens, and external characters because 
these analyses utilised different sets of locations 
and sample sizes. 
Skull, dentary and dental characters 
Sex. No characters were significantly influenced by 
sex alone, although least interorbital breadth and 
mesopterygoid fossa breadth were weakly 
associated with sex (F 187 = 4.828; P= 0.031 and F 187 
6.091; P=0.016, respectively) (Table la). 
Location. All characters, except C'C width and M 2 
breadth were very significantly (P<0.001) 
influenced by location alone (Table la). 
Interaction. There were no significant interactions, 
although there was a weak interaction between sex 
and location for mesopterygoid fossa breadth 
(F 987 = 2.429; P= 0.016) (Table la). 
External characters 
Sex. A number of characters representing wing size 
were influenced by sex alone, with females being 
larger than males. These characters were forearm 
length (F 19 ,= 11.784; P=0.001); digit 3 metacarpal 
length (F 191 = 11.059; P= 0.001); digit 3 phalanx 1 
length (F 191 = 7.668; P= 0.007; digit 4 metacarpal 
length (F,,= 9.448; P= 0.003); and digit 5 
metacarpal length (F 19] = 9.713; P=0.003) (Table lb). 
Location. All characters were significantly (P<0.001) 
influenced by location alone. The lack of significant 
interaction between sex and location indicates that 
the relationship between wing size of male and 
females was consistent for all locations. 
Mean, standard deviation, minimum and 
maximum values and sample size for each locality 
are presented in Table 2 for skulls, dentary and 
dental characters (2a) and external body characters 
(2b) for all characters examined for the locations or 
location groupings determined in this study to 
have taxonomic significance. Values for males and 
females are combined for both the skull, dental and 
dentary characters and for the external characters. 
Although most of the wing measurements for 
females were larger than those of the males, these 
differences were generally greatly exceeded by the 
differences between islands or groups of islands. 
Multivariate analyses 
In the subsequent DFA of skull, dentary, dental 
and external characters, males and females were 
combined. This is appropriate for the skull, dentary 
and dental characters because they were not 
significantly influenced by sex. It is less satisfactory 
for the external characters because many of the 
wing measurements were influenced by sex. For 
this reason, the location groupings (possibly 
representative of putative taxa) were selected on 
the basis of the skull, dentary and dental 
characters. The external characters were then 
placed into location groupings as for the skull and 
other characters. It was considered that the loss of 
distinction between the recognised island 
groupings, or taxa, based on their external 
measurements, would be minimal because location 
differences far exceeded sex differences for most of 
the external characters (Table lb). 
All locations - skull, dentary and dental characters 
DFA was run on a reduced set of 10 selected 
characters (listed in Table 3a) for all locations. 
Three groupings of locations were apparent. These 
were: 
(i) The Lesser Sunda Group (Java I., Kangean I., 
Nusa Penida I., Sumbawa I., Moyo I., Flores I., 
Lembata I., Pantar I., Alor I., Wetar I., 
Yamdena I., Timor I., Savu I., New South 
Wales); 
(ii) The Western Australian Group (Western 
Australia, Northern Territory, Queensland, 
Seram, Papua New Guinea and Solomon 
Islands); and 
(iii) The Victorian Group (Victoria, South 
Australia). 
A DFA using the above 10 characters and these 
three a priori groupings extracted two highly 
significant Functions. Function 1 explained 72.2% 
of the variance and Function 2, 27.8%. 
A total of 97.6% of individuals were classified to 
their correct group. Misclassifications were in the 
Lesser Sunda Group, with one specimen from this 
group allocating to the Victorian Group and two 
specimens to the Western Australian Group. 
Function 1 separated both the Lesser Sunda 
Group and the Victorian Group from the Western 
Australian Group (Figure 2a). The characters 
