This paper describes a model of adaptation to remapped auditory localization cues that is based on previous decision-theory models of psychophysical performance. The present model extends earlier work by explicitly assuming that past experience affects subject perception and by quantifying how training causes subjects' responses to evolve over time. The model makes quantitative predictions of total sensitivity, bias, and resolution for subjects involved in experiments investigating spatial auditory adaptation. One assumption of the model is that subjects cannot adapt to nonlinear rearrangements of localization cues, which is consistent with previous experimental reports in both audition (ShinnCunningham, Durlach, &Held, 1998b) and vision (Bedford, 1993). The model assumes that, in spatial adaptation experiments, subjects learn to interpret a continuous internal decision variable differently than normal; they do not learn to associate discrete stimulus-response pairs. This view is consistent with previous analyses of results from experiments investigating adaptation to visual rearrangement, as well as with the McCullough effect in vision (Bedford, 1993(Bedford, , 1995.There are many studies in which the effect of altering, or rearranging, sensory localization cues is investigated. In most such studies with vision, the perceptual effects of viewing the world through prisms that rotate the visual field are investigated. In the majority of such studies with audition, the question of how mean localization judgments change when the interaural axis is effectively rotated is investigated (see, e.g., Day & Singer, 1967;Kalil & Freedman, 1967;Mikaelian, 1969Mikaelian, , 1972Mikaelian, , 1974Recanzone, 1998;Willey, Inglis, & Pearce, 1937;Young, 1928). (For a review of auditory adaptation, see ShinnCunningham, Lehnert, Kramer, Wenzel, & Durlach, 1997; for a review of adaptation studies in general, see Welch, 1978.) In all such studies, naive subjects mislocalize sources in a direction predicted by the transformation imposed on the sensory stimuli reaching the eyes or the ears. If the subject is given appropriate training, however, these localization errors decrease with time. In most studies of sensory rearrangement, this training is achieved by allowing the subject to interact freely with the environment while observing the transformed sensory stimuli (e.g., by viewing their hand as they reach for a stimulus or by lis-
33tening to sounds as they walk toward or away from a sound source). To date, models of adaptation to both auditory rearrangements and rearrangements in other modalities have set out to describe qualitatively the changes in response error that occur with training. In general, these models (see, e.g., Ebenholtz, 1970Ebenholtz, , 1973Welch, 1978;Welch & Warren, 1980) do not allow quantitative predictions of changes in response error, nor do they address how resolution might be affected by changes in the adaptive state of the subjects. Rather, these models enumerate the factors that are thought to influence lo...