Summary In this experiment, the role of linguistic context in verbal production has been investigated. The context consists of a list of 10 nominalizations of either the subject or the object of the action. The subject is asked to supply, as rapidly as possible, an incomplete phrase. The results have shown that the completion has the same structure as the previous linguistic context, i. e. subject or object nominalizations. The authors have interpreted the results within the light of the distinction between the deep and surface structures, the only distinction which enables to differentiate nominalizations with an identical surface structure (as it was the case in the reported experiment) and which suggests that distinct responses are given...