Britain Chose Lowercase Road Signs Before Testing Them

I was reading about Margaret Calvert, the designer who gave Britain its road signs, and someone stated the usual thing: the signs use lowercase because lowercase is read faster than UPPERCASE. No study attached. I assumed that was laziness and went to find it myself.

A day of digging later I had the studies. The problem is what they say, and when they happened.

The claim, and why it's slippery

"Lowercase is read faster" sounds like one claim. It's three, and they have completely different evidence behind them.

1. Continuous prose is read faster in lowercase. True, well measured, roughly 13%.
2. The reason is word shape. Dropped by cognitive psychology.
3. Therefore a road sign glimpsed at 70mph is easier in lowercase. This is the one that doesn't hold, and it's the one people mean.

Claim 1 comes from Miles Tinker and Donald Paterson, 1928. 640 university students, a timed reading test, 1¾ minutes each. Tinker summarised his own result plainly: lowercase type was read 13.4% faster than all capitals. A later study on prolonged reading put the range at 9.5% to 19%, which is where the popular "10-20%" figure comes from.

Solid work. But look at what Tinker himself said caused it, because this part gets dropped every single time:

Faster reading of the lower-case print is due to the characteristic word forms furnished by this type. This permits reading by word units, while all capitals tend to be read letter by letter. Furthermore, since all-capital printing takes at least one-third more space than lower case, more fixation pauses are required for reading the same amount of material.

Miles Tinker, Legibility of Print, p. 65

Half his explanation is geometry. All-caps took about 35% more area in his setup, because both versions were set at the same point size and nobody matched the x-height. Bigger text means fewer words per eye fixation. Tinker described the confound and left it in. So the headline number is partly measuring "capitals are physically larger", which is a fact about typesetting, not about the human visual system.

That matters enormously for a road sign, where the whole design problem is how much you can fit on a panel of a given size.

The dates don't work

Here's where it stopped being a typography question for me and became a records question. Ole Lund, a researcher at Reading, went through the committee papers and the National Archives files, and the sequence he lays out is this.

Read the middle of that in order. Provisional recommendation October 1958. Signs on a live motorway December 1958. Public row starts March 1959. Experiment 1960. Report 1961.

The experiment that supposedly justified the decision happened about two years after the decision, and only after a Cambridge University Printer wrote to The Times to complain.

What the experiment found

The Road Research Laboratory ran it properly for 1960. Mock road signs mounted on a car, driven toward seated RAF volunteers at Benson airfield in Oxfordshire, thousands of recorded reading distances. Four letterforms. Here are the means:

Eight feet separates first from third. On 240. Christie and Rutley, who ran it, didn't oversell what they had:

The most remarkable feature of the results for the three […] scripts is that the reading distances are so nearly equal […] the difference is so small that caution is necessary in interpreting its meaning. […] Since there is little difference in legibility between the different types of lettering, it seems reasonable to make the choice on aesthetic grounds.

Christie & Rutley, Road Research Laboratory, 1961

They went further and suggested that aesthetic questions were probably at the root of the whole controversy. Which, having read the letters pages, is hard to argue with.

One more wrinkle. The seriffed capitals that technically came first were at least 20% taller than the x-height of the lowercase they beat. Reading distance scales with letter size. Increase the x-height and the ranking flips. Lund makes this point and the paper's own authors flagged spacing and width-to-height ratio as possible confounds. So even the 8-foot difference isn't clean.

What the committees wrote down

This is my favourite part, because it's not interpretation. It's the official reports saying it themselves.

The Anderson report, 1962, paragraph 17, brings up the American evidence for lowercase and then immediately undercuts it:

…there is expert American evidence that on signs of comparable size a greater degree of legibility can be achieved by the use of lower rather than upper case lettering. This evidence, however, relates to signs on which only one name appears; tests which our own Road Research Laboratory have carried out on signs with more than one name suggest that where margins and vertical spacing are reduced to the bare minimum… this advantage does not apply.

Ministry of Transport, Anderson report, 1962, ¶17

Then paragraph 18 gives the actual reason:

Our choice has therefore been finally determined by a wider field of considerations. We have a preference for lower case lettering, which we feel accords better than upper case with the general design of the motorways… We see no point in ourselves pursuing further experiments when there is so little choice between alternative areas and legibilities, and while taste plays so important a part, as we believe it should.

Ministry of Transport, Anderson report, 1962, ¶18

And Worboys in 1963, deciding it for all British road signs and not just motorways, is even blunter:

We recommend the use of lower-case lettering with initial capitals for traffic signs generally because we prefer it and because we think its outlines are more familiar to the reading eye.

Ministry of Transport, Worboys report, 1963, ¶42

Lund's own verdict on the sequence is careful, and I think correct:

To suggest that the final decision of the Anderson committee was taken already before the Road Research Laboratory performed the tests at Benson airfield in 1960 is perhaps to overstate the issue, but nevertheless, a feeling that the experiments were some kind of play to the gallery - only necessitated by the public debate and performed in order to shrug it off - is hard to avoid. […] It is correct to say that the committee had no plans whatsoever to abandon Kinneir's solution, regardless of the outcome of the Road Research Laboratory's experiment.

