The Narrative Hypothesis

Let's assume that a proposed experiment was carried out. A certain amount of memory is then said to exist of the external speech. One could then attempt to compare internal and external speech to determine if the assertions generalize. Fine. Let's assume they do.

We know might want to study how a person's representations of a problem change as a result of the internal speech having addressed an issue. In an informal way, it might be quite obvious to most people that if the narrator covers an issue (say, what to eat for lunch) that the representations will change as a result. This may not be a given, though. It is important to poke and prod this. The implantation of false memories may border this area.

The studies being done on synthetic memories are important because of the question of how we represent many of our long-term memories. I have come to believe that one useful role that the IOS stream plays is that of enhancing and organizing our memory for the abstract. Just as our internal narrator might state that a seven was rolled, so it may be easier to recall that reduced information set, rather than the full visual image. We may be better at recalling that the roll was a seven than the fact that a three and a four were rolled. If a test subject was asked to state out loud the sum of the roll, and then later asked for the sums, he might well be accurate. Whereas if a test subject was asked for the individual die rolls, the accuracy might be significantly less. If the test subject states the individual rolls, he might remember it better. Similar to the statements out loud, the internal speech might also help. If it is confused with another task, memory of the dice-rolling events would probably be less.

Scope of the Narrator

The narrator improves Man's ability to live his mental life outside the here and now. It is one of the gifts of language. This is not to say that animals can't think about things not in front of them. It's just harder. And the ability to maintain trains of thought are greatly improved with a narrator.

I'm coming to believe that this characterization of the narrator, of what it can and can't do, is very important in the analysis of human cognition. Obviously animals have a certain level of cognition, and this serves as a base from which to characterize that which is not the narrator. Mostly it is the higher level concepts, of time, distance, social relations, etc. that get left out if you don't have language working for you.

One might wonder about the cognition being done that the narrator misses. Clearly there is a lot, both on the animal side and on the language side. Yet the product of the narrator should serve to sum up much of the language side's activity. It should only convey a minor portion of the animal side. How often do you practice statements like "I am driving down highway 5"? Yet somehow you remember the activities taking place. It may be that this narration helps to document the activities of the day. But it is much more likely that it documents the abstraction being manipulated, as the individual makes sense of the world.

The narrator's transcript is one of the useful products of thought. Yet we may be focused on it based on the fact that this is the product which is most easily conveyed. At some point if we are able to show the visual interpretations and nightly dreams, this may change.

Because the range of internal narrative statements are as complex as speech itself, it may be hard to analyze the range of mental states being covered. A quick scan through a thesaurus allows one to get a sense of the breadth of single words or terms. And after an analysis of grammar, one can see that there must be deep structures which get manipulated when a sentence is formed. It would be neat to hypothesize a pushdown automata integral to the process of sentence construction. We might also attempt to relate this to dyslexia or Freudian slips.

What role does rhyme and clanging play in the speech rehersal process? Freudian slips often are based on rhymes. Similar sounds and the retrieval of thecompiled voice activation sequence. Practiced utterances associated by how they sound (versus the muscles used to created them).

Do we know what we want to say before the words begin to come out? And what of changes that might get made as the words are being uttered? We have a production center and a recognition center which are both happily at work, serving as the two obvious parts of the narrator's kingdom. One would think that deep structure exists somewhat before production, and that deep structure exists after decomposition.

(also comment on narrative cognition in Koko the gorilla)

(comment on the genesis of the personal story)

Creating Internal Observed Sentences

One thing that I've come to believe in is the concept of the Internal Observed Sentence (IOS). Consider that the narration we have discussed is broken up into units. The IOS is, then, the basic unit of narration. It reverses the classic statement that "a sentence conveys a complete thought".

In one sense, the generation of IOS sequences does little to resolve the more basic problem of how cognition gets done in the brain. We can, however, focus our attention now on the problem of how a single IOS unit comes to be. In doing so, we divide the general problem into processes which lead up to an IOS unit, and those that consist of sequences thereof.

In order to speak a single sentence, what are the necessary steps that are needed to produce it? There is the rendering of the sentence as speech. There is the syntactic and morphological rendering of the of the abstract thought. There are the component elements and their relationships. There is the planning and selection of this thought as the one we wish to render.

