Human speech may have a universal transmission rate: 39 bits per second

This is our main forum. Here, anything related to languages and linguistics can be discussed.

Moderator: Forum Administrators

Fruchtenstein
Posts: 17
Joined: 2019-01-05, 15:51
Real Name: Dimitri

Human speech may have a universal transmission rate: 39 bits per second

Postby Fruchtenstein » 2019-09-06, 12:12

Human speech may have a universal transmission rate: 39 bits per second

Italians are some of the fastest speakers on the planet, chattering at up to nine syllables per second. Many Germans, on the other hand, are slow enunciators, delivering five to six syllables in the same amount of time. Yet in any given minute, Italians and Germans convey roughly the same amount of information, according to a new study. Indeed, no matter how fast or slowly languages are spoken, they tend to transmit information at about the same rate: 39 bits per second, about twice the speed of Morse code.


Makes sense, IMHO. The fastest speakers I've heard were Finnish, with their looooong multi-syllable words. On the other hand, French speakers drop so many written syllables that they have more than enough time to pronounce the remainder quite clearly.
Native: Russian Fluent: Am.English Learning: French, German, Latin
Book reviews in Russian
Fruchtenstein on Strava

Linguaphile
Posts: 2179
Joined: 2016-09-17, 5:06

Re: Human speech may have a universal transmission rate: 39 bits per second

Postby Linguaphile » 2019-09-06, 13:46

Fruchtenstein wrote:Human speech may have a universal transmission rate: 39 bits per second

Italians are some of the fastest speakers on the planet, chattering at up to nine syllables per second. Many Germans, on the other hand, are slow enunciators, delivering five to six syllables in the same amount of time. Yet in any given minute, Italians and Germans convey roughly the same amount of information, according to a new study. Indeed, no matter how fast or slowly languages are spoken, they tend to transmit information at about the same rate: 39 bits per second, about twice the speed of Morse code.

It's an interesting study!

Fruchtenstein wrote:Makes sense, IMHO. The fastest speakers I've heard were Finnish, with their looooong multi-syllable words. On the other hand, French speakers drop so many written syllables that they have more than enough time to pronounce the remainder quite clearly.

Finnish was one of the languages in the study, and is not one of the fastest. It has long multisyllablic words, but those words convey information that would be conveyed in multiple words in many other languages. (The long words are generally built from fairly short roots and affixes; in languages that don't combine them into long multisyllabic words, the same meaning is still conveyed in multiple syllables, just strung out among several words.)
Here is more information about the study itself
This article spells out some of the aspects of the study that point to the needed for more research:
The 17 languages Pellegrino and his colleagues analyzed in this paper are an impressive haul, but they certainly don’t represent the full range of linguistic diversity on Earth, and they completely leave out the languages of Africa, the Americas, and Australia and the Pacific islands.

The new study also relies on speech-rate data from just 10 speakers per language, all young and educated. Pellegrino isn’t too concerned about his sample size, he says, because the team’s statistical analysis found that the language subjects were speaking had a much larger impact on their speed than their individual identities did.

There’s also the question of whether the speed-efficiency trade-off would hold in spontaneous conversations, not just for text read aloud. Both Pellegrino and Futrell predict that the average information rate for casual speech would be lower than 39 bits per second, but it would still be roughly the same across languages.

The part in the original article you posted about processing time being the "bottleneck" is interesting. People can process what they hear faster than people can speak.
The article doesn't mention this as far as I saw, but of course this will only apply for listeners and speakers who are fluent in the language. When a person is learning a new language and people seem to be speaking too fast, that's only because their processing time for that language isn't up to speed yet. They may be mentally translating to another language, searching for words, relying on context to fill in gaps, etc. which take additional time.
I have known for years that the best way to improve my listening comprehension in a language is to read, read, read... the more I read, the faster I get at reading in a particular language, my mental processing time for that language speeds up, and eventually I'm reading at a speed that is roughly equivalent to spoken language - at that point I can suddenly understand spoken language much more easily. This is not what the study is about though - the processing time referred to in the study refers to how quickly the speakers can process the information and speak it.

