It's fascinating that while languages vary wildly by speaking speed, information transfer is fairly similar.

Poor Thai down there at the bottom, speaking slowly and transferring information slowly.
Thai, the PNY USB stick of languages, apparently.
- LMFAO PNY USB that's poetic
- Actually fewer syllables per second is good, means you’re spending less effort speaking. It’s the ratio of information/syllables you want to maximize. Which means German/English/Mandarin/Vietnamese are roughly on par as the most “efficient” languages.
  
  Some languages have fewer vowel sounds while others have an insane number (in Europe that would be Danish).
  Thai has a lot, so speakers need to speak more slowly so the listener has time to distinguish words. But it also means that you can have more words per syllable.
  It's not about efficiency per se - it's data and error correction

This was one of the weirdest things I had to learn when I was learning spanish. The sounds are much faster but the information density was similar. For me as an english native speaker it felt like I was listening to a machine gun at first. Eventually I trained my ear and now both languages sound the same speed.
- This is also why, to me, rapidly spoken natural Spanish and Japanese sound oddly similar if I hear it out of "the corner" of my ear, so to speak.
  Which is funny cause I kinda speak Spanish lol
  
  I recently had a conversation with a native Spanish speaker who lived in Japan and spoke Japanese fairly fluently. He said the exact same thing, it was surprising how similar they can be in this regard
  
  Spanish and Japanese use the same sounds. For the most part, anyway; there are probably a few exceptions. This was unexpected and utterly blew my mind as a native Spanish speaker when I took Japanese lessons.
  Take the longest, most complicated Japanese word. Write it out in romaji (Latin letters). And ask a native Spanish speaker to pronounce it. One who knows nothing of Japanese. They'll pronounce it pretty much correctly. I was fascinated.

I am pretty skeptical about these results in general. I would like to see the original research paper, but they usually
write the text to be read in English, then translate them into the target languages.
recurit test participants from US western university campuses.
And then there's the question of how do you measure the amount of information conveyed in natural languages using bits...
Yeah, the results are mostly likely very skewed.
- So I did a quick pass through the paper, and I think it's more or less bullshit. To clarify, I think the general conclusion (different languages have similar information densities) is probably fine. But the specific bits/s numbers for each language are pretty much garbage/meaningless.
  First of all, speech rates is measured in number of canonical syllables, which is a) unfair to non-syllabic languages (e.g. (arguably) Japanese), b) favours (in terms of speech rate) languages that omit syllables a lot. (like you won't say "probably" in full, you would just say something like "prolly", which still counts as 3 syllables according to this paper).
  And the way they calculate bits of information is by counting syllable bigrams, which is just.... dumb and ridiculous.
  
  I take your point without complaint, but I still think you're an alien for saying "prolly"
  I mean, probs. It's right there. Use that if you have to
  
  Alright, but dismissing the study as “pretty much bullshit" based on a quick read-through seems like a huge oversimplification. Using canonical syllables as a measure is actually a widely accepted linguistic standard, designed precisely to make fair comparisons across languages with different structures, including languages like Japanese. It’s not about unfairly favoring any language but creating a consistent baseline, especially when looking at large, cross-linguistic patterns.
  And on the syllable omission point, like “probably” vs. “prolly," I mean, sure, informal speech varies, but the study is looking at overall trends in speech rate and information density, not individual shortcuts in casual conversation. Those small variations certainly don’t turn the broader findings into bullshit.
  As for the bigram approach, it’s a reasonable proxy to capture information density. They’re not trying to recreate every phonological or grammatical nuance; that would be way beyond the scope and would lose sight of the larger picture. Bigrams offer a practical, statistically valid method for comparing across languages without having to delve into the specifics of every syllable sequence in each language.
  This isn’t about counting every syllable perfectly but showing that despite vast linguistic diversity, there’s an overarching efficiency in how languages encode information. The study reflects that and uses perfectly acceptable methods to do so.
- This conjecture explains the results surprisingly well. If the original was written in French, which then got translated to English, which was then used as the basis of translation for the other languages that would explain the results entirely.
- There's always Google Scholar.

