Abstract
Background: Although uniform stimulus magnification often compensates for eccentricity dependent sensitivity loss, crowding is frequently cited as a refutation of magnification theory. However, if one assumes multiple sources of eccentricity-dependent sensitivity loss then changes in crowding with eccentricity may be characterized simply in terms of non-uniform magnifications with eccentricity (Latham & Whitaker, 1996, OPO). Method: In three experiments we measured size thresholds for relative target/crowder separations of 1.25, 1.70, 2.32, 3.16, 4.31, 5.87, 8.00 and ∞ times target size. The sizes of target and crowders were varied uniformly to find the size eliciting threshold-level performance. Thresholds were measured at 0, 1, 2, 4, 8 and 16° in the lower visual field. The three tasks were grating orientation discrimination (Latham & Whitaker, 1996), T orientation discrimination (Tripathy & Cavanagh, 2002, VR) and letter identification (Pelli et al. 2007, JoV). We plotted target size at threshold as a function of separation at threshold. Results: In all cases size thresholds at fixation were independent of target/crowder separation. In other words, there was no effect of crowding at fixation. At all other eccentricities log(threshold size) decreased roughly linearly as log(separation) increased until asymptote was reached, at which point size thresholds were independent of separation. However, the rate at which threshold size decreased with separation increased with eccentricity and in some cases reached a point at which a critical separation was achieved; i.e., separation at threshold was independent of target size (Pelli, 2008, COIN). Conclusions: Although there are systematic changes in the size/separation curves from fovea to periphery there seems to be a qualitative change between fixation and periphery: at fixation size thresholds are separation independent and at the furthest eccentricities separation thresholds approach size independence. Contrary to our expectations multiple linear magnifications seem inadequate to characterize the data.