William Tanksley, Jr wrote:
"Circuit" is perhaps a misleading term for what's going on -- but the
Stroop effect is what I was indeed referring to. My point is that when
you attempt to read a token that has both color and textual
information, you can't read both in one glance -- you have to look
once at the color, then look again at the text, then carefully think
about the two in isolation to make sure you don't mess them up, then
combine them to make sure you extract the correct meaning. It's a
slow, serial process that's inherently highly prone to cognitive
error.

Not sure I entirely agree with you, but I'm grateful for you bringing up
the topic of color perception.

Let me say why I feel I'm entitled to an opinion here. I worked in the
Human Factors Lab at the time Hursley was developing the IBM 3279 color
display. This essentially showed characters in color. The HF Lab
extensively tested color perception in connection with the device. Peter
Robertson was our color expert and he was skeptical about there being much
cognitive benefit to color except in certain well-defined tasks.

I say "cognitive" because there is evidently a strong affective benefit, eg
"red-for-danger", looks pretty, etc.

Moreover any productivity gain from using color was swamped by a host of
contributing factors (applied psychologists call them "variables") like
vision defects, device imperfections, ambient lighting, stress, Stroop
effect, various optical illusions... the list went on and on. Worse, this
differed widely between individuals and Pete had to perform extensive
ophthalmology on his test subjects to control these variables.

This wasn't what the developers wanted to hear, so we had to do some very
good science to hold our ground. Pete (if memory serves) was of the opinion
that any performance benefit from color could be entirely replicated on a
black+white screen with a font-change, eg bold, underlined, italic,
boxed... and that it was less messy to do so.

He found only two tasks which color improved significantly:

1. Drawing the eye to a given position on the screen
2. Blocking related text.

You can see an example of support for each task 1 and 2 at:
http://www.jsoftware.com/jwiki/IanClark/colortask

However this performance gain rapidly disappears with "noise", e.g. if you
didn't use color sparingly, or colored the screen to support two or more
tasks simultaneously, or used more than two colors. (Task 3 at the above
link)

This is not good news to proponents of syntax coloring. There is a danger
we start using a fashionable technology because we believe (without
foundation) its advantages are well-established. We get used to it and feel
bad when we have to do without it. Not unlike tobacco, which actually
causes (via its withdrawal) the tense feelings we use it to alleviate.

I've had to maintain a lot of APL+Win code in a syntax-colored editor.
Without consciously reminding myself of the above, I used to turn off all
coloring except for local and global variables, colored differently.
Everything else went monochrome. The result supported the 2 tasks of
finding all instances of local (global) variables, i.e. tasks 1 and 2, and
nothing else.

There is a case for coloring code simply to support the task in hand, to
change it when the task changes and to use only 2 colors, 3 at most. What
we're beginning to see instead is a candy box on the working screen with no
evidence for much benefit accruing.


On Thu, Apr 11, 2013 at 6:00 AM, William Tanksley, Jr <[email protected]
> wrote:

> Ian Clark <[email protected]> wrote:
> > William Tanksley, Jr wrote:
> >> Humans don't process color with the same circuits that process text
> > Doesn't the Stroop Effect
> > http://en.wikipedia.org/wiki/Stroop_effect suggest they do?
> > (Tightly-linked circuits, at least.)
>
> "Circuit" is perhaps a misleading term for what's going on -- but the
> Stroop effect is what I was indeed referring to. My point is that when
> you attempt to read a token that has both color and textual
> information, you can't read both in one glance -- you have to look
> once at the color, then look again at the text, then carefully think
> about the two in isolation to make sure you don't mess them up, then
> combine them to make sure you extract the correct meaning. It's a
> slow, serial process that's inherently highly prone to cognitive
> error.
>
> Therefore,
>
> -Wm
> ----------------------------------------------------------------------
> For information about J forums see http://www.jsoftware.com/forums.htm
>
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm

Reply via email to