Despite this lag, saccade performance remained unaffected even when the saccade target appeared only during the time in which the gain field incorrectly reflected pre-saccadic rather than post-saccadic eye position (i.e., 50 to 150 ms after the end of the previous saccade). The authors reason that if an inaccurate eye-position gain field is used to compute saccade target location, then saccade behavior should also be inaccurate. The authors’ striking observation of normal saccade performance despite inaccurate eye-position

signals therefore provides evidence that gain fields are not—indeed cannot be—utilized in computing target locations for eye movements. If gain fields are not updated rapidly enough to be used in neural computation, what is the alternative model? A signal indicating a change in eye position could be delivered to GW-572016 concentration LIP and the updated vector

computed in some other manner. It is clear that receptive fields are remapped (Duhamel et al., 1992; Colby and Goldberg, 1999). Nevertheless, the alternative to the gain field model has only been characterized in phenomenological terms; a remaining challenge is to develop it into a mechanistic model (Mauk, 2000). The specific version of the double-step task used by Xu et al. (2012) differs from the classic paradigm in an important respect that may have influenced check details their behavioral results. As previously Phosphatidylinositol diacylglycerol-lyase mentioned, in the typical double-step paradigm, two saccade targets are presented sequentially in time with a distinct temporal gap between them. This design eliminates the presence of allocentric spatial cues that subjects could use to help localize the final saccade target. For example, if both saccade targets in Figure 1 are presented simultaneously, then subjects

might simply memorize the spatial relationship between A and B (e.g., B is to the right of A). After completing the initial saccade to A, subjects can then simply generate a saccade vector (A→B) that matches the stored allocentric representation of A and B. Indeed, Dassonville et al. (1995) demonstrated that the presence of allocentric spatial information during target presentation reduces (although does not completely eliminate) standard localization errors in the double-step task. It is then potentially problematic that Xu et al. (2012) employ a stimulus configuration that seemingly provides exactly this kind of allocentric spatial cue. In their version of the paradigm, both of the saccade targets (as well as the initial fixation target) were simultaneously present on the screen for a full 75 ms before the monkey was instructed to move. This additional spatial information could potentially improve accurate spatial localization performance and thereby mask mislocalization effects due to inaccurate eye-position signals. It could also explain why the findings reported by Xu et al.