Line graphs are the Swiss army knives of data visualisation. They can be almost anything… which is both good and bad.
Line graphs are slow to interpret
Many graphs serve one clear purpose. Take the five graphs below:
Even without labels, it’s clear what role each of these graphs serves:
- Pie chart—components of a total
- Thermometer—progress toward a goal amount
- Speedometer—percentage of the largest possible value
- Histogram—distribution of values
- Box plot—statistical summaries of several datasets
In other words, if I’m presented with one of the graphs above, I have an immediate head start on interpreting it. If, instead, I’m presented with a line graph, I’m forced to read the axes labels and limits first.
Deciphering text is the slow way to intake information. Shape is fastest, then colour, and only then text. This so-called Sequence of Cognition, popularised by Alina Wheeler, is something marketers need to know about.
Of course, marketers play to a different audience than scientists. Marketers have to entice people who don’t care, whereas scientists tend to have captive audiences. Someone reading a scientific paper, for example, has already demonstrated some desire to learn. Except, maybe they haven’t. Maybe they’re not reading the paper, but rather skimming it.
If you’re like me, there have been numerous times when you’ve chanced upon a potentially relevant scientific paper, and to judge whether you’ll actually read it, you skim over the figures. And this is when line graphs fail!
Having to stop to read the labels and axes negates the whole point of skimming. As Jonathan Schwabish notes in his book Better Data Visualizations: “Reading a graph is not like the spontaneous comprehension of seeing a photograph. Instead, reading a graph has more of the complex cognitive processes as reading a paragraph.”
Line graphs are seldom pretty
Andy Cotgreave asks whether the ultimate goal of graphs is communication or engagement?
His short answer: “Getting people to engage is sometimes as important as building the cognitively most valid method.” Although he refers to line graphs as the “purer way” to display data, he recognises that they’re also forgettable. They don’t pique the reader’s curiosity in the way that more elaborate visualisations can.
To summarise his point, Cotgreave quotes Giogia Lupi: “Beauty is a very important entry point for readers to get interested about the visualisation and be willing to explore more. Beauty cannot replace functionality, but beauty and functionality together achieve more”
At this stage, you might be ready with the following argument: “sure, eye-catching aesthetics are great for graphs intended for a broad audience, but line graphs suffice when scientists are communicate with other scientists”
Not exactly… as I’ll show.
Line graphs aren’t emotive
“There’s a strand of the data viz world that argues that everything could be a bar chart. That’s possibly true but also possibly a world without joy.” That’s a take from Amanda Cox, statistician-turned-graphics editor at the New York Times.
Although Cox said “bar chart” not “line graph”, she’s implying that a spartan approach to visualisation is often a poor choice, as is true in my less-than-joyful example below.
Consider the 40-year time series below of sea ice extent in the Arctic Ocean in September (its seasonal minimum):
Until a few years ago, this was my only sense of the state of Arctic sea ice. I’d seen the figure many times; it’s routinely used by scientists to remind everyone of the dire situation. But it was only when I saw similar data presented in a more visceral way that I finally appreciated the magnitude of the problem:
But let’s keep using line graphs
I’ve spent this whole post hating on line graphs. And yet I still use them all the time. They’re often either the best choice or the only sensible choice.
But when I do make line graphs, I aim to overcome some of their limitations described above.
Consider a quantity that fluctuates about zero—a temperature anomaly, say. By default, its graph might look like this:
Only by reading the label can the reader see that the quantity plotted is an anomaly. Whereas with the approach below, this is immediately obvious:
Or consider using other preattentive attributes to save effort for the reader. For example, in my field of oceanography, it’s common to plot how currents change with depth. An obvious choice is the standard line graph below on the left. Changing this to a series of arrows gives the reader clues about what quantity is being shown before they’ve even read the axis labels.
With multiple lines, line graphs can become a tangled mess. By default, plotting software often distinguishes successive lines with arbitrary colours:
There’s nothing necessarily wrong with presenting data like this. Except there’s no focal point. When deciding which of the eight lines to look at, the reader may give up and look at none of them. To quote Jonathan Schwabish again: “How many complex line charts have you seen and just immediately skipped over?”
Forgoing some detail, together with a simple change in colour scheme, makes the figure more inviting, accessible, and focused:
Another way to avoid a tangled mess of lines is to introduce them one-by-one. This isn’t possible in a scientific paper, but can be used to great effect otherwise. A widely shared visualisation from 2015 was simply a graph with eight lines. But by showing only one line at a time, the authors tell a compelling story about the cause of climate change.
Even if there’s only one line, the reader may still not know where to look. Again, consider creating a focal point. This time with annotations.
There are no set rules on when to use line graphs
I’m being intentionally neutral in calling line graphs both the best and worst way to visualise data. There are arguments for either position, an ambiguity that is well captured in a Datawrapper post:
We can solve the problem “Simple charts are boring” with “We need other chart types.” Or we can solve it by using simple charts — with all their advantages — but letting them show more relevant and interesting data and making them look better.