Although the North American Breeding Bird Survey (BBS) is a principal
source of information regarding populations of most North American bir
d species, many features of the survey complicate analysis of populati
on change. Correlation studies based on BBS data cannot be used to def
ine cause and effect relationships unambiguously. Recently, Bohning-Ga
ese et al. (1993) presented an analysis of population trends in insect
ivorous songbirds using data from the BBS. They concluded that predati
on has played an important role in influencing population trends. We r
eview aspects of the analysis methods for estimating population trends
(e.g., observer effects, data subset) and for associating mean trends
with species attributes (e.g., confounding of attributes). Using alte
rnative analyses of the same BBS data, we demonstrate that the evidenc
e that predation is associated with population declines is weaker than
they suggested. Based on our analyses the only factor among those tes
ted that was consistently associated with population trends was migrat
ion status (short-distance migrant/resident versus long-distance migra
nt) during the period 1978-1987. Also, we present evidence that the ha
rsh winters of the mid-1970s severely depressed populations of short-d
istance migrant species and may have been responsible for the observed
associations between migration status and population trends.