Font Readability Test for Printed Material

John M. Pye - December 2001

Objective

I needed to select a readable font for use in the PDF version of the Southern Forest Resource Assessment Draft Report, released November 23^rd, 2001. While many visitors might read the PDF's online, most are expected to print it out first. Therefore readability in printed form was the important criterion.

Background

A scan of the literature on readability of different fonts showed that some studies favored serif fonts over non-serif for the relatively high resolution output achievable with today's printers, but advice for specific, readily available fonts was lacking. We wanted a font that looked good when printed both on laser printers and lower quality ink jet printers, but was not so exotic that it would create production or publishing problems.

Method

In reply to elected to run a quick usability test, printing a page full of text selected from the report, but formatted with one of nine different combinations of type face and font size. Readers were asked to locate selected words on the page, with faster completion times taken as the measure of readability. User aesthetics were not considered for this test although the formatting selections I chose to compare at least partly reflect my subjective judgment of what constitutes an attractive appearance.

Four readers were selected for the test, representing a literate group but with individuals not particularly familiar with the discipline the text was discussing.

Readers were given one sample page at a time plus its corresponding list of four target words. The target words were drawn from each quarter of that page from top to bottom, in order. With only a few exceptions, target words differed between pages. The test required that the four words be found and circled on the page with the effort timed. Each page had a different combination of typeface and font size, and page order was changed for each reader. Each reader was tested on each combination of typeface/font size.

The type faces chosen for the test were Arial, Bookman Antigua, Palatino, Garamond, Georgia, Palatino, Tahoma, and Times New Roman. All are readily available on commonly configured PC's including standard USDA Forest Service PC's. Arial and Tahoma are sans serif fonts, the remainder are serif. Font sizes ranged from 10 to 14, with particular typeface-size combinations selected to represent reasonable choices for readability. In some cases leading was also adjusted to 14 pt. Pages were all printed on a 1200 dpi laser printer but readers were given photocopies of the original to simulate lower quality printing.

Reformatting the example document altered the specific portion used for each full-page sample, but examples all were of a similar portion of the document. More importantly, reformatting altered the number of words on a page. To adjust for variations in the number of words needed to be scanned, I divided the seconds needed to complete each page by the number of words on that page to yield a "words scanned per second" metric. Because different readers had different overall speeds, for each reader I ranked the different typeface/font combinations by words per second. The conclusions below are based on rankings for each font/size combination, averaged across readers. I got similar conclusions by averaging the words per second metric across readers.

Results

Every reader achieved their fastest times with the Georgia 11 pt combination. Three of the four readers placed Arial 11 pt as their second or third fastest-reading font, with Arial 10 pt less readable for all four readers. The remaining font combinations were generally inferior. Garamond 12 pt (14 pt leading) was highly variable among the readers. For two readers this combination ranked in the top three for speed, but it was ranked eighth for the other two readers. Times New Roman 12 pt was ranked third for one reader but did poorly with the other three. Tahoma, designed for the screen rather than print, was in everyone's bottom three. Bookman Antigua 11 pt (14 pt leading) did considerably worse than the same typeface and font size with normal leading.

Table 1. Readability scores for different typeface, size and leading settings.

typeface	size	line leading	words/ page	words/ sec	aver. rank
Georgia	11	14 pt	485	16	1
Arial	11	single	567	12	3
Bookman Antigua	11	single	467	11	4
Arial	10	single	622	10	5
Garamond	12	14 pt	465	8	6
Palatino	11	single	381	10	6
Times New	12	single	339	9	6
Bookman Antigua	11	14 pt	429	7	7
Tahoma	11	single	441	5	8

Conclusions

Georgia was found to be a very readable choice for printed material, with the added benefits of being widely available, attractive, and not overused. Arial proved a reasonable second choice, surprising given that it's a sans serif font. Such simpler fonts are usually thought be at a disadvantage to serif fonts for printed media. Note however that Arial must be at least 11 pt. If you expect the same publication to be read heavily online as well as in printed form, Arial would be a good choice. At the other extreme was the other sans serif font, Tahoma, it is clearly a poor choice for readability. Times New Roman also showed a surprisingly low ranking despite its reputation as a well-designed typeface and certainly one with which readers are familiar.

Conclusions are more difficult beyond these four typefaces. Garamond worked well for some readers, but very poorly for others. Bookman Antigua's readability was good under standard spacing but poor for slightly wider spacing.

Future Analyses

Four readers is a small sample size, it'd suggest two to three times that size for more reliable conclusions. There is also the issue of selecting the candidate formatting options. Document designers can choose from a large number of typefaces plus 10 to 14 pt sizes for the generally accepted range of body text. Line spacing offers yet another adjustment that apparently can affect readability. This test is by no means comprehensive, nor does it untangle the interactions of the formatting options available. Lastly, the testing method employed could be someone more rigorous. In future I would select the same number of words for each offering, all on its own same printed page, with replicate pages for each font combination. This system would allow you to offer only one target word per selection, selected randomly from the page (but still in a grammatically and logically correct context).