How do speakers plan utterances? Horton and Keysar (1996) suggested that speakers initially plan utterances without regard for their intended addressee. However there are a number of concerns over the methods, which they used to test this. They used a between-subjects design, a confederate listener, simulated co-presence and little roleswapping in a referential communication game. In the current experiment 24 speakers described hidden targets to real listeners in the presence of a context shape that was either visually co-present to the pair or privileged information to the speaker. The time pressure with which speakers initiated their utterances was manipulated. Speakers were more likely to rely on context related descriptors when context was shared by their listener and increasingly so when they were under no time pressure to initiate their utterances. Findings suggest that with shared visual co-presence and increased role-swapping speakers take the real needs of a listener into account when planning utterances. Speakers may not be as egocentric as Horton and Keysar (1996) suggested. Successful dialogue involves cooperative interlocutors. Speakers are thought to adhere to the “co-operative principle” and Gricean Maxims, which dictate that they attempt to provide listeners with unambiguous, honest and “suitably informative” utterances (Grice 1975, cited in Horton & Gerrig, 2002). Much effort in psycholinguistics has been spent recording instances where speakers accommodate the listener at the lexical (Brown & Dell, 1987; Levelt, 1981), prosodic (Brennan & Williams, 1995; Schober & Brennan, 2003;) and syntactic levels (for a review see Lockridge & Brennan, 2002, Temperley, 2003). However, debate remains concerning “when and how” the words, intonation and form of speaker’s utterances come to reflect a listener’s needs (Lockridge & Brennan, 2002).