Xref: utzoo alt.hypertext:802 comp.cog-eng:1913 comp.graphics:17018 comp.multimedia:269 comp.software-eng:5228
Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!usc!elroy.jpl.nasa.gov!decwrl!amdcad!jetsun!pyramid!athertn!hemlock!mcgregor
From: mcgregor@hemlock.Atherton.COM (Scott McGregor)
Newsgroups: alt.hypertext,comp.cog-eng,comp.graphics,comp.multimedia,comp.software-eng
Subject: Re: Images vs. Text
Message-ID: <34941@athertn.Atherton.COM>
Date: 2 Apr 91 22:56:29 GMT
References: <10292@pitt.UUCP>
Sender: news@athertn.Atherton.COM
Reply-To: mcgregor@hemlock.Atherton.COM (Scott McGregor)
Followup-To: alt.hypertext
Distribution: na
Organization: Atherton Technology -- Sunnyvale, CA
Lines: 90

In article <10292@pitt.UUCP>, grefen@sun14.cs.pitt.edu (Gregory
Grefenstette) writes:

>I'm doing work on hypertext visual interfaces and I
>would like to be able to prove what seems evident to me,
>that is,  that people can scan information presented
>in a visual image-based form FASTER than in a plain
>textual form.

I think I understand what you mean, but how you state your
hypothesis probably oversimplifies a key point. Visual image
information has more degrees of freedom in conveying information
in a limited area of constant size, vs. text.  It is therefore
conceptually possible to convey more information in that limited
area using visual means than textual means.  One important
difference is that text is read sequentially, whereas a visual
image may be processed more holistically.  In this respect a 
viewer may get the "big picture" faster than the reader, but
apprehending all the small visual specifics in detail might take as much
time as the reading the text.  But it might seem faster to the viewer
since they get some benefit sooner.  (For instance, looking at
a sales graph might quickly get across the general trend and
some notion of the variation from month to month, but finding the
exact amounts for each month in a graph could take as long or longer
than referring to a table). 

The notion of greater degrees of freedom giving the *capability* for
greater information density is important because whether that capability
is turned into to a real difference is dependent upon the skill of the
visual designer.  A bad visual design may actually convey less information
or take longer to convey it than a textual representation. (E.g.
a graph of the %change in monthly growth of sales actually visually obsures
information about the absolute values of sales).   Obviously,
bad textual representations can  also obscure information, but it seems
that given present educational systems in the US, people have more 
practice developing writing skills than visual design skills before
coming to the job market. So it is important in addressing the superiority
of visual presentation to textual to consider the backgrounds of the
producers.   This is one of the reasons why companies are often surprised
when they use "graphics, icons, windows..." in their designs but people
don't necessarily find them easy to use--often the people making the
selection of the "graphics, icons, and window" weren't trained visual
designers but just garden variety software developers.  Meanwhile, another
company with trained designers may make a big splash with their UI,
composed of the same graphics, icons and window primitives but arrayed in
a more thoughtful manner by experienced visual designers, (often even despite a
poorer underlying representation created by less skilled programmers).

An additional complication is that considerable information can be
apprehended not by explicit mention in the text, or visual display
but by a conventional reference to a body of widely known information.
This information can thus be elicited rather than directly conveyed, but the
effect on information density is effectively the same.  The CYC project
at MCC, and other investigations into natural language processing and
"common sense" reasoning are applicable here.  Much of this work is
specifically aimed at linguistic (typically textual) elucidation of 
conventional wisdom  I am not aware of any similar studies of this effect 
for visual information conventions though conventional use of
principles of composition, color theory, and contrast theory is widely
known and discussed in depth in Tufte. 

> But having little psych or human factors background,  
> I don't know of any references that I
> could use,  or even where to begin looking.

With little HCI background, you might want to start with some things
that are more accessible to the layman, such as Edward Tufte's book the
"Visual Display of Quantitative Information", or his other book, "Envisioning
Information". Don Norman's book, the "Psychology of Everyday Things", also has
some useful lay information on the cognitive limitation of human perception.
Another accessible and relevant book for the layman is "How to Lie with
Statistics".  

The Tufte books are filled with useful references to more technical
references, and you may also find the various publications of
and conference proceedings sponsored by ACM SIGCHI society to be
of use to you.  A reference librarian should be able to help you
find any of these books or journals and to narrow your search to
particular domains of interest.  You might also find information on
this sort of thing in other related areas, such as statistics.  For
instance, I gave a paper on applying principles of graphic design
to the creation of business graphics was presented at a Statistical
Analysis System (SAS) Users Group International (aka SUGI) conference
many years ago. You might find other nuggets of information of this
sort in various proceedings of such conferences.  Again a refernce
librarian should be of help.

Scott McGregor
Atherton Technology