What's in an alt text?

Over the past few months, we’ve been helping one of our clients improve the accessibility of their online resources, by writing alt text for the thousands of images that can be found throughout.

What we hadn’t quite taken into account at the onset of this journey was the shear scale of work involved in this project, and over 200,000 words later, it’s fair to say we’ve learned some lessons along the way.

To put that number into context, since June we've written more words than such literary classics as Dune, Little Women (parts 1 and 2), and every book in the Lord of the Rings trilogy. And we know what we’d rather be reading… (it’s the alt texts, obviously).

But before we start, what actually are alt texts? Short for ‘alternative text’, alt text is a short-form description that provides context and meaning to digital images, primarily for anyone who cannot view them. Although still somewhat underrepresented in digital media, alt text has become increasingly more common in recent years, not just in formally published media, but also on social media – where it has partly evolved from its original aim of improving accessibility, and can now often be seen giving additional meaning and context for the purpose of humour or storytelling. One such example can be found in a recent review of criminally weak tea on Sarah Millican’s X page.

However, in an educational publishing context, writing alt text presents a unique challenge, because there cannot be any room for misinterpretation or whimsy. This is particularly true in STEM subjects, where diagrams are nearly always a vital part of the exposition, rather than being used illustratively, and they are often also highly technical in nature. They must be explained accurately, and in a way that is appropriate for an audience that may not already fully understand the subject.

And that’s just the start of it – there are countless other considerations to be made when writing these texts for STEM images, and in this post we’re going to explore a few of the most common challenges we came across.

Giving it away

One difficulty with writing alt text for images that form part of quizzes, tests or exercise questions, is that you cannot write it in such a way that gives away the answer. Take this diagram, for example, which forms part of a quiz question about an experimental setup.

‍

At first, this diagram might seem simple to describe for any budding chemist. It's a standard setup showing a a tripod and gauze over a Bunsen burner, with a beaker resting on top. In the first image, the Bunsen burner is lit, and a colourless solution fills the beaker. In the second image, the Bunsen burner is off, and there is a dark powder in the beaker, in place of the where the liquid previously sat.

Pretty simple description, right? Now try again, except taking into account the fact that the question asks students to name each piece of equipment. Which means, of course, that you cannot use the following words: Bunsen burner, gauze, or beaker.

Here was our attempt:

On the left of the image is a thin metal tube standing on a wide rounded base, with fire at the top of the tube. It also has a gas supply tube near the base. This structure is labelled A. A tripod allows a transparent glass container labelled B, with liquid inside it, to sit above the fire on a sheet of criss-crossing lines of grey material, labelled C. To the right is exactly the same image, except the liquid inside the container is not there, and a small amount of powder is left inside the container. The fire is also turned off.

Tip: Screen readers typically announce that there is an image, so you don't need to start with 'An image of', or similar.

Up to interpretation

Another challenge in STEM-based images comes when trying to write alt texts for graphs or other statistical diagrams. A natural human response when presented with a graph is not just to look at it, but to also interpret it. Where are the relevant points plotted? Do they follow a particular trend? What do the axis labels tell us?

But what if the question is asking students to interpret the graph themselves? You can't just tell them that a graph follows a particular trend, because that might give the answer away. Conversely, you may not be able to just list all the plotted points, because the alt text would become long and impossible to follow.

There is quick solution to this, and it really comes down to a case-by-case basis. The key thing to always consider is, if you were the student, what would you find most helpful?

Tip: When writing alt texts, you are writing primarily for a screen reader software, rather than a human. So you don't always need to follow standard written styles, for example when writing coordinates. "The point 4 740" might be more accessible than "The point (4, 740)", even though visually it is harder to read, because when reading out coordinates you wouldn't typically add a pause between numbers.

‍

Order matters

When looking at an image, our eyes will subconsciously move around the image in a particular order. This order and direction will depend on the image, and whether it has a specific shape or features that draw our gaze in a particular direction.