It is this last that is the most difficult, vague, and intriguing of the set. The others are ones that either have been tackled or are in the process of being tackled. This "what to do next" problem probes several basic philosophical areas. Free-will, scheduling, division of attention and other mental resources, the role of serendipity, etc. all come to mind.

Can the conscious mind effectively program itself? Can it put itself into various modes, whereby a mental agenda can be followed? What are the forces that buffet the direction of the conscious window? What IOS will be next produced?

If we postulate that there are a few general basic kinds of selection mechanisms, what might they be? First might be to logically follow an existing train of thought. One of the ways to do that is to ask the general association store for something which is similar to a set of active concepts. In effect, "what does this remind you of".

Introspection

Consider the idea of dividing the problem in two. The first part is the magic of how an IOS unit is generated. The second part is how several sequential IOS units are used to solve a larger puzzle. Introspection is valid on the second, and not the first. IOS units are basically visible for recall. The process by which an IOS is generated is not available for recall.

The IOS unit is the item being globally observed. It is the item to which the rest of the system responds.

We are clever entities, capable of creating grammatical sentences, full of meaning. These are the results of more complex thought processes, but as the results, they serve to cache answers that we would otherwise have to re-derive. They also serve to remind us of the images, sounds, and smells of times long since past. Give me the essence of a lifetime, and I will give you a text which has a chance to survive an eon. So maybe the right thing for me to do is to write a text, and then encode that text for long life.

Let's for a moment think about meaning. What makes one set of thoughts or experiences more significant than another? Why should one record A and not B? The record becomes important when there is something objective about the observation. That it is useful or significant to the reader and not just to the writer. We could videotape everything (and might one day do so), but it doesn't serve to distill meaning. Where are the good parts? The stuff that might mean something?

A Limited Channel

The IOS stream is a limited channel. Just as it is necessary to serialize what comes out as speech, so too is it necessary to convert the constellation of other brain activity into a linear IOS stream. Realizing this, one should also be able to put an upper limit on the speed of this stream. One can ask whether this stream is always on a a given rate, or whether there are factors which speed up, slow down, or even interrupt it.

I might state, for example, that the creation of the text in front of you now was in fact created from my IOS stream, but that in having to type it, I am unable to create the IOS units at the speed I might normally create them at if I were not attempting to write them down.

It is also worth examining the manner in which the larger semantic construct gets assembled on its way out the door. On occasion in the middle of a sentence I will forget where I am and be forced to read the previous stream again in order to reform the rest of the sentence. Clearly editing also raises its head in this context.

Sub-Vocalization

There's this whole area of speed reading, and the associated theories which go with it. Although it may not be scientifically very reputable, the idea of sub-vocalization during reading is one that borders on this whole internal speech issue. Most of us first learned to read by reading aloud. The challenge of converting letters written on a page to sound was primary, and only as a secondary effect were those sounds made to make sense. It is with proficiency in the area of reading aloud where we then began to read quietly, silently to oneself. The speed reading courses seek to short-circuit that vocalization process, in an attempt to reduce the time required to comprehend the text. Certainly it is possible to skim a text, focusing on keywords.

(comment on things said out loud versus internal propositions)

(comment on activity in the language areas during free-running periods)

Convertibility

We acknowledge that much of the cognition in the mind has little to do with the processing of language. There are vast areas of cortex devoted to problems of vision, manipulation, planning, and real-world modeling. As we stated at the beginning, the internal generation and observation of sentences should be considered a pulse, a distant reflection of a heart beat.

A corollary of the non-exclusivity of this IOS mode of thought is convertibility. Convertibility between the sentence and other representations is an important aspect. If one experiences some event, one wants to be able to describe that event. Similarly, if one composes a sentence, one wants to then be able to mentally construct a visual synthesis of that sentence.

The building of stories is seen here. Let's say you observe the rolling of dice. You want to then construct a sentence which reflects the outcome of the roll. "He rolled a 7" one might say. The visual scene might be two dice, one with a three and the other a four, slightly at an angle to one another. A synthesis of the spectrum of processing has made the sentence possible, yet the utterance has defined the salient quality of the event. It is not only the whole visual processing regieme, including shape from shading, 2 1/2 D processing, solid modeling. It is also the sense of what is important about the scene, how the dice combine, and what numerical quantity is represented. Then one selects words and make the syntactic construct, having it rendered in a series of potential muscle movements necessary for speech.