[overlordalex]

They're considered fillers that mark a continuance on part of the speaker so that the other participants know that they will continue speaking. Different languages not only use different sounds, but in many cases continuation markers can also be "normal" words from the language. In English some examples would be "well", or "yes" which can be pure filler words to bridge to the next utterance.

That being said, what is a word anyways? You could argue that they're well understood units of language that convey meaning, which I would argue is pretty much a word

[philipov]

You make a good point. I looked it up and it's in MW and even scrabble recognizes it as an interjection. I think to be a word, at a minimum an utterance would have to conform to the phonemic structure of a language, but that's a low bar, and 'uh' passes. I guess it's a word!

[eru]

> [...] would have to conform to the phonemic structure of a language [...]

I don't think that's a good criterion, because it would exclude a lot of loanwords.

[thaumasiotes]

It won't exclude anything. Loanwords can only be expressed in the phonemic structure of the language, because the speaker isn't capable of doing anything else.

[eru]

> [...], because the speaker isn't capable of doing anything else.

Huh? Japanese people have the same vocal tracts as Germans who have the same vocal tracts as Egyptians. Just because someone's native languages don't have a particular sound or sound combination doesn't mean they can't learn.

That's especially true for sound combinations: a Turkish speaker might find a word that doesn't follow Turkish vowel harmony a bit weird, but would have no trouble pronouncing it even without any training.

And even if a particular speaker can't produce a certain word, they can still recognise it as a word when someone else uses it in the context of their language.

As an example, most English speakers I know can't pronounce 'kn' like in Germany 'Knie' or 'Knecht'. But I can say "Knie is German for knee." and that is a sentence of five words. Or "The Knesset is the unicameral legislature of Israel."

(Just to be clear, English has words like knee or knight that are spelled with kn, but the k is silent. If you ask English speaker to pronounce "Knie" the German way, they tend to introduce something like a Schwa between the k and the n sounds.)

See https://ell.stackexchange.com/questions/151054/why-is-the-k-...

[philipov]

A Japanese speaker can learn to speak German, but that's not the same as a loan word. Japanese is a great example for how loan words are modified to conform to the phonemic structure of a language because it absorbed so much english in the past century.

"Fight" is a word in english, but the loan word in japanese is pronounced "faito" because the language demands that all words end in either a vowel or syllabic 'N'. Japanese people are capable of saying english words without ending them in a vowel, but then they're speaking english, not japanese.

The word for salmon roe in japanese is 'ikura'. This is a loan word from russian: 'ikra' for caviar. But because of the Consonant-Vowel structure of the language, a 'u' was added when the word was borrowed.

One of the funniest examples is when a word is borrowed by japanese speakers and then gets translated back into english, the translators won't always return it to the original form. The name Lily would be pronounced "Riri" in japanese (japanese speakers might not even notice the difference between R and L because they're the same phoneme), and when it's translated back into english, it might come back as "Really". This has been a source of consternation for video game players before.

Likewise, when american english speakers borrow a word from a language with a trill or rolling R, they make it conform to the phonemic structure of the language by changing to a retroflex R.

And these sorts of examples are found everywhere. Every language has a particular structure for how valid words can be formed, and speakers of the language will modify foreign words to fit the sounds they're trained to emit unless they're consciously trying to speak a different language.

[thaumasiotes]

> but the loan word in japanese is pronounced "faito" because the language demands that all words end in either a vowel or syllabic 'N'.

By the way, it's worth observing that that is a requirement of the kana writing system, but it's not a requirement of the language. [It also isn't a requirement of the kanji writing system, in which a symbol can indicate any arbitrary sequence of sounds, but that system is difficult to use for purposes of indicating pronunciation.] There are circumstances in which high vowels are entirely deleted, the most prominent example being the ordinary pronunciation of です /desɯ/ with no final vowel at all.

The fact that "fight" gets borrowed as "faito" also looks like an artifact of the spelling system - /o/ is not a high vowel and can't be deleted, but in native Japanese there is no tu syllable - that space in the syllabary is occupied by the affricated tsu, so loanwords that originally ended in /t/ or /d/ are given the final vowel /o/ instead. Illiterate Japanese might or might not interpret the English sound of "fight" the same way.

[thaumasiotes]

> The name Lily would be pronounced "Riri" in japanese (japanese speakers might not even notice the difference between R and L because they're the same phoneme), and when it's translated back into english, it might come back as "Really". This has been a source of consternation for video game players before.

It's also a source of consternation for people who consume Japanese media. There is a particularly funny example in the franchise Ah! My Goddess, in which the demon Maaraa (named after the Buddhist demon whose name in Sanskrit is Maara) gets "translated" into English as "Marller".

It's easy to see how a Japanese person went from the Japanese name to the English one - long Japanese vowels are taken as indicating English rhotic vowels, and then the fact that it would be strange for a rhotic vowel to be followed by an R hints that the Japanese R should be interpreted as an English L - but it's not exactly well-motivated by the actual name of the demon. (And while it might make sense to Japanese speakers attempting to make English out of Japanese, it makes much less sense to English speakers attempting to do the same thing - we hear "maaraa" as "mara" and view the Japanese derhotification of our rhotic vowels as a mistake, not an equivalence.)

[thaumasiotes]

I'm not sure what you think you're arguing. All loanwords in every language are pronounced using the phonemic structure of the borrowing language, the one that is being used, because that is the only possibility. The fact that someone is theoretically capable, after years of study, of learning to pronounce a foreign word, is not relevant in any way.

[eru]

> All loanwords in every language are pronounced using the phonemic structure of the borrowing language, the one that is being used, because that is the only possibility.

That's not true at all. Have a look at eg how Turkish borrows from English.

Turkish borrows eg the word 'computer' wholesale without modification. Despite that word violating Turkish vowel harmony. (Unless you want to tell me that vowel harmony is not part of Turkish 'phonemic structure'?)

Similarly lots of German borrowings from English in recent decades don't adjust to German 'phonemic structure'. Eg German typically pronounces st at the beginning of a word like sht, but that only happens in the borrowing 'Stress' for some speakers. (See the IPA on https://de.wiktionary.org/wiki/Stress for evidence.)