84

I do scientific research, and I noticed for a lot of plots, the different curves are represented with different styles, but they are all black. For instance, one curve may be a solid line, one may be a dashed line, one may be a dotted line with triangular markers, etc.

However, this makes it hard for me to differentiate between the curves. Color is much more effective for helping people distinguish between data (and I’m taking a data visualization class that has research to prove it).

So... can I do this in my research article? I imagine most people are keeping their plots black and white because of printing restrictions on color, but at the same time they include colors in all kinds of other figures in their document. Also, how many people actually read printed articles anymore? For scientific work, I look all of it up online.

Are there good reasons for me to stick with this backwards black-and-white scheme?

  • 4
    actually, I read all my articles on paper. I love to scribble on the margins. But, truth be told, if there's any figures in the paper, I'm gonna have it open on my screen and look at the pics on my computer. – penelope Oct 24 '13 at 9:09
  • 20
    "Color is much more effective for helping people distinguish between data" except for people who are color blind. – StrongBad Oct 24 '13 at 9:24
  • 17
    That's why you use color-blind safe colors. You can calculate how colors will appear for different kinds of color-blindness. – James Oct 24 '13 at 10:33
  • 13
    Many journals charge extra for colour plots. – Jack Aidley Oct 24 '13 at 13:34
  • 14
    Every serious guide I read while preparing my thesis suggested colors where possible but mandated that figures be interpretable without them (i.e. for fully color blind readers or when the articles has bee printed in B&W or xeroxed), and I expect to stick to that rule essentially forever. – dmckee Oct 24 '13 at 23:39
63

Hail to Colorbrewer

I use Cynthia Brewer's website Colorbrewer to pick color schemes. It's also embedded in ggplot2 which is a common choice of graphing package in R. It allows users to specify colorblind-safe and photocopy-able color schemes. It's meant to be for map making, but the idea works for visualization all across the broad.

enter image description here

Other color specialists

For graphs that are used in more eye-catching media such as reports, posters or infographics, you may consider taking some advice from fashion and interior designers (See the graph below proposed by Japanese Shigenobu Kobayashi). Each combination can conjure a certain type of feeling, color schemes that seem absurd may work in other settings.

enter image description here

Works on data visualizations

You mention there is research backing up the use of color, which I do believe there must be. Though for that, I'd just like to share another point of view. Bill Cleveland, whom I consider the guru of visualization in science, advises that color should actually be avoided. In his book, he lays out this hierarchy:

enter image description here

I may not go thus far to suggest color is worse than area and volume, but I do agree that positioning along non-aligned scales is much better than color. In other words, if I have two curves to show, I'd opt for paneling them onto to their own coordinate then put the graphs side by side, or plotting them together using line with different styles or grey scales over using colored lines.

The baseline for me is, the added colors should have their dimension (aka variable) to represent. If there is anything that does the same job, remove the colors because now the colors are "non-informative ink," as described by Tufte.

Human eyes are prone to visual tricks

I do agree that colors catch attention, but just because it gets attention does not mean people can distinguish them well and can isolate the information well. Particularly, colors interact; just by pairing with some different colors, the same color can look different. For instance, here is an example borrowed elsewhere. The brown and the orange tiles pointed by an white arrow are actually of the same color.

enter image description here

Why care about black and white?

And you asked "how many people actually read print articles anymore?" I would think that there are a lot. Schools in developing countries may not have color printer, and not every student has access to computer and the Internet. Their only way to learn about a journal article may just be through photocopies. Even in the US, with the rapid adoption of tablets, I still see a lot of journal clubs and meetings relying on B&W photocopies.

Now when I go to deliver any workshops in other countries, I always photocopy a few sample pages of my handout to make sure they are still legible. And Brewer's work has been a lifesaver for me.

Closing remark

In a nutshell, I embrace data visualization and the increasing acceptance of colors thrills me. Though in the process I believe we should remind ourselves to be humble and do not create works that very disproportionally benefit people who are riding high at the technology tide. John Tukey laid the groundwork of some graphical exploratory data analysis, Charles Minard made a visual with rich amount of variables like 150 years ago and yet still considered to be pretty bad-ass by today's standard. How many colors did they use? Just one.

