Do they just speak faster? Do the Indian words/pronunciation flow better/faster than English does? And they are simply trying to match the cadence?
Do they just speak faster? Do the Indian words/pronunciation flow better/faster than English does? And they are simply trying to match the cadence?
One way of classifying languages is grouping them into stress-timed, syllable-timed and “mora”-timed languages.
Stress timed languages (like English) are ones where the time between stressed syllables is roughly the same. Take the phrase “I went to the store with my friend John”. Most native English speakers will stress “went”, “store”, “friend” and “John”. It might not be a big difference, but you’ll notice the “to the” between “went” and “store” is rushed, and that there’s a sort of gap between “friend” and “John” since both are stressed. (Also, if you were to modify that slightly and say “I went to the store with my friend named John”, the time between “friend” and “John” wouldn’t change much at all, you’d just slip “named” into that gap.)
Many Romance languages are seen as syllable-timed, where each syllable takes the same amount of time. In French that phrase is “Je suis allé au magasin avec mon ami John”, that’s 14 syllables, all roughly the same timing. In Spanish it’s “Fui a la tienda con mi amigo John”, 12 syllables. Unless you’re really drawing attention to one of the words, every syllable there gets roughly the same timing.
Japanese is mora timed, which is pretty similar to being syllable timed, except that when you encounter double-letters they double the length of the syllable. So, “Just a moment please” is “Chottomatte kudasai”, where the syllables with double-t letters take twice as long. The cities Tōkyō (two syllables), Ōsaka (three syllables) and Kawasaki (four syllables) all take the same amount of time to say because the “ō” symbol means that letter gets double the length of the standard “o”.
The 4 most widely spoken languages in India are Hindi (way out in front with 44% of the population speaking it as a first language), followed by Bengali, Marathi and Telugu (with about 6-8% each) The first 3 are all Indo-Aryan languages, and Telugu is a Dravidian language. The 3 Indo-Aryan languages are considered to be syllable-timed and Telugu is considered to be mora-timed.
IMO, what makes Indian-inflected English seem fast is that they’re adopting the syllable / mora timing from their primary language and using it in English. That means they spend less time on syllables / words that English speakers would stress and more time on the un-stressed syllables. The overall timing of what they say is probably similar, but in evening out the length of the syllables, they take time away from the syllables that other English speakers naturally slow down to stress. Since you tend to notice the stressed words more, since they’re rushed it seems like the entire sentence is rushed.
>BestOfLemmy
Thank you, that was a good and interesting start of the day
This is a fantastic explanation. Thank you.
Great explanation, thank you!
I remember seeing a linguist doing research into the actual timing of long Japanese vowels and finding that they weren’t actually double the length, more like 1.5 times as long (or 1.7 or something like that). I’ll have to see if I can find the article or paper again.
Yeah, that makes sense. It seems hard to lengthen a vowel out like that unless you’re actually chanting or something and are keeping to a specific rhythm.
Ok, so I heard a thing a long time ago about information density in languages, and that there’s a specific amount of information conveyed per second which is pretty consistent across languages, even when the number of sounds is higher or lower. Which means that a single word in English, for instance, would convey more information than a single word in Hindi.
Is there anything to that? Or was that just nonsense?
Someone posted a link to just that topic here. Apparently almost all languages transmit about 39 bits per second of data. Italians use 9 syllables per second, Germans only about 5-6, but both convey the same amount of information per second. But, not all syllables are equal. Japanese has about 5 bits per syllable, English has about 7 bits per syllable. The most information dense language per syllable is apparently Vietnamese with about 8 bits per syllable.
Apparently though, the bottleneck is the brain. The end result seems to be that languages that have fewer “bits of data” per syllable say those syllables more quickly, and the ones with fewer bits of data per syllable say those syllables more slowly, so that the average is about 39 bits per second no matter what the language.
Having said that, I often listen to podcasts sped up to 1.5x speed, and I listen to podcasts while doing other things, so I guess the bottleneck is probably on the sending side rather than the receiving side.
Podcasts, being prerecorded and edited, don’t really fit this model. It’s more for a conversation with a back and forth where both interlocutors don’t know ahead of time what the other person will say. So they need to observe/listen, reflect while also coming up with answers and putting effort into being properly understood. So basically the natural context in which inter human communication evolved.
Does anyone know how the amount of information is actually derived? The article just says “researchers calculated”
They were vague about it, but they said something about converting it to computer code. I would guess they just wrote it out as ASCII text and counted how many bits of ASCII equivalent they transmitted. (Of course this ignores intonation and emphasis, but I’d guess they did ignore those.)
If that’s really what they did, it’s stupid. First, you need to find a translation for every language to ASCII, which will wildly skew the results. Second, there are many ways to express the same concept, which all vary wildly in length. Take “Hi”, 2 letters, which means exactly the same as “How are you doing?”, 14 letters.
It’s similar, but not exactly the same by any stretch. But, yeah, it’s not a perfect method. But, there probably isn’t a perfect method. How would you decide what “1 unit of information” is?
I wouldn’t, because I have no knowledge in the field. But since the paper hinges upon that exact definition, and “They were vague about it”, this raises the biggest red flag I’ve seen in science yet.
This is true.
I don’t think that’s the right interpretation. There are words in English that would require sentences to be made for each if conveyed in a different language. But the same is true vice-versa.
Have a look at subtitles for movies from one language to any other. Translators struggle conveying what should be paragraph long sentences of context behind a single word for one language. Do not get me started on double speak.
Oh, interesting. I hadn’t considered that there would be variances in information density within a language, but that makes sense; “truth” is a very loaded concept that means a lot of different things in context, even though it’s only one syllable; but on the other hand “authenticity” is five syllables but carries with it a meaning that is a subset of the definition of “truth.”
I guess that’s why subtitling is even possible in different languages; if there were languages with vastly less information density than the source language, they’d need a whole screen just for the captions.
Fairly nonsense If anything I’d say it’s the other way around – there are lots of words in Hindi/Malayalam that you need 5 or 6 English words to describe
It’s not nonsense. Information density isn’t about number of words. It’s about duration and complexity of communication. And it is fairly consistent across all languages. Some languages take 3 words to say something the other can say in one, but those 3 words probably take a similar amount of brainpower and time to communicate as the one word.