What's the evidence? ​Deciding between words and tables

May 2018

Facts

Looked at the Super Rugby results lately? A bus timetable, perhaps, or the TechCommNZ subscription rates? Chances are, you’d have looked at a table. How should technical communicators decide when to use one?

Martin Cutts, editing director of the Plain Language Commission, asked writers for alternative ways of explaining the information on paying for hall hire below. Why not have a go yourself? Use words, tables, or whatever you please. Solutions in a sec, but no cheating!

Table listing five formats. The first preference was a table, winning 11 out of 35 votes and rating 17 out of 20 for clarity. The next preference was a question and answer format, winning 8 votes and rating 18 for clarity.

Hiring fees of £100.00 or less must be paid in full at the time of booking. In the case of all hiring fees in excess of £100.00 the hirer agrees to pay £100.00 deposit at once on all bookings valued up to £200.00 or 50 per cent of the total hire fee as a deposit on bookings on or over £200.00 in value. The balance of the cost of each hiring must be paid at least 21 days before the date of the event (Cutts, 2013, p. 180).

Image by Mediations.

Believe it or not, there’s a standard on using tables ( British Standard 7581 A guide to the presentation of tables & graphs ), but we won't discuss it as it is 27 years old. Instead, we’ll review recent research, then look at tips on choosing and designing tables.

Reviewing recent research

There has been research comparing the effect of tables and text on memory and reading speed (cited in Cardoso, Leite, & de Aquino, 2016), but that was a few decades ago. More recent research tells us tables can sometimes be good, especially if they are well designed, but users may not read everything.

Tables can be good for presenting information. Back to hall hire payment. The writers suggested re-writing the information as:

  • paragraph text
  • bullet points
  • questions and answers
  • a table.

When Cutts took the suggestions to a focus group, 11 of the 35 people favoured the table, followed by 8 who liked the questions and answers. This was even though they found the question and answer style slightly clearer, rating it 18 out of 20 for clarity, against 17 out of 20 for the table (Cutts, 2013). By the way, what do you think of reading these results in a table instead?

Preferred format for payment information

Table 1: Focus group evaluation of different formats for presenting payment information. The table was clearly favoured. Adapted from Cutts (2013).

But tables aren’t always better. Nearly 300 professional accountants in Brazil were shown how many people entered and left a store in a 12-minute period (Cardoso et al, 2016).

Sicily

Image by kirkandmimi.

They were shown the numbers in one of several formats. For example, as text:

In the first minute, 9 people entered and 8 exited from the store. In the second minute, 10 people entered and 5 exited from the store.

Or in a table:

Cardoso experiment showing people entering and exiting

Or in line or column graphs. But let’s not get distracted by those since we’re only comparing text and tables.

Participants were then asked:

  • In which minute is there the largest number of people entering the store?
  • In which minute is there the largest number of people exiting the store (Cardoso et al, 2016, p. 5)?

If you think – as I did – that tables would be more helpful than text for this exercise, well, surprise! There was no difference in the accuracy of answers. The researchers believe this finding ‘can be explained by the fact that both display symbolic information (numbers)’ (Cardoso et al, 2016, p. 10). I asked one of the authors what they meant but did not hear back.

Tables can be useful when well designed. There’s global research, spanning more than a decade, about summarising healthcare information on diagnostic tests. Hundreds of participants, mainly researchers and health professionals, provided feedback in working groups, workshops, and formal user testing.

Some of the findings are too specialised to mention here, but the following are more generally applicable (Mustafa et al, 2015):

  • Tables had the thumbs up. ‘Almost all participants preferred summarising the results of systematic reviews of test accuracy in tabular format rather than plain text’ (Mustafa et al, 2015, p. 1).
  • User testing is important. Some early respondents wanted a simple table with certain information, but when the authors tested the format, ‘participants unanimously noted that they did not prefer it’ (Mustafa et al, 2015, p. 6).
  • Headers providing context were appreciated.
  • Generic comments were deemed unnecessary.
  • Footnotes had mixed reviews. Many participants did not read footnotes. They felt critical information should be in the table, not ‘hidden’ in footnotes. Footnotes should be concise, appear on the same page as the table, and called ‘explanations’ or ‘clarifications’ [ sic] instead (Mustafa et al, 2015, p. 9).

People are more likely to read the first few lines (never mind what they say). Researchers at the University of Minnesota recruited 203 participants through a local magazine. This eye-tracking study involved browsing nutrition labels of 64 common American foods, such as pizza, crackers, and soups (Graham & Jeffery, 2011).

Typical nutrition label. First five rows show number of servings, serving size, calories per serving, total fat, and saturated fat.

Image by U.S. Food and Drug Administration.

Findings suggest consumers read only the top five lines of nutrition labels, although participants said they were equally likely to read the top and bottom of the label (Graham & Jeffery, 2011).

They authors caution this may not reflect real-life shopping behaviour. They don’t comment on the demographics, but I’d like to. Participants averaged 42 years of age, were predominantly female, and two-thirds had at least college education. Their high education level suggests that whatever the reason for reading only a few lines, it wasn’t literacy.

In summary. So the research confirms tables can be useful for looking up information. At worst, as Cardoso et al’s store entry and exit research showed, they are no worse than text. And as seen from Cutts’ payment exercise and Mustafa’s research on diagnostic tables, they may be preferred. Even then, users may mainly read information at the top of the table, so provide context, keep notes concise, and do user testing.

