George Miller [Magical Number 7 +/-2] Part 3

From: Sam Heighway (skh197@soton.ac.uk)
Date: Mon Mar 16 1998 - 14:57:59 GMT


    Here's what I got out of the article:

    According to Miller, the Magical Number 7 only applies to
one - dimensional objects, probably because they only have
one variable aspect, as opposed to things such as faces which
have several variables. Therefore we can recognize
thousands of different faces.
    
    Experiments be Klemmer & Frick, and Beebe-Centre, Rogers
& O'Connell have shown that two - dimensional objects, such
as a point in a square, have two capacities, in this case one
for horizontal and one for vertical position. If the capacity
for one of those dimensions is 3.25 bits, then it would be
expected that the two together would equal 6.5, but in fact
the addition of the second capacity only gives us 4.6 bits,
so it is not perfect when a second dimension is taken into
account.

    However, two - dimensional objects are still a long way
from faces. Pollack & Ficks conducted an experiment in which
noises were presented with 6 variable qualities - frequency,
intensity, rate of interruption, on-time fraction, total
duration and spatial location. There were 5 values for each
variable, so in total there were 56, or 15,625 (?) different
tones. The information was transmitted at 7.2 bits, and
participants could identify up to 150 different categories.
This greatly exceeds the Magical Number 7, plus or minus 2.

    It is interesting to note that the variables do not have
to be independent of each other to increase the channel
capacity. Basically, increasing the amount of variables
increases channel capacity, but actually decreases accuracy.

    Miller argues that in evolutionary terms, humans have
survived and evolved because they have a little bit of
information about lots of things, rather than vice versa, and
this would seem to be preferable.

    As far as speech is concerned, Pollack & Ficks did a test
where tonal stimuli had 8 variables, but each was only
binary, ie. it had only two different values. About 120
different recognizable sounds were identified. With human
speech, it is not yet known whether the limit in the number
of variables we have is due to a limit in processing them or
in producing them.

    The summary of this section that Miller gives is that
people become less accurate at identifying stimuli when they
have to judge several variables at once.
    
I hope this hasn't confused anyone too much! See you at 3:30,
SAM



This archive was generated by hypermail 2b30 : Tue Feb 13 2001 - 16:24:20 GMT