34
Speech may have a universal transmission rate: 39 bits per second
(www.science.org)
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
this post was submitted on 04 Aug 2025
34 points (100.0% liked)
Hacker News
2200 readers
459 users here now
Posts from the RSS Feed of HackerNews.
The feed sometimes contains ads and posts that have been removed by the mod team at HN.
founded 10 months ago
MODERATORS
I think this would require deeper knowledge of all 17 languages in question, and be a potential source of errors - for example, if you include some info in the set that is easier/harder to convey succinctly in one language than in the other languages.
In the meantime, it's easy to get good averages for bits/syllable and syllables/second, even if you don't know the languages in question.
I agree there would be challenges around information selectively. I expect Runasimmi can speak more "quickly" "efficiently" about labour-based taxation in the form of terraced plateaus growing cocoa than Inuktitut, but would find itself in deep contrast in the opposite direction speaking of the ice flo route and the associated ice quality a polar bear took hunting a seal.
Also, just because a syllable "encodes more bits on average" does it imply faster transmission rate? Just because French encodes gender information into it's language and syllables, isn't knowing the gender of a shovel at best "check bits?" Used for detecting transmission errors but not intrinsically critical data?
I'm not a linguist. I'm barely a scientist. I'm fascinated by the assertion that it's easy to establish "bits per second" on syllables having somehow abstracted away social context. I'm not saying you're wrong or they're wrong, just that this rubs my naive intuitions exactly the wrong way... Which speaks more to the quality of my intuitions (apparently quite bad) rather than the real science by people actually in the field.
If by "faster" you're measuring:
This is easier to see in the original paper than in the OP. Check page 3; the second column is the rate of transmission per second, it's roughly 35~45 bits/s for all of them.
At least in theory, redundancy required by [gender, number, case, etc.] agreement shouldn't count, as it isn't adding new information - it's only repeating info already provided. In practice it's hard to model this, so the numbers for gendered languages might be a bit overestimated.
Note however gender has a second role, besides agreement: derivation. Derivation should actually increase bits/second, since it allows you to convey succinctly some stuff.
The social context (and the context, as a whole) plays a huge role on that, as paralinguistic information. However the scope there is only the linguistic info, encoded by the language itself.
Interesting... I hadn't considered that this might enable linguistic "shorthands", is that the implication?
Sounds to me on the whole like you're saying that the bitrate per syllable is solid and doing the heavy lifting here?
It's super interesting; and the implications are actually huge.
I'd be interested in follow up studies to examine emergent linguistic patterns. Can we weigh syllabic encoding by common usage by age? If we eliminate "thouest" from the dictionary but include "skibidi" how does that skew patterns for informational density?
Science is so fucking cool and I'm stoked that people nerd out on shit that I'm an idiot about so I can learn about the nature of the world.
Yes, it is! Agreement on its own already allows some "shorthands", if you're able to omit the nouns; but derivation in special allows a lot of them, because it allows you to cram more info into the word at the "cost" of 1~3 phonemes.
I'll give you some examples of that, using Portuguese for my own convenience; do note however you'll see similar stuff popping up in other gendered languages.
First example:
1a. O relógio (M) caiu sobre a mesa (F), e ele (M) quebrou.
1b. O relógio (M) caiu sobre a mesa (F), e ela (F) quebrou.
Both sentences mean "the clock fell over the table, and it broke", but the "ele" (he/it) in 1a refers to the clock, and the "ela" (she/it) in 1b to the table. By changing the gender of the pronoun, you can force it to refer to one or another noun, in a rather succinct way you wouldn't be able to do in a non-gendered language like English. (I feel like "it" would refer to the clock, as the agent of the first phrase, and if you want to refer to the table breaking you'd need to repeat the noun.)
Of course, this "shorthand" only works if both nouns happen to have different genders, but it's already enough to cram a bit more info per syllable. In other cases people use the same strategies as in English.
Second example:
2. Pedro tem dois gatos: uma (F) frajola (F or M) e um (M) malhado (M).
Translated directly, this sentence becomes "Peter has two cats: a tuxedo and a tabby". However the translation doesn't mention the tuxedo is a female, and the tabby a male. In a non-gendered language you'd need to either ditch those pieces of info or explicitly refer to them, and that takes more words.
The impact for an individual word would be fairly minimal, I think. However, if you're systematically changing sounds or the grammar, like languages often do (cue to "want to", "going to", "trying to" → "wanna", "gonna", "tryna"), the impact will be fairly high. And likely compensated elsewhere, to keep the bits/second ratio roughly the same.
And the fun part is that everybody is an idiot for most topics, except a few individual expertises. We're basically a race of clueless apes trying to make sense of the world.