viewers naturally expect to find forest-related objects. Trees are interpreted as trees, snow as snow and branches as branches. If someone instead announces that a face is hidden in the picture, the viewer receives a new expectation and begins searching for face-like arrangements. A simple instruction can therefore change what receives attention even though it does not alter a single pixel.
Hints work for the same reason. Telling someone exactly where the hidden face is located can make the puzzle almost trivial because attention immediately moves toward the relevant region.
