Previous research has suggested that exposing a witness to a ‘misleading’ facial composite can affect identification accuracy. In practice it is the process of constructing a facial composite that may have the most effect on witness memory. In the present study, participants were asked to construct a facial composite but were actually presented with a pre-prepared sequence showing a ‘misleading’ composite being constructed. Participants were then shown either an immediate or delayed target-absent identification task, containing a foil similar to the misleading composite. The results of this study are discussed in terms of the malleability and durability of witness memory.