Figure 1.

(a) schematic diagram of our task. Results are shown in (b–d) – see text for description. PSFs were estimated by probit analysis. Observers denoted by initials show authors, remainder are naïve volunteers. All error bars show 95% confidence limits except for (d) where the lower 90% are also shown (longer horizontal) as here we have a clear directional prediction. Results collapsed across observers are shown in (d) denoted by “All”. Here error bars are derived from the t distribution (t(4) = 4.43, p = 0.01), elsewhere errors bars were determined using bootstrapping.