Although the North American Breeding Bird Survey (BBS) is a principal source of information regarding populations of most North American bird species, many features of the survey complicate analysis of population change. Correlation studies based on BBS data cannot be used to unambiguously define cause and effect relationships. Recently, B?hning-Gaese et al. (1993) presented an analysis of population trends in insectivorous songbirds using data from the BBS. They concluded that predation has played an important role in influencing population trends. We review 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 alternative analyses of the same BBS data, we demonstrate that the evidence that predation is associated with population declines is weaker than they suggested. Based on our analyses the only factor among those tested that is consistently associated with population trends is migration status (i.e., short-distance migrant/resident vs. long-distance migrant) during the period 1978-1987. Also, we present evidence that the harsh winters of the mid-1970's severely depressed populations of short-distance migrant species, and may be responsible for the observed associations between migration status and population trends.