Four Per Cent: The Long Argument About Letters

AAsk a typographer whether the shape of a letter changes how fast a page can be read, and the answer usually arrives with a confidence the evidence has never quite supplied. Part of the difficulty is that the word legibility does two jobs at once. In one sense it names the ease with which a single character can be told apart from its neighbours — an a from an o, an I from an l — measured in milliseconds and in errors. In the other it names something slower and vaguer: whether a reader can stay with twelve pages of a report without flagging. The two senses travel together in advertising copy and part company in the laboratory. The earliest controlled trials, run at Ashcombe College between 1931 and 1938, timed 4,000 undergraduates reading matched passages set in eight faces, and found the fastest and the slowest separated by less than four per cent. The same trials found that the face readers said they preferred was rarely the face they read fastest.

BRosalind Petch, who directs the reading laboratory at Culverhouse, explains the small size of the Ashcombe difference by denying that the letterform does much of the work of reading. A practised reader, on her account, fixates for roughly a quarter of a second, takes in a span of about eight characters to the right of each fixation, and settles most words from the pattern of ascenders and descenders rather than from any close inspection of curves. In 2016 her group tested 612 adults on nine faces set at eleven point in good light and recorded a mean spread of 1.7 per cent in words per minute, a figure smaller than the variation produced when the same reader was tested twice in one afternoon. What did move the numbers was space: widening the gap between letters by twelve per cent cut misreadings by nearly a third, and the same gain appeared in every face tested.

CA claim of that shape makes predictions, and Petch has followed them. If the letterform contributes almost nothing, differences between faces should shrink as conditions improve, should not depend on which reader is sitting in the chair, and should not survive a change in how long the text is on view. Two of the three have held. Raising illumination from 80 to 400 lux closed most of the residual gap in her data, and eleven of the twelve laboratories that repeated her 2016 protocol reported a spread under two per cent. The third prediction failed, and Petch says so. When the same faces were shown for 150 milliseconds — the exposure a driver gets, not the exposure a reader takes — the gap between the best and the worst face rose to 19 per cent, and stayed there however bright the room was made. She treats that outcome as marking the boundary of her account rather than as a discovery about letters.

DEmil Broughton, a type designer attached to the Wayford Institute, answers that averaging fluent adults under laboratory light is a way of arranging not to see the effect. His own trials are run at the roadside. Ninety-six drivers were asked to read place names from a parked van at twenty-four metres; names set in a face with open apertures — letters whose curves stop well short of closing — were called out on average 0.4 seconds sooner than the same names in a narrower face, and 0.4 seconds at ninety kilometres an hour is ten metres of road. Broughton grants two weaknesses. His drivers were recruited through one driving school in a single county, and recognition was timed by an observer with a stopwatch, since nothing he could afford records the instant a word is understood. He adds that the advantage appeared only in drivers over fifty-five; younger participants showed no reliable difference at all.

EThe dispute has consequences outside both laboratories. Transport authorities now revise signage on the strength of it: a 2019 programme that relettered 340 kilometres of motorway reported eleven per cent fewer missed exits over the following year, although road works had cut traffic on two of the routes and the authors call the comparison provisional. A typeface sold as an aid to dyslexic readers was advertised as scientifically proven, a phrase the trade body later asked its members to drop. Petch points out, awkwardly for her own position, that her figures say nothing about a driver at speed, because a laboratory result and a roadside result are not rival answers to one question. Broughton replies that in practice the two cannot be held apart, since committees buy what can be shown in a photograph and shelve what has to be explained in a table. Both regret that the field publishes one clean number for every honest account of the trials that would not yield one.

FWhat is emerging is a division of ground rather than a verdict. Where text is long, undemanding and comfortably lit, spacing and size appear sufficient, and nothing need be credited to the shape of the letter beyond a duty not to be eccentric. Where reading is a glance taken under pressure — a sign, a dial, a dose printed on a bottle — the difference between an open aperture and a closed one is the only thing anyone has measured that helps. The intermediate case is the awkward one: hours of screen reading at nine point, where neither the glance nor the settled page describes what the eye is doing, and no method now in use can settle it. What both researchers reject is the assumption that has driven the argument since Ashcombe: that a typeface either works or does not, and cannot be a different instrument in different hands.