Baubock el al. • CLIMATE CHANGE DOES NOT AFFECT PROTANDRY 
21 1 
unusual years. However, for nearly all species, 
less than 2°/c of birds were captured on the first 3 
or last 3 days of banding. The exception was the 
American Goldfinch (Spinas tristis) for which 5 c 7c 
of birds were caught within the last 3 days of 
banding, indicating that we may have missed the 
tail end of migration for this species. We believe 
this is unlikely to have substantially influenced 
the results. 
We used linear regression to characterize 
temporal changes in mean arrival dates for males 
and females of each species, difference in mean 
arrival dates between males and females, and 
the ratio of females to males. These regressions 
indicated whether males or females arrived 
earlier, later, or no differently over time, and 
whether the amount of protandry or the ratio of 
females to males changed over time. 
We used reverse stepwise regression to test the 
relationships of the mean arrival dates of males 
and females, differences in mean arrival dates, 
and the female-to-male ratio of each species with 
three climate variables: (1) mean temperature in 
Manomet, Massachusetts in the 2 months prior to 
the average arrival date for each species, (2) 
the NAO index, and (3) the ENSO index. We 
included a lagged temperature term (the temper¬ 
ature in the 2 months prior to migration during the 
previous year) when analyzing changes in lernalc- 
to-male ratios as a way to test the effect of the 
previous year's temperature on sex ratios. 
We rated the seven species according to 
several ecological characteristics, including num¬ 
ber of clutches, territoriality, feeding and breed¬ 
ing behavior, breeding habitat, and length of 
migration (Table I). Our categorizations were 
based on banding data at Manomet Center for 
Conservation Sciences (population changes, 
mean arrival date, female-male ratio, average 
arrival, and average wing-chord length) and a 
review of relevant literature. Wc also calculated 
the average size of males and females of each 
species based on measurements of wing length, 
and calculated the difference between males and 
females. We evaluated the importance of these 
(actors by grouping species by each characteristic 
and used ANOVA to compare the effect of each 
v ariable on each group. We considered all results 
w here P < 0.05 to be statistically significant. Wc 
used sequential Bonferroni correction to guard 
against Tvpe-I error (Rice 1989) in cases of 
multiple tests. 
RESULTS 
Protandry. —Six of seven species studied had 
significant protandry; males arrived earlier than 
females by 3.6 days on average each year (paired 
/-test, two-tailed P < 0.001) (Table I). Only 
Baltimore Orioles had no significant difference 
between the average arrival times of males and 
females (paired /-test, two-tailed P = 0.57). 
However, the amount of protandry varied widely 
among the species studied: male American 
Redstarts arrived 2.1 days earlier than females 
on average (paired /-test, two-tailed P < 0.001), 
whereas male Black-and-white Warblers ( Mnio - 
til la vuria) arrived 5.S days before females on 
average (paired /-test, two-tailed P < 0.001). 
None of the ecological variables examined- 
winlering range, breeding range, number of 
clutches per season, population status, average 
sex ratio, and mean difference between male and 
female arrival dates significantly explained vari¬ 
ation in the amount of protandry observed. We 
tested whether males have significantly longer 
wings, and are larger, than females. Males were 
significantly larger than females in all species 
studied (paired /-test, two-tailed P < 0.001) 
(Table I). However, the magnitude of size 
difference did not explain the amount of protandry 
observed l / = 0.30, P — 0.52). 
Migration Times. —Neither males nor females 
of any of the species we examined arrived 
significantly earlier over time as measured by 
linear regression after correcting for the number 
of tests (// = 14). The difference in arrival times 
between males and females also did not change 
significantly over time for any species during the 
period studied (Fig. 1). 
Male American Redstarts, after correcting for 
multiple tests, were the only group for which 
arrival dales w'ere significantly related to any 
climate variable with males arriving earlier in 
warmer years (reverse stepwise regression. P = 
0.002) (Table 2). However, prior to correcting for 
multiple tests, seven of 14 tests had relationships 
with P < 0.05. which is unlikely to have been 
caused by chance alone (y 2 - 59.7. df = 1, p < 
0.001). suggesting the sequential Bonferroni cor¬ 
rection may he overly conservative. Five of the 
seven species, before the correction, had a pattern 
where arrival dates for one sex were related to a 
climate variable, while arrival dates of the other sex 
were unrelated to climate i Table 2). Despite these 
differences, the arrival dates of males and females 
