David R Badcock, J Edwin Dickinson, Jason Bell,Serena Cribb, Hugh Riddell

Purpose:
To examine the specific hypotheses thatlocal shape coding is sensitive to contour curvatureand that multiple shape-templates are involved inprocesses leading to global integration of contourinformation in human vision

Method:
Two-interval forced-choice (2IFC) psycho-physics is used to determine the thresholds for shapedeformation in radial frequency (RF) targets (circledistorted by periodic modulation of the radius). Threeconditions are compared (sinusoidal, negative-, andpositive- rectification of the modulator).2 × 2IFC methodology is used to compare the thresh-old for modulation detection to that required toidentify different patterns. Identical thresholds implyseparate labelled lines for the shapes. Conditions:1) detection and identification of the modulatortype, 2) detection of modulation in multiple radial-frequency-component contours (Multi-RF) and thethreshold for identifying their mirror image equiva-lents and 3) detection and identification of modulatorsof different periodicity.

Results:
Psychophysical observers (n = 4) used toshow that modulator rectification supports globalcontour integration but not identification at thatthreshold. Integration is obtained with multi-RF con-tours but identification of mirror image alternativesrequires more modulation. Detection and identifica-tion thresholds can be identical for RF patterns varyingin modulator frequency.

Conclusion:
Modulator rectification in RF contourschanges curvature but not the results: curvature con-tinuity isn’t essential for integration. Integration waspreserved with multi-RF contours but mirror imageswere not identified at detection threshold: thesepatterns are not represented by different templates.The polar angle between two points of maximumcurvature was sufficient to allow identification atdetection threshold: patterns could be labelled for thismeasure.