At the end of the day, I would probably argue that design trumps all. It's not about color or non-color, it's about if they are used efficiently to maximize information transfer and minimize noises. At this moment, for me, lines, dots, shades along grey scale, and empty spaces are my staples; colors are my spices.

  • 3
    @Federico, because there isn't black-grey-white color blindness, for copying purpose you can uncheck "colorblind-friendly." As for colorblindness, the most common type is green-weak colorblindness, which can be worked around by avoiding putting red and green on the same chart. For the other cases, if you provide both colored and a photo-copier grey scale versions, it should be fine. – Penguin_Knight Oct 24 '13 at 16:57
  • 1
    @DavidKetcheson loading it into an image editor, selecting one of the orange/brown areas, and watching it change color as you drag it to the other is an even more interesting effect. – Dan Neely Oct 24 '13 at 20:52
  • 3
    @Penguin_Knight ok, unchecked "colorblind-friendly", and I was left with a single set of three colors, one of them clearly too pale to be used for a line in a plot (unless you also change the background color, which has its set of problems in print). This may be ok for maps, but I don't find it as good as you're advertising for coloring a plot, which was the original question. – Federico Poloni Oct 25 '13 at 8:38
  • 1
    @Federico, glad that you're experimenting with it! Please also check out the "diverging" categories because there you'll get a few more color schemes with three color (high, neutral, and low.) Time to time I may modify the scheme (for instance, if the color is too pale, try factor down the RGB code by the same number, say, 0.8, and you can have a darker theme to work with.) followed by a photocopier test by myself. And you're correct that the site is for cartographer, but the schemes has been giving me some bases to work on. – Penguin_Knight Oct 25 '13 at 12:21
  • 1
    con'd: Also, I believe by no means those are the only schemes. If we read Brewer's research articles, she also details how color spectra were chosen (a bit technical but nice read nonetheless). From that work it is possible to enrich the schemes by ourselves. All in all, the difficulty of finding a color scheme that is photocopier friendly should not be understated; it's a risky business, and line styles or paneling may be a better alternative. – Penguin_Knight Oct 25 '13 at 12:23
44

When making plots, I like to use reasonably subtle colours and combine them with different line or marker styles (e.g. blue triangles and red circles, or a black solid line and a blue dotted line). If printed in black and white the figure is still easily readable, and the colour version makes things a little more differentiable.

Stick to dark or pale colours, and they will enhance the appearance and readability of your document.

Do:

enter image description here enter image description here enter image description here

Don't:

enter image description here enter image description here enter image description here

  • 14
    Use of the colorbrewer palattes is a near defacto standard in cartography now, and is becoming widely accepted in other fields. You can see the online tool also has color-blind safe and scales that can stand black and white photo-copying. – Andy W Oct 24 '13 at 12:03
  • 1
    Excuse my ignorance, but why is the "don't" color scheme not preferred? I'm trying to understand the reasoning behind all the suggestions here. For instance Sigmaplot uses a default scheme of black, red, green, etc. lines, which would go against all the advice herein. – che_kid Oct 25 '13 at 2:05
  • 5
    @che_kid - Bright colours are attention-grabbing, jarring, distracting, and unattractive. None of those properties are ideal in a published paper. Yes, the majority of plotting programs use RGB coordinates of [255,0,0], [0,255,0] and [0,0,255] for their default red, green and blue (and have similarly bright intermediate colours). But, aesthetically, that's usually an extremely poor choice. – Moriarty Oct 25 '13 at 2:13
  • 4
    Used correctly, bright colours can be very effective as well as attractive. To accentuate one curve, for instance, do use a bright colour (and use dull colours for the others). I think the advice in this answer is way too generalised. – Konrad Rudolph Oct 25 '13 at 9:57
  • 2
    @che_kid: The "don't" is also bad because it's very difficult for people with red-green color blindness to distinguish those colors! – aeismail Sep 15 '14 at 17:30
18

Why might be reasons for you to stick to black and white:

  • Some legitimate scientific journals, even in the beginning of this 21st century, still have publishing fees for color figures and do not allow the combination of “color online and B/W in print”. It's sad to realize, but I know some in my own field.

  • Having color figures allows you to have a higher density of information on a given figure. Yes, it sounds like a good thing and it usually is, but it is a double-edge sword. First, because you may end up making figures that are just too dense and difficult to interpret (if the caption requires more than 3 sentences, you may have a problem). Second, because if you do not choose your colors well, it may be actually less readable than a black and white figure.

