The reviewer is mistaken. The actual text reads " [the apolar cooperative contour] produces no effect [as opposed to no feedback] back down at image level, because there is no contrast signal available across the contour to generate the brightness percept." In other words the top-down feedback from the apolar contour image is indeed fed-back down to the image layer, but its influence only becomes visible when there is a contrast across that boundary to trap the diffusing brightness signal.

This concept is demonstrated perfectly clearly in figure 17 (a-c) as compared to figure 17 (d-f). The computational algorithm for those two examples was exactly identical, the only difference between those simulations being the pattern of the input image. One produces a modal contour, whereas the other does not, but both produce an amodal contour.

Although the apolar cooperative image does not contain any information about what the colors should be at the image level, it does contain the information that there may be a color contrast across that boundary, and therefore feedback of this information to the image level is quite properly expressed as a boundary signal that would block the diffusion of brightness signal across that boundary, if such a brightness signal should be present.