Though communicative goals are an important element in language production, few studies investigate the extent to which these goals might affect the form and content of referring expressions. In this study, we directly contrast two tasks with different goals: identification and instruction giving. Speakers had to refer to a target building nearby or further away, so that their addressee would distinguish it between other buildings (identification) or give route directions and use the same building as a landmark (instructions). Our results showed that irrespective of goals, the referring expressions consisted of the same types of attributes, yet the attribute frequency and formulation differed. In the identification task, references were longer, contained more locative and more post-nominal modifiers. In addition, referential choices were influenced by the visual distance between the speaker and the target: when the speaker observed the target from far, references were longer and contained more often locative modifiers.