O’Brien and Ritchison • NON-BREEDING ECOLOGY OF SHRIKES 
361 
four centrally located rectrices using dental floss. 
Radio-marked shrikes were tracked using a 
receiver (Model TR-4; Telonics Inc., Mesa, AZ, 
USA) and a 2-element Yagi antenna (Telonics 
Inc.. Mesa, AZ, USA). 
We observed the hunting behavior of shrikes 
from 1 January to 31 March 2005 and 1 
November 2005 to 31 March 2006 with all 
observations between 0800 to 1600 hrs. We 
recorded perch type (e.g., tree or utility wire), 
perch height (estimated using previously mea¬ 
sured heights, e.g., height of fence posts), time 
spent on perches, whether the shrike attacked or 
gave up (flew from perch without initiating an 
attack) and, for attacks, the outcome (successful, 
unsuccessful, or unknown) for each hunting 
attempt. We attempted to identify prey to the 
lowest taxonomic level possible for successful 
attacks. 
All shrike locations were marked on aerial 
photographs (Kentucky Office of Geographic 
Information 2004) to generate territory maps. 
We obtained fewer locations of shrikes without 
transmitters and, therefore, all analyses of territo¬ 
ry size and composition were based on data from 
radio-marked shrikes. The six most frequently 
used hunting perches (n = 3 continuous and 3 
isolated) were selected in each territory and their 
characteristics were compared to six randomly 
selected, apparently unused perches. Random 
perches were selected by placing a numbered grid 
over each territory, using a random numbers table 
to select six grids, then, at the center of those 
grids, selecting the nearest apparently suitable 
perch. 
Hunting perches were categorized as either 
continuous (utility wires, fencerows, woodlot 
edges, and tree-lines; >20 m in length) or isolated 
(isolated trees, shrubs, and snags). Continuous 
perches were analyzed by placing four 20-m 
transects parallel to the perch. Each transect was 
separated by a distance equal to one-half of the 
perch height (e.g., for a 2-m tall fenceline, the first 
transect was I m from the fence and all transects 
wer e I m apart) and. along each transect, we 
sampled vegetation at 5-m intervals. Isolated 
perches were characterized using four transects 
radiating from the perch at angles of 36, 72. 108, 
and 144 (i.e., a 90 area in the direction the 
shrike most often faced). Transect lengths and 
distances between sample points (n = 5) were 
based on perch height (i.e., one-half perch height). 
F °r example, if an isolated perch was 4 m tall. 
sample points were at 2-m intervals beginning 2 m 
from the perch (total transect length = 10 m). 
Transect directions relative to a perch were based 
on the direction a shrike most often faced when 
hunting. A direction was chosen randomly, if 
there was no directional preference, using a 
random-numbers table to select a starting point 
(i.e., isolated perch) or flipping a coin (i.e., 
continuous perch). 
Characteristics of vegetation were described 
following James and Shugart (1970). We mea¬ 
sured litter depth, vegetation height, foliage cover, 
and type of ground cover (bare ground, grass, 
herbs, or litter) at each point. Foliage cover was 
recorded by counting the number of hits (stems or 
leaves within I cm of a 1.5-m pole) at intervals of 
<0.5. 0.5-1, and 1 — 1.5 m. A densitometer was 
used to quantify ground cover. 
Habitats in territories of radio-marked shrikes 
(n = 7) were categorized as hay field (>95% 
grasses or herbaceous plants), pasture (>95% 
grasses or herbaceous plants and grazed by cattle), 
old field (S5% and ^50% woody vegetation with 
the rest grasses and herbaceous plants), woodlot 
( wooded area at least 3 trees wide by 3 trees long 
with tree spacing <6ml, and crop field (com or 
tobacco). The number of potential perch sites and 
impaling sites in territories was ascertained by 
counting the number of shrubs and saplings 
(woody plants >1 m and ^4 m in height) and 
trees (woody plants >4 m in height), and by 
measuring the total length of utility wires, fence 
rows, and tree-lines (row of trees and shrubs with 
a canopy ^5 m wide) or woodlot edges (wooded 
area with canopy ^ 5 m wide), 
Territory (n = 7) boundaries were delineated 
using the locations of hunting perches plotted on 
aerial photographs (Kentucky Office of Geo¬ 
graphic Information 2004). We constructed poly¬ 
gons by connecting the outer-most points and 
areas were estimated using Terrain Navigator Pro 
7.0 Kentucky (Maptech Software, Billings, MT, 
USA). Territory characteristics were compared to 
those in randomly selected, apparently unused 
areas. Seven unused areas (n = 3 in Madison 
County and 4 in Garrard County) each encom¬ 
passing 85 ha (the mean size of territories in our 
study) were randomly selected. Unoccupied areas 
were selected using aerial photographs (Kentucky 
Office of Geographic Information 2004). Each 
image (46.5 km 2 ) was assigned a number and 
selected using a random-numbers table. Areas 
were further divided into 85-ha squares by 
