A spatial description often includes a target object that is being searched for, a reference object that serves as the starting point for the search, and spatial terms that describe the spatial relation between the two objects. It has been frequently assumed that speakers select reference objects that are salient relative to the available contrast set. Of interest to the current research is determining which salience dimensions speakers attend to during the process of reference object selection. In two experiments, participants were asked to provide spatial descriptions for given targets in displays that included two reference object candidates. These candidates were manipulated along conceptual, perceptual, and spatial dimensions to determine which are computed by speakers during reference object selection, and the relative importance of each type. Results suggest that while perceptual salience was considered most important by participants, all three forms of salience were computed during reference object selection.