Activator proteins bind to enhancer DNA elements and stimulate the initiation of transcription. It has been proposed that activators contact general initiation factors at a promoter, and evidence for such direct interaction has been obtained. Studies of transcription in vitro, however, have suggested that activators might function through an intermediary molecule(s) distinct from the general factors. In the first of these studies, we exploited the finding that one activator could inhibit transcription stimulated by a second activator (activator interference or 'squelching'). This inhibition, which is attributed to competition between the activators for a common target factor, could not be relieved by addition of a large excess of general initiation factors, suggesting that the target for which activators compete is distinct from these factors. Similar conclusions came from the observation that TFIID's expressed from cloned genes fail to replace partially purified 'natural' TFIID fractions in supporting activation, evidently because they lacked some component present in the impure fractions. While these lines of evidence for a novel 'mediator' of activation were negative, we also showed that a partially purified fraction from yeast would reverse activator interference. This positive effect of a presumptive mediator provided an assay for its activity, but its role in activation was still only inferred. We now present direct evidence for a mediator which is required for stimulation of transcription in vitro by the activators GAL4-VP16 and GCN4, but which has no effect on transcription in the absence of activator protein.