I conducted 167 unlimited-distance point counts in central Texas in Fe
bruary 1987 and January and February 1988 to determine how count durat
ion, time of day, site type (woodland perimeter, woodland interior), w
inter date (before or after 14 February), and year were related to the
winter detection probabilities of 13 species. Within a species, diffe
rences in detection probability among the levels of a factor helped id
entify the environmental (e.g., site type) or methodological (e.g., co
unt duration) conditions under which one could maximize both detectabi
lity and point-count sample size. Among the levels of time of day, the
detectabilities of each of several species were comparable. Thus, ins
tead of restricting censuses to the morning hours as is common in bree
ding-season studies, researchers could sample and obtain comparable ra
tes of detection for one of these species throughout the day. An impor
tant benefit would be larger point-count sample sizes, which would gen
erally improve the statistical power of subsequent analyses. Similarly
, detection probabilities did not differ among 5-, 10-, and 15-min cou
nts for several species; one could capitalize on this similarity by us
ing 5-min counts to maximize the number of counts during a fixed time
interval. Comparisons of detection probabilities among the levels of s
ignificant interaction effects demonstrated that a variety of levels,
involving combinations of main-effect levels, yielded like detection p
robabilities. Such combinations could be scheduled to maximize simulta
neously both point-count sample size and detection probabilities. Beca
use additional point counts and improved detectability would tend to y
ield data that are more accurate, these strategies are likely to incre
ase the accuracy of ecological inferences based on winter point-count
data.