h2g2 The Hitchhiker's Guide to the Galaxy: Earth Edition

(for Clive): Daniel Dennett is an American philosopher who wrote, among other works, 'Consciousness Explained', an excellent book which examines consciousness from a blend of philosphical, and partly from a computing perspective - ideal for an old electrical engineer / computer guy like myself. He's determinedly anti-dualist.
I heartily reccommend the book, though for someone *entirely* unaware of the subtle and precise meaning of some computer terminology, I suppose it might prove a little harder to understand, or easier to misunderstand, though I'm sure the major points would shine through.

Back to the debate :

I'm sure the linguistic component of consciousness must be of central importance. There certainly seem to be processes in the brain whose function is to analyze thoughts and convert them into some sort of language-like data.

For example, when programming, I'll sometimes look at a fragment of code trying to work out what it does, and a thought of the form "Oh I see, *that* function is an XXXX" will cross my mind. If I had to describe it in English, 'XXXX' may mean something like 'check-for-an-error-event-and-trigger-some-recovery-function-if-one-has-occurred function', but there's no *word* associated with the thought, it's almost like there's an unnamed token (but one with a very specific meaning) being used in place of a nameable word. Effectively, the thought may be more like "Oh I see, *that* function is an '____'", except that I know what the space means.

Presumably, it is the normal linguistic classification processes that generate the unnamed token '____' from my analytical thoughts about the piece of program I'm reading.
The odd thing is that, apart from the lack of a name, the '____' has all the mental *feel* of a proper word or phrase, and I seem to be able to manipulate and reason with the token in the same way I would with nameable words.
Actually, that may be slightly misleading - from my consciousness-as-a-literate-semi-observer perspective, maybe I should say that when I'm commentating on my sub-verbal reasoning, there's some persistent or repeatable element of my thoughts which is reliably classified by my consciousness as '____' whenever it occurs, in the same way that the word 'chair' would result from my classification of thoughts involving various kinds of chair.

It seems to me that such a tokenising mechanism may be the key to language - all that is required to add is an association between a sound or symbol and an as-yet unnamed token, whether that token codes for some physical object, or some action, process, thought or emotion.

This is rather like the pan-demon theory of language. I can't remember the details but essentially the demon is what Potholer refers to as a token and meaning is derived from a collection of demons. I must read up on this again - what does Daniel Dennett have to say about forgetting?

I don't remember, so it couldn't have been important.

In fact, I'm not sure he said much about it at all, but from a neural network (hardware) perspective, forgetting is simply a result of failing to reinforce old memory traces whilst laying down new memories. There's no need for an explicit mechanism.

(I would try and make that more intelligible, but I suspect any significant expansion is likely to be several screens long)

Do links in neural networks decay or are they overwritten? I had been led to believe that neither of these things occur and that the links do in fact remain. The lack of reinforcement would make them less likely to be active. This is not a theory of forgetting, it is a theory of why we may not remember very well.

In an organic brain, there will be some loss of connections due to natural cell loss. Clearly, many human memories are extremely persistent, so the rate of loss is very low, and/or there are efficient mechanisms for refreshing important information.

In learning new information, I believe either the strengths of existing connections may be modified, or new connections are made. Even if those two processes may be biologically different, as far as their effects are concerned, they are essentially similar, and they will *generally* lead to lessening the effects of previously learned information. However, when new information throws new light on existing knowledge, it's quite possible that the traces of old memories will be reinforced by the subsequent learning

I'm wondering what you mean by a theory of forgetting - if it's not a matter of simple decay of memories, is it some deliberate process to remove the memory traces relating to a collection of information?

I am suggesting that 'forgetting' is in fact no more than an inability to recall something at a given moment and that all input is permanently stored (barring injury or disease)and can be recalled under appropriate conditions - hence the ability of people to recall extraordinary detail under hypnosis and to remember childhood experiences with some accuracy in old age.In other words, there is no theory of forgetting because we do not forget.

Many memories do last a long time, and we can often remember apparently unimportant memories with extraordinary detail, but I'd disagree with the idea of *all* input being stored permanently.

To an extent, it does depend how we define 'input'. In fact, there's a great deal of unavoidable filtering of raw sensory input before it ever reaches the higher-level areas of the brain. From a design point of view, even storing every piece of partially processed data would be extremely inefficient. A large amount of processing before storage would be required in any feasible system.

Likewise, I think it would be inherently impossible for a brain to remember its own entire history of previous activation (thoughts, emotions, etc). There simply wouldn't be enough space.

Additionally, I'd argue that memory is not simply a matter of passive recall, but of active reconstruction. Repeated inaccurate reconstructions can lead to memories of particular events changing over time. It is undeniable that people not only forget items of data, they also subjectively mis-remember them, and can have false memories of events that have never happened. That can't just be a matter of a faulty recall mechanism.

On the topic of hypnosis, some hypnosis subjects allegedly regress to past lives, but I don't think the population at large accepts that as definitive proof of reincarnation.

As an experiment, I'd be extrememly interested to see an example where someone was videotaped as a small child (for example, at a 5th brithday party or some other potentially memorable event), never saw the tape, and was hypnotised 10 or 20 years later and requested to recall the event in detail. I'm sure they'd remember some details well, but I'm certain that some memories they were confident of would prove to be inaccurate.

My apologies. The pandemonium model (Selfridge,1959) was a feature recognition theory related to general visual stimuli, not specifically to language. I think, however, that the feature processing theory may apply equally to audible inputs.