The Nature of the “Given”

Carnap claims that the “problems of the so-called given or primitive data” are purely verbal: one answers it by specifying what kinds of words occur in observation sentences. Ayer appears to strongly disagree, but there is one important point of contact, one which Carnap probably would have viewed as primary. Carnap speaks of specifying the words that occur in what are specified to be observation sentences, while Ayer talks about adopting a certain language: according to Ayer, we can describe what we observe in the language of material things, that of sense data, that of “multiple appearances,” and so forth; according to Carnap, we can take "chair" to be an observation term or "chair sense data," or "chair as seen from a particular location sense data." That is pretty much the same point that Carnap was concerned to make: there is no matter of fact about whether we directly observe material things or sense data, it is just a matter of, as Carnap would put it, adopting a linguistic convention, or, as Ayer would put it, of choosing a language. Carnap emphasized that to the exclusion of a very important point, which is the source of the disagreement Ayer is emphasizing: it is not a matter merely of convention and language which observation sentences we endorse. That is an empirical matter dependent on what we are given in awareness. Carnap didn’t adequately take that into account. Nonetheless, there are important overlaps between Ayer and Carnap, overlaps that Ayer’s exposition tends to hide.

According to, for example, Russell, there can be sense data presented of which we are not aware. Also, our expectations can change the way we see the very same type of sense data, and two sense datums can appear to be identical, even when they are not.

Here are the examples:

• We can look in a cluttered drawer and not notice what we are looking for even though, once it is pointed out, we can see that it was in plain view. That is taken to show that the appearance of what we were looking for was part of the initial sense data, even though we didn't see that appearance.

• One can arrange that a black piece of paper and a white one reflect the same amount of light (shine a light on the black one, and leave the white one in shadow). If one hides them behind a screen with two holes in it, they appear identical. But, under more normal circumstances, the black one continues to look black, and the white one continues to look white.

• One can produce three green tiles, A, B, and C, tiles that are very close in hue, so that A and B appear to be the same color, B and C appear to be the same color, and yet A and C appear to be different colors. One can tell exactly the same story I just told about the tiles about their appearances.

If one takes sense data to be physical, something like, what is shining on the retina, these examples are perfectly good. They, in fact, are informative for the scientific psychology of perception. However, since Ayer is adopting sense-data language precisely to abstract from any theory of what causes what we are aware of, this misses Ayer’s point. He denies that these descriptions are appropriate to the sense data. The book is called Foundations of Empirical Knowledge. The idea is that we start with what is the fundamental data on the basis of which we obtain our empirical knowledge and clarify the process by which we obtain that knowledge, and knowledge of retinas, light, memory, and so forth clearly comes after how things look to us.

For the purposes of Ayer's project, what the examples show is that

• We can sometimes be aware of different sense data when in the presence of what we take to be the same objects from the same perspective, and we can sometimes be aware of things in memories of sense data that are not in the original sense data.
• Sense data is not only a function of the objects we take the sense data to belong to, but also of our memories, expectations, surrounding sense data, and so forth.
• We can experience sense data of objects A, B, and C so that the sense data in the presence of both A and B are sense data of the same type, the sense data of B and C are of the same type, while the sense data of A and C are of different type. Given the previous item, there is nothing surprising about that.

None of those facts are of central importance. We will, however, in giving an account of how we take there to be objects on the basis of our sense data and taking the sense data to belong to certain objects, to make sure that our theory is compatible with those phenomena.

There is nothing wrong, even according to Ayer, with Russellian sense data in themselves. However, such "sense data" is not suitable to play a foundational role in the foundations of empirical knowledge, and so Ayer has no use for them. Ayer needs incorrigibility above all.

There is no such thing as sense data

Last time, we argued that there is no such thing as a proposition. That is not to say that we don't express, believe, and so forth, particular propositions, but rather that there is no general kind, Proposition, to which they all belong. "They" are a heterogeneous lot, and we should not expect informative generalities.

That is not unusual: we don't expect informative generalities about food either. "Food" is just whatever we eat, we eat lots of different kinds of things, and there are a lot of interesting things to say about those various kinds. It is only, when we try to say things about all the kinds simultaneously, we are forced to lapse into uninformative, analytic generalities.

The same is true of sense data, the rules for making assertions about sense data, the kinds of sense data (color, visual, auditory, and so on— a heterogeneous list), and other related things. The available generalities are all analytic. If we want to say anything informative, we need to get down to cases.

