Ergonomics and neurology for interface designers

[Eric S. Raymond]

David Kastrup <address@hidden>:
> I don't think that the numbers you throw in here carry a lot of meaning.
> As a musician, I certainly have to be able to produce and recognize runs
> with individual notes shorter than 0.17 seconds.
> 
> The speed typing record is at 216 words per minute.  That's words, not
> letters.
> 
> So whether or not those kinds of delay turn out relevant in practice
> very much depends on which tasks with what kind of interactivity they
> appear in.  Blanket musings about some "speed of thought" are
> meaningless.

As you say, context matters a lot.  The 0.17sec ergonomic latency threshold
is specific to human-computer interfaces; it was discovered by Jef Raskin during
the early design studies that led to the Macintosh interface, and has been
experimentally confirmed pretty solidly.

I'm a musician too, and music does indeed have time granularity finer
than the reflex-arc time.  This is possible (as is speed typing and
pitching baseballs) because humans have the ability to compose action
sequences with finer time granularity and ship them to a limb for
execution.  In effect, we do downloadable motion subroutines.

(By the way, other animals - even higher primates - are not very good
at this.  There is an interesting and plausible theory that the
ability developed in early hominids as an adaptation for throwing
rocks at small game, well before we became tool-using cursorial
hunters of large game. And that the same action-buffering circuitry
was later recruited for pattern recognition in language, music, and
mathematics - getting good at throwing, in effect, pre-adapted us for
abstract intelligence.)

Music fools us. The ear can hear with finer time resolution than
spinal-reflex-arc time, so we think musical patterns like 1/32
drumbeats are being generated by an action/reaction process that loops
faster than humans are actually capable of.  In reality, a skilled
musician is not controlling every motion through an action/reaction
loop - he or she is shipping those precomposed sequences.

What humans cannot do is take sensory input, process, and then *react*
faster than spinal-reflex-arc time!  Raskin's basic discovery was that if you
throw a mockup of your application start controls on the display, you have
a minimum of 0.17 seconds to finish initializing the real controls before
a human is capable of noticing and then trying to do something.
-- 
                <a href="http://www.catb.org/~esr/";>Eric S. Raymond</a>