This paper presents a proposal for refining the methods of semantic description based on corpus collocational data. The central tenet of our proposal is that the conventional description of collocation as a bipartite structure, along with the binary contrast of node and collocate, is inadequate for explaining certain forms of interaction observed in patterns involving more than two lexical items. Our contention is that, in addition to node and collocate, a third category (co-collocate) is required to account for those cases in which the cohesiveness of a collocational pair is not self-contained and the node-collocate attraction shows dependency on an element external to the pair itself. After explaining some necessary theoretical distinctions, we will present a step-by-step description of the methodology. Then we will illustrate how it can be applied to the analysis of co-collocational patterns found in the vicinity of the word consequence.