Summary1. Numerical studies of the relationship between birds and their habitat are important because they provide understanding of the impacts of natural and human factors on avian diversity. However, collinearity between explanatory variables and spatial autocorrelation can hamper the detection of key environmental factors underlying birdenvironment relationships identified by traditional regression approaches. This study utilized two alternative statistical methods to address these difficulties in biodiversity modelling. 2. We examined bird abundance patterns, spatial structure and relationship to environmental factors in an agricultural-forest mosaic landscape in Finland. We used data from 105 grid squares each 25 ha in size. Using variation partitioning and hierarchical partitioning methods, we determined the independent and joint effects of habitat cover, landscape structure and spatial variables on the total number of bird pairs and that of agricultural bird pairs. 3. The explanatory variables highlighted as important predictors of bird patterns by the two methods generally coincided well. The total number of bird pairs was negatively related to agricultural land, and positively to cover of forests and landscape heterogeneity. However, a clear majority of the explained variation in bird patterns was related to the joint effect of predictors, and the independent contributions of predictors were small. The univariate importance of landscape heterogeneity decreased greatly if the habitat cover variables were considered simultaneously. 4. Most of the explained variation in the number of agricultural bird pairs was related to the joint effects of the explanatory variables. In addition, the independent effect of habitat cover variables was considerable and agricultural birds showed a positive relationship with semi-natural grasslands. 5. Synthesis and applications . Variation partitioning and hierarchical partitioning approaches provide deeper insights into bird-environment relationships than traditional regression methods. This is particularly so when they are employed in a complementary manner. Our results indicate that a major part of the spatial structure in bird patterns in agricultural-forest mosaics can be caused by the clumping of habitats either preferred or avoided by birds. Moreover, at a scale of 25 ha, the abundance of bird pairs is not necessarily related to landscape heterogeneity as such, but depends more on the distribution of the most important habitats for birds.