558 
THE WILSON JOURNAL OF ORNITHOLOGY • Vol. 123, No. 3, September 2011 
71° 31' W). The station encompasses 33 ha of 
deciduous and early successional forest patches 
fragmented by abandoned agricultural fields. The 
fields have undergone a variety of maintenance 
since 1956 when banding began. Woodlands are 
currently dominated by American ash (Ftaxinus 
americana), red maple (Acer rub rum ), and several 
species of oaks (Quercus spp.). Shrubs and 
dominant fruiting species in the understory and 
edges along canopy openings include common 
greenbrier {Smilax rotundifloria), arrowwood 
viburnum (Viburnum dentatum ), oriental bitter¬ 
sweet (Celastrus orbiculatus), European privet 
(Ligustrum vutgare ), and fox grape (Vitis lab- 
rusca). American pokeweed (Phytolacca ameri¬ 
cana), Virginia creeper (Parthenocissus quinque- 
folia), and common winierberry (Ilex verticilluta) 
are present in isolated patches in the understory 
and canopy openings. 
Banding Protocols .—We summarize 45 years 
of banding records between 1960 and 2007, 
except for 1995-1997 when the station was 
closed. Banding was conducted each day between 
7 August and 31 October, except during inclement 
weather (steady rain, high wind, below freezing 
temperatures). D. L. Kraus operated four mist nets 
(30-mm mesh, 12-m length) for 5 hrs each day 
beginning at sunrise from I960 to 1994 with nets 
checked at 30-min intervals. We operated 10 mist 
nets daily using previously established protocols 
from 1998 to 2007. We recorded date, net 
location, time of capture, and classified age and 
gender of each captured bird. Unbanded birds 
were fitted with serially-numbered aluminum leg 
bands from the USGS, Bird Banding Laboratory. 
We included only initial captures and excluded all 
recaptured and previously banded birds from 
analyses. 
Climate Data .—We used data collected daily 
since 1893 at the University of Rhode Island 
Weather Station. <5 km from KWRS, to assess 
variation in the local climate. We used the 
maximum daily temperature ( C) and averaged 
this over the banding season (7 Aug-31 Oct) each 
year to calculate mean autumn temperatures. We 
used the mean autumn temperature to assess 
annual changes in temperature over the 48-year 
time span covered in the study. 
Data Analysis .—We first described the direc¬ 
tion and nature of the relationship between mean 
passage date and year for each species. We then 
used an information-theoretic approach to rank 
competing candidate models and examine the 
relative importance of average autumn tempera¬ 
ture and capture rates for explaining annual trends 
in mean passage dates. 
We categorized species as a long-distance 
migrant if it primarily overwintered south of 
North America, or as a short-distance migrant if it 
mainly wintered in southern North America. We 
analyzed data from 11 species of long-distance 
migrants and eight short-distance migrants (Table 1). 
We only considered a species for analysis if at least 
five individuals were captured in a given year. 
Thus, years with rt < 5 captures were excluded for 
that particular species. We included a species only 
if >50% of banding years between 1960 and HOOT 
were retained and if the species did not have >5 
consecutive years of data when there were <5 
captures annually (with the exception of Blackpoll 
Warbler [Dendroica striata]). Capture effort dif¬ 
fered among years (fewer nets were operated prior 
to 1994, and nets were not operated during 
inclement weather), and effort varied unpredictably 
among years; banding was conducted over a 
similar range of dates during all years. We 
calculated the mean Julian date of capture for each 
species to assess annual variation in autumn 
migration passage dates. However, total number 
of captures in a particular season depended on 
capture effort (number of nets operated and number 
of hrs ot operation of each net), and we calculated 
annual capture rate for each species as the number 
of birds captured per 100 net hrs (1 net hr equals 1 
net operated for I hr) between 7 August and 31 
October. 
We used polynomial regression (PROC REG; 
SAS Institute 2002) to evaluate linearity of die 
relationship between year (independent variable I 
and three potential dependent variables; mean 
date of capture, mean autumn temperature, and 
annual capture rate. We transformed year data by 
converting each datum to its deviation from the 
mean (Xj — T) to address mulu’collinearity issue- 
associated with polynomial regression (Tabachnik 
and Fidel I 2001). We used an information- 
theoretic approach to examine if linear, quadratic, 
or cubic functions best described these relation¬ 
ships for a given species. We calculated second- 
order Akaikc's Information Criterion (AlCjand 
used Akaike weights (w,) to rank models (Bum- 
ham and Anderson 2002). Our objective was to 
assess patterns over time and we were not using 
these models for prediction, but we accepted the 
model with the highest Akaike weight as the best 
model. AIC,. model selection results indicated that 
