Tuesday, June 3, 2008

The Analytic-Synthetic Distinction and The Knowledge Argument

Note to the reader: You do not need to read any of my previous posts to understand this one, assuming you are at least somewhat familiar with Frank Jackson's knowledge argument. That said, this post is a sort of appendix to my last post, "Testability, Omniscience and The Knowledge Argument" (essentially a long email I sent to Professor Torin Alter), which was a follow up to my first entry on the knowledge argument. So, if you're not familiar with the knowledge argument, you may want to start there.

Professor Alter has promised to respond to my lengthy email when he has a little more time. I must again comment on the extraordinary generosity he has shown me, since he has no professional or personal obligation to respond to my emails. While I am awaiting his response, I offer the following elucidation of my views through a brief discussion of the analytic-synthetic distinction.

Traditionally, analytic statements are said to be those which are true by virtue of their meaning. That is, analytic statements are those which are true "by definition." For example, "all unmarried men are bachelors" is a common example of an analytic statement.

Synthetic statements, on the other hand, are not necessarily true. They are contingent on facts about the world. For example, "John is a bachelor" may or may not be true, depending on any number of facts about the world.

The distinction here seems simple enough, yet has generated a tremendous amount of philosophical debate. The distinction becomes problematic when we try to establish some rule for deciding whether or not a given proposition is analytic or synthetic. It seems that analyticity is established merely because we say it is, without any rules to determine if our judgment is valid or invalid. The philosopher W. V. O. Quine famously called this fact to our attention, and claimed that all knowledge was in fact synthetic to one degree or another.

Many philosophers today agree with Quine, and say that the difference between analytic and synthetic statements is not absolute, but only a matter of degree. However, a good number disagree, following the lead of Rudolf Carnap, who accused Quine of missing the point. According to Carnap, there is an absolute distinction between analytic and synthetic propositions; we just cannot produce any evidence to demonstrate it. Carnap reasoned that evidence can only be gathered to support synthetic statements, not analytic ones, and so the lack of evidence cannot be held against the analytic-synthetic distinction.

I think we can overcome this disagreement by reformulating the analytic-distinction as a difference in how we test propositions. It is thus a difference in types of knowledge, and not types of propositions.

We call a proposition analytically true when its truth is determined solely by its relation to other propositions. Thus, "all married men are bachelors" is true, given that our language defines a certain set of propositions as "true" which regulate the use of the terms "married men" and "bachelors."

Synthetic propositions, on the other hand, point us away from our language, and towards non-propositional experiences. Thus, synthetic knowledge involves testing propositional knowledge against non-propositional information.

All propositions can therefore be both analytic (defined as true with respect to a set of propositions) and synthetic (defined as true with respect to some non-propositional knowledge), because the difference here is not defined by the propositions, but by the way they are tested. We decide if knowledge is synthetic or analytic by testing for how the person (or system) who possesses the knowledge justifies their belief. If a belief is justified solely with respect to propositional knowledge, then we call the knowledge analytic. If it is justified by relationships between propositions and non-propositional abilities, then we call it synthetic.

The implication here is that there are neurological and behavioral correlates which can point us in one direction or another, and so there is a fact of the matter when it comes to distinguishing between analytic and synthetic knowledge.

Let's consider the case of Mary again.

Facts are recognized and defined by their discursive learnability, and nothing else. Thus, when we consider Mary in her black-and-white room, we can say that any particular scientific fact can be communicated to her through her black-and-white television. Yet, for the knowledge so learned to be synthetic, and not simply analytic, Mary must be able to test it against non-propositional experiences. It is this ability Mary gains upon leaving the black-and-white room, though she need not absorb any new facts in the process. Inside the room she can learn any particular set of facts about color vision, including facts about any relevant phenomenal characteristics; but that knowledge remains analytic until she leaves the room.

Mary should be able to test some of the information from the televised lectures non-propositionally, even if she cannot test the information about color vision directly. Indeed, for Mary to function in the room at all, she must have synthetic knowledge.

One might claim that all of the facts about color vision can be a priori deduced from Mary's synthetic knowledge. Thus, her knowledge of color vision would be synthetic, and not merely analytic. This might be argued from some sort of holistic principle of scientific knowledge, where every scientific fact can be logically derived from every other fact, though that would require that Mary's scientific knowledge of color vision be complete. I argued against the possibility of complete knowledge in my previous two entries on the knowledge argument.

Still, valid or not, I do not think the holistic argument counters my argument here. Mary still gains a new set of abilities upon leaving the room, and is able to test her propositional knowledge in new ways, even if she had some synthetic foundation for justifying her beliefs about color vision inside the room.

In conclusion, we have three types of knowledge. On the one hand, we have the difference between propositional and non-propositional knowledge. Propositional knowledge can then be sub-categorized in terms of analytic and synthetic knowledge.

Non-propositional knowledge is measured in terms of abilities which have nothing to do with language. This is the most basic form of knowledge, and while it requires anticipation, it does not involve abstract thought. Propositional knowledge, on the other hand, is always about predictions. With analytic knowledge, we require the ability to make predictions about the proper use of language without reference to non-linguistic phenomena. With synthetic knowledge, the abilities involve regulating the use of language with respect to non-linguistic phenomena. The three types of knowledge are defined, and thus distinguishable, in terms of abilities.

When Mary sees colors for the first time, she gains non-propositional knowledge which she can then use to test her propositional knowledge. Thus, her potential for synthetic knowledge is broadened. However, any facts she so learns were already available to her inside the black-and-white room, even if only as analytical knowledge.