Good headings help, bad headings hurt

I’ve been on the road recently, teaching my ‘Editing that Works’ workshops to teams of web content providers in a government department. ‘Choose what to say,’ I urge them. And do it like this:

  • apply temporary headings to your text, making sure each heading is a full sentence
  • review the headings to see whether they’re any use to your readers
  • cut out the bits that readers don’t need
  • and don’t bother doing any editing until you’ve done your cutting.

We spend a lot of time in the workshop writing headings and cutting. Then writing more headings that reflect what we’re left with. Do you see a bit of a heading theme developing here?

Researchers tested the effectiveness of headings

So I was brought up a bit short when a friend asked me my opinion about a recent article in Technical Communication, the journal of the Society for Technical Communication. Bartell, Schultz and Spyridakis, researchers at the University of Washington, tested the effects of headings in comparable print and online materials.

  • Bartell, A, L., Schultz, L. D. and Spyridakis, J. H., (2006) “The Effect of Heading Frequency on Comprehension of Print Versus Online Information” Technical Communication, Volume 53 Number 4, November 2006

They chose similar articles on two types of arthritis, prepared four versions: no headings, low frequency (one heading approximately every 300 words), medium frequency (every 200 words) and high frequency (100 words). To give you a benchmark, we’ve just hit 204 words in this article. So in the high frequency condition, we would have already had two headings. I’ve had one, which makes it ‘medium frequency’.

The team chose to ask engineering students to read the articles, then answer questions on them without referring back. The task was deliberately set up as

(a) a reading task where the material would be both unfamiliar and not very interesting and

(b) a ‘reading to learn’ task, where the objective is to master and remember material.

Did you spot that reading to learn uninteresting material is rather unusual? Most of us are ‘reading to do’, trying to achieve something. Or, if we are ‘reading to learn’, then it’s generally because we are somewhat interested in the topic. But let’s not worry too much about that for the moment and continue with the research.

paper and pen alongside an online article on a laptop
In research, paper was easier to learn from than online material

The results: paper is easier to learn from

Overall, paper scored better. Not a great deal better. Students who used paper achieved a mean score of 9.70 whereas students who studied online achieved a mean of 8.88 – that’s about 8% better on paper.

But if we look at the standard deviations (that is, the statistical measure of the spread of results), then it’s noticeable that the results for paper were more tightly grouped: 1.84 versus 2.59. So the top end of the students studying from paper (one standard deviation above from the mean) was a score of 11.54. And the top end of students studying online was: 11.47. That’s only 1% difference.

Now look at the ‘low end’. One standard deviation below the mean for paper is 7.86, the same for online is 6.29. That’s 40% difference.

So what I think is happening (my view, not taken from the paper) is that good students survive the transition from paper to web better than poorer students. The good ones get knocked back a tiny bit: the poor students take a big hit.

The results: too many headings hurt as much as no headings

When the researchers analysed the findings for the impact of headings, they got mixed results.

On paper, the frequency of headings didn’t make much difference. The researchers don’t give us the actual figures but the graph they publish shows all four heading conditions grouped rather closely around that mean comprehension score of 9.70.

But online, the difference is quite obvious. The order is:

  • Medium frequency headings (about every 200 words): the best
  • Low frequency headings (about every 300 words): in the middle
  • High frequency (about every 100 words) and No headings: tied for the worst, with ‘No headings’ slightly ahead.

Be careful with your headings

So yet again, we find that unthinkingly following a guideline is no good.

It’s not enough to say: ‘Write headings’. They have to be good headings, and placed thoughtfully in the text.

Brush up your headings

If you’d like more advice about writing better headings, try Ginny Redish’s explanation: Add useful headings on the website

This article is a version of one that first appeared in Usability News

Image, Traditional paper and pen vs electronic document, by Alisa Cooper, creative commons