Spatial localization without visual references

To explain the veridical percept of the spatial ordering of objects and the generation of eye movements to peripheral targets, Lotze (1885 Microcosmos. Edinburgh: T. & T. Clark) proposed that there is a position label (local sign) for each retinal element. To estimate the precision of local sign information, we measured absolute localization thresholds at various eccentricities in the nasal visual field, in the complete absence of visual references. To eliminate perception of the visual surround, observers viewed a large display screen through a neutral density filter (2.0 log unit) in a dark room. The fixation target was extinguished at various times (interstimulus intervals or ISIs) prior to the onset of the test stimulus. In general, our results show that localization thresholds are proportional to the target eccentricity at all ISIs. At each eccentricity, localization thresholds are elevated after the extinction of the visual reference compared to thresholds when the reference is present. However, relative to the referenced threshold, unreferenced thresholds are elevated by a greater proportion at smaller eccentricities than at larger eccentricities. Our threshold vs ISI data can be adequately modeled on the basis of an intrinsic positional uncertainty, which increases with eccentricity, and additive and multiplicative sources of noise. The additive noise appears to reflect primarily the increasing scatter in eye position when the fixation target is extinguished. Our model's estimate of intrinsic positional uncertainty in the isoeccentric direction appears to reflect primarily the intrinsic positional uncertainty of the peripheral retina (the local sign), being very similar to cumulative cone position uncertainty and to the spacing between ON-P beta ganglion cells. In the isoeccentric direction, the estimated precision of the local sign mechanism across eccentricities is slightly better than the precision of saccadic endpoints, suggesting that noise in the motor system must also contribute to the scatter of saccadic endpoints in the isoeccentric direction. Interestingly, in the radial direction, we find a surprising similarity in our observers' positional uncertainty and the precision of saccadic endpoints.