244 
THE WILSON JOURNAL OF ORNITHOLOGY • Vo/. 123. No. 2. June 2011 
clutch size. New eggs observed in each nest were 
marked with non-toxic indelible ink to indicate 
first, second, and third-laid eggs, and denoted as 
A-, B-, and C-eggs. The colony was left 
undisturbed during incubation of the study 
clutches (8-27 Oct 2005). A different methodol¬ 
ogy for nest selection was used in 2006 and 2007 
due to the logistical constraints of the area. Study 
nests with 1-3 eggs were selected in 2006 (« = 
66) and 2007 (n = 47) during late incubation (16 
Oct-9 Nov 2006, 4-17 Oct 2007) based on 
observations that suggested imminent hatching 
of al least one egg, such as small cracks in the 
shell or a star-pipped egg. Sampling in 2006 was 
conducted after hatching had begun, and 14 days 
later in the season than in 2007, Nest selection in 
the second year started at onset of hatching. Eggs 
were marked based on signs of hatching to 
indicate presumed first, second, and third hatching 
order, and denoted as A-, B-. and C-eggs. Only 
nests near the periphery were selected in 2006 and 
2007 to minimize disturbance in the inner areas of 
the colony. Each egg was measured (max breadth 
and length, nearest 0.01 mm) using digital calipers 
(Mitutoyo. Kawasaki, Japan). Egg volume (V) 
was calculated in cm- as V = 0.000476 x length 
X breadth 2 (Bolton 1991). 
Nests were visited daily during the hatching 
period of the marked clutches (27 Oct-2 Nov 
2005. 17 Oct-9 Nov 2006. 4-24 Oct 2007) to: (1) 
ascertain hatching date, (2) estimate incubation 
ength (only in 2005), (3) identify hatching order. 
( ) examine extent of hatching asynchrony, and 
(5) calculate hatching and breeding success. 
Incubation length for individual eggs in 2005 
was defined as the time elapsed between layins 
and hatching date of the same egg. Laying date 
was defined as the date a new egg was first 
observed in the nest. Synchronous hatching was 
defined as hatching of two eggs within the same 
24-hr period and producing chicks without overt 
physical or behavioral differences. Hatching 
success was defined as the proportion of marked 
eggs that hatched successfully in three-egg 
clutches, and breeding success was the number 
f ebeks produced per nest in three-egg clutches, 
ggs in all years w'ere considered lost or 
depredated if not found after the initial visit, or 
f found broken. Eggs which did not show signs of 
hatchmg 6 days after hatching of the las, eg g in 
he clutch were considered hatching failures. 
Statistical Analyses .—The exact laying dates 
ould not be ascertained in 2005 (nest visits every 
second day), and differences in incubation period.; 
between egg-types should be interpreted with 
caution. Study nests in 2006 and 2007 were 
selected upon hatching of the first egg in the 
clutch, and length of incubation could not be 
calculated. Hatching success was calculated in ali 
three seasons but breeding success was calculated 
only in 2006 and 2007 because of our failure Ic 
monitor all eggs within clutches in 2005. The 
observed clutch size upon nest selection in 200c 
and 2007 may have been different from the 
original size at the end of laying due to egg loss 
and predation. Considering this potential source or 
bias, hatching and breeding success in any year 
were calculated using only three-egg clutches, and 
hatching intervals between eggs from the same 
nest were not calculated in two-egg clutches from 
2006 and 2007. 
Data from eggs measured in three-egg clutches 
in 2006 and 2007 (151 eggs from 55 clutcho 
were used for a general description of egg size and 
to investigate egg size variation within and among 
clutches. Egg volume is a composite of egg 
breadth and length, and it was used in the 
investigation of factors affecting egg size in 
three-egg clutches. Generalized linear raised 
models (GLMM) were built using the function 
"Inter” in the *Tme4” package (Bates and 
Maechler 2009) of R (R Development Core Team 
2009). Variables investigated were "hatching 
order” (first, second, third), “year” (2006. 
2007), and their two-way interaction. Nest 
identification number (NEST ID) was included 
as a random effect to account for correlated 
variation among eggs in the same nest (Pinheira 
and Bates 2000). Model building was based on 
maximum likelihood (ML) and the final model 
was fitted by restricted maximum likelihood 
estimation (REML) (Zuur et al. 2009). /’-values 
were calculated by Markov Chain Monte Carlo 
simulations in R using the “coda” package 
(Plummer et al. 2009) due to unresolved issues 
regarding estimation of the degrees of freedom 
associated with fixed effect coefficients in mixed 
models (Bates 2006). The significance of the 
models was assessed using Akaike's Information 
Criterion (A1C; Akaike 1974). Variables were 
removed unless they reduced the AIC by mom 
than two units when included (Burnham and 
Anderson 2002). 
Source ol variation in egg measurements 
(breadth, length, and volume within and among 
clutches) was estimated by variance component 
