The Connecticut warbler (Oporornis agilis) is an uncommon Neotropical migrant that breeds in the north-central United States and south-central Canada. Breeding populations of this species are reported to be declining. I analyzed habitat and landscape at three spatial scales (buffer radii of 100 m, 500 m, and 1,000 m) for 86 sites within 28 forest stands in northern Minnesota for Connecticut Warblers sampled over an 18-year period. I regressed combinations of habitat variables with two response variables, Connecticut Warbler abundance (the total number ever recorded at a site or stand) and Connecticut Warbler frequency (the number of years recorded out of 18 years) using a zero-inflated negative binomial distribution and logistic regression, respectively. A subset of models with #1;AICc ≤ 4.0 was retained and model-averaged predictions were calculated for each combination of buffer size and response variable. When comparing model-averaged predictions to observed data, the best models were those using Connecticut Warbler frequency at the 1,000 m buffer (r2 = 0.52). These models were used to create a map of predicted Connecticut Warbler breeding habitat in the two national forests sampled. At the 1,000 m scale, Connecticut Warblers were positively associated with large, simple patches of upland coniferous and lowland black spruce forest, and were negatively associated with upland deciduous forest.