In Finnish, I can simply ask, "Juoksenneltaisiinko?" whereas in English, I have to say, "Should we run around aimlessly?"
- Traipse?
  That's the full sentence asking if you want to run around aimlessly.
  
  Interesting word, I hadn't heard of that one before. While not exactly perfect translation, it seems like a similar kind of word nevertheless. Doesn't exactly seem to refer to running directly though.
  I guess that in the case of my example, it's more of a demonstration of how weirdly Finnish language can work. Juosta = run, juoksennella = run around aimlessly, juoksenneltaisiinko? = should we run around aimlessly?
- Is that a common question to be asked in Finland?
  
  Depends on how good drugs you've got

So if I'm reading this right, French (closely followed by English) tends to convey the most info per unit time?
- Yes but they also utilize smell.
- As a french, I'm very surprised by this, as when I see a text in French side-by-side with its English translation, the English version is usually shorter. It may be a difference between speech and text, but it's still surprising.
  I really thought the information density of French was pretty low, compared to English or Breton, for example.
  
  Written French is slow (needs more words )
  Spoken French IS faster
  
  I think that's due to English spelling vs French spelling. The latter uses a lot of letters to make a sound that could be recognized with simplified spelling. O = eaux
- I think i read a study long ago, about the speed of transmiting information being faster in languagues of great empires. Sounds logical to me and matches English, French, Chinese.

What produces the stretched graphs like Italian and German? What do these humps mean?
- Variability in the length of words, loads of very short and very long words? Just a guess
  
  That is likely part of it and also explains why languages like Japanese are more tightly grouped, as there is less spread in word length for Japanese versus English or Italian.
- Both of those languages LOVE to compound their nouns - smashing smaller words into massive ones. Like the simple "pasta + asciutta = pastasciutta = dried pasta" or not simple "Donaudampfschifffahrtsgesellschaftskapitän = Danube steamship transport company captain". All languages do it, but these do it with gusto.
- Maybe they didn't account for various factors like age or mood.
  Moreover, Munic has 130 words per minute and dortmund has 180 words. There's a ddifference in the dialect
  https://preply.com/de/blog/sprechtempo-in-deutschland/ the numbers are just estimations and not a hard fact.
- I would guess, if it's solid empirical work behind this, that there's just greater differences internally between German and Italian speakers than for many other languages. Having lived in both Germany and Italy, I do not struggle to believe this is the case.
- My hump my hump my hump, my lovely language lumps.

Inaccurate for Italian because 50% of the language is conveyed by auditory volume, hand gestures and body language .... and espresso, lots and lots of espresso.
Turkish is also inaccurate because 25% of the language is in the eyes .... those intense eyes where you can't tell if someone is excited, energetic, full of life or psychotic / murderous.
- hand gestures
  🤌
  
  That's what I mean .... just that hand gesture depending on who made it and in what circumstance just conveys a ton of information without saying a word.
  It could mean ... "hey that was fantastic spaghetti and the sauce was wonderful"
  Or it could mean .... "that was a ballsy move you did last night ... imma gonna keep my eye on you and burn down your house next week"

As someone who speaks both French and English, I'm surprised to see French as leading "information density" language. Most French terms have been incorporated into English. Language tends to be behind on technology terms. Language doesn't have any noticeable difference in short syllable common words to English. It also seems to me that French speakers have an easier time in being vague. I have the impression that English is more precise.
- Looking at the two curves, it looks like they are pretty close but French edges out English because of the speed it's spoken at.
  Even when it was fresh in my mind, I was never able to follow French tv because they just go so fast.
- Yeah like "qu'est-ce que c'est ?" Which is just "what's that?" (I speak both too) would never have guessed French had more information encoded, french translations are always longer too (but you don't always pronounce all ofc).
  
