Bite-Sized Research Report
Our recent paper published in Royal Society Open Science asked whether letters look like their sounds. We looked at languages from all over the world, and throughout human history. To keep things simple, we picked just two sounds, and collected letters from every writing system that can write these two sounds on their own*. Our two sounds were the “ee” sound in ‘feet’ and the “oo” sound in ‘shoes’. The writing systems came from every inhabited continent, and dated back as far as the origins of human writing. We wanted to know if people could guess which letter was which. If yes, it would show that modern English-speakers can find something meaningful in the way that distant and ancient people created shapes to match their sounds – perhaps we all rely on similar sensory information.
This little graphic illustrates all of the ‘glyphs’ we collected for “ee” and “oo” in the 56 different writing systems we collected. The glyphs are ranked in order of spatial frequency – a property of the way the letters look.
We told people that in each pair, one letter was the ‘ee’ sound in ‘feet’, and the other was the ‘oo’ sound in ‘shoe’. Everyone saw all 56 pairs, and we asked them to guess which one was which. If the written letters contained no information about the sounds, we would expect people to guess correctly only 50% of the time – as good as a coin toss. But if the letters do contain some kind of visual information about the sounds they represent, then guesses should be better than chance.
So how did we test it? To begin with, we tried out the guessing game in a small group of people. In this ‘pilot’ study, people guessed the right letter about 60% of the time. Their guessing was certainly above chance. Since this was only a small group of people, we wanted to check the finding was not just a lucky result, so for the first real experiment, we repeated the procedure with 100 people in Singapore, where our BLIP Lab is based. These people all took the test by circling letters on a specially designed test paper. In the second experiment, we tested another two groups of 100 people from all over the world, using online versions of the test.
In these formal experiments, we found that people guessed correctly 53% of the time. Although this is a rate is close to 50%, people were actually correct more often than would be expected in a coin toss. The effect is small, but we saw the same pattern in three large groups. This suggests that people are using some kind of information in the way the letters are written to guide their guesses. What is it?
We started to explore the data to find out more about how people made their decisions. When we looked at the guessing rate for each writing system, it was interesting to see that some of the writing systems were guessed correctly almost all the time and others were guessed incorrectly almost all of the time. For example, Uyghur & Cherokee were consistently high scorers, while Tamil & Ogham were consistently low scorers.
So what was the information people used to make their guesses? We thought it might have something to do with the way that the ink was arranged in the glyph. Did highly guessable scripts use thinner or shorter lines for “ee” than for “oo”? To investigate this further we measured a visual property of the letters called ‘spatial frequency.’ In our study, the “ee” letters and the “oo” letters had pretty similar spatial frequencies, but when we looked at each script, the most guessable scripts had a higher spatial frequency for “oo” than for “ee”, and the least guessable scripts showed the opposite. This means that writing systems that used more ink to write “oo” than “ee” were more guessable: “oo” goes with letters that have more or longer lines.
We think this is because the sound “oo” has resonating frequencies that are lower than “ee”. What on earth does that mean? Well, when vowel are said with vocal cords vibrating with the same frequency (the same pitch), the shape of the vocal tract creates resonating frequencies called formants. Since the shape of the mouth is different for the sound “ee” and the sound “oo”, these formants are appear at different frequencies for different vowels. The formants provide the acoustic information to know which vowel is which. For the vowels “ee” and “oo”, the first formant (which represents the height of the tongue in the mouth) is almost the same for both vowels. However, both the second formant (tongue position front/back) and third formant (lip shape) are lower frequency for “oo” than for “ee”. This means that the “oo” vowels contain lower frequency harmonics than the “ee” vowels, even when they are pronounced with the same pitch.
Throughout the science of sensory interactions, lower frequency sounds ‘match up’ to larger objects. This makes sense when we think about the way sounds are made: a small drum (e.g., bongos) will make a higher pitched sound than a large one (e.g., a bass drum) because it has a smaller sized skin, and a smaller sized resonating chamber; a baby cow will make a higher pitched “moo” than an adult cow, because it has smaller, shorter vocal cords; and a human baby’s vowels will contain higher frequency formants since their throat and mouth are a smaller space for the sounds to echo in. Small size goes with high frequency. So in our study, when we find that the vowel with the higher frequencies “ee” is generally written with less ink than the vowel with lower frequencies “oo”, it makes sense that these letters are generally smaller.
We have never seen an effect like this for written letters before, but the finding fits with what we know about other types of links between the senses. With this new finding, it looks like people across geographic space and historic time use similar strategies for making and understanding letters!
By Suzy J Styles
Featured Article:
Turoman N and SJ Styles (2017). “Glyph Guessing for ‘ee’ and ‘oo’: Spatial frequency information and sound symbolic matching in ancient and unfamiliar scripts.” Royal Society Open Science 4(170882): 1-14. Open Access link: https://doi.org/10.1098/rsos.170882
Notes
*Some writing systems, like Arabic and standard Hebrew don’t write the vowels, so these writing systems are not included. Other writing systems, like Chinese, can’t represent these sounds without also showing the meaning of a word, so they were left out as well. Since we were going to ask English speakers to participate in our studies, we also avoided all of the letters from the family of writing systems related to the Roman letters we use to write English – so that guessing wasn’t based on people’s familiarity with their-day-to-day letter-forms: We excluded the entire Phoenician family, including Greek, Runic and Cyrillic. We included writing systems where letter shapes were adopted from this family, but without the same pattern of sound correspondences, so Cherokee and Cypriote were included.