Observation Is Theory Laden

Today's topic is that observation is theory-laden. That is a point that should have been made much earlier, around the time that we discussed holism. The reason that Godfrey-Smith postponed it is that he wants to advocate a naturalistic approach to the problem.

There are many versions of the point that observation is theory laden. They are all intended to pose problems for simple empiricism. In each case, the idea is that not only does what we observe influence what we believe (that is, our theories, paradigms, ...), but that what we believe influences what we observe. It's a two-way street. That poses a challenge to every form of empiricism from that of the ancients through that of the logical empiricists. The idea on which simple empiricism is based, that what we observe is independent of and prior to what we believe, is wrong.

To defend the idea that science has a large empirical component, Godfrey-Smith attacks the arguments that science is theory laden. I think that that is, in general, a mistake: observation is very much theory laden. We need to find a way to describe why, nonetheless, experience is important to science. But I'm not going to burden you with my philosophy of science. I'm just going to point out problems with Godfrey-Smith's arguments.

First, what are the ways in which observation is theory-laden?

I'll start with the uncontroversial ways and work toward the disputed ways.

1. What we observe is, in important ways, determined by our interests and what we believe.

It is a good idea to look for things we aren't sure about, but that isn't so easy given that our theories and interests suggest where to look. If you only look at what you know your theory will explain, you're never going to "test" your theory. Popper made related points.

Examples:

• We focus more on what we are interested in. We know more about human disease than kangaroo disease.
• Our theories tell us what is worth looking for. Before the Copernican theory, no one tried to measure the distance to the planets.
• One's theory tells where to look for an explanation. (Feyerabend) At various times people have attributed hysterical pregnancies to
  • impregnation by the devil
  • early childhood experiences
  • hormonal and neurochemical imbalances

2. How we describe what we observe is, in important ways, determined by what we believe.

Standard Example

Hansen talks about Copernicus and Tycho looking in the same direction early one morning. Tycho sees the sun rise; Copernicus sees the horizon fall.

We often have no way of describing observations that is neutral between theories. That can be a serious problem while comparing theories, since observations must often be described in ways that differ for different theories. That casts doubt on whether descriptions of observations, including the observation sentences beloved of logical empiricists, can be used to decide between theories. Descriptions of experiments and observations do not occupy neutral common ground.

Probably more philosophers than not have concluded that that is a problem that can be worked around, though it does pose a difficulty for logical empiricism.

3. The ways in which we are careful depend on our theories.

We need to avoid errors in science. What kind of errors you think need to be avoided depends upon your theory. This is starting to pose a serious difficulty for the idea that we can use observations to decide between theories.

Example

Before Einstein's theory of special relativity, it was believed that the universe was filled with ether. Theories of light predicted an "ether wind" caused by the motion of Earth. Scientists set out to measure it, devising an ingenious instrument called an interferometer for the purpose. After each failure, a more sensitive interferometer was tried. The instruments were so sensitive that the vibrations caused by a horse trotting by outside overwhelmed them. To shield from vibrations, an interferometer was built in a building in a quiet part of a college campus and placed in the basement, surrounded by bedrock. The experiment still failed: no ether wind was detected. The set-up was criticized for blocking off the ether wind by surrounding the interferometer with rock; the experiment had a fatal flaw.

What counts as a good, careful experiment is often different for different theories. As a result, a single experiment can�t be used to compare two theories. Experiment does not provide a neutral common ground between theories.

4. What counts as a result depends on our theories. It is only when a theory tells you that a result makes sense that you believe it.

Instead of an argument, I'll give an examples:

• Famine reveals incredible shrinking iguanas
• Special relativity predicts that there is no ether wind, and, as a result, the "failed" experiments that served as an example for the previous item became a great success: they were now "crucial experiments" that "proved" special relativity. The "fatal flaw" has been more or less forgotten—since, according to the new theory, special relativity, there is no ether wind, there is no need to worry about blocking it.

This is serious. It is pretty much an alternative way to put Kuhn's point about anomalies. The simple empiricist view is wrong. Experiments never prove theories. (That may be an exaggeration, but not as big a one as you might expect.)

I'm not claiming that Kuhn's explanation is correct. I'm just saying that solving the problem is harder than Godfrey-Smith acknowledges. It is always possible to reject an experiment, instead of rejecting a theory it appears to contradict.

5. Not just high-level interpretation and experiments, but what we just plain observe (that is, see) is not neutral. It depends on what we believe.

Everything we see is seen according to our interpretation.

Examples

The duck-rabbit

• Looking through telescopes and microscopes takes practice. At first you see nothing, then eyelashes, then the barrel. Once you finally manage to see anything at all, it doesn't look like anything much until you learn to pick out what is expected: a cell wall with a thin membrane around each cell, a nucleus, an elliptical galaxy, and so on. Once you learn to see such things, you see them just as directly and automatically as you see tables and chairs. (Feyerabend)
• A bird identification field manual with drawings is much better than one with photographs: a drawing emphasizes the features you are supposed to notice.
• Biologists thought for decades that humans have 48 chromosomes when in fact we have only 46. Now that that is known, it is easy to see that the extra pair was just the result of breakage.

That is a huge and important effect that Godfrey-Smith attempts to deny. It is an important part of why Kuhn says shocking things like "switching paradigms is like a religious conversion." The whole world, quite literally, looks different. Empiricism must take account of the fact that adherents to different views see differently. The idea that even observation in the sense of ordinary sense perception provides common, neutral ground is an oversimplification.

6. Even at the preverbal level of perceptual processing, what we perceive is influenced by prior experience and knowledge.

Example

The Müller-Lyer Illusion

The blue line looks longer to most people, but most of them nonetheless recognize that it isn't. As in the previous point, what we observe is theory laden: we know to correct in certain circumstances.

The present point is different: With practice, the illusion actually disappears. The two lines look the same length. Artists learn to judge colors in a way few of us can. Backpackers learn to judge long distances in a way few of us can. Clothing-store clerks can tell someone's waist size at a glance. Practice and probably belief modifies the very way we perceive. Since what practice we have itself depends upon belief and theory, that means that even the most basic aspects of perception are theory laden.

• Your auditory system is designed to pick up phonemes of your native language.

• Here is an example in which what you see is changed by knowledge:

Free JavaScripts provided by The JavaScript Source

---

Godfrey-Smith argues that the last version of theory ladenness and perhaps the one before (5 and 6) aren't real, by giving Fodor's argument about perceptual modules. Fodor says the our perceptual processing is made up of hard-wired processes, and they in some cases make systematic errors, but

1. There is nothing we can do about them
2. They are the same for everyone

Fodor uses the Müller-Lyer illusion as an example, and we have seen that it doesn't work. It may well be that it is true that we have some hard-wired modules, but we also have parts of the perceptual mechanism that are modified by experience, belief and expectations.

---

One reason Godfrey-Smith argues against theory ladenness seems to be that he thinks that the strong forms of theory ladenness are an obstacle to empiricism. It does torpedo simple empiricism, the idea that experiment, or at least observation, or at least sense data is neutral with respect to our theories, and so it can help decide between them.

Given that that is false, how can experiment and observation decide between theories (or paradigms, ...).

1. Experiments must be designed with who they are trying to convince in mind, just as arguments must be: one must experiment from common ground.
2. Experiments must be designed so that the data are easy to observe.
3. How can we resolve the question of how theory laden various things on the list are? Naturalism provides an answer: by using scientific methods. The psychology of observation and the sociology of agreement are both relevant.

-- ShaughanLavine - 19 Oct 2005