Theory-loadedness and a theory of concepts

The following was originally written in May 1990, as a final paper for a course taught by Fred Dretske in the philosophy department at Stanford University, dealing with the issue of whether perception is "theory-loaded".

During the course, we have discussed various theories of perception, and their implications for the issue of the theory-loadedness of observation. We have not, however, devoted similar attention to theories of concepts. I submit that, to understand the issue, an examination of theories of concepts is crucial. In this paper, I will take for granted a direct realist view of perception, and show how different views of concepts would lead to different conclusions on the issue of theory-loadedness.

The problem

Granted that many observations do depend on a theory, the question of theory-loadedness is: do there exist some observations that are independent of theory? If there aren't, then knowledge can never be grounded in observation. Any instance of cognitive perception will depend on some theory; if that theory is justified by other instances of cognitive perception, then those will depend on another theory; and so on, leading either to an infinite regress or a circle, never allowing an observational foundation for knowledge.

At first, simple perceptual judgments, such as "this table is brown", may seem like the obvious candidates for a theory-independent foundation; on the direct realist view, we directly perceive objects, and their attributes such as color, independently of any theory. Closer examination, however, reveals that things aren't so simple. A statement like "this table is brown" does more than report your perception of the object and its attributes; it also involves conceptual knowledge, for example your knowledge of the concept "brown". By calling the table "brown", you are saying that it is, in some respect, similar to other objects --- objects which have different shapes and different colors, but there's some similarity in the colors which justifies classifying all of them under the concept "brown"; this is prior knowledge, a theory, which you needed to have before you could have made this statement.

How, then, can we find a theory-independent foundation in observation? Clearly, your theory of concepts will determine your approach to this problem.

Realism

Let us first look at the most ancient theory of concepts, dating back to Plato: realism [1]. This is the view that our concepts, such as "brown", refer to some metaphysical essence of "brownness", which exists in reality independently of us --- either in some other dimension (Plato's version) or in all brown objects (Aristotle's version).

Realism implies what we can call a "discovery", as opposed to a "construction", model of concept-formation [2]. When we form a new concept, its meaning is determined, not by the definition we give it, but by the objective kinds of things it picks out in the world; we do not impose concepts on reality, but rather, get them from reality.

Realism usually also implies a rationalistic approach to knowledge --- the idea that conceptual knowledge is not based on perception. Perception only brings us in touch with concretes, not with abstract essences (to use the classic example: you can see particular men, but you can't see "man" as such; or, to use the example from section 1: you can see specific shades of brown, but you can't see "brown" as such); so cognitive contact with these essences must be established by some other means. This is Plato's basic reason for adopting his theory of innate ideas, as presented in the Meno.

Of course, rationalism explicitly rules out any possibility for observations independent of theory. A rationalist answer to our "this table is brown" problem is that you really do need prior knowledge to use the concept "brown", and that you had to obtain this knowledge by means other than observation (in Plato's version, for example, you need to have seen the essence of "brown" before you were born, so that the brown table can evoke this memory).

Some realists do reject rationalism, and hold an empiricist view --- that all knowledge comes from observation. Two recent books that try to reconcile realism with empiricism are:

(1) Gibson's The Ecological Approach to Visual Perception. To achieve this reconciliation, Gibson uses the (to say the least) highly implausible theory, that abstractions are directly perceived, similarly to concrete entities and attributes.

(2) Kripke's Naming and Necessity. In Kripke's view, we form a concept thru our perception of some of its instances, and the concept's meaning is determined by the real essence those instances share with other instances (an essence that can then be discovered thru further observation and experiment, e.g. discovering that water is H₂O).

