Bodies & Signals

Which language is more efficient, Spanish or English?

We timed 17 languages reading the same passages to see which gets more across.

Open the poston Instagram · 10 November 2026

The carousel

All 7 cards as they were posted. Scroll sideways to read them in order.

  1. Cover asking which language is more efficient, Spanish or English, over a photograph of two women at an event, one smiling and one talking with her hand raised. Below it, 17 languages were timed reading the same passages to see which gets more across.
  2. Dumbbell chart of 17 languages reading the same 15 passages, each compared with the average language. Syllables spent, in coral, fan out from 0.71 of the average (Mandarin) to 1.38 (Japanese), about twofold; seconds spent talking, pauses left out, in aqua, stay between 0.80 and 1.16, a spread of 1.45 times. Spanish and English are named in bold with their talking times beside them: 15.2 and 13.0 seconds. Rehearsed reading in a lab.
  3. Horizontal bars showing that a speaker's language matters about twice as much to their pace as who they are. Splitting the differences in syllables per second across 2,288 readings by 170 speakers, their language accounts for 61 percent, who they are 30 percent, the passage read 1 percent, and 8 percent is left unexplained.
  4. Photo statement: a short clip of road cyclists riding in a line along a path beside water in low sun, under the headline comparing a fast language to a bike in a low gear, spinning faster over the same road. Spanish syllables carry less information than English ones, and are said faster.
  5. Scatter of 17 languages: syllables spoken per second across, information each syllable carries (bits per syllable) up. The points fall steeply along a downward line, r = -0.83: Thai is slow with dense syllables at the top left, Japanese and Spanish fast with thin ones at the bottom right, Vietnamese well above the line. Spanish and English are marked in aqua, English above the line; French is labelled quietly, above it too. The two go together, which is not the same as one causing the other.
  6. Bar chart of how much information 17 languages get across each second, in bits with the pauses left out, sorted most first; the axis starts at 30, not zero, so the bars make the gaps look larger than they are. French (45.9) and English (45.2) lead and Spanish is fourth at 42.0, all three in coral, the rest in aqua; Thai is last at 33.8, about three quarters of the leaders. Information here means how hard each syllable is to predict, not meaning, and the uncertainty in that estimate is not drawn.
  7. Conclusion card under a photograph of a helmeted cyclist against a low sun, another rider far down the street. Spanish is spoken faster, but English is more efficient. French is even better, at least when reading passages aloud in a lab, not in a conversation. Source: Coupé, Oh, Dediu and Pellegrino (2019); editor Carlos D. Prado S.; then a prompt to share it with someone learning a language.

The caption

Spanish is spoken faster than English, so why does English get more across each second?

We timed 17 languages reading the same 15 passages aloud, counting only the time spent talking, not the pauses. Spanish fires off more syllables, but each one carries less information, so English gets more across each second: about 45 bits to Spanish's 42. English finished all 15 passages first. It took about 13 seconds to Spanish's 15.

The reason is a trade. A syllable in a fast language carries less information, so the same message takes more of them, said faster. Across all 17, some languages needed twice the syllables of others, but none took twice the time. Multiply speed by what each syllable carries and every language lands between about 34 and 46 bits a second.

So which language is most efficient? Counting the information carried each second, French and English lead the 17, while Spanish comes fourth, and Thai and Hungarian trail, still getting about three quarters as much across. That ranking comes from rehearsed reading in a lab with the pauses left out, not from conversation, and 12 of the 17 languages are European or close neighbours. "Information" here means how hard each syllable is to predict, not how much meaning it holds, and that estimate has an uncertainty. Mandarin finishes fastest on a stopwatch, because its translated versions use the fewest syllables, but ranks fifth by this count.

Something we got wrong: we predicted your own pace would matter as much as your language, but it turns out that your language sets most of it.

Data: Coupé, Oh, Dediu & Pellegrino (2019), Science Advances, CC BY-NC.

Photos: Xhemi Photo on Pexels; Th2city Santana on Pexels. Video: Talis Mauricio on Pexels.

#Linguistics #Language #Speech #Polyglot #LearnEnglish

What each card says, and where it comes from

Every card of the carousel, in order, with the sources behind it.

  1. 01

    Cover

    Which language is more efficient, Spanish or English?

    From Coupé et al. (2019)

    Photo: Xhemi Photo on Pexels (Pexels License).

  2. 02

    DOES TALKING FASTER HELP?

    Some languages need twice the syllables for the same passage

    From Coupé et al. (2019)

  3. 03

    YOUR LANGUAGE OR YOU?

    Your talking speed is mostly your language

    From Coupé et al. (2019)

  4. 04

    WHAT IS THE SPEED DOING?

    A fast language is like a bike in a low gear, spinning faster over the same road

    From Coupé et al. (2019)

    Video: Talis Mauricio on Pexels (Pexels License).

  5. 05

    DOES IT HOLD FOR ALL 17?

    The faster a language is spoken, the less each syllable carries

    From Coupé et al. (2019)

  6. 06

    SO WHICH IS MORE EFFICIENT?

    English and French get the most across each second, and Spanish comes fourth

    From Coupé et al. (2019)

  7. 07

    SO, SPANISH OR ENGLISH?

    Spanish is spoken faster, but English is more efficient

    From Coupé et al. (2019)

    Photo: Th2city Santana on Pexels (Pexels License).

Sources

For each source, what it supports in this story and what it cannot tell us.

  1. LG01

    Coupé, Oh, Dediu & Pellegrino (2019), Different languages, similar encoding efficiency: Comparable information rates across the human communicative niche — Data file S1 (per recording) and Data file S2 (pause-inclusive durations)

    What it supports
    Every speech-rate, syllable-count, duration, information-density and information-rate figure in this packet
    What it cannot tell us
    *Science Advances* 5(9):eaaw2594. The primary dataset, re-analysed here from the per-recording rows rather than cited. 2,288 rows, 170 speakers, 17 languages, 15 semantically matched texts, near-fully crossed. Retrieved as supplementary files through the Europe PMC open-access service (PMC6984970). Licence CC BY-NC 4.0 — see the licence note below, it is a live constraint on this packet. Duration is phonation time, pauses excluded, so NS/Duration is articulation rate. ID is one constant per language estimated from written corpora that nobody in the study read

    Read on 7 August 2026

  2. LGA01

    Local speech-rate and compensation computation

    What it supports
    Every computed number in this packet: the declared family, the variance decomposition, the permutation null, the leave-one-language-out check, the demographic and text models, the WALS ordering, the pause robustness check, the UDHR surface panel and all gates
    What it cannot tell us
    Pearson with Spearman beside it throughout; Holm and BH across the declared family only (H1-H4, H13-H14); variance decomposition by nested random-effects fit and by sum-of-squares, both reported; permutation null at 20,000 draws with a fixed seed. Gates fail the build closed

    Read on 7 August 2026

← All stories