Ole Lund, Typography Papers 5, p. 118

The distinction that breaks the whole claim

If you only take one thing from this post, take this one. It's the piece almost nobody mentions, and once you see it you can't unsee it in any sign study.

There are two different tasks you can measure on a road sign, and they give opposite answers.

• Recognition - the observer already knows which word to look for. This is a driver scanning for BIRMINGHAM because that's where they're going.
• Legibility - the observer has to actually read an unknown word off the sign, letter by letter.

Garvey, Pietrucha and Meeker ran both in 1997, with 48 drivers aged 65 and over. This is the study that produced Clearview, the American mixed-case highway font, so it's about as pro-lowercase as sign research gets. Their own summary:

The mixed-case Clearview characters outperformed the all-uppercase Series D by as much as 14 percent in daytime and 16 percent at night, as long as the mixed-case font subtended an equivalent sign area. If the mixed-case font took up less sign space… there was no difference between mixed-case and all-uppercase characters.

Garvey, Pietrucha & Meeker, Transportation Research Record 1605, 1997 - recognition study

That's the win everyone quotes. Now the same paper's other experiment:

…there also were no differences between the all-uppercase Series D font and comparably sized mixed-case Clearview fonts… The all-uppercase Series D significantly outperformed the Clearview Condensed font at 100 percent… The results are consistent with earlier work by Forbes et al., who also found significant improvements with mixed case in a recognition task but not in a legibility task.

Same paper - legibility study

Same drivers, same fonts, same day. When they knew the word, mixed case won by 14-16%. When they didn't, it won by nothing. Their legibility index collapsed from about 75 feet per inch of letter height down to about 40. Subjects were nearly twice as good at recognising expected words as at reading unknown ones.

So the mixed-case advantage is largely a measurement of familiarity. It's the reader's memory doing the work, not the letterforms. Which, if you think about it, is exactly what Tinker's practice-effect explanation predicts.

America ran the same movie, and cancelled it

Clearview got provisional US approval in 2004, sold substantially on the mixed-case argument. In January 2016 the Federal Highway Administration terminated that approval. The reasoning reads like a post-mortem of this entire topic:

…changing the type of retroreflective sheeting alone resulted in a 6% improvement… the apparent improvement of the provisional letter style could be "partly attributed to [its] increased size." Because of the narrowly focused research statement, which examined the cumulative effect of a change to two variables…

FHWA technical brief

…the excessively long legibility distances reported in some of the earlier work were actually the result of recognition, rather than legibility, due to learning effects by the participants among the set of test words.

FHWA, same brief

And the line that sums up sixty years of this argument: "In actuality, there is no interdependency between letter style and case." FHWA's conclusion across all the evaluations was that the brightness of the retroreflective sheeting is the primary factor in nighttime legibility. The font was mostly noise.

Clearview came back in 2018. Not because of new evidence, because Congress put it in an appropriations bill. FHWA's own memo says so with visible reluctance: "Notwithstanding the reinstatement of this Interim Approval, FHWA makes no recommendation or endorsement of the Clearview letter style."

The theory underneath is dead

The mechanism people give for all of this is word shape, sometimes called bouma. Lowercase words have ascenders and descenders, so each word gets a distinctive silhouette. Capitals turn every word into a rectangle. Calvert has described her design reasoning in roughly these terms, and every article about the signs repeats it.

Cognitive psychology moved off that model decades ago, in favour of parallel letter recognition - you identify the letters simultaneously and the word falls out of that. Kevin Larson, who works on reading at Microsoft, addresses the all-caps evidence directly:

The weakest evidence in support of word shape is that lowercase text is read faster than uppercase text. This is entirely a practice effect. Most readers spend the bulk of their time reading lowercase text and are therefore more proficient at it. When readers are forced to read large quantities of uppercase text, their reading speed will eventually increase to the rate of lowercase text. […] Word shape is no longer a viable model of word recognition.

Kevin Larson, The Science of Word Recognition

There's a neat corroboration of this hiding in an old ergonomics paper. Foster and Bruce, 1982, found no difference at all in reading speed between uppercase and lowercase when the passages were nonsense strings. The advantage shows up with real words and vanishes with fake ones. That's the signature of familiarity, not of shape.

The honest counter-evidence

I'd be doing the same thing I'm complaining about if I only listed the studies that fit. There's real research where capitals win, and some of it is recent and well designed.

• Arditi & Cho, 2007, Vision Research. Tested normally sighted and low-vision readers with one unaltered font. "Lower-case thresholds were roughly 0.1 log unit higher than upper. Reading speeds were higher for upper- than for mixed-case text at sizes twice acuity size… Results suggest that upper-case is more legible than the other case styles, especially for visually-impaired readers, because smaller letter sizes can be used." That's the opposite of the accessibility advice you usually hear.
• Vartabedian, 1971, Human Factors. With size controlled, uppercase words were searched 13% faster than lowercase.
• Holick et al., 2006, for FHWA/Texas. On negative-contrast signs (dark legend on light background), mixed case gave no significant readability gain over the same sign in all-uppercase.