At first glance, this seems to answer our "this table is brown" problem; we had formed the concept of "brown" by perceiving some instances, i.e. some brown objects; those perceptions are the theory-independent base of our knowledge of the concept; so when we say "this table is brown" --- which means: "this table shares some real essence with other objects that we've called brown" --- our judgment does depend on a theory, but that theory was itself formed by previous, theory-independent perceptions, i.e. our first few perceptions of brown objects. [3]

Closer examination of Kripke's theory, however, shows that it does not really explain how a concept like "brown" can be formed. Contrary to Kripke, our perception of an object, and the fact that this object bears some real relation to other objects (on Kripke's view, a relation of sharing an essence) are not enough to form a concept. For example, Kripke says that "The original concept of cat is: that kind of thing, where the kind can be identified by paradigmatic instances" (p. 122); but how is this different from giving a name to some specific cat? In that case, too, we perceive the cat; the cat bears the same real relation to other cats; and we've designated it by a word; so what is it that makes the word refer to all cats in one case but only to one cat in the other case? In the same way (returning to our original example) what is it about those first observations of brown objects that makes us form the concept of "brown", rather than just invent a name for a certain brown object? Kripke does not answer this question --- and I submit that, on his realist premises, he can't possibly have an answer to it; and without such an answer, he has not explained how the knowledge, that we need for making a statement such as "this table is brown", could have been obtained by theory-independent perception. [4]

It seems, then, that realism and empiricism can't be reconciled. Realism logically leads to rationalism --- and, as a result, does not allow for theory-independent observations.

Nominalism

The traditional alternative to realism has been nominalism. This is the view that concepts, such as "brown", are names that we give to groupings of objects which, in reality, do not have anything in common --- groupings which are either arbitrary or based on some vague "resemblances".

As opposed to the discovery model of concept-formation, implied by realism, nominalism clearly leads to a construction model. Since the concept is not based on reality, there is nothing to discover; the concept's content has to be created by the definition we give it, by the user's criteria for including objects under it.

As a result, even though nominalism is usually associated with an empiricist approach to knowledge, this empiricism has to break down when it comes to the meaning of concepts. In explaining how we come to learn the meaning of a concept such as "brown", nominalists generally choose either conventionalism --- the view that the meaning of concepts is a result of social convention, which we learn primarily by learning language --- or innatism --- the view which goes full circle, starting with the opposite premises from Plato but arriving at the same conclusion: that the meaning of all our concepts is pre-determined at birth, by the innate structure of our minds. [5]

Whichever of these possibilities we choose, we again lose all hope of ever finding a theory-independent observational foundation for our knowledge.

Requirements for a solution

Neither of the two traditional theories of concepts can defend the existence of observations independent of theory. If we look back at the reasons for this, we can see what would be required of an alternative theory if it is to solve this problem.

(1) Empiricism. Clearly, if we are to find a foundation for our knowledge in theory-independent observations, then the meaning of the concepts we use must be based in observation. A theory of concepts, to solve the problem, must therefore be consistently committed to empiricism.

(2) A discovery model of concept-formation. If a theory is to explain how our conceptual knowledge can have a foundation of theory-independent observations, then it must hold that we get the meaning of concepts from reality, not supply it ourselves. At the same time, the theory must avoid the pitfalls of realism and of the rationalism it implies; the meaning of concepts must come, not from some abstract essence, but from the concrete attributes that we can perceive.

(3) Finally, a theory should identify some intermediate cognitive stage, between sense-perception and cognitive perception, and show how this stage can be independent of theory and can lead to the formation of concepts. We have already seen, in section 1, that cognitive perception, since it includes a conceptual judgment, can't be fully independent of theory. On the other hand, sense-perception can not, by itself, lead to conceptual knowledge; as we've seen in section 2, those realists who claim otherwise fail to make their view plausible. An intermediate stage, if its nature and role are precisely explained, can provide a way out of this dilemma.

One recent theory does fulfill these requirements: the Objectivist theory of concepts. [6] In the rest of this paper, I will briefly describe the Objectivist theory, and discuss its implications for the issue of theory-loadedness.

The Objectivist theory of concepts

Rand's brief summary of the process of concept-formation is:

Similarity is the relationship between two or more existents which possess the same characteristic(s), but in different measure or degree. The process of concept-formation consists of mentally isolating two or more existents by means of their distinguishing characteristic, and retaining this characteristic while omitting their particular measurements --- on the principle that these measurements must exist in some quantity, but may exist in any quantity. A concept is a mental integration of two or more units possessing the same distinguishing characteristic(s), with their particular measurements omitted. (p. 83; page numbers refer to the 2nd edition)

The key to concept-formation is the ability to be aware of objects as units:

When a child observes that two objects (which he will later learn to designate as "tables") resemble each other, but are different from four other objects ("chairs"), his mind is focusing on a particular attribute of the objects (their shape), then isolating them according to their differences, and integrating them as units into separate groups according to their similarities. ....

