FACTORS LIMITING LARGE-SCALE LOCALIZATION

The mechanisms mediating relative spatial localisation in the visual s ystem are still unclear. There is a growing amount of evidence that th is capability is not merely limited by the processing of the front-end visual system. Models of localisation should, therefore, include high er-level processing stages. A careful study of the sources of error in localisation tasks may further our understanding of the nature of the se processes. A study is reported in which the possible role of higher -order processing in relative spatial localisation is explicitly addre ssed. For this purpose the error sources of threshold performance were investigated for two similar relative-spatial-localisation tasks: two -dot separation discrimination and two-dot orientation discrimination. Fovea-centred stimuli with large dot separations were used. The front -end processing for these stimuli is probably identical in both tasks. Hence, differential effects of the variation of the experimental para meters on threshold performance for both tasks may reveal the characte ristics of the higher-level processing involved. The effects of dot se paration, stimulus orientation, and experimental procedure (single-sti mulus binary forced choice versus two-alternative forced choice) on th reshold performance for both tasks are reported. The results show that thresholds for both tasks increase proportionally with dot separation . However, separation-discrimination thresholds are always significant ly higher than orientation-discrimination thresholds. Thresholds for s eparation discrimination are independent of stimulus orientation. In c ontrast, orientation-discrimination thresholds show an oblique effect: thresholds are consistently lower for horizontal stimuli. Both tasks also show a different dependency of threshold behaviour on the experim ental procedure. For a horizontal stimulus orientation, separation dis crimination is better with an explicit (physical) reference standard, whereas orientation discrimination is better with an implicit referent . These differential effects cannot be explained by any of the known c haracteristics of the front-end visual system. They suggest that large -scale spatial-localisation performance is probably limited at a proce ssing level at which spatial relations are explicitly represented.