pumpkinstrokemarrow

Been thinking about the business of being able to read words with jumbled interiors (see above). Observation 1: the brain holds information about words we know about. So there is an entry, for example, for the word 'cow'. Observation 2: when the brain sees the word 'cow', it accesses that entry. So the question is, how does it find that entry? A computer might simply search all its entries until it finds a match - computers being quite good at questions like: does 'abd' equal 'abc'? And tasks like: search this list until you find 'abd'. This computer having avoided the issue of turning a picture of 'abd' into the text of 'abd'. If the number of entries gets large it might go to the bother of having an index. So if the string we are interested in starts with an 'r', then search the entries for that letter, rather than searching all the entries. Such a wheeze usually depends on having all the entries for r in the same place, differant from that for entries for s. Observation 3: on the evidence of our anecdote, the brain knows about first letters, last letters and possibly length. Perhaps it indexes its entries by these attributes. So the picture of the word comes in. Maybe length is the easiest attribute so that is the first cut. Then it goes to first letter - again easier to get at than an intermediate letter - and that is the second cut. And so on. Maybe it pulls out the intermediate letters in parallel, pushing through the decision tree at the letters pop out, the easier to identify popping out first. This would explain why it does not care about order. Observation 4: it can't depend much on the shape of the word as jumbling the letters will jumble the shape. So the algorithm must go for letters.

Having missed the downstairs exhibition at the National Gallery by two days, went to have a peek at 'Bacchus and Ariadne' instead, our having had a copy for many years now. First item, a lot bigger than I remembered. About twice the size in fact - or four times the area. Second item, it looked much squarer than I remembered. Although when we got home our copy was the same shape as the real thing. Third item, the composition looked much more spacious - and much better for it. Without having the two side by side, can't be sure why this was, but the framer having framed over the outer half inch of the print might be the answer - although not clear why he would have done such a thing. Copies usually come with a generous margin. Fourth item, I always forget how much of the quality of the colour is lost in a copy. Maybe this is to do with colour being a more complicated quantity than a number on a scale. Colour built up with layers of semi-transparent paint with a white ground (assuming that is how this picture was built) is not going to be the same as a shade built up with coloured dots. Although that said, I do not have much of a clue as to how the colour in a copy applied to hardboard thirty years ago would have been built up. Colour lithography? Digital not invented at that time. Picking up an earlier thought, must see if we can get a decent jig-saw of this picture. How much would I learn about a picture which has been part of the furniture - and much looked at - for a long time from such a process?