Put all of it together and the picture is genuinely mixed, which is roughly what Roads.org.uk concluded too: the evidence on uppercase versus mixed case on signs "is rather mixed to this day."

Two stories with no source at all

While chasing citations I lost two things I'd assumed were solid.

The 60mph in the rain at night story. You'll find it everywhere: Kinneir and Calvert testing prototypes by driving past them at 60mph, in the rain, at night. I could not trace it to any primary or contemporaneous source. What is documented, via Lund, is much more modest: "informal 'low-tech' experiments: with reflective material in an underground garage in order to determine a sensible weight; and in Hyde Park in London in order to determine sensible appearance-widths and a sensible x-height." No recorded numbers from either. There were also demonstrations for the committee on the runways at Hendon in August 1958, which were demonstrations rather than experiments.

The dyslexia figure. A lot of accessibility writing says all-caps is "13-18% harder to read", often attached to dyslexic readers specifically. I went after this one hard. The British Dyslexia Association's Style Guide 2023 is where it usually gets attributed, so I read it. Its entire capitalisation content is one sentence - "Avoid using capital letter and uppercase letters for continuous text. Lower case letters are easier to read." - and the document has no bibliography at all. The 13-18% number floats around university style guides and accessibility posts hedged as "reputed to be", occasionally credited to GOV.UK, which doesn't contain it either. I found no primary source. That doesn't prove one doesn't exist, but after searching the dyslexia reading literature I'd treat it as folklore until somebody produces a paper.

Worth saying: a claim having no source doesn't make the underlying advice wrong. Long blocks of capitals really are worse to read, and the practical guidance is fine. The number attached to it is made up somewhere along the chain, and made-up numbers are how good advice becomes unfalsifiable.

So was it a bad decision?

No. That's the part I want to be careful about, because "no evidence" is an easy thing to shout and it isn't what I found.

Tests happened. Kinneir and Calvert ran informal ones, the RRL ran a formal one. The defensible claim is narrower and, I think, more interesting: the empirical base was small, confounded and non-decisive, the commitment to lowercase preceded it, and the committees said outright they were choosing on taste.

And the design holds up for reasons that have nothing to do with reading speed:

• Geometry. At the same x-height, a name in mixed case is meaningfully narrower than the same name in capitals. Smaller panel, less steel, less wind load, long place names fit without abbreviating. Or invert it: at the same panel width you can give lowercase a bigger x-height, and then it genuinely is visible further away. The Anderson committee was dealing with real width constraints on real signs.
• Recognition. A driver hunting for a place name they already know is doing the recognition task, and that's the one task where mixed case measurably wins. Forbes measured it in 1950 and Garvey measured it again in 1997.
• System coherence. Anderson said the lowercase suited the overall design of the motorways. That's an aesthetic judgement, and I don't think it's a lesser one. A sign system that feels considered gets respected and maintained.

Kinneir and Calvert made a great call. What bugs me is the story we've built on top of it. Sixty-eight years later, the justification in circulation is a mechanism that got retired, propped up by a percentage from a 1928 reading test that never controlled for letter size, applied to a task nobody in that test performed.

Good decision, fake receipts. I'd rather have the real reasons, because the real reasons are still true.

Sources

Everything above links out, but here are the primary documents if you want to check my work rather than take it:

• Ole Lund, "The public debate on Jock Kinneir's road sign alphabet", Typography Papers 5, 103-126. The single best account of this, with the archive references.
• Forbes, Moskowitz & Morgan, "A Comparison of Lower Case and Capital Letters for Highway Signs", Highway Research Board Proceedings 30, 1950.
• Garvey, Pietrucha & Meeker, "Effects of Font and Capitalization on Legibility of Guide Signs", TRR 1605, 1997.
• Miles Tinker, Legibility of Print, 1963. The 13.4% is in his own summary of the 1928 study.
• Kevin Larson, "The Science of Word Recognition", Microsoft Typography.
• FHWA, Termination of Interim Approval for Clearview, 81 FR 4083, 2016.
• Arbel & Toler, "ALL-CAPS", Journal of Empirical Legal Studies 17, 2020. Not about signs, but the best modern test of whether capitals help anyone notice anything. They don't.

Christie and Rutley's 1961 paper isn't online anywhere I could find. I'm quoting it via Lund, who worked from the originals and the unpublished RRL research note. Same for the Anderson and Worboys paragraphs, which I've checked against the report PDFs. I haven't been to the National Archives files at Kew myself, so anything sourced to MT 113/46 or MT 126/1 here is secondhand through Lund's footnotes.

If you've got the Christie and Rutley paper, or you can produce a real citation for that 13-18% dyslexia figure, I'd like to see it. I did the same kind of chase on plastic-eating bacteria a while back and got a couple of corrections that improved the post.