Environmental conditions, dispersal lags, and interactions among species are major factors structuring communities through time and across space. Ecologists have emphasized the importance of biotic interactions in determining local patterns of species association. In contrast, abiotic limits, dispersal limitation, and historical factors have commonly been invoked to explain community structure patterns at larger spatiotemporal scales, such as the appearance of late Pleistocene no-analog communities or latitudinal gradients of species richness in both modern and fossil assemblages. Quantifying the relative influence of these processes on species co-occurrence patterns is not straightforward. We provide a framework for assessing causes of species associations by combining a null-model analysis of co-occurrence with additional analyses of climatic differences and spatial pattern for pairs of pollen taxa that are significantly associated across geographic space. We tested this framework with data on associations among 106 fossil pollen taxa and paleoclimate simulations from eastern North America across the late Quaternary. The number and proportion of significantly associated taxon pairs increased over time, but only 449 of 56 194 taxon pairs were significantly different from random. Within this significant subset of pollen taxa, biotic interactions were rarely the exclusive cause of associations. Instead, climatic or spatial differences among sites were most frequently associated with significant patterns of taxon association. Most taxon pairs that exhibited co-occurrence patterns indicative of biotic interactions at one time did not exhibit significant associations at other times. Evidence for environmental filtering and dispersal limitation was weakest for aggregated pairs between 16 and 11 kyr BP, suggesting enhanced importance of positive species interactions during this interval. The framework can thus be used to identify species associations that may reflect biotic interactions because these associations are not tied to environmental or spatial differences. Furthermore, temporally repeated analyses of spatial associations can reveal whether such associations persist through time.