Knowing 10 September 2004

Carl Zimmer described some interesting research in a recent blog entry . It seems that people who are unable to recall any of the events of their past lives are still able to identify which of a list of words best describes them as people: although their explicit knowledge of their own autobiographies has disappeared, they still have self-knowledge in a different form. It is suggested that two different brain systems are involved. This research might possibly shed a chink of light on the debate about whether, and in what sense, we actually have selves, but it also raises the thorny question of different ways of knowing things. We often talk about knowing things as though knowledge was a straightforward phenomenon, but it actually covers a range of different abilities – look at the following examples. 1. I know what the capital of Ecuador is. 2. I know where the keys are. 3. I know how to sign my name. 4. I know that zebras don’t wear waistcoats

The first example is the case in which an explicit fact has been memorised – possibly even a fixed formula (“The capital of Ecuador is Quito.”). This is perhaps the easiest form of knowledge to deal with in a computational way – we just have to ensure that the relevant string of characters (“Quito”) or the appropriate digits are stored in a suitable location. There’s relatively little mystery about how you can articulate this kind of knowledge, since it has probably been saved in an articulated form already: it was words on the way in, so it’s no surprise that we can provide words when it’s on the way out.

In the second case, things are slightly less clear. It’s unlikely, unless you have a really bad key-losing problem, that you have memorised an explicit description of the place where they are: however, if you need to produce such a description, you would normally have no particular difficulty in doing so. A reasonable assumption here might be that the relevant data on the position of the keys are still stored somewhere explicitly, (on some sort of map, as co-ordinates, or perhaps more likely, as a set of instructions like those on pirate’s treasure maps, telling you how to get to the treasure/keys). The question of how you are able to translate this inner data into a verbal description when you need one is less easily answered, but then, the process of coming up with a description of anything is not exactly well understood, either.

The third case is a bit different. The importance of ‘knowing how’ as a form of knowledge was emphasised by Gilbert Ryle as part of his efforts to debunk the ‘Ghost in the Machine’. It could legitimately be argued that the difference between ‘knowing how’ and ‘knowing that’ is so great that it makes no sense to consider them together - but both do allow some stored knowledge to influence current behaviour, so there is at least that broad similarity. You sign your name without hesitation, but describing how to do it (unless you happen to be called ‘O’) is challenging. To describe the required series of upstrokes and downstrokes would require careful thought – you might even have to watch yourself signing and take notes. The relevant data must be in your brain or your hand somewhere, but you are hardly any better off when it comes to putting them into words than anyone else who happens to be watching. Presumably the relevant data are still stored somewhere in your brain. Perhaps they are just in a different part of it, or otherwise less accessible: but it seems likely that at least some of them are held in a form which just doesn’t translate into explicit terms. There may be a sequence of impulses recorded somewhere which, when sent down the nerves in your arm, results in a signature: but there need be no standard pattern in the sequence which symbolises ‘downstroke’ or anything else.

The fourth case is the most difficult of all. Most of the things we know, we never think about or use. We never asked ourselves whether zebras wore waistcoats until Dennett proposed the example, but as soon as we heard the question, we knew the answer. This vast stock of common-sense knowledge (Searle refers to it, or something very like it, as ‘the Background’) is crucial to the way we deal with real life and work out what people are talking about: it’s the reason human beings don’t generally get floored by the ‘frame problem’ – unanticipated implications of every action - the way robots do. It surely cannot be that all this kind of knowledge is saved explicitly somewhere in the brain, however vast its storage capacity. In fact, there are good arguments to suggest that the amount of information involved is strictly infinite: we know zebras are normally less than twenty feet tall, normally less than twenty-one feet tall, and so on.

That last argument suggests a better explanation – perhaps key pieces of information are stored in a central encyclopaedia, and the more recondite conclusions worked out as necessary. After all, if we know that zebras are less than twenty feet, simple arithmetic will tell us that they are less than fifty, without the need to store that conclusion separately. The trouble then is that there simply is no general method of working out relevant conclusions from other facts: formal logic certainly isn’t up to the job. I’ve discussed this further elsewhere , but it seems likely to me that part of the problem is that, as with the third case, the information is probably not recorded in the brain in any explicit form. To look for an old-fashioned algorithm is probably, therefore, to set off up a blind alley.

What about the two forms of self-knowledge we started with? It looks as if we are dealing with a loss of the kind of knowledge covered by example 1 above, while type 2 is retained. If so, the loss of memory might be less disabling than it seems. The patients in question would not be able to tell you their address, but perhaps they might still be able to walk to the correct house “without thinking about it”? They might not be able to tell you their own name, even - but perhaps they could sign it and then read it.