Old sevens

At some point yesterday I wanted to know what was meant by chunks of memory and was directed, via Wikipedia, to reference 1. One of those rare days when one reads something which is seriously old with serious profit. I believe the paper is something of a classic and so I share a few snippets.

First, we have another magic seven, something which I comment on from time to time. See for example, reference 3. While Miller ends with 'the seven wonders of the world, the seven seas, the seven deadly sins, the seven daughters of Atlas in the Pleiades, the seven ages of man, the seven levels of hell, the seven primary colours, the seven notes of the musical scale, and the seven days of the week'.

But the magic sevens with which Miller is more concerned are our difficulty in classifying one dimensional variables like length or pitch to more than seven bins and our difficulty in holding more than seven things in short term memory. Maybe he overdoes things a bit, despite allowing the phenomenon of perfect pitch (which involves more than fifty bins) and at reference 2 it is shown than one can get a lot better at the first task with training, but the basic point being made is solid enough.

I associate to my days of doing slump tests on wet concrete, when, after a bit of practise, I used to claim that I could measure the slump by eye to within a quarter of an inch - with slumps typically going from nothing up to a couple of inches. Which is eight bins, one more than the magic seven.

Miller goes on to speculate why it is that we can only do seven bins, while we can remember lots of faces and lots of words. The fact that these things have lots of dimensions, rather than just the one, is clearly part of the mix here. To speculate a little about the role that naming things might play. To introduce the bit of jargon 'recoding' in this connection - a timely reminder of the link between signal processing and what the brain gets up to when it processes, for example, signals from the retina.

He introduces the concept of 'chunks' in the context of Morse code. Chunking the raw signals up into groups of dots and dashes makes it a lot easier to remember a sequence of said dots and dashes and Morse code operators fairly quickly stop thinking or working in terms of individual dots and dashes.

I associate to the bit in Koestler's 'Darkness at Noon', where the novice prisoner, the main character of the story, has to learn the tapping code used by prisoners to communicate between cells, and rapidly graduates from thinking in terms of the basic tapping codes to thinking in terms of words, albeit being spoken rather slowly. A story which predates Miller's paper by fifteen years or so.

The paper has a rather old-fashioned tone. Miller writes in clear and attractive English. He shows courtesy to colleagues. The diagrams and statistics are kept under control. All qualities which are missing from much of what is written by his successors today.

Reference 1: The Magical Number Seven, Plus or Minus Two: Some Limits on Our Capacity for Processing Information - George A. Miller – 1955.

Reference 2: Learning in a unidimensional absolute identification task - Jeffrey N. Rouder, Richard D. Morey, Nelson Cowan, and Monique Pfaltz - 2004.

Reference 3: https://psmv2.blogspot.com/2015/08/7-up.html.