Thursday, January 13, 2011

Thought and Meaning

Ken Aizawa and Richard Menary are having a virtual debate about the book "Bounds of Cognition" by Ken and co-author Fred Adams. The issue du jour is whether thoughts have "derived content"; ie, is their "content" (or "meaning") established as a matter of social convention. Menary says yes, Ken says no.

Menary's argument appears to be, in essence, that a thought S is just an unvocalized sentence S and thereby has the meaning of S when S is expressed - vocalized, written, printed, signed, etc. Ken disputes that the thought S is literally the sentence S, but is instead a representation of S (in something he calls "mentalese"). Specifically, Ken says:

Thinking in English is not a matter of sound streams bearing derived content being found in the brain. ... [There is no] token of [a sentence in, say, English] that bears derived content occurring in the brain.

I agree with Ken on the bottom line but disagree with him about the token. I consider that when you dig deeper into what we mean by a "sentence" and it's "meaning", you find that when one is thinking S, there actually is a "sentence" S in the brain - but it has no "meaning".

To be as precise as possible, I'm going to try to define terms explicitly and use some symbolic notation. I'll also try to make assumptions - of which there are necessarily many - explicit.

Start with my definition of "sentence":

A sentence is a sequence of words in a language that conforms to the syntax of that language.

First, consider the difficult concept of the "content" of a sentence (which I'll henceforth call its "meaning"). First, an assumption that I believe to be common to Aizawa and Menary:

"Derived meaning" is a matter of social practice within a specific community.

Informally, this just saying that in normal conversation among members of a community that share a common vocabulary and its use, the meaning of a sentence is a common interpretation of it that "makes sense" to most members of the community. Eg, The "The harbor is beautiful today." would make sense to most English speakers - ie, would have meaning - but "My automobile has a cold." would not. On the other hand, a sentence that has meaning for members of a narrowly specialized technical community may seem gibberish to non-members. (Sokal showed that unfortunately, the converse may be true as well.)

So, another assumption:

The "meaning" of a sentence S within a community C is a function of both S and C, represented henceforth as M(S,C).

Deciding on a codomain for M(S,C) is itself a difficult task that fortunately can be ignored for the present purpose.

Now, consider what is involved in expressing - in particular, vocalizing - sentence S. Vocalizing S can be envisioned as a pattern of motor neuron activity N(S) that becomes manifest in movement of vocal chords, lips, et al. For the present purpose, we need not address the details of the process by which N(S) gets converted to sound waves and those get transformed into neural activity in the hearer's brain. It suffices to note that in essence, neural activity N(S) at the speaker gets mapped into a pattern of neural activity at a hearer. Call that neural activity at the hearer H(N(S)).

It suffices for our purpose to note that one of the products of H(N(S)) is the phenomenal interpretation of the heard sentence - what it "sounds like", it's aural mental image or qualia. Call the subset of H(N(S)) that is necessary and sufficient to produce this phenomenal experience P(H(N(S))).It should be noted that the nature of this subset of H(N(S)) is, as far as I know, completely unknown, including whether or not it is an empty set (ie, phenomenal experience could be in some sense illusory).

The hearer's interpretation of the meaning of S must be determined by H(N(S)), in accordance with - by assumption - the agreed upon interpretation within a community C. And in order for this meaning to be the same as the speaker's intended meaning, both speaker and hearer must be members of C.

Now consider what happens when a person has the thought S. My impression is that we really don't know. So, in order to proceed assumptions must be made. Up to this point, my description has been, I think, mostly uncontroversial - mainly because it is so high level and vague as to afford little substantive to attack. That now changes dramatically. But keep in mind that the objective is to construct a plausible argument, not to prove anything.

The present interest is in thoughts that are manifest in sentences, at least some of which are accompanied by verbal mental imagery, ie, the phenomenal experience described as "talking to oneself". I now need the following assumption:

When a thought is manifest in a phenomenally experienced sentence S, the neural activity pattern that initiates production of the phenomenal experience is essentially the same as N(S). That is, it's an unexpressed version of the neural activity pattern that results in S being vocalized.

I can in no way justify this claim other than to note the obvious parallels between speaking to another and to oneself. One has to accept it as a plausible possibility - or not.

With that assumption - which amounts to assuming the presence in the brain of an unvocalized neural activity pattern equivalent to N(S) - call it N'(S), we can turn to the phenomenal aspect of a thought. A neural activity pattern analogous to H(N(S)) is needed. Which leads to another assumption:

To produce the verbal mental imagery of a thought S - that "voice in the head" - N'(S) is essentially fed into the aural processing stream. This produces a neural activity pattern equivalent to H(N(S)) - call it H'(N(S)). And as in the case of a hearer of a vocalization of S, there is a subset of H'(N(S)) that produces the phenomenal experience - call it P'(H'(N(S))).

I say "essentially" to emphasize that the actual implementation is of no concern here. It suffices that there be a plausible description of how this might be effected.

All of this is intended to show that there is a plausible way of viewing the process of producing a thought S as completely equivalent to that of producing a vocalized S. Ie, that in an understandable and reasonable sense:

"Thinking in English is not a matter of sound streams bearing derived content being found in the brain."

However, it remains to attach meaning to the thought S. Analogous to attaching meaning to the neural activity pattern at a hearer, we can argue that it must be a function of H'(N(S)). But what about C? In the case of vocalizing S, C must be common to speaker and hearer. But in the case of thinking S, "speaker" and "hearer" are the same and thus trivially have C in common. But that clearly is not the sense in which meaning is "a matter of social practice".

One can argue that meaning can be attached to the thought S by intending to vocalize S in the presence of members of a community C. But that also misses the point of determining meaning by social practice - doing so ensures that a speaker's intended meaning is the shared one. One can think they have the meaning of a sentence in a community right, but until expressing it and getting a reaction one can't be sure.

So, I would argue that the thought S is - in the sense described here - a sentence S in the brain but that it's meaning is undefined until S is vocalized within a relevant community and it's intended meaning is verified by acceptance within that community.

No comments: