Realistic borrowing into a conlang from natlangs

[Trebor]

Since there's only a single pair of consonants in your language that distinguishes voicing -- /t/ vs. /d/ (which notably isn't distinguished in many Finnish dialects) --, I think you might me able to get away with simply adjusting the voicing of consonants in loanwords to Kosi phonology, as you have done in your examples. There would probably be some amount of allophonic variation anyway, for example in clusters or depending on the presence or absence of stress in a particular syllable. (This gives me an idea: If you have the latter kind of allophony, e.g. /k/ tends to be voiced [g] intervocalically in unstressed syllables, then maybe you could change the location of stress in some loanwords depending on the voicing of the original word.) Another thing that helps with not gaining a systematic voicing distinction is that your language seems to be located in a rather peripheral area, so while its speakers would of course be exposed to some amount of Russian and other European languages, the pressure to pronounce words from these languages accurately won't be as strong as it would be in a more cosmopolitan place like, say, the city of Helsinki.

All important points. There isn't any voicing-related allophony in Kosi at the moment, except when consonants end up immediately adjacent to one another. In plosive-, affricate-, and fricative-initial clusters, a difference in voicing is resolved with progressive assimilation: /bs/ -> [bz], /sv/ -> [sf]. In nasal- and liquid-initial clusters, such a conflict is settled with regressive assimilation: /ms/ -> [m_0s], /rt/ -> [4_0t].

Fricative+plosive combinations have raised a dilemma for me, because (as I recall reading a long time ago) Russian has regressive voicing assimilation when obstruents collide. Kosi /avstrali/ <- Russian [afst4alija] 'Australia' would have to be pronounced [avzd4ali] and /vintovka/ <- [vintofka] 'rifle' would have to be pronounced [vintovga]. I'm not so fond of this idea.

When borrowing into Kosi, I have also been able to take advantage of (what I recall reading was) Russian's word-final devoicing rule: Kosi /bodvik/ <- Russian /podvig/ [podvik] 'feat/adventure', /deviz/ <- /deviz/ [dEvis] 'motto', /fotosintes/ <- /fotosintez/ [fotosintEs] 'photosynthesis', /kas/ <- /gaz/ [gas] 'gas - state of matter', /baraS/ <- /baraZ/ [baraS] 'barrage', /mjateS/ <- /mjateZ/ [mjatES] 'mutiny/rebellion'.

How would you employ allophony and stress shifts to handle Russian /p/, /g/, /f/, /z/, and /Z/ when they appear in a range of different environments? (The loanwords in the first post of this thread should be numerous enough to cover the different possibilities. If you could use some of them to illustrate what you envision, the issue should become clearer for me.)

[Trebor]

I don't think it'd be a problem to drop affixes like -atsija. Not sure if it's attested anywhere, but there are plenty of languages (English and German included) that prefer to abbreviate long words than prefer them, and if your conlang is one of those, then that's three syllables that can be lost. Could become a regular pattern.

That's great. I haven't encountered a language where Latinate words in -tion were borrowed with the final suffix skipped entirely--does anyone know of such a case?

What instances are you aware of in which English and German shortened loans systematically?

And if most speakers are bilingual it might be more likely for only the root to get borrowed. Didn't that happen to some extent with English and Latin?

Maybe. But for your real-world example, it's hard to say, because "most [of its] speakers [have never been] bilingual" in Latin.

[Pole, the]

Since there's only a single pair of consonants in your language that distinguishes voicing -- /t/ vs. /d/ (which notably isn't distinguished in many Finnish dialects) --, I think you might me able to get away with simply adjusting the voicing of consonants in loanwords to Kosi phonology, as you have done in your examples. There would probably be some amount of allophonic variation anyway, for example in clusters or depending on the presence or absence of stress in a particular syllable. (This gives me an idea: If you have the latter kind of allophony, e.g. /k/ tends to be voiced [g] intervocalically in unstressed syllables, then maybe you could change the location of stress in some