What is required for a signal to be qualified as a 'grouping' tag?

In our commentary paper (Dry, Kogo, Putzeys, & Wagemans, 2010) on 'The utility of image descriptions in the initial stages of vision: a case study of printed text' (Watt & Dakin, 2010a), we raised a few concerns about the approach to perceptual grouping proposed by Watt and Dakin. Specifically, we argued that the 'overlap tokens' resulting from convolution with Gabor filters do not reflect the global configuration of the image and hence that they would not be able to reproduce cases of context-sensitive perception such as illusory contours in the Kanizsa image. In their reply to our commentary, Watt and Dakin (2010b) showed that the responses of their model do reflect the illusory contours. In the present brief commentary paper, we explain why their data are problematic. The crucial problem is that illusory contours are not mere borderlines drawn in the gap between the pacmen. Instead, the perception of the illusory contours corresponds to the existence of border-ownership signals, which reflect the global configuration of the image.