  • The rendering and perception of colors in printed materials and in video projected presentations is not at all the same. Thus, if you want to provide top-notch material, you might need to make two versions of your figures, which is extra work.


Mortiarty has said that the choice of colors is crucial. This is very true (and under appreciated by many), both for rendering and perception. I would advise you to read a book about color theory or color design (this one is the one I read to strengthen my understanding of these things).

To given an example, the software I love for drawing figures (Grace) has an almost unusable default color palette (left below), but with some practice I could change it to a palette that is quite nice to use (right below):

        enter image description here          enter image description here

Look in particular at how the basic colors have been desaturated: red, blue, green, and yellow.

  • The ROOT default palette is pretty grotty, too, though it is not a patch of the abomination that was the PAW default palette. – dmckee Oct 24 '13 at 23:42
9

Basically, the reason is as you suggested: historically, it has been very expensive to print in color, which made it unattractive for most research groups to do so.

Now, however, it is increasingly common for most distribution of journal articles to be in PDF form, which can just as easily use color as not. Many journals will let you do "color online" for the same cost as black-and-white. However, if you are planning to use colors, then they should be used in an intelligent manner. Arbitrarily using colors in a haphazard manner can make understanding even harder than using black and white. (Don't forget that many people are color-blind!)

  • 2
    Being exactly the same cost in PDF doesn't stop journals charging extra for colour. – Jack Aidley Oct 24 '13 at 13:34
  • 1
    Typically color charges are only if the print version is in color. – aeismail Oct 24 '13 at 15:40
  • +1 for the link from which I found this blog on information visualization and the "Colors for data scientists" web-application. – Alexey Popkov Sep 15 '14 at 16:22
3

Some might claim that having black and white is better because it's cheaper to print and keeps the focus on the content.

But I disagree with using black and white and think that the content is better supplemented with color. There is research that if your mood is changed positively when learning (with our 5 senses), humans retain more information. (Example: don't you retain more when you eat dinner with your friends in a new restaurant versus sitting at your desk and talking to someone on the phone?) This is not true if the brain becomes conditioned to the environment (or restaurant in my example). So the environment should continually change to increase one's ability to learn. Thus, you should have a web interface to keep changing the theme to something new each day they visit the research. If it's on paper, then you're S.O.L.

You could keep it in black and white if it's on paper and just spray some good smelling perfume on the paper. That may allow others to retain more with smell than with color, because smell has more of an impact on the brain than color.

Sight
Hearing
Touch
Taste
Smell

Another way would be to put it in a card that plays classical music. When listening to classical music while learning, you will retain more. Hearing is on the list of 5 senses.

I found this the other day that might help you.

https://kuler.adobe.com/create/color-wheel/

Then click the "Explore" hyperlink. And finally click the "Most Popular" hyperlink and the "Most Used" hyperlink.

I'm not sure which one is better.

You will find the work of many Adobe geniuses that might make your research a lot easier, rather than re-inventing the color wheel.

2

Some journals will ask you to pay a per-figure fee for your print figures to appear in color. If color really enhances the figure, then this is worth it. But if, say, you're just using a monochrome red colormap instead of grayscale, it's not.

2

Many journals will either charge for the use of color (especially those that still appear in print) and a few I've run across outright ban its use unless it's essential for an understanding of the figure (some photographs, etc.).

That being said, there's no reason you can't use color if you're willing to pay the fees, publish in online journals, use color in electronic supplements, etc.

One essential thing I find, that I always check as a reviewer, and find appallingly uncommon in color figures: make sure they're still understandable in black and white. It doesn't have to be as good, but someone printing out your work on their office printer should still be able to understand what you did. Distributed in PDF or not, I can't read things for comprehension on a screen, so I print them out - I know a fair number of colleagues who do likewise. And I have yet to encounter a figure that needed printer unfriendly colors to be understood.

protected by Alexandros Jan 2 '18 at 20:23

Thank you for your interest in this question. Because it has attracted low-quality or spam answers that had to be removed, posting an answer now requires 10 reputation on this site (the association bonus does not count).

Would you like to answer one of these unanswered questions instead?

Not the answer you're looking for? Browse other questions tagged or ask your own question.