Landers et al. • SIZE DIMORPHISM IN WESTLAND PETRELS 
721 
molecular analysis using the CHD gene of the 
avian sex chromosomes (Kahn et al. 1998), 
Morphometric Measurements. —Seven body 
measurements typically collected in ornithologi¬ 
cal, including seabird, studies (Guicking el al. 
2004, Bourgeois et al. 2007. Thai man n et al. 
2007) were taken (always from the left appendage 
for bilateral traits). These were: (I) head length 
(HL) from the cerebellum roof ( = supraoccipital) 
to the edge of the feather implantation on the 
culmen, (2) bill length (BL) - exposed culmen 
from the tip of the hook to the edge of the feather 
implantation, (3) minimum bill depth (MBD) = 
minimum bill thickness of upper and lower 
mandibles measured vertically. ( 4 ) tarsus length 
iTL) — metatarsus length from the depression in 
the angle of the intertarsaJ joint to the base of the 
last complete scale before the toe diverges. (5) toe 
length (MTL) = middle toe length from the first 
scale of the middle toe to the base of the nail on 
this toe. (6) wing length (WL) = maximum 
flattened chord), and (7) body mass (BM). All 
measurements were made with digital slide 
calipers to the nearest 0.1 mm except wing length, 
for which a stopped wing-rule was used (to the 
nearest 0.1 cm), and mass where a handling bag 
and Pesola balance were used (measured to the 
nearest 5 g). We measured birds as a team to 
maximize consistency where one researcher (TJL) 
held the bird while the other (TED) took the 
measurements. Several repeat measures were 
taken lor the first few birds to familiarize 
ourselves with the procedure (only the last set of 
measurements for each bird was used for 
analysis), after which all subsequent birds were 
measured only once. 
Statistical Analyses. —Wc performed univariate 
/-tests of each morphometric character, and 
calculated the percentage of dimorphism between 
males and females (based on the molecular data) 
for each character measured as f(x,„ - X/)/xj\ X 
100. where x,„ and fy are mean values for males 
and females, respectively (Holmes and Pitelka 
1968). We used Pearson product-moment corre¬ 
lations to examine relationships among individual 
morphometric characteristics. We assessed wheth¬ 
er the seven morphological characters collectively 
differed between females and males, using a one¬ 
way multivariate analyses of variance (MAN- 
OVA) with gender as the categorical factor and 
morphological characters as the dependent vari¬ 
ables. 
Linear discriminant function analyses (DFA) 
were used to identify which characteristic best 
differentiated females and males, and how 
accurately gender could be classified using the 
resulting canonical classification function. We 
performed a standard DFA using all characteris¬ 
tics, so they could be examined conjointly, and to 
facilitate comparisons with published studies on 
avian size dimorphism. We performed stepwise 
DFAs (both forward and backward) to evaluate 
the reliability of the results of the standard DFA. 
The results of all DFAs were similar, and we 
report only the canonical function from the 
standard DFA, as this was most comprehensive. 
We tested the effectiveness of the standard 
DFA by a jackknife procedure, in which each 
individual was classified based on a discriminant 
function formulated when the focal individual was 
removed and the remaining individuals were used 
to calculate the function (Quenouille 1956, 
Lachenbruch and Mickey 1968, Genovart et al. 
2003). The resulting canonical classification 
functions may be used to classify the gender of 
additional Westland Petrels by adding a bird’s 
morphometric measurements from the field into 
the female and male functions with the higher DF 
score predicting its gender. Partial eta-squared 
( tv) values are also reported, which are defined as 
the proportion of total variation attributed to the 
factor, excluding other factors from the total non¬ 
error variation (Pierce et al. 2004). All statistical 
tests were conducted with STATISTICA 9 
(StatSoft 2009). 
RESULTS 
Molecular analysis identified seven females 
and 30 males in our samples (binomial test with 
random expectation 50%: P = 0.098). There was 
considerable variation in the magnitude and range 
of the differences between males and females in 
the seven morphological characteristics; males 
were consistently larger in all traits (Fig. 1). 
Unpaired /-tests revealed that males were signif¬ 
icantly larger in head length, minimum bill depth, 
and body mass with gender explaining 35, 42, and 
11% of the variation of these characteristics, 
respectively (Table I). Several characters were 
positively inter-correlated within subjects (Ta¬ 
ble 2) as expected for size dimensions; however, 
there was no strong multicollinearity as r was at 
most 0.58 for any correlations. Generally, larger 
morphometric features (e.g.. wing and head 
length) were related with body mass, and features 
of the head and bill were associated. Overall, 
