VisiBone

Web Palette Color Distinction Test

Sat, 30 Jan 1999

Dear Webmasters and Designers,

Thanks very much for the careful responses to this survey from the Webdesign-L and VisiBone mailing lists.  Here are the results, displayed somewhat crudely by operating system, which seemed about the only way to differentiate answers that mattered.  I don't know if brands mean much any more (e.g., as heard on public radio, "Have you driven a Volvo lately?") but here are the monitors represented, almost as many brands as respondents:

Here is the original question, reworded slightly so the answer tallies make sense: "Do these four reds appear different to you?  If they don't, you'll see one wide red rectangle.  If they do, you'll see four tall red rectangles up against one another."

R
FF0000
RRP
FF0033
RRO
FF3300
LHR
FF3333
pixel.gif (807 bytes) pixel.gif (807 bytes) pixel.gif (807 bytes) pixel.gif (807 bytes)
Obviously: Mac Mac Mac Mac NT Win95 Win98
Subtly: Mac Mac
Barely: NT Win95 Win95 Win95 Win95 Win98 Win98 Win98
Not at all: NT NT NT Win95 Win95 Win95 Win95 Win95 Win95 Win98

How about these four greens:

G
00FF00
GGT
00FF33
GGS
33FF00
LHG
33FF33
pixel.gif (807 bytes) pixel.gif (807 bytes) pixel.gif (807 bytes) pixel.gif (807 bytes)
Obviously:
Subtly: Mac
Barely: Mac Mac Mac Mac Mac NT Win95 Win95 Win98
Not at all: NT NT NT NT Win95 Win95 Win95 Win95 Win95 Win95 Win95 Win95 Win98 Win98 Win98 Win98

Or these four blues:

B
0000FF
BBA
0033FF
BBV
3300FF
LHB
3333FF
pixel.gif (807 bytes) pixel.gif (807 bytes) pixel.gif (807 bytes) pixel.gif (807 bytes)
Obviously: Mac Mac Mac Mac Mac NT Win95 Win98
Subtly: Mac Win95 Win95 Win98 Win98
Barely: NT Win95 Win98
Not at all: NT NT NT Win95 Win95 Win95 Win95 Win95 Win95 Win98

And how about all these eight "blacks":

K
000000
OWR
330000
OWY
333300
OWG
003300
OWC
003333
OWB
000033
OWM
330033
OG
333333
pixel.gif (807 bytes) pixel.gif (807 bytes) pixel.gif (807 bytes) pixel.gif (807 bytes) pixel.gif (807 bytes) pixel.gif (807 bytes) pixel.gif (807 bytes) pixel.gif (807 bytes)
Obviously: Mac Mac Mac Mac Mac Mac NT NT NT Win95 Win95 Win95 Win98 Win98 Win98
Subtly: Win95 Win95 Win95 Win98 Win98
Barely: NT Win95
Not at all: NT Win95 Win95

For contrast, the following four shades of red should be obviously different to everybody:

LHR
FF3333
LRP
FF3366
LRO
FF6633
LFR
FF6666
pixel.gif (807 bytes) pixel.gif (807 bytes) pixel.gif (807 bytes) pixel.gif (807 bytes)

 

Clearly, Mac's tend to do a better job than PC's.  No surprises there.  But Mac's are not perfect either. See the greens: nobody reported they could see obvious distinctions among all of them.

One technical point on the differences between the answers in red, green and blue, and how it's easy to draw the precisely wrong conclusions.   Note that the greens are much more likely to appear identical than reds or blues.   That's not a crime in green representation, it's in red and blue.  Notice that the green samples differ only in their red and blue values -- the greens all use a green value of FF and red and blue values of 00 or 33.  Several reported that within a set of four rectangles, some shades appeared identical while others did not.  Based on all of this I conclude:

Green levels of 00 versus 33 are the most distinguishable by far
Red less so
Blue least of all

I tend to believe this distinction only represents eyeball physics, and that the crimes somewhere in the driver/monitor system are uniform across the three colors.

Summary

Now there are many aspects to this survey and the conclusions that would make a statistician blanche in horror.  The sampling is from a unique group, with voluntary participation to boot.  The questions and answers are terribly subjective.  Maybe some big magazine with a big budget can do it "right" with representative samples and sensitive color test equipment.  I'm certainly not able or willing.  But here's a little data and here's where I think it points.

It's apparent there is some seriously widespread nonlinearity on the way from the numbers to the eyeballs among the low (dim) values in the 216 color web-safe palette.  The results of this survey imply that choosing between 00 and 33 for RGB color codes will make a less than obvious difference for 70% of the people who view them.  For 40%, there will be no difference at all.  I still don't have anything in the way of an explanation for this bizarre state of affairs.   I'm sure there's an engineer or two somewhere who's responsible.  If anyone has clues, please let me know.

When I design a web site I certainly can't afford to ignore a random 40% of the users.  Until monitor technology is much more advanced, or unless the audience is anything other than captive and homogenous, I suggest we're in a disappointing situation:  We really only have 125 colors to choose from.

The 1-3-letter codes are the VisiBone Anglo-Centric Color Code (VACCC).  I made them up for web palette colors. To compare other colors, feel free to use the online color lab.

-- Bob Stein

COLOR  HTML   LAB SWATCH CODES   HTML TAGS CHAR   REGEXP CC POP      faq  bulk  press  about  home  stein@visibone.com
    color wheel  color blindness  links  articles  tnar  contact
PRODUCTS  FREE COLOR FREE HTML FREE  ETC
2002 VisiBone