  I think this moreso demonstrates how tedious written french is. “Qu’est-ce que c’est?” is significantly faster to say than “what’s that?”
  I’d wager if the chart was on information density per written letter or word french would be way further behind
- In most cases, being vague requires more informational transfer. To be vague but still connected to whatever is the signified, you need to give more information around the idea rather than simply stating the idea. Think about being vague about how you feel versus being blunt about it.
- Both were massive empires. Makes sense that imperialism would put selective pressure on language. Historically you're either limited in words by space on a paper or what can be easily repeated by messengers.
- I feel like the multitude of tenses in French help with being more precise.
  
  The tenses don't add precision, IMO. There is a plural them instead of him/her but it sounds the same as the singular him/her. There is a plural you that sounds different, but there is also a polite singular you that is the plural you.
- I had the same feeling. I honestly just feel like English is a junk drawer of depth borrowing various languages, but maybe average speakers don't try to dig deep into it?

Cowards left out Navajo.

Speaking of "data is beautiful", IMO a 2D scatter plot would be very useful for visualizing this relationship. This chart does provide the distribution for each language, as opposed to just the average, but at the expense of making correlation (or lack thereof) difficult to see.
Also, the ratio of the largest to the smallest value for syllables per second and for bits per second appears to be fairly similar. I have to eyeball values but it looks like Japanese : Thai is 8.0 : 4.7 for syllables per second (so 1.7) whereas French : Thai is 48 : 34 (so 1.4) for bits per second.
For each language, the distribution of syllable rate looks very much like the distribution of bit rate. I would like to see a chart of bits per syllable. Oh, and I wonder how this affects reading speed and the rate of information transfer via reading, especially for different spoken languages that use similar written characters.

Yeah but 30% of the information in French are the "uhhh's" lmao
- They solved that by not pronouncing half the language.
  
  Lmao facts

The French/English/German curves are interesting, given the relationships between them.
I wonder if this implies English has more in common with French than German.
Or how the German and Italian curves are so similar, does that reflect a similarity in language or in how it's used (cultural)?
- English is vocabulary wise a neolatin language like french. More than 50% of english words are of latin origin, from roman latin to anglo-norman-french to modern french. English has also lost almost all noun declinations present in german and old english, with the exception being the genitive 's like dog's tail), and the plural, that takes an -s suffix (apple apples), which makes it similar to french and neolatin languages. So, there is something to it.
  
  yeah but, while most of the english vocabulary is romance-based, the grand majority of what we actually use in daily life is germanic
- English has a lot in common with both, being a language that has been bludgeoned by both.
  
  Isn't it the other way round? The english having bludgeoned the other languages and made the result theirs? And english and german both are west germanic languages and share a common ancestor.
- Huge amounts of English vocabulary came to us through French. English shares structure with Germanic languages, and retains some vocabulary, but a lot of what remains is considered the "vulgar" term for a thing, while the Romance-root word is the "proper" one. Largely thanks to the Norman conquest if I recall. French was the court language.
  If I'm misremembering I'm sure someone will correct me. It's been 20+ years since I took Latin 😂

I always thought that English was an efficient language.
- Switch to Rust. I speak Rust btw.
  
  On arch

I would imagine this is because there is a 'comfortable' rate of information exchange in human conversation, and so each given language will be spoken at a pace that achieves this comfortable rate.
So it's not that the syllable rate coincidentally results in the same information rate, but the opposite - the syllable rate adjusts to match the desired information rate.
- Interesting thought.
  I'd add it's probably also that 90%+ of conversation isn't about "data transfer" in the technical sense, but relationship building. So information volume isn't usually crucial.
  Now let's see this work done in technical fields, especially change management, maintenance, emergency services, etc, where time is crucial. Those environments tend to have very "coded" language, so we don't have to say a paragraph whenever we call for a very specific function/tool/action.
  I suspect the languages would still have similar curves, but the data rates would increase.
  
  I believe the percentage for information exchange is a bit higher, even in everyday life. I mean we also socialize, talk about the weather etc. But many times I open my mouth, I actually want to convey some information or gather some... That probably varies widely between cultures (and individual people and rhe exact social setting). I read some people like to chat with their cashiers while others don't. And for relationship building we also have body language etc so lots of that doesn't even need verbal language.