Choosing tables

So how do we actually know when to use tables? Cutts advises us ‘the best presentation would depend on the document’s purpose and the readers’ purpose. A question and answer layout might be seen as friendlier than a table, but friendliness would be more relevant in a sales leaflet than a legal contract. Readers wanting to get the answers to specific questions might perform better with a table, while those wanting to remember what they had read might perform better with questions and answers’ (Cutts, 2013, p. 182).

Writer Josiah Fisk believes ‘for summarizing text, it's hard to beat a bullet list, with a few bolded words at the beginning of each bullet giving the topic or key message of that bullet’ (Fisk, 2017).

Try a table:

  • for large amounts of information or complex information. ‘Tables convey large amounts of numerical data easily, and they are often the only way to present several variables for a number of items . . . . Although tables lack the visual appeal of other kinds of graphics, they can handle much more information.’ (Markel, 2012, pp. 317–321).
  • Tables are good for words too, being ‘especially useful for depicting complex data, either in numbers or in words’ (Woolever, 2008, p. 96).
  • for reference information. ‘Tables are usually best for when you want to look something up. They're good at steering you quickly to what you're looking for’ (Fisk, 2017).
  • to compare items against a standard (Woolever, 2008)
  • to show if … then … choices (Woolever, 2008)
  • as an alternative to related bullet lists (Malamed, n.d.), for example in table 2:

Two-column table showing four categories and the trusted brands. For example, in the breakfast food category, the brands are Sanitarium, Hubbards, Vogel’s.

Table 2: Selected categories of New Zealand’s most trusted brands in 2017. Anyone know why the survey didn’t cover essentials like coffee and beer? Source: Reader's Digest New Zealand www.trustedbrands.co.nz/results.asp

Designing tables

Go ahead and use tables if they’re the best tool, but bear in mind that they don’t always print well or display well on different screens. They can also be inaccessible to those using screen readers if not properly designed.

If you opt for tables, remember to:

  • arrange information to be compared in columns rather than rows, and in logical order (Markel, 2012; Woolever, 2008)
  • show row or column totals if they would be helpful (Markel, 2012)
  • draw only as many lines (rules) as necessary (Markel, 2012; Woolever, 2008)
  • keep layouts simple. Merged cells, nested tables, and other elaborate layouts can confuse and disorient those using screen readers and cause problems for responsive design on small devices. ‘Use the simplest table configuration possible’ (WebAIM, n.d.).
  • make tables only as wide as necessary: ‘the reader should be able to scan across a row easily’ (Markel, 2012, p. 323)
  • mark up tables for the web with appropriate HTML tags, to enable those using screen readers to ‘see’ columns and rows. Otherwise someone trying to read a timetable could be confronted by this:

Unformatted text-only box with text separated only by commas. The text is a jumble of row and column headers, days, times, room numbers, and instructor names.

Figure 1: A timetable as it sounds to someone using a screen reader, if the table is not marked up properly. Ugh! Image © WebAIM https://webaim.org/techniques/tables/

To learn about designing accessible tables for the web, see WebAIM’s Creating accessible tables.

Being flexible

Let’s give the last word on choosing the right format – words, tables, or something else – to Cutts: ‘There’s no rulebook. You have to be flexible and experiment a bit to see what works. And, ideally, you should test your decisions with typical readers’ (Cutts, 2013, p. 178).

References

Cardoso, R. L., Leite, R. O., & de Aquino, A. C. B. (2016). A graph is worth a thousand words: How overconfidence and graphical disclosure of numerical information influence financial analysts accuracy on decision making. PLOS ONE, 11(8), e0160443. Retrieved from http://journals.plos.org/plosone/article?id=10.1371/journal.pone.0160443

Cutts, M. (2013). Oxford guide to plain English (4th ed.). Retrieved from http://kupdf.com/downloadFile/5905b05bdc0d601764959efc

Fisk, J. (2017). Re: How people read text when it is in table form [Online forum comment]. Retrieved from http://www.linkedin.com/groups/158634/158634-6299805080157855746

Graham, D. J., & Jeffery, R. W. (2011). Location, location, location: eye-tracking evidence that consumers preferentially view prominently positioned nutrition information. Journal of the American Dietetic Association , 111(11), 1704-1711. Retrieved from https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3267975/pdf/nihms336079.pdf

Malamed, C. (n.d.). 6 alternatives to bullet lists . Retrieved from The Learning Coach website: http://theelearningcoach.com/media/graphics/alternatives-to-bullets/

Markel, M. (2012). Technical communication. (10th ed., pp. 638-643). Boston, MA: Bedford/St. Martin’s.

Mustafa, R. A., Wiercioch, W., Santesso, N., Cheung, A., Prediger, B., Baldeh, T., Carrasco-Labra, A., BrignardelloPetersen, R., Neumann, I., Bossuyt, P., Lelgemann, M., Bühler, D., Brozek, J., & Schünemann, H. J. (2015). Decision-making about healthcare related tests and diagnostic strategies: User testing of GRADE evidence tables. PLOS ONE, 10(10), e0134553. Retrieved from http://journals.plos.org/plosone/article?id=10.1371/journal.pone.0134553

WebAIM (n.d.). Creating accessible tables. Retrieved from http://webaim.org/techniques/tables/

Woolever, K. (2008). Writing for the technical professions (4th ed.). New York, NY: Pearson Longman.