Collinearity improves alignment Response to Keeble and Hess (1998)

reported that collinear carrier orientation had no effect on the alignment of Gabor patches. This null conclusion, part of a broader paper on orientation effects, is not necessarily fully jus-tified––(a) we failed to replicate it, (b) it is not particularly well supported by their own data, (c) predictions of the effect suggest it is likely to be small, and (d) it is unwise, in view of accepted modes of scientific inference, to draw any strong conclusions from a null result; at best one can estimate the size of the effect within a certain confidence interval. Previously, we have found that collinearity improves alignment accuracy Aiming to investigate this apparent contradiction between our results and theirs, we attempted partial replication of Keeble and Hess (1998). We tested the same stimuli in a sample of 20 observers. There was a highly significant (p < 0:0002) 20% reduction in thresholds when the patches were collinear (mean threshold 3.2 arcmin) rather than orthogonal (mean threshold 4 arcmin). Our result shows that 2D localisation is not based on envelope information alone. Moreover, the results for the different observers (including those from Keeble & Hess) are well fit by summation of variances from envelope and carrier cues, assuming a carrier-only threshold of 4.5 arcmin––little worse than the 4 arcmin envelope threshold. How do we align Gabor patches? We might use their centre, the outside envelope, or––when the internal carriers are aligned––the relative offset of the bars inside the patches. The dominant theory has been that we use some measure of central tendency, and that the accuracy with which we can perform this alignment depends entirely on the scale or space constant of the Gaussian It has also been claimed that the relative offset of the bars inside the patches is never used for their alignment measured the acuity with which the central of three collinear patches could be aligned with the outer two, over a range of contrasts and spatial frequencies. The single observer (RFH) then repeated some of these conditions for a configuration where the central patch was orthogonal, and this second set of data is shown to roughly overlay the first set. data show a small effect of collinearity in two or more observers, these authors nevertheless maintain ''the fundamental result that thresholds are the same in the aligned and orthogonal conditions'' (Keeble & Hess, 1999). Keeble and Hess (1998) found an effect of …