The investigation was carried out in Dera Ghazi Khan and Multan districts which are located along the Indus river system. Presence/ absence data were collected from 160 crop fields of wheat and cotton during 2005. We used classification as well as ordination methods to analyze the data structure. The main emphasis was on classification, ordination being used in part to check whether the classification results reflect in an adequate way the main floristical gradients in the data set, and also to detect relations between some environmental factors and the composition and structure of weed vegetation. The major axes brought by the ordination were related to soil characteristics and anthropogenic factors. The application of multivariate statistics {detrended correspondence analysis (DCA) and cluster analysis} allowed an interpretation of weed species spatial distribution and assemblage. From the results of classification (cluster analysis) samples from wheat and cotton fields could be divided in to distinct groups (plant associations) based on floristic composition. We have shown that there is a clear relationship between physical and chemical features of the substrate and the association defined by numerical analysis which surely shows that these groupings of weed species are not arbitrary assemblage. In this respect, soil moisture and texture seems particularly important in shaping the weed communities. Again, the scatters of the ordination diagrams demonstrate the lack of boundaries between weed vegetation types and species distribution: some species were restricted, but the majority had very wide ranges across the associations. Although, the difficulty of describing vegetation boundaries may be general one for weed vegetation yet these series of associations make good sense and may be the most useful of any land-use application. However, weed ecologists must continue to attempt the difficult definition of hazy boundaries of weed vegetation.