Michael Morgan is not attempting to define or explain consciousness and in the end is hostile to those who wish to. This book focusses on the visual system and how we process 2D and 3D information in the brain. It is charming and packed with insights. This was for me, a useful review of the visual system with new information.
Michael writes in almost a folksy way with wonderful little asides that I liked. For example, the parenthetical translation from Latin here: "The powerhouse for our eye movements lies in a region of the midbrain called the superior colliculus (little hill)." p. 43. And "saccades after an old French word for the movement of a cart with square wheels" p. 141. He has evidently wandered much of the former British Empire with references to the U.S., New Zealand, Australia and Canada in fun little asides, such as talking about representing images with Gabor patches: "The landscapes of images, it turns out, are boring and repetitve, more like East Anglia or the American Midwest than Switzerland". p.111.
I may be overstating this, but Morgan makes the case that analog computers (animal neurons an synapses) are qualitatively different and hence, digital computers will never be able to model animal neurons precisely. In the end of the book he takes jibes at many aspects of the attempts to understand consciousness, taking a direct jab at Koch and his Quest for Consciousness - even if names are not named. p. 188 "They call this (rather pompously) the search for 'neural correlates of consciousness'", which Koch abbreviates as NCC. He is not as intellectually diligent as Koch, but, less daunting to read.
Morgan uses a lot of examples from Viz while talking about maps. And maps are the name of the game in our visual system. Long quote from p. 149/150
The conclusion is that there are several maps of space in the brain. To begin with we have the relatively simple sensory maps in areas like the primary visual cortex (V1), which are little affected by whatever plans we might have to act on the basis of the information they provide. These maps tell us about the position of objects relative to our direction of gaze, nothing more. They are gradually transformed by the addition of information about eye and limb position into other maps, closer to the planning of movement. Some of these transformations take place in downstream structures like the colliculus, but they also happen in the multiple maps of the parietal lobe. The parietal lobe contains a bewildering variety of specialised maps, which seem to continue their number with further research. As two authorities on the parietal lobe recently put it: 'The space around us is represented not once but many times in the parietal cortex'. But we should not get carried away by the idea of thousands of maps with difference reference frames. Keeping the maps in register is an important problem, and to a large degree the visual system solves this problem by the simple device of hanging on to an eye-centered frame of reference for as long as possible.
V1 - an integrated map for line angles, direction of movement (p.101) However, there is not atempt to organize this information, which was disappointing.
The "icon buffer" lasted several tenths of a second. A flash to cue attention can come after the stimulus and still be effective. p. 159
7 ways to portray depth - without binocular vision p.47:
Lot's of nice color plates - with binoculor rivalry very well illustrated.
V2 has a recognisable map of the visual field not unlike that of V1, but less precise. Each cell has a large receptive field - p. 41
V5 p.58 "some recent evidence .. for maps of binocular parallax"
LIP p.145 - lateral intraparietal area (monkey) is a strong candidate for an eye movement map. "If a monkey is planning to an eye movement from A to B, nerve cells in LIP begin to fire in a location corresponding to B, even before the movement is made."
"frontal eye field" - motor maps to produce eye movements.
1) shape from shading ( and confusing shadows with shading? ) anyway, it works on a small scale at least.
2) perspective distortion of geometry
3) texture - change of scale
4) occlusion
5) upward sloping ground plane
6) familiar size - why architectural renderings have people in them, man is the measure of all things
7) atmospheric perspective - aka depth attenuation of OpenGL fog.
August 2006