Carving up the world

by Neil Rickert

It is said that we carve up the world at its seams.  I doubt that there are any seams.  We carve up the world in ways that are easy enough and that we find useful.  But those requirements — that it be easy enough and that it be useful — underdetermine how the world is to be carved.  So it is a matter of pragmatic decision making.

As we saw in my last post, carving up the world is what gives us the entities that we can talk about and is what allows us to say true things about the world.

I should say at the outset, that carving up the world isn’t an entirely conscious and deliberate activity.  Much of the work is done behind the scenes by our perceptual systems.  So, in part, this post is related to how perception works.  So when I talk about us carving up the world, I am not restricting this to conscious activity.

Why it is hard

We cannot just look around and see what are good ways of carving up the world.  To be able to look around and see, then what you are looking at has to have a lot of detail.  But the detail that we see gets there because of how we carve up the world.  So we cannot presuppose that it is available before we do any carving.

Here’s a way to think about it.  Suppose you are blindfolded and in a dark room.  You feel around, and find some obstacle.  You step back.  Then you try again.  And again you find an obstacle.  And it seems to be the same obstacle.  So now you can mentally divide (carve) the room into what is to the left of the obstacle and what is to the right of the obstacle.  Next you probe around in the part to the left of the obstacle to see if you can find more detail.

The idea is that you find some detail that you can find again.  Use that to divide.  Then repeat for each of the parts of that division.  And that allows you to fill in more detail.

Our vision system must work something like that.  The eye moves around in motions called saccades.  So the part of the world that affects a particular retinal sensor is, in effect, moving around or probing the environment.  If that eye movement crosses a boundary, that will result is a sharp change in the signal received by the retinal sensor.  That’s pretty much like locating an obstacle in the room, except it is done by retinal sensors instead of by our arms.  And that, in turn, allows dividing the environment into what is to the left and what is to the right of that boundary.  And then further probing allows further dividing and adds detail resulting from the way that we have divided up (or carved) our immediate environment.

The problem for vision, is that our vision system doesn’t know where in the world it is facing.  So the stimulation of a retinal cell, by itself, tells us very little about the world.  We move as we look around.  And we rotate our heads.  These all add to the difficulty of making sense of a retinal cell being stimulated.  By using a method of scanning for boundaries in the world, the vision system becomes less affected by our motion and by our uncertainty about our position.  Instead of our motion being a problem, the vision system uses motion (with saccades) as a tool for locating boundaries.  And it uses those boundaries for anchoring us to the world, for identifying things relative to the world even though we are moving around.

If we open our eyes and start looking around, the vision system is going to first locate the easiest to find boundaries.  And then it can find other boundaries as it adds detail in the repeated dividing or carving.  This happens very rapidly for us, so we don’t notice it.  For a newborn infant learning to use its eyes in an unfamiliar world, we should expect this to be a slower process.

Because of our reliance on boundaries, we tend to carve up the world around boundaries.  And, as a consequence, we tend to carve it into what we see as objects.

Criteria and transducers

When we first find a useful way of carving up our immediate environment, how can we be sure that we do it about the same way the next time?  It would seem that we need to identify the parts using some sort of criteria.  The criteria might depend on characteristics such as shape, color, etc.  And then on the next occasion where it is useful, we can apply those same criteria to recognize the same parts that we have previously carved out.

J.J. Gibson, in his work on perception, has argued that we use what he calls “transducers” to recognize objects.  So there might be a cat transducer that picks out cats, a chair transducer that picks out chairs, etc.  A transducer would work by applying criteria.  So my suggestion that we apply criteria is similar to what Gibson argued.  Gibson argues that our perceptual systems tune these transducers to invariants.  So we don’t memorize exactly what a cat looked like the first time.  We find characteristics of that cat that we can detect, even when the cat moves around or squats down.

The thing about using criteria, or using transducers, is that they won’t pick out or recognize just a single thing.  If the criteria that my perceptual system uses can pick out the  family cat, then it will probably also pick out some other cats in the neighborhood.  So we aren’t actually picking out individuals.  Rather, we are picking out categories — the category of all things that fit the criteria that we are using.  And I see this as the reason that categories are so important to us.


In this post, I have discussed the basic ideas behind carving up the world.   In an earlier post, I indicated that how we carve up the world needs to be a social convention, so that we all do it in similar way.  But I did not get into how we can have such carving conventions.  That will be a topic for a future post.

2 Responses to “Carving up the world”

  1. I find a lot of relevant overlap here with a previous post of mine on Predictive Processing.

    I think you outlined quite well how some of these processes come together to create an ontology of some kind. Nice post!



%d bloggers like this: