Update March 2012
See my expanded critique of Colin Wheildon’s legibility research.

Back in 1998 when Times New Roman was still widely used on the web, my then boss made sure we always designed our web sites with Arial, as she hated the look of serif fonts on the web. Was it the case that sans serif fonts were more legible, or was it just a matter of taste?

In 2003 as part of my master’s degree I reviewed over 50 empirical studies in typography and found a definitive answer.

Introduction

An argument has been raging for decades within the scientific and typographic communities on what seems a very insignificant issue: Do serifs contribute to the legibility of typefaces, and by definition, are sans serif typefaces less legible? To date, no one has managed to provide a conclusive answer to this issue.

Part 1 provides typographical definitions.

Part 2 reviews the evidence for and against the legibility of serif and sans serif typefaces.

1. Definitions

Legibility vs. readability

Legibility is concerned with the very fine details of typeface design, and in an operational context this usually means the ability to recognise individual letters or words. Readability however concerns the optimum arrangement and layout of whole bodies of text:

An illegible type, set it how you will, cannot be made readable. But the most legible of types can be made unreadable if it is set to too wide a measure, or in too large or too small a size for a particular purpose.(Dowding 1957, p.5; in Lund, 1999 )

Typographical features

There are many elements in the design of a typeface which can contribute to its legibility.

Serif / Sans Serif

“Serifs” are the small finishing strokes on the end of a character. “Sans serif” fonts do not have these small finishing strokes.

Point size

Point size is perhaps the element most used to describe the legibility of a type face, but it can also be the most deceptive. Point size is a legacy from the letterpress system, where each letter is held on a small metal block. The point size actually refers to the size of this metal block, and not the actual size of the letter. The letter does not have to take up the full area of the block face, so two fonts with the same nominal point size can quite easily have different actual sizes. ( Bix, 2002)

X-height

X-height refers to the height of the lower case “x” in a typeface. It is often a better indicator of the apparent size of a typeface than point size ( Poulton, 1972 ; Bix, 2002 ).

Counters

Counters are the “negative spaces” inside a character. They are also good indicators of the actual size of the type.

Ascenders and descenders

Ascenders and are the vertical strokes which rise above the body of a character or x-height. Descenders are strokes which fall below the baseline of the x-height.

2. Evidence

Overview of legibility research: serif vs. sans serif

There are plenty of studies that show no difference between the legibility of serif and sans serif typefaces ( Tinker, 1932 ; Zachrisson, 1965 ; Bernard et al., 2001 ; Tullis et al., 1995 ; De Lange et al., 1993 ; Moriarty & Scheiner, 1984 ; Poulton, 1965 ; Coghill, 1980) ).

There are some high profile studies which claim to show the superiority of serif typefaces ( Robinson et al., 1983 ; Burt, 1959 ; Weildon, 1995 ) but these have been soundly criticised on points of methodology. ( Lund, 1997, 1998, 1999 ). See this post for a more detailed critique of Weildon’s research.

Particularly interesting is the case of Sir Cyril Burt, well known in psychology circles for being accused of fabricating his results. It turns out that he is likely to have continued this deceptive behaviour in his typographical work ( Hartley & Rooum, 1983 ).

Unfortunately, many researchers, typographers and graphic designers continue to cite Burt and Weildon uncritically, meaning that many of the informal resources on typography found on the web today continues to propagate unsubstantiated claims on the utility of serifs.

Most disappointing however, is that in more than one hundred years of legibility research, researchers have failed to form a concrete body of theoretical knowledge on the part that serifs may play in legibility ( Lund, 1999 ). Nor have they managed to make their work sufficiently known in the typographic community ( Spencer, 1968, p.6 ).

Arguments in favour of serif typefaces

Serifs are used to guide the horizontal “flow” of the eyes; The lack of serifs is said to contribute to a vertical stress in sans serifs, which is supposed to compete with the horizontal flow of reading ( De Lange et al., 1993 )

These are the most common claims when trying to make a case for the utility of serifs. However, serifs cannot in any way be said to “guide the eye”. In 1878 Professor Emile Javal of the University of Paris established that the eyes did not move along a line of text in one smooth sweep but in a series of quick jerks which he called saccadic movements ( Spencer, 1968, p. 13 ; Rayner & Pollatsek, 1989, pp. 113-123 ). Unfortunately many graphic designers and typographers continue to use this rationale for the existence of serifs, due to a lack of communication and cooperation with the research community.

Serifs are used to increase spacing between letters and words to aid legibility

Serifs are not required to control letter and word spacing – in fact, serifs would be woefully inadequate for this purpose. In traditional letterpress systems, spacing is achieved with small pieces of metal inserted between the letters, and by the spacing between the letter form and the edge of the print block. Spacing is even easier to manipulate with modern computerised typesetting equipment. ( Sassoon, 1993 ; Rubinstein, 1988 )

Serifs are used to increase contrast (and irregularity) between different letters to improve identification

Well established research has shown that whole words can be recognised just as quickly as letters during an eye fixation and that single letters can be identified quicker when embedded in a word. Such a ‘Word superiority effect’ would indicate that serifs are not needed for distinguishing between single letters ( Reynolds, 1979 ).

