Reply to Popple and Levi ( 2002 )

we showed that alignment performance in a three-Gabor alignment task was impaired when patches which were oblique to the stimulus configuration were employed, rather than patches which were orthogonal or collinear. As part of this study we showed that for a wide range of configurations (all vertical, all horizontal , inner vertical with outer vertical, inner horizontal with outer horizontal) thresholds were the same within the limits of experimental error, thereby verifying and extending the results of Hess and Holliday (1992) and Kooi, De Valois, and Switkes (1991). The key point is that because thresholds are the same for collinear and orthogonal conditions the mechanism mediating this task cannot be primarily dependent on the positions or orientations of the internal structure of the patches. In other words, performance cannot simply be mediated by the responses of a large oriented linear filter overlying all three patches. This result, and the similar thresholds found using bull's eye and Gaussian patches instead implicates an envelope-based mechanism. Popple and Levi (2002) purport to have shown that there is, averaged across many subjects, a small improvement in performance for the three-patch alignment task where the patches are collinear compared to one instance when they are orthogonal, and take us to task for having violated ''accepted modes of scientific rea-soning''. There are some instances in scientific enquiry where establishing that a quantity or difference is identically equal to zero is of deep significance. One such is the question of whether the neutrino has a non-zero mass, because even a very small mass for the neutrino would bear on the issue of whether the Universe is closed (Perkins, 1982). There does not appear to be a particularly vital issue at stake in the result of Popple and Levi. It is well known that in the case of closely spaced micropatterns there is a large effect of micro-pattern properties on alignment performance, and that this effect falls off rapidly with increasing separation (Levi & Waugh, 1996). It is therefore rather obvious that at any separation a cue can be available which is dependent on internal patch properties, in addition to the envelope-based cue. An interesting question is how the fall-off in first-order cue utilization with separation can be modelled using linear oriented filters (Whitaker & MacVeigh, 1991). There are a number of methodological problems which make it difficult to take the study of Popple and Levi (2002) as …