Further evidence for global orientation processing in circular Glass patterns

Further evidence for global orientation processing in circular Glass patterns We recently published a paper demonstrating that concentric Glass (1969) patterns are processed by global, linear summation of concentrically organized orientation information throughout large receptive fields about 4.4° in diameter (Wilson, Wilkinson, & Asaad, 1997). Dakin and Bex (2002) have recently challenged our results and claim to find no psychophysical evidence for global concentric orientation summation in circular Glass patterns. As we show in this letter, the situation is quite the opposite: Dakin and Bex (2002) have actually replicated our major evidence for global concentric orientation summation in Glass patterns almost exactly, although they missed this fact by not fitting power functions to their data. The solid black circles in Fig. 1 replot mean data for the four subjects in our original study (Wilson et al., 1997). In this experiment the Glass pattern was no-tionally divided into 12 pie-shaped wedges, only some of which (4, 6, or 12) contained signal dot pairs, while the remaining segments were filled with noise dots at the same mean density (6%) as the signal sectors. Thresholds (75% correct) were then measured for discriminating patterns containing signal dots from random dot patterns as a function of the area filled with signal (see original article for further details). All data were well fit by a power law with an exponent close to)1.0, the mean exponent for the data being)0.91. Thus, we found almost perfect linear summation of concentric orientation information throughout an area which we estimated from other experiments to be about 4.4° in diameter. Similar linear summation results for a motion analog of Glass patterns have been reported by Morrone, Burr, and Vaina (1995). We then constructed a computational neural model that provided an excellent fit to the data and was compatible with properties of V4 units in ma-caques discovered by Gallant, Braun, and VanEssen (1993). Both the large pooling area and the almost linear global summation of orientation information are incompatible with any reported properties of V1 units. Dakin and Bex (2002) also conducted an area sum-mation experiment with concentric Glass patterns. Instead of restricting the signal to pie-shaped wedges, however, they placed signal and noise in alternate rings or vertical bands of varying width (see their Fig. 6). The open symbols in Fig. 1 plot the mean results for SCD (an author) and TM from panels e and f of their figure. They …