From: skr@hagg.psy.ed.ac.uk (Simon Rushton)
Subject: Re: TECH: Networked VR under 100ms?
Date: 16 Sep 1996 09:39:18 GMT
Message-ID: <51j786$e84@scotsman.ed.ac.uk>
Organization: Edinburgh University


william r. cockayne (wrc@mbay.net) wrote:

: This number actually has nothing to do with technical limitations; it is
: classically the human threshold for noticing change (ie, discrepancies).
: If something occurs below the 100 ms limit, the human will not notice. If
: something takes longer than 100 ms to occur, then the use may notice a
: discrepancy.
[..] 
: seem to remember that Card & Newell provide a good explanation of this
: idea.

from what i remember Card & Newell try to calculate a figure for an
appropriate frame rate based upon their "human information processor"
and instead of using this magic 100msec no. they have calculated
instead decide to half it and go for 50msec or 20hz!  (this is
probably an unfaithful recall but i don't have a copy of the book to
hand.)

i would query the choice of 100msec as an appropriate figure.  if
you take a look at some figures on the perceptual systems you will
find that the "magic number" for humans determining whether two
events are simultaneous is 30-40msec.  also if you look at visual 
search tasks (determining whether or not a target is present in an 
array of elements or whether there is a "discrepant" element) then 
presentation times of hundreths of seconds have been used.

so i'd go with Rych's statement that the latency that can be tolerated
depends on the task domain (and the implementation).

simon

skr@hagg.psy.ed.ac.uk (Simon Rushton)