For example, take this diagram of the cross-section of a plant stem. Rather than ordering the letter labels top to bottom, we found it made more sense to go with a slightly adapted 'outside inwards' approach. This better suits the order in which it makes sense to hear about the structure, rather than the order in which you might find yourself looking at it:

A green circle labelled A, with a dark green border, labelled B. The green circle has a paler green inner circle, labelled pith. A number of ovals lie half in A and half in the pith, arranged around the circumference of the circle. The boundary between A and the pith, inside an oval, is labelled cambium. The end of the oval in A is labelled C, and the end of the oval in the pith is labelled D.

Tip: Alt texts should, ideally, also be considered when creating the image in the first place, especially when ordered labels are used like in the example above. This one was created with the labels A to D given in the order that they would be described.

Science is hard!

Sometimes, the image you're trying to describe is just, well, difficult!

If you want to see what we mean, try describing this next image accurately and concisely to somebody who can't see it.

Although the goal is to always keep alt text as short as possible, sometimes we had to resign ourselves to the fact that it would be long text (sometimes used as a 'more complete' version of alt text, in addition to the shorter summary) or nothing. This was one of those occasions, and the beast of an alt text we found ourselves needing to write is below. And this, perhaps, raises further questions about accessibility in educational publishing, and how much reliance we should be putting on highly technical images in the first place. We certainly don't envy the students finding themselves on the other end of these headphones...

Three graphs all sharing the same horizontal axis, which is days of menstrual cycle starting with day 26 and 28, then going from 2 to 28, then from 2 to 4. The 3 vertical axes are hormone level in nanograms per millilitre, estrogen level in pico grams per millilitre, and thickness of endometrium. F S H levels rise to a peak of around 750 nanograms per millilitre during menstruation, then fall again, with a smaller peak around 600 nanograms per millilitre on day 5 of the cycle, before falling to around 250 on day 10. Another peak occurs just before ovulation, on day 12, with a value of around 600. F S H levels then steadily fall until the next menstruation. L H levels remain around 150 nanograms per millilitre for most of the cycle, but sharply peak at around 900 nanograms per millilitre on day 13. In the first part of the cycle, the follicle is starting to develop, then ovulation occurs on day 14. A corpus luteum forms around day 24. Estrogen levels gradually rise to a peak of around 300 picograms per millilitre on day 13 of the cycle. This then drops sharply to around 150 on day 16 of the cycle, before rising slowly to a second maximum of 200 on day 22, before falling to 50 as the cycle restarts. Progesterone levels, measured in nanograms per millilitre, remain consistently low, at around 1, for the first 14 days of the cycle before rising. Progesterone levels peak at 8 on day 19 of the cycle, before dipping slightly at day 20 and peaking again at 8 on day 22 of the cycle. Progesterone levels then rapidly fall to around one on day 26 of the cycle. The thickness of the endometrium rises steadily from day 2 to day 28, before rapidly breaking down and the cycle restarting.

You may also notice some other quirks of STEM subjects in the above text (assuming you've made it this far!). There isn't one standard as to how screenreaders interpret certain styles, and so FSH might need to be written out with spaces to avoid it being pronounced as the word 'fish' but without the 'i'. But conversely, the letter 'A' might be pronounced like the word if it's separated out, and so initialisms containing 'A' are better off written as a single word. And similarly, units that your brain interprets naturally on paper, such as ml^–1, will need to be written out in full as 'per millilitre'.

What's the alternative?

Well, to be honest, we don't know! As a team, we've learnt a great deal throughout this project, but we're also cognisant of the fact that the world of accessibility is fast-evolving, as the industry becomes incrementally better at making resources that work for everyone.

We'd love to hear from anyone else who has thoughts or experiences of writing alt text, and specifically for STEM subjects. Or, if you have resouces that need making more accessible and you'd like us to help, then do of course get in touch.