Bayliset al. • CAPSAICIN AS A NEST PREDATION DETERRENT 
521 
We left untreated eggs without any modification, 
and these were placed bare in the imitation nests. 
W e placed powder chili-treated eggs in the nest and 
chili powder was sprinkled directly over the eggs at 
a volume of 5 mL of powder per nest. Powder 
paprika-treated eggs were placed in nests and 
paprika was sprinkled over them, also at a volume 
of 5 mL powder per nest. Adhesive-treated eggs 
had either chili or paprika powder directly adhered 
using raw albumen and vitellus from store-bought 
Domestic Chicken ( Gal Ins gal l u.s) eggs. This was 
chosen as an adhesive because, once dry, it is hard 
and weatherproof, and because egg contents are a 
plausible substance inside natural nests. Imitation 
eggs were immersed in a well-mixed bowl of 
albumen and vitellus, buried in a container of 
powder, removed from the container, and allowed 
to air dry at room temperature. 
Blocking Experimental Design. —It was not 
logistical^ feasible to use larger numbers of nests 
for sufficient statistical power concurrently within 
the geographic confines of ARBG. and we used a 
Randomized Complete Block Design (RCBD; 
McIntosh 1983. Montgomery 2005) with three 
blocks. We placed the three blocks sequentially 
over the same region within the study area. Blocks 
each contained 20 nests, including four of each of the 
live treatment types. We placed these nests in a 
random order and randomized within each block. 
Randomization Scheme. —We established a 
'ransect line through the New Zealand Bush 
section of ARBG and used random numbers 
(generated in Microsoft Excel) to assign distance 
along and from the transect for nest placement. 
' Ve randomized the order of treatments within 
each block. We placed nests as close as possible 
10 the pre-allocated site wherever it was not 
possible to place a nest precisely following the 
protocol (for example, due to the presence of a 
■>rnall pond). 
Data Collection.—We left nests in the field for 
1 1 days during the austral breeding season 
I spring/summer: Nov 2009-Feb 2010), which is 
‘‘bout equal to the incubation period (10-12 days) 
* or introduced Common Blackbirds (Tardus 
writ la) in Australia and New Zealand (Magrath 
1992). We considered an egg to be depredated it it 
had tooth marks (in the case of imitation eggs) or 
"as broken (in the case of real eggs) or removed 
'Dm the nest. The color contrast between the 
Painted exterior and the while plaster-of-Paris 
interior of the imitation eggs allowed tooth marks 
to be easily identified (Fig. 1). We recorded 
predation events for each study egg. allowing 
statistical analysis of partial-nest predation (Hau¬ 
lier 1998). 
We visually inspected depredated eggs and 
compared the predator's marks with tooth imprints 
from skulls of museum specimens of introduced 
rats (R. ratios and R. norvegicus). ermine, possums, 
and house mice (Mas musculus) in New Zealand 
(Boulton and Cassey 2006). We could easily 
identify predators to genus using this method, 
although differentiating between the two rat species 
was not feasible. We used the width of the tooth- 
gouges and the presence of paired furrows in the 
imitation eggs (presumably carved by paired rat 
incisors, which are much less prominent in possums 
and ermine, and much narrower in mice) as key 
attributes in the identification. 
Statistical Analysis—We checked the data for 
any effect of egg type (imitation or real) on 
survival time using nonparametrie Wilcoxon 
signed-rank tests. We performed Kaplan-Meier 
survival analysis, following Venables and Ripley 
(2002), on the number of days of survival per egg 
with block as a blocking factor. We tested for 
differences in hazard function between treatment 
groups using the 'efron' technique (Venables and 
Ripley 2002). We conducted two separate tests. 
Wc first looked for significant differences in 
hazard rate between any two treatment groups 
once egg type and block were considered using a 
two-tailed test. Second, based on a priori 
reasoning, we performed a one-tailed, single 
comparison test to test the hypothesis that 
adhesive capsaicin-treated eggs should experience 
a lower hazard rate than eggs of other treatment 
groups once egg type and block were considered. 
There was technical pseudoreplication in this 
experiment, as it was necessary to treat all eggs 
within the same nest in the same manner. Thus, 
we also analyzed the mean survival time (i.e.. the 
mean number of days that eggs survived) within 
each nest as the response variable, as this response 
metric bypasses the non-independence of eggs 
within nests. We considered nests as independent 
data points in the second analysis, although 
nearby nests may have been visited and depre¬ 
dated by the same individual mammals. We 
modeled the effects of block on mean survival 
time of nests using ANOVA, and tested the 
residuals from the ANOVA model using a 
Wilcoxon signed-rank test to examine if the mean 
survival time of adhesive capsaicin-treated eggs 
exceeded the mean survival time of the other 