A unit is an existent regarded as a separate member of a group of two or more similar members. .... Note that the concept "unit" involves an act of consciousness (a selective focus, a certain way of regarding things), but that it is not an arbitrary creation of consciousness: it is a method of identification or classification according to the attributes which a consciousness observes in reality. (pp. 6--7)

Rand uses the term "unit" for a member of a similarity group, because, in her theory, it is deeply connected to the mathematical concept of a unit of measurement. What makes two objects similar is that their differences are regarded as quantitative, as different measurements of the same characteristic, while the differences of those two objects to some third object are regarded as qualitative. Difference, on this theory, is more basic than similarity; similarity is a kind of difference --- a difference regarded as quantitative, relative to another, qualitative difference.

Note that awareness of differences as qualitative or quantitative is relative to the context. For example, when seeing patches of brown, green and white, the brown and green can be seen as similar (i.e. as quantitatively different), both of them qualitatively different from the white. However, when seeing two patches of brown and a patch of green, the brown and green will be seen as qualitatively different (this is when the concept of "brown" is formed). In yet another context, the differences between two brown patches can themselves be seen as qualitative, e.g. if you see two patches of light brown and one of dark brown, and are forming a concept for "light brown".

This theory does successfully combine empiricism with a discovery model; it makes the concept a result of the concrete attributes of objects, and of selective attention on those concrete attributes, leading to awareness of quantitative vs. qualitative differences. The process of measurement-omission is the next step --- once you've become aware of a characteristic existing in two objects in different degrees, you can create a mental integration of all objects (including those that you've never seen) that have the same characteristic to any degree.

The theory identifies not just one, but several intermediate stages between sense-perception and cognitive perception. Selective attention on specific attributes of an object goes beyond simply seeing the object; awareness of quantitative vs. qualitative differences between various objects goes further; measurement-omission goes even further. And yet, none of these stages reaches full cognitive perception; none requires a conceptual identification of what we are perceiving.

When, in the process of concept-formation, man observes that shape is a commensurable characteristic of certain objects, he does not have to measure all the shapes involved it nor even to know how to measure them; he merely has to observe the element of similarity.

Similarity is grasped perceptually; in observing it, man is not and does not have to be aware of the fact that it involves a matter of measurement. (p. 14)

So, returning to the "this table is brown" question: when we make an observation and express it by the statement "this table is brown", we need some previously acquired knowledge, such as our knowledge of the concept "brown". This knowledge was acquired by previous perceptions, which included: sense-perception of several brown objects, and of some objects of different colors; selective attention on the attribute of color; awareness of the similarity (or quantitative difference) among the brown objects; and measurement-omission, classifying all the brown objects under the concept "brown". All of these steps were performed without the need for any theory, purely on the basis of what we perceived.

The Objectivist theory therefore provides precisely the foundation we need to get out of the theory-loadedness catch, and, for the first time, makes possible a full answer to this issue.

footnotes

[1] For the rest of this paper, I am using the term "realism" to refer to the realist theory of concepts, not to the realist theory of perception.

[2] These two terms are introduced by David Kelley inThe Evidence of ths Senses, p. 225.

[3] Note that, both on Gibson's and Kripke's view, we learn the meaning of concepts from what we referred to in class as "sense-perceptions", i.e. perceptions of objects. In this way, even if all "cognitive perception", or perception of facts, depends on some theory, sense-perceptions can still provide the theory-independent foundation.

[4] The above criticism of Kripke is taken from David Kelley and Janet Krueger, "The psychology of abstraction", Journal for the Theory of Social Behaviour, March 1984.

[5] A typical contemporary example of a conventionalist is Quine, for example in Word and Object. A typical example of an innatist is Fodor, for example in The Language of Thought.

[6] The theory is described in detail by Ayn Rand, in Introduction to Objectivist Epistemology. Its relation to other contemporary theories and issues is discussed by David Kelley in "a theory of abstraction", Cognition and Brain Science, Summer/Fall 1984.