For example: When you see a photograph of someone with a venerable castle visible some distance away in the background, both images represented in two dimensions on the same flat piece of card, you do not see the person as several hundred feet tall posing alongside a castle, nor as normal sized but with a small model castle immediately over their shoulder. Your brain does not simply accept the data with which it has been presented, it interprets it; in this instance probably correctly. It makes an assumption, based on what it knows of the world, and tells you that the castle is a really big building which only looks small because it is far away. It was not told this by the data on the piece of card, though; the photograph, in itself, did not make this clear. The photograph actually showed a castle which, in relation to the person sharing the picture, was tiny.

A far more flamboyant demonstration of this is the wonderful Ames room illusion. Here, a carefully proportionately distorted room is built. The illusion is viewed from one point, through a small aperture in one end wall of the room. From this viewpoint the room appears absolutely ordinary. The wall to the left of the viewer, though, is actually much longer than the one to the right and gradually becomes higher, maybe reaching twelve feet. The wall to the right is much shorter and tapers to under six feet high. The far right corner of the room is much closer to the viewer than the left. The walls, the floor and ceiling, the door and the (fake) windows, are all distorted to match such that, when viewed from the proper vantage point, the room appears to be a run of the mill, rectangular living room. The proportions seem absolutely normal, viewed from that point. Two apparently rather short women stand talking in the far left hand corner. They are much the same size. Then one moves a few paces across the room and stands against the far wall, to the right of the viewer. As she does this she grows, grotesquely, before our eyes. Standing by the right wall she now utterly dwarfs her companion only a short distance from her - she must be at least nine feet tall. Her hair brushes the ceiling and she stands slightly stooped to avoid banging her head on it. Her size almost doubled, just like that, right in front of us! Moving back to the side of the other lady she contracts to the same size as her, and they chat on.

Our mind had two impossibilities to choose between. Either the room was crazily distorted, or the woman, like Alice, just grew and grew. From the point we have been obliged to look, though, the room seems absolutely as all living rooms always are, so the only solution is that we do, indeed, have a woman who can smoothly and rapidly change her size just by walking about. Sherlock Holmes once told the long-suffering Dr Watson that ‘When you have eliminated everything else, whatever remains, however impossible, must be the truth’. We obviously ‘see’ whatever the brain decides must be the truth, however ridiculous that might be, all else having been eliminated as even more impossible.

And the truth is that visual information is hardly ever simple, clear, well-defined and unambiguous. In most real life circumstances, as researchers into vision have found to their cost, visual information is disjointed; a fiendishly complex and intricately mixed jumble of colours, intensities, angles and surfaces, and, to make matters considerably worse, often varying from one moment to the next. Most of the images we are presented with (and which we decipher so easily and accurately) are imprecise, with huge quantities of detail much of which is irrelevant, much novelty, much confusing context and often considerable movement. Equivocal stuff! How on earth do we make sense of it all?

Vision – is it data-driven or concept driven? Clearly the mind must be driven at least in the direction of the correct decision as to what it should see by the actual visual information itself, by the incoming data per se, by the data arriving from the eyes. If this were not so, of course, we would never be able to distinguish anything from anything else, we would not be able to tell a sow from a sofa, by sight alone. The mind's choice as to what is seen is, therefore, at least partly what we can call data-driven, arrived at by force of incoming data. This is bottom-up processing, whereby the raw data at the ‘bottom’ of the process is deemed to be what decides outcome.

However, we can think along a bit and see that our mind has to analyse data and impose a structure on it in order to reach a meaning. Our mind must act on the data if only to reduce it to manageable, understandable amounts. Our mind, though, must do more than just separate relevant from irrelevant data, indeed unless it analyses raw data & begins to form hypotheses about what its meaning might be how will it be able to tell which data is relevant? In considering figure 4.1 you will recall that, despite a fair amount of visual information, you were unable to reach any meaning because you lacked a good hypothesis but that once you had such a hypothesis to test against the incoming raw data you ‘saw’.