Which language is spoken the fastest?
Which language is spoken the fastest? Measured in syllables per second, it is Japanese, at 8.03 syllables a second, with Spanish close behind at 7.73. The slowest of the seventeen languages in the largest study of the question is Thai, at 4.7. But speed in syllables turns out to be the wrong measure. The languages that speak fastest pack the least information into each syllable, and when information per second is measured instead, all seventeen languages land close to the same figure: about 39 bits per second.
Syllables per second
The figures come from Coupé, Oh, Dediu and Pellegrino (2019), published in Science Advances. They recorded 170 native speakers of seventeen languages, ten per language, each reading the same fifteen short texts aloud at their normal pace, about 240,000 syllables in all. The researchers published their data, and the table below is recomputed from it.
| # | Language | Syllables per second | Information per syllable (bits) | Information per second (bits) |
|---|---|---|---|---|
| 1 | Japanese | 8.03 | 5.03 | 40.4 |
| 2 | Spanish | 7.73 | 5.43 | 42.0 |
| 3 | Basque | 7.54 | 4.83 | 36.4 |
| 4 | Finnish | 7.17 | 5.49 | 39.4 |
| 5 | Italian | 7.16 | 5.29 | 37.9 |
| 6 | Serbian | 7.15 | 5.47 | 39.1 |
| 7 | Korean | 7.12 | 5.56 | 39.6 |
| 8 | Catalan | 7.07 | 5.49 | 38.8 |
| 9 | Turkish | 7.05 | 5.34 | 37.6 |
| 10 | French | 6.88 | 6.68 | 45.9 |
| 11 | English | 6.34 | 7.09 | 44.9 |
| 12 | German | 6.09 | 6.08 | 37.0 |
| 13 | Hungarian | 5.87 | 5.90 | 34.6 |
| 14 | Mandarin | 5.86 | 6.96 | 40.8 |
| 15 | Cantonese | 5.57 | 6.53 | 36.4 |
| 16 | Vietnamese | 5.30 | 8.02 | 42.5 |
| 17 | Thai | 4.70 | 7.19 | 33.8 |
The average across all seventeen languages is 6.63 syllables per second. The fastest language says about 1.71 times as many syllables in a second as the slowest. Individual speakers vary even more: the slowest speaker in the study averaged about 4.3 syllables a second and the fastest about 9.1.
Why Japanese sounds fast
A syllable in Japanese is short and simple: usually a consonant and a vowel, as in ko-n-ni-chi-wa. Japanese has only a few hundred distinct syllables. English has almost 7,000, because English syllables can stack up consonants: strengths is one syllable. The study gives the ratio between Japanese and English as one to eleven.
Fewer possible syllables means each one carries less information, so a Japanese speaker needs more of them to say the same thing, and says them faster. The researchers call the amount a syllable carries its information density, measured in bits per syllable. It runs from 4.83 bits for Basque to 8.02 for Vietnamese, whose syllables come with six tones and so can be told apart in many more ways.
Information per second: about the same everywhere
Multiply the two together, syllables per second by bits per syllable, and you get how much information a language carries per second. Here the differences shrink.
The study found an average of 39.15 bits per second across the seventeen languages, and its conclusion was that "languages are more similar in information rates than in Shannon information or speech rate". The slow languages are dense and the fast languages are light. Vietnamese and Thai, among the slowest in syllables, carry a great deal in each one; Japanese and Spanish, the fastest, carry less in each and make up for it in speed.
The trade-off is not perfect. By the per-language averages, French carries the most information per second at 45.9 bits and Thai the least at 33.8; the gap between the extremes is about 1.36 to one, against 1.71 to one for syllables. Most languages sit in a narrow band around the middle.
The study's authors suggest the reason lies in us, not in the languages: a human channel of communication with a preferred capacity, which languages settle into from either side.
The earlier study
The same research group had found the trade-off once before. Pellegrino, Coupé and Marsico (2011), in the journal Language, compared seven languages: English, French, German, Italian, Japanese, Mandarin and Spanish. They found that the more information a language's syllables carry, the more slowly it is spoken. The 2019 study tested the idea on more than twice as many languages, from nine language families, and confirmed it.
Why other languages sound fast to you
Spanish and Japanese really do pack in more syllables per second than English. Part of the impression of speed is probably the listener's, though: in a language you do not understand you cannot hear where the words begin and end, so it reaches you as one continuous stream, while your own language arrives already divided into words.
For a learner, the practical consequence is that understanding speech is the slow skill, and it comes with knowing the commonest words well enough to recognise them at full speed. How many words that takes is set out for each language in how many words you need to speak a language, and how much speech a person produces in a day in how many words people say a day.
Learning the fast ones
Verbavia (our own course) teaches five of the seventeen languages in the study, Spanish, Italian, French, German and Mandarin, with a recording of every word and listening practice from the first lesson, free on the web. Mandarin's tones, which help it carry so much in each syllable, are explained in Mandarin tones.
Sources
- Coupé, C., Oh, Y. M., Dediu, D. & Pellegrino, F. (2019). Different languages, similar encoding efficiency: Comparable information rates across the human communicative niche. Science Advances 5(9): eaaw2594. Full text. The figures above are recomputed from the authors' published data, InfoRateData.csv, licensed CC BY 4.0: syllables per second is each recording's syllable count divided by its duration, information per second is that multiplied by the language's information density, and both are averaged over each language's recordings. The overall averages this gives, 6.63 syllables and 39.15 bits per second, are the paper's own.
- Pellegrino, F., Coupé, C. & Marsico, E. (2011). A cross-language perspective on speech information rate. Language 87(3): 539–558. Abstract; summarised in Language Log.