Serifs are used to bind characters into cohesive ‘word wholes’

The simple Gestalt created by spaces between words would be enough to bind letters into ‘wholes’. Furthermore, other features such as character ascenders and descenders should have a much greater effect on word recognition than serifs ( Poulton, 1965 ).

Readers prefer body text set in serif typefaces, so they must be more legible

Many studies conducted in the past did indeed find a preference for serif typefaces ( Tinker, 1963 ; Zachrisson, 1965 ). However, Tinker commented that perceived legibility was due to a great extent to familiarity with the typeface. 40 years ago sans serif typefaces were not as common as they are now, and if these studies were repeated, it would not be surprising to find completely different results. Indeed, more recent studies have shown that computer users prefer sans serif typefaces for body text online ( Boyarski et al., 1998 , Bernard et al., 2000-2001 , Tullis et al., 1995 , Reynolds, 1979 ).

What is important to bear in mind is that in almost all legibility studies, reader preference or perceived legibility tends to be inconsistent with user performance ( Lund, 1999 ).

Serifs are used for body text because sans serif causes fatigue

It is often claimed that reading large amounts of body text set in sans serif causes fatigue, but there is no evidence to support this, as measuring fatigue has not been a concern in the vast majority of legibility research comparing serif and sans serif typefaces.

Furthermore, “no satisfactory objective method of measurement has been devised. Subjective assessments of fatigue are subject to modification by a great many factors which may be totally unrelated to the experimental situation”. ( Reynolds 1979, p313 )

Arguments in favour of sans serif typefaces

Serifs are just an historical artefact

This could be true to a great extent, especially since claims attempting to justify serifs in retrospect have been less than convincing.

Many researchers attribute the origin of serifs to the Romans, some claiming that “Roman masons … terminated each stroke in a slab of stone with a serif to correct the uneven appearance made by their tools”. ( Craig, 1980; in Bix, 2002 ). Others state that “design by brush before execution in stone gave rise to … tapering serifs at the terminals of many strokes”. ( Bigelow, 1981; in Rubinstein, 1988, p10 ).

What ever their origin, serifs have been around for so long that perceived legibility is very likely to have been affected by familiarity – readers tend to rate as more legible the typefaces they are most used to ( Tinker, 1963 ; Zachrisson, 1965 ).

Sans serif are better on the web

Although studies of screen reading show no difference between reading from screen and from paper ( Dillon, 1992 ; Bernard, 2001 ), there could be some validity to this argument.

When typefaces are digitised for use on computers, the letter forms have to fit within a relatively small pixel grid, often leading to what are called the “jaggies” ( Rubinstein, 1988 ). Many web professionals such as graphic designers claim that this relatively low resolution cannot render effectively enough the fine finishing strokes of serif typefaces, and that sans serif typefaces lend themselves more naturally to being digitised, and come out cleaner and thus more legible.

However, this has not been borne out by recent evidence ( Bernard, 2001 , Boyarski et al., 1998 , Tullis et al., 1995 , De Lange ), that shows no difference in legibility between serif and sans serif font on the web.

Sans serif is better at small sizes. Sans serif fonts survive reproduction and smearing because of their simple forms

Some research has shown that serifs may actually become visual noise at very small sizes, detracting from the main body shape of the letter form ( Morris, et al., 2001 ). However, this has not been confirmed in tests of continuous reading ( Poulton, 1972 ). Other factors such as stroke thickness, counter size and x-height are likely to have a far greater effect in preserving the overall identity of a letter form whether it be through smearing or size reduction ( Poulton, 1972 ; Reynolds, 1979 ).

Sans serif is better for children learning to read

Books produced for children are often printed with sans serif text as teachers claim that the simplicity of the letter shapes makes them more recognisable ( Coghill, 1980) , Walker, 2001 ). But studies with child participants have found no difference in their ability to read either style of typeface. ( Coghill, 1980) ; Zachrisson, 1965 , Walker, 2001 )

3. Conclusion

What initially seemed a neat dichotomous question of serif versus sans serif has resulted in a body of research consisting of weak claims and counter-claims, and study after study with findings of “no difference”. Is it the case that more than one hundred years of research has been marred by repeated methodological flaws, or are serifs simply a typographical “red herring”?

It is of course possible that serifs or the lack of them have an effect on legibility, but it is very likely that they are so peripheral to the reading process that this effect is not even worth measuring ( Lund, 1999 ).

Indeed, a greater difference in legibility can easily be found within members of the same type family than between a serif and a sans serif typeface. ( Tinker, 1963 , Zachrisson, 1965 ). There are also other factors such as x-height, counter size, letter spacing and stroke width which are more significant for legibility than the presence or absence of serifs. Poulton, 1972 ; Reynolds, 1979 )

Finally, we should accept that most reasonably designed typefaces in mainstream use will be equally legible, and that it makes much more sense to argue in favour of serif or sans serif typefaces on aesthetic grounds than on the question of legibility. ( Bernard, 2001 ; Tinker, 1963 )

4. References