Question Are patterns of weed co-occurrence structured or random? Do these patterns differ at the field and landscape scales? Are co-occurrence patterns related to the effect of environmental filters and/or competition? Location France. Methods Weed communities were surveyed in 3008 4-m2 plots, i.e. 32 plots per field in 94 winter-wheat fields located in the same landscape. We estimated patterns of species co-occurrence (C-score) at the field and landscape level. Additional characteristics of weed communities (degree of habitat specialization, competitive ability) and sampled fields (within-field and among-fields environmental heterogeneity) were used to interpret observed co-occurrence patterns at both scales. Results Non-randomness was detected in <20% of the fields; these fields were characterized by higher within-field environmental heterogeneity and contained a higher proportion of competitive species. This subset of fields partly drove the pattern of co-occurrence of weed species at the landscape scale. After removing this effect, specialist species were found to be aggregated and species that had an intermediate degree of habitat specialization were segregated, despite the lack of marked environmental gradients across the studied landscape. Conclusions Patterns of weed co-occurrence differed at the field and landscape scales. Weeds co-occurred mostly randomly within fields but were in some instances segregated as a result of environmental heterogeneity and possibly weedweed competition. At the landscape scale, aggregation of specialist species and segregation of intermediate species are likely to result from variations in crop management among the sampled fields.