Friday 21 May 2010

Reading Group - Gibson (1979) Chapter 5 Part 2

Chapter 5 Part 2

In this part of Chapter 5, Gibson talks about how various relations between objects can be specified by the ambient optic array. This is important because image-based theories assume that we have to make a lot of inferences to detect, for example, that one object is partially obstructing the view of another object. In contrast, Gibson once again goes looking for regularities in the ambient optic array that can distinguish one relation between objects from another.

The change between hidden and unhidden surfaces: Covering edges

A strange property of our experience is that we perceive the layout of the environment, not just the projected surfaces. This means that we have some idea that hidden objects persist. It isn’t that we have theories about the persistence of objects. Rather we perceive the layout of an environment over time and get information about the continuation of surfaces via the optical specification of edges that separate surfaces. An important component of this experience is what Gibson calls “the principle of reversible occlusion” (p. 77). This means that things that go out of sight when we travel in one direction come back into sight when we travel in the opposite direction. This principle applies to all scales, from the very small to the very large.

Projected and unprojected surfaces

An occluded surface is simply a surface that is out of view. Occlusion happens because surfaces are usually opaque (see chapter 2) – that is they reflect and absorb light – and because the environment is fairly cluttered (see chapter 3) with objects, enclosures, and so on.

From a particular point of observation, surfaces that have a visual solid angle in the ambient optic array are projected. Surfaces that don’t have a visual solid angle in the optic array are unprojected. If a surface is moved very far away, its visual solid angle will reduce to a point. If a surface is turned away from the point of observation, its visual solid angle reduces to a line. If a surface goes out of view by our moving through the environment so that the surface is behind something (or by it moving behind something) then the visual solids angle is wiped out. This last point is worth taking some time over.

Going out of and coming into sight

It’s important to distinguish going out of and coming into sight from disappearing and appearing. When something goes out of sight, it is nonetheless observable from somewhere else. If we reverse our motion, the thing will come back into sight. But, when something disappears, we can no longer see it, no matter where we go. When something disappears, its visual solid angle vanishes suddenly from the optic array. But, when something goes out of sight, its visual angle changes gradually with respect to our motion.

Self-occlusion and superposition

Think of a book (or any object). We can see some, but not all of its surfaces. Books are self-occluding in that part of it will always obstruct the view of the rest of it. The book also occludes things in the environment. Maybe it’s covering up part of the desk on which it rests, or maybe it’s blocking my view of the rest of the room when I hold it up to read it. Self-occlusion and superposition are general properties of objects. It’s reasonable to think of objects as having near sides and far sides. As we move the object (or move about it) the near side becomes the far side and vice versa. We see the object move via perspective transformations. These are perfectly reciprocal, so that the near side of an object turns into the far side just at the same time as the far side becomes the near one.


Imagine observing a solid object sitting in front of some surface (e.g., a ball in front of a wall). As you move from side to side the wall is progressively covered and uncovered by the edge of the ball. Because the surface of the ball is unchanged, we know that the ball is in front of the wall. And, because the surface of the wall is gradually deleted and accreted, we know that the wall is behind the ball.

The information to specify the continuation of surfaces

Now imagine you’re sitting having your cereal in the morning. On the counter sits the box of cereal, and next to it sits a pitcher of milk. From your seat you can see both the box and the pitcher. If you stand up and move around you might find that at certain points the box occludes the pitcher. When this begins to happen you will be able to see the entire facing surface of the box, but only some of the facing surface of the pitcher. If you drew a picture of the outline of the pitcher at this point it would look pretty odd. One side of the outline would look like a normal pitcher, but the other side would terminate in a straight line, where the box obstructs your view. When this happens, how do you know that the pitcher continues behind the box? Why doesn’t it seem like the pitcher is broken or transformed? Gibson suggests that the key to perceiving continuity in object structure is the progressive shortening of textures or forms. In this example, the pitcher doesn’t suddenly disappear. Rather, the solid visual angle is gradually deleted. This specifies a continuous object surface. The same logic explains how we perceive the continuation of background surfaces as well. In the above ball / wall example, the ball occludes a portion of the wall. However, we still know that this portion of the wall exists, even though we can’t see it.

The case of very distant surfaces

When we see something go out of sight at the horizon (say, a car on a very flat road), the solid visual angle is gradually deleted. But, the way in which it is deleted is different from the process described in the previous section. That’s because the horizon isn’t an occluding edge (at least for terrestrial objects). Instead, it’s the location at which a distant object goes out of sight. When the pitcher was occluded by the cereal box, the visible parts of its surface looked just like the surface of an un-occluded pitcher. But, as an object recedes into the horizon, we perceive every part of that object becoming smaller and smaller, all at the same time.

These insights are very important because they show how complex relationships between objects in the environment can be specified. Those with a cognitive bent are quick to find jobs for the brain to do. But, Gibson shows that a vast amount of this work is done directly by perception. This is good because perception is much less likely to be fooled (or flat-out incorrect) than conception. Concepts are important and necessary to a lot of what we get up to, but they are also probabilistic and messy.

Until now we’ve taken it for granted that light can be structured, without really considering the detail of how this might occur. Gibson tackles this question in the remainder of Chapter 5.

No comments:

Post a Comment