15 
Taxonomy of Rhinolophus simplex 
Table 5 Canonical variate functions coefficients for the three subspecies of Rhinolophus simplex (R. s. simplex, R. s. 
parvus, R. s. subsp. nov.). Rhinolophus s. keyensis unallocated. Standardised values, followed by (in brackets) 
unstandardised values (a) skull and dental; (b) external characters. 
Table 5a 
Character 
Function 1 
Function 2 
I.M,L 
1.2760 
(5.8790) 
-0.9938 
(-4.5787) 
CM 3 L 
-0.8948 
(-4.4169) 
1.0576 
(5.2205) 
M 3 M 3 W 
0.1955 
(1.1858) 
0.8975 
(5.4438) 
CB 
0.4427 
(3.8282) 
0.2631 
(2.2755) 
NIB 
0.2770 
(1.6894) 
-0.4586 
(-2.7968) 
CONSTANT 
-42.6169 
-27.1418 
VARIATION 
75.4 
24.6 
EXPLAINED (%) 
Table 5b 
Character 
Function 1 
Function 2 
TIB 
0.4414 
(0.6271) 
-0.4069 
(-0.5781) 
SV 
0.9602 
(0.6010) 
0.6795 
(0.4506) 
PES 
0.3907 
(1.1571) 
-0.8443 
(-2.5005) 
D4P1 
0.4708 
(1.3476) 
0.4592 
(1.3145) 
VSH 
0.6294 
(3.3753) 
0.0508 
(0.2722) 
TV 
-0.5209 
(-0.2470) 
0.1439 
(0.0682) 
BSB 
-0.5409 
(-2.5337) 
0.3179 
(1.4891) 
CONSTANT 
-50.4099 
-6.9463 
VARIATION 
85.4 
14.6 
EXPLAINED (%) 
was selected on the basis that they provided values 
that minimise Wilk's lambda. The plots of the 
discriminant function 1 and 2 based on this 
reduced set of five characters (lower, tooth row 
length, IjM 3 L; upper maxillary tooth row length, 
C'M 3 L; outer M 3 M 3 width, M 3 M 3 W; cranium 
breadth, CB; and nasal inflation breadth, NIB) 
produced very similar plots to the above analyses, 
and so only these are presented and discussed 
below. 
The DFA produced two significant canonical 
functions. Function 1 explained 75.4 percent of the 
variance and function 2, 24.6 percent (Table 5a). A 
total of 100 percent of individuals were correctly 
classified to their appropriate subspecies. The plot 
of function 1 and 2 (Figure 4a) clearly separates the 
subspecies: simplex, parvus and subsp. nov. with the 
unallocated keyensis grouping with simplex. 
Function 1 separates all three allocated subspecies 
clusters and function 2 partially separates R. s. 
simplex from both R. s parvus and R. s. subsp. nov. 
and completely separates the R. s. parvus and R.s. 
subsp. nov. clusters. 
The characters loading most heavily (>0.8) on 
function 1 and presumed important discriminants 
between the three allocated subspecies, were lower 
tooth row length, I,M 3 L, and upper maxillary tooth 
row length, C‘M ! L (Table 5a). Characters loading 
most heavily on function 2 (>0.8) and presumed 
particularly important in discriminating between 
R. s. parvus and R. s subsp. nov. included, in 
addition to the above two characters, outer MTVP 
width, M 3 M 3 W (Table 5a). 
Externals. The DFA for the R. simplex subspecies 
was first run using the reduced set of 17 characters 
and using islands as the a priori groupings. When 
these islands were then grouped to represent the 
three allocated R. simplex subspecies (see above), 
the configuration of the taxon clusters in 
discriminant function space was similar to that 
produced above. Flowever, because the number of 
characters was larger than the number of 
individuals in one taxon group (R. simplex parvus, 
12) a reduced set of seven characters was selected 
(tibia length, TIB; snout to vent length, SV; pes 
length, PES; digit 4, phalanx 1 length, D4P1; 
vertical sella breadth, VSB; tail to vent length, TV; 
basal sella breadth, BSB) using the method for 
skulls above; this produced similar DFA plots to 
those produced using the 17 characters. Only the 
DFA based on this set of seven characters are 
presented and discussed below. 
The DFA produced two significant canonical 
functions. Function 1 explained 85.4 percent of the 
variance and function 2, 14.6 percent (Table 5b). A 
total of 100 percent of individuals were correctly 
classified to their appropriate subspecies. The plot 
of functions 1 and 2 (Figure 4b) clearly separates 
the subspecies, with the unallocated keyensis again 
grouping with R. s. simplex. Function 1 separates 
all three subspecies clusters and function 2 
separates R.s. parvus and R.s. subsp. nov. 
