Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!wasatch!cs.utexas.edu!uunet!ginosko!rex!ames!zodiac!meridian!marcel
From: marcel@meridian.ads.com (Marcel Schoppers)
Newsgroups: comp.ai
Subject: Paper not recognized
Message-ID: <8850@zodiac.ADS.COM>
Date: 18 Aug 89 16:53:46 GMT
Sender: news@zodiac.ADS.COM
Reply-To: marcel@ads.com (Marcel Schoppers)
Distribution: na
Organization: Advanced Decision Systems, Mt. View, CA (415) 960-7300
Lines: 24
Keywords: chess, classification, features
Summary: how many features identify a chess position?


My thanks to the half-dozen-or-so people who responded to my previous
request; the pointers were useful and interesting, but did not include
the paper I was looking for...

> ... The paper reported on
> a psychology experiment that set out to determine the number of features
> used by chess masters to recognize and/or classify random chess positions.
> After a position had been set up (and not shown to the master), the master
> got to ask yes/no questions until he thought he knew what the position was.
> I think I remember that on average, it took about 70 questions to classify
> a position. But I don't remember the author(s). Something in the back of my
> head says it might have been Chase & Simon, or Nievergelt, or perhaps even
> deGroot.

It wasn't DeGroot (1965); his experiment was, how much of a seen position could
be remembered. It also wasn't Chase & Simon (1973); their experiment was to
identify featural chunks. Nor do the papers I've seen even refer to the result
I'm after, which implies that the paper I want is probably after 1973. I have
yet to track down Simon & Gilmartin (1973) which is not the paper I want but
may have a reference. Could it have been Nievergelt, and if so, when? Any more
clues are welcome. 

marcel@ads.com