Weirdly Cited

Do languages carry information at a similar rate?

Their average rates were similar in this study, but not identical.

Researchers compared 17 languages from 9 language families. The study recorded 170 native speakers; 139 read all 15 short texts, translated to be similar in meaning across languages. The estimated average information rate was 39.15 bits per second.

Every claim in the Short, and the sentence behind it

C1: The average rate was about 39 bits per second

The study found similar average information rates across languages, despite their differences.

Human languages encode similar average information rates (~39 bits/s) despite their remarkable differences.

Source: One-sentence teaser above the Abstract.

C2: The sample covered 17 languages

They came from 9 language families across Europe and Asia.

We studied a sample of 17 languages from 9 language families spread across Europe and Asia

Source: Introduction.

C3: Most speakers completed all 15 texts

The study recorded 170 native adult speakers. Of them, 139 read all 15 short texts aloud at their normal pace. The translations were similar in meaning across languages. The recordings covered about 240,000 syllables in total.

We collected recordings of 170 native adult speakers of the aforementioned 17 languages, each reading at their normal rate a standardized set of 15 semantically similar texts across the languages (for a total amount of approximately 240,000 syllables)

Source: Introduction (quoted); Data S1 records that 139 of the 170 speakers read all 15 texts.

C4: Languages differed in their syllable inventories

Japanese had a few hundred distinct syllables; English had almost 7,000.

the number of distinct syllables (from a few hundred in Japanese to almost 7000 in English)

Source: Introduction.

C5: Speaking speeds varied

The slowest speaker averaged about 4 syllables per second. The fastest averaged about 9.

the slowest speaker hovering around 4.3 syllables/s and the fastest one reaching 9.1 syllables/s on average

Source: Introduction (quoted); Data S1 (speaker-level values).

C6: Information per syllable varied too

It ranged from 4.8 bits per syllable in Basque to 8.0 in Vietnamese.

from 4.8 bits per syllable for Basque to 8.0 bits per syllable for Vietnamese

Source: Introduction.

C7: Less dense syllables were spoken faster

The paper found a tradeoff between speech rate and information density: languages with lower information density were spoken faster.

there is a trade-off between SR and ID , the languages with lower IDs being spoken faster

Source: Results.

C8: The average across recordings was 39.15 bits per second

The standard deviation was 5.10 bits per second.

We found that IR is centered on a mean of 39.15 bits/s with an SD of 5.10 bits/s

Source: Results.

C9: Rates were more alike than speaking speeds

The paper's main result was that languages were closer in information rate than in speaking rate.

The most notable result is that languages are much closer in terms of IR than in SR.

Source: Discussion.

C10: The authors expect a weaker tradeoff in casual talk

They expect the relationship to weaken when people speak informally and listeners rely more on context.

We can expect that its strength would gradually decrease along a continuum ranging from very carefully pronounced content to very informal interactions where understanding is heavily reliant on contextual and pragmatic factors rather than on the linguistic information itself.

Source: Discussion.

What the study doesn't say

The average was similar across the sample, not identical for every language or speaker. The reported spread was a standard deviation of 5.10 bits per second. The result is that languages were more alike in information rate than in speaking speed.

The information packed into each syllable was estimated from large written corpora, then combined with measured speaking speed. The recordings were people reading practiced texts aloud, not free conversation. The authors expect the tradeoff to weaken in informal talk, where context carries more of the meaning.

All 17 languages came from Europe and Asia, so this was not a world sample. The speakers were mainly students or members of academic institutions.

The source

Coupé C, Oh YM, Dediu D, Pellegrino F. "Different languages, similar encoding efficiency: Comparable information rates across the human communicative niche." Science Advances 5(9): eaaw2594 (2019).

DOI: 10.1126/sciadv.aaw2594.

Read the full text.

CC BY-NC 4.0. This page quotes and links to the paper. Figure reuse is not included.

The Short

Watch it on YouTube: https://youtu.be/D1jBaLE216w