In the process of writing chapter 2 of Finding Ghosts in Your Data, I used quite a few images to tell a story. I’ve used many of these images in other contexts, but ended up needing to make changes to increase contrast and deal with the lack of color. Forthwith are two examples.

## Example 1: Two Colors

First, I have this example of the Gestalt principle of good continuation:

In a full-color environment, this is absolutely fine: blue and orange are great contrast colors, as there’s no risk of someone with color vision deficiency. But there can be a problem in print, where images tend to be grayscale in an effort to save money. Here’s what the same image looks like in grayscale, thanks to the excellent Coblis tool.

Yes, there’s a difference, but it’s hard to tell because the contrast is relatively low. How low? Well, we can calculate that with a contrast ratio tool. According to SnagIt, the lighter color is #969696 and the darker #6e6e6e. Plugging that into the contrast ratio tool and we have a contrast ratio of 1.72, which is not great. Granted, that tool is for text on a background rather than images like this, but that only changes the “Is this good enough?” recommendation, not the contrast ratio itself. Meanwhile, if I change the orange dots to black, the image looks a bit more like this in grayscale:

Now the contrast ratio is 4.11, making this much easier to distinguish, especially on a white or near-white page.

## Example 2: Lots of Colors

The second example shows off the principle of similarity, in which we naturally associate similar-looking things together. In this case, we have a series of dots of three separate colors, and if I show you the image, your first inclination is to believe that the dots of the same color are related.

This image works fine in full color, but in grayscale?

In this case, dark grey and blue are really difficult to distinguish. It’s #6e6e6e versus #595959, which leads to a contrast ratio of 1.37. Our eyes can differentiate them, but only with difficulty and not necessarily 100% of the time. If I went to print with the image above, it’s possible that nobody would have seen it until it was too late, and then I completely lose the point I was trying to make about similarity.

To fix this, we have to worry not only about the colors themselves, but also their contrast ratios. Here’s the image which is going to go to print…one of these years…

In grayscale, it looks like:

We already know that our hue of blue translates to #6e6e6e and black is #000000, for a contrast ratio of 4.41. From #6e6e6e to the pale yellow’s #dbdbdb, we have a contrast ratio of 3.68. And from #dbdbdb to #000000 we have a ratio of 15.16. Now all three shades are easily distinguishable in our print image and people can focus on the story, not fighting to discern the image.

## Coda: Why Is This Important?

You can’t always control the medium in which people will see your work. For example, in last week’s In the Papers, I covered an interesting article but noted one big flaw when reading the paper on my grayscale tablet: the images. Here’s an example of one image:

In full color, the image makes sense. On my tablet? Not so much. Call that a flaw of my reader if you will, but if there’s a reasonable chance someone will read or print your work in grayscale, it makes sense to check your images beforehand.