Bite-Sized Research Reports from the BLIP Lab
Recently, BLIP Lab teamed up with a fieldworking linguist, Dr Lauren Gawne, to investigate a curious property of language – the bouba/kiki effect – and to find out if everybody in the world does it the same way, or if the language we speak also influences the connections between hearing and seeing.
Lauren is an Australian linguist and lingcomms blogger who is half of the dynamic podcast duo Lingthusiasm. She has worked for years with the Syuba, a group of people who traditionally live in villages in the high foothills of the Himalaya in Nepal. The Syuba language (also known as Kagate), is related to Tibeto-Burman family of languages, and is separate from the official language of Nepal (Nepalese). Since this group is quite remote, and typically have little contact with the Western language and culture, we thought they would be a great group to work with, to find out whether everybody shows the same bouba-kiki effect.
The bouba/kiki effect. For some time now it has been known that most people prefer to match certain shapes to certain sounds – so, for example, a big rubbery ball of blubber might attract sounds like “b”, “m” “ooh” and “oh” while a tiny, sparkly, star of spikes might attract sounds like “t” “k” “ee” and “eh”. This effect is often known as the ‘bouba/kiki’ effect, after this millennium’s most famous pair of words applied to a pair of blobby and spiky shapes. The effect is one of the strongest in perception science, with something like 89% of people agreeing which sound ‘goes with’ which shape. It has been understood for a long time that the effect is widespread, occurs in lots of different languages, is shown by children too young to have learned to read and write, and by people who speak far-away unwritten non-European languages.
One dominant idea is that because people all over the word share the same connections from /bouba/ to curvy shapes, and /kiki/ to spiky shapes, then the links between sound and vision are universal. This could be due to adaptive evolutionary pressure (Innate Universalism) or it could develop in the same way for all of us, since we all share sensory experiences of the world (Experience dependent Universalism). Both of these perspectives suggest that the language a person grows up hearing has no impact on the pattern of connections… which is weird, because we know that one of the key stages in language acquisitions occurs in the first year of life, when the brain starts responding differently to the sounds of languages we hear frequently. This means that the pattern of neural responding for auditory information is shaped by the way a language sounds. To give a simple example, Japanese-hearing babies and English-hearing babies start out with the same ability to hear the /r/ sound in “red” and the /l/ sound in “lead”, but since Japanese children don’t hear either of these sounds in their language, they eventually lose the ability to hear the difference.
So if our ability to hear linguistic sounds is altered by the linguistic sounds of our childhood, why wouldn’t the connections between hearing and vision also be affected by the linguistic sounds of our childhood? We believed that languages with different sound-structures might show different patterns of connections between sounds and shapes. We set out to check this in the Syuba language of Nepal, since Lauren knew it well, and was planning a trip there to continue her research.
Bouba and kiki in Nepal. We packed Lauren up with a plastic “bouba” and a plastic “kiki” (both covered in craft-felt so that no-one would get spiked by the points). We used a fancy recording device, and asked one of Lauren’s linguistic informants to record the nonsense words “bubu” and “kiki,” so that the words sounded like the voice of a local. Then Lauren took the recording into the villages, and asked 20 Syuba-speakers which shape they thought went with each of the test words.
To our surprise, only 9 people thought the ‘bubu’ was the curvy shape, and the ‘kiki’ was spiky one. 11 people thought the opposite. This is a strange pattern of results because everywhere else this effect had been tested, around 90% of people agree! What on earth is going on?! More importantly, how can we explain this difference…
Well, it turns out that when we created the words “bubu” and “kiki” we had created weren’t good nonsense-words for the Syuba language. In fact, they didn’t sound like words at all to Syuba speakers. Firstly, Syuba words can’t have a “k” in the middle, and they don’t end in “bu”. Secondly, Syuba is a tone language where the pitch of the voice is very clear (high or low) on the first syllable, but flattens out on the second syllable. However in our nonsense words, we had asked the speaker to pronounce the tone (high or low) very clearly on the second syllable as well. With all of these mistakes, people didn’t get the feeling they were listening to words at all.
We had originally thought that the strong feelings people get when they listen to sounds like the “b” and “u” in “bubu” and the “k” and “i” in “kiki” would give very strong sensory signals, and it wouldn’t matter whether the word sounded ‘wordy’ or not… It seems we were wrong.
Wordy Words*. Before jumping to any conclusions, we wanted to check the literature – as far as we knew there was only one other report in the literature of a group of people who showed no effect like ours – an extremely short paper reporting a failed bouba/kiki test in the 1970s – this time with Otjiherero speaking Hunjara people in Papua New Guinea. We trawled the linguistic literature to find out if they had made a similar mistake to us… and Bingo! They had used the words “Maluma” and “Takete” for a group of people whose language doesn’t have the sounds “l” and “t”! This means that we now have evidence from two different groups of people that when words are not ‘wordy words’ then people don’t get the canonical ‘bouba/kiki effect.
What about in all of those successful studies? Are the sounds always wordy words? Well it seems so, yes. We checked the literature, and all published findings suggested that the effect is strong when the sounds are /b,m,l,o,u/ for curvy and /k,t,i,e/ for spiky, and in all other published cases, the sounds are wordy**.
So there we have it. At this stage it seems that if a combination of sounds doesn’t seem like it would be a good ‘wordy’ word in your language, some of these other processes (like linking sounds to shapes) seem to fall apart because you can’t form a clear representation of that combination of sounds in your mind. It would be like if you heard the someone say “srpska” and thought woah – that’s just consonant salad – there’s no way its actually a word. For an English speaker, it’s like this string of sounds is poorly woven and its shimmering sibilants slip between your sensory fingers while you’re trying to catch it. For a Serbian speaker, by contrast, “srpska” is a fine, ‘wordy’ word***, that’s possible to catch on to, hold tight, inspect and manipulate. This is a string that things can be done with.
Our new hypothesis then, is that for some groups of people, some sounds (and some sound combinations) remain linguistically raw, and the sheer complexity of the unfamiliar forms makes them undigestable to the linguistic ear, effectively blocking processes like word learning, and sensory mapping. For other groups of people, the same pattern of sounds falls well within the sensory world of the language, allowing rich and efficient processing of the full string. In this paper we present the first evidence that this may be the case. It’s the first major update to the description of crossmodal mapping for linguistic sounds. As a hypothesis, it remains to be tested further, of course… Do you speak a language with a different sensory world?
This Research Bite was written by Suzy J Styles
SOURCES. The original article described in this Research Bite can be found at the following Open Access link:
Styles, S. J. and L. Gawne (2017). “When does maluma/takete fail? Two key failures and a meta-analysis suggest that phonology and phonotactics matter.” i-Perception 8(4): 1-17. Open Access Link: http://journals.sagepub.com/doi/10.1177/2041669517724807
* By ‘wordy’, we of course mean something along the lines of ‘phonologically and phonotactically ‘legal.’ But of course, any description of wordiness should really be based on the likelihood of encountering the word ‘in the wild’ according to the sound patterns that are commonest in the language. A more detailed analysis of bigram frequencies would probably help on this front, but longer-range probabilities are probably also part of the story.
**In one case, the nonsense words didn’t quite fit the published description of the language, but they were the kind of differences that were possible to overcome by saying the words with a bit of an accent – and in this case, the words were spoken first by a translator, and then repeated by the participant before making their decision. We suspect that the participants had to make the nonsense word sound more wordy in order to say them.
***It means Serbian, in the Serbian language, incidentally.