User avatar
linguoboy
Posts: 23263
Joined: 2009-08-25, 15:11
Real Name: Da
Location: Chicago
Country: US United States (United States)

Re: Human speech may have a universal transmission rate: 39 bits per second

Postby linguoboy » 2019-09-06, 19:32

I've been meaning to delve into the methodology when I have the time because I would think one of the most difficult parts of this sort of study would be coming up with a sound crosslinguistic method of determining the amount of information conveyed by a particular morpheme.
"Richmond is a real scholar; Owen just learns languages because he can't bear not to know what other people are saying."--Margaret Lattimore on her two sons

User avatar
Osias
Posts: 8411
Joined: 2007-09-09, 17:38
Real Name: Osias Junior
Gender: male
Location: Vitória
Country: BR Brazil (Brasil)
Contact:

Re: Human speech may have a universal transmission rate: 39 bits per second

Postby Osias » 2019-09-07, 3:24

They'll probably just typing the words in Unicode and counting the bits (including blankspaces)
2017 est l'année du (fr) et de l'(de) pour moi. Parle avec moi en eux, s'il te plait.

Linguaphile
Posts: 2179
Joined: 2016-09-17, 5:06

Re: Human speech may have a universal transmission rate: 39 bits per second

Postby Linguaphile » 2019-09-07, 4:32

linguoboy wrote:I would think one of the most difficult parts of this sort of study would be coming up with a sound crosslinguistic method of determining the amount of information conveyed by a particular morpheme.

Osias wrote:They'll probably just typing the words in Unicode and counting the bits (including blankspaces)

LOL. No, their methodology was much more complex than Unicode bits, enough so that I don't really understand it without taking more time to read it carefully. For example, here's a section that seems relevant to the question of amount of information transmitted per morpheme (they have chosen to use syllables rather than morphemes):
For each language L, we computed information-theoretical metrics derived from Shannon’s seminal theory to estimate the average amount of information transmitted per syllable. More precisely, we estimated the first- and second-order entropies of the syllable distribution. [...]
The second-order entropy is the main information index used here, and it is thus denoted by ID. It refers to conditional entropy where the context in which each syllable occurs is taken into account. We characterized this context as the identity of the previous syllable or a null marker for syllables occurring word initially (thus, no bigrams span across word boundaries)
ID=−x,yp(x,y).log2(p(x,y)/p(x))

Well that's not going to display correctly (although I had some fun trying to make it work, LOL). I will have to read through the whole methodology in more detail later, and will probably understand it only slightly better afterwards. :mrgreen: In any case it's certainly an interesting study and like all interesting studies, leads to more questions and the need for further research.

vijayjohn
Language Forum Moderator
Posts: 24180
Joined: 2013-01-10, 8:49
Real Name: Vijay John
Gender: male
Location: Austin
Country: US United States (United States)

Re: Human speech may have a universal transmission rate: 39 bits per second

Postby vijayjohn » 2019-09-07, 5:17

Oh God. I'm pretty sure this is a computational linguistics paper. In other words, bunk.

User avatar
md0
Posts: 7548
Joined: 2010-08-08, 19:56
Country: FI Finland (Suomi)

Re: Human speech may have a universal transmission rate: 39 bits per second

Postby md0 » 2019-09-07, 8:46

Why are they not counting features per morpheme or some metric like that? :hmm:
"If you like your clause structure, you can keep your clause structure"
Stable: Cypriot Greek (el-cy)Standard Modern Greek (el)English (en) Current: Standard German (de)Elementary Finnish (fi)
For fun: Legacy: France French (fr)Japanese (ja)Standard Turkish (tr)

vijayjohn
Language Forum Moderator
Posts: 24180
Joined: 2013-01-10, 8:49
Real Name: Vijay John
Gender: male
Location: Austin
Country: US United States (United States)

Re: Human speech may have a universal transmission rate: 39 bits per second

Postby vijayjohn » 2019-09-07, 16:36

The way they describe their approach reminds me more of computer science papers (i.e. NLP papers written by computer scientists) than computational linguistics papers, honestly. Computational linguists usually seem to be better at explaining what they mean when they write.


Return to “General Language Forum”

Who is online

Users browsing this forum: Saim and 1 guest