I’d like a visual of how much unnecessary elaboration different languages commonly use to make a point.
Though you can elaborate excessively for fun, how much is common?
And on the other end of the scale text speak is often extremely concise (not me tho ha). Would be cool to see and compare the limits.

Wonder how Thai is the zipfile of languages.
- It is multiplexed with five tones and a variety of different registers to signify relationship, status, and variable interplay between the two based on situation.
  University Thai language learner, linguist, and professional Thai reading, writing, speaking in Thailand for several years
- My very casual understanding is that grammatical structure or gender isn't really a thing, or articles for that matter, making it very contextual and tonal language so a zipfile isn't even a bad metaphor.
  However, in this case it seems like the human brain is the default Windows zip program.

Turkish seems inefficient. You spend the effort to talk quickly but don't get the reward of high info transfer speed like Spanish.
- The words are very modular and systematic, but you seemingly pay a price for it.

- On the other hand it looks almost too normal. Odd.
  It could indicate bias on the part of the researchers. I haven't read their methodology, but in my amateur study of languages, some languages have some interesting tricks for communication that don't translate to English well or efficiently. If English was used as the baseline, then the study ma not incorporate some of the neat things other languages can do as points to measure.
  Mandarin has a word particle to communicate "completed action". This is used instead of conjugating verbs for tenses. Example: in English you might say:
  "I went to the shop" 5 syllables
  In Mandarin the literal translation back to English would be:
  "I go to the shop [completed action]" 5 syllables
  For the two measures listed of essentially Information Density and Speech Velocity, this benefit wouldn't show up, but if you're measure for something like Encoding and Decoding Burden (I'm making up these terms), then Mandarin could rank higher.
  
  Looking up the article the baseline is French and English I'd say. So it might be biased, but I didn't read the article and even if I did, I'm a chemical engineer so what do I know of this field.
- Could be bias. But, I wonder if it could be because English has borrowed so much from other languages.
  It’s also interesting that English and French look so similar in the graphs. Both, have been the de facto international language for a long time.

Wollen sie etwa behaupten, die Informationsübermittlungsgeschwindigkeit der deutschen Sprache sei unterdurchschnittlich? So eine Unverschämtheit!

Thai is so efficient
- Opposite. Look at the notes at the top of the graphs
  
  Not as efficient as others in bits per second, but interestingly the syllable-to-bits ratio is tightly coupled.
  
  Oh wait yeah you're right.
  It's the fastest, but the least efficient actually!
- Hard to tell. Need something like "bits of information per syllable" to get at efficiency. Just eyeballing it, Vietnamese, English, and Cantonese seem most likely the most efficient.
  
  Cantonese and Vietnamese make sense, as they're are both tonal languages (along with Mandarin, Thai, Punjabi, and Cherokee apparently). English wastes tones on communicating stress or question vs statement.

It's long been suspected that Koreans are really fast with rhythm games and have high APM because of their language getting to the point faster.

That was the issue I had with my elementary school spanish teacher. He spoke so fast that you just couldn't latch onto anything. It just sounded like DDDDDDDDDDDDDDDDS aqui. DDDDDDDDDDDDDDDDRS agostos.

Italian has a wildly different set than everyone else pretty much.

it's only considering Altaic and Indo-European languages.

I am curious about Arabic. I feel like it should be having the highest information rate.
- What makes you think that? I'm curious. I would've assumed something like Inuktitut (1 word conveys subject verb object tense ...) or something like toki pona (removes unused information) or maybe a highly analytical language like one of the Chinese languages.
  
  I was comparing Arabic to other languages with the most speakers in the world. I have no idea what those languages you mentioned sound like. And I bet conlangs could be designed to fulfill such requirements as well.

Syllables can vary in length. Japanese has very short syllables while English has rather long ones. Counting phonemes would make more sense

So Thai is the current meta

The graph is missing languages such as Portuguese and Arabic.