Questions

• The adoption of sense datum language is, Ayer says, a matter of convention. But (113–114) Ayer says "it does not follow from this that the propositions which are intended to describe the characteristics of sense data are true only by convention. For sense-data [sic] can have properties other than those that belong to them by definition." How are the two claims compatible?

Since it isn't clear what "characteristics" are, it seems clear that one might take some "characteristics" to be purely conventional. For example, Ayer seems to take the characteristic that no sense data includes anything of which we are not aware to be a matter of convention. There are, however, characteristics of sense data that are not a matter of convention. Here are some examples:

• • What sense data we are aware of is not under our direct conscious control.
  • Our color sense data are mutually exclusive: nothing can appear to be both red all over and green all over.
  • Color and touch sense data are not exclusive: something can seem to be both red all over and smooth all over.

• Does Ayer make a point about how we see sense data? Does our brain have some mechanism, or anything like that?

It is critical for Ayer that descriptions of sense data are merely descriptions of that of which we are aware in what are apparently acts of perception. He is therefore careful not to tie sense data to any mechanism. For example, on 135, he says that even if we apparently perceive through extra-sensory perception, the experience can be reported in sense-data language. He says that precisely to emphasize what his answer is to your question. He doesn’t particularly believe in ESP, he just wants to make the point that his theory of sense data abstracts from any question of mechanism. Presumably, if you asked him what he thought the mechanisms typically are, he would give much the same sort of answers you might be inclined to, in terms of brain, neurons, retinas, light, …. But we believe those things as a result of what we have made of our sense data, not the other way around, and that is what is important to him.

• What is the relation between sense data and phenomenology?

There is one sense in which sense data are part of one pretty standard account of the phenomenology of perception. However, even in that sense, phenomenology is much broader, and includes the phenomenology of belief, emotion, and a long list. There is a philosophical school called phenomenology too, and there the use of the term is very different, and has to do with performing special sorts of psychological gymnastics.

It is a familiar fact, traditionally associated with gestalt psychology, that how things look depends a lot on our expectations. Ayer therefore concludes that the sense data we experience in a given set of material circumstances is not fully determined by those circumstances, but also by prior learning, expectations, cultural conditioning, belief, emotion, … . Russell, for example, wanted a notion of sense data that was epistemologically prior to all of those things, what we “really just see.” That led to him to say that the same type of sense data can appear different in different circumstances, and so forth. Ayer bites the bullet: there is no way to get at such pure phenomena except as a result of lots of philosophical theorizing, and so he doesn’t attempt to build anything like that into his notion of sense data. He needs to explain how it is that our sense data lead us to believe that there is something purely perceptual behind it.

• Does Ayer ever give a brief definition of sense data?

No. He says how he intends to use sense data terminology, and he explains why he can't give a direct defintion. He gives what I would call a contextual definition of the term "sense data," not an explicit definition. Here is the contextual definition:

Why can't Ayer give an ordinary, explicit definition of sense data? Because explicit definitions explain a term in terms of other, antecedently understood terms. Unfortunately, in our usual material object language, all the relevant terms apply either to material things or to mental states. Since sense data are neither material things nor mental states, and since sense data are not of some third kind, there is simply no ordinary vocabulary available to make it possible to give an explicit definition. Ayer thinks that the idea that one ought to be able to give an explicit definition, or at least the idea that sense data are things among others, has led to a lot of philosophical errors and confusions, including, most importantly, the use (most prominently by Berkeley) of the argument from illusion as an argument for idealism. Since Ayer wants, in part, to explain how we come to take some sense data to belong to material objects and other sense data to belong to mental states and even how we come to take there to be such distinct kinds of things as material objects and mental states, he can hardly presuppose them in defining sense data. His project is, nearly, to go the other way around: to define material objects and mental states in terms of sense data.

• Is the status of sense data supposed to be like that of the average family?

I'm not sure what "status" is supposed to be. Nonetheless, it is reasonable to take both to be contextually defined, not a matter of our ordinary experience or usage. But, and this is a big but, our knowledge of the average family is presumably all indirect and inferential, while the important bits of our knowledge about sense data are supposed to be incorrigible and empirical.

-- ShaughanLavine - 14 Feb 2003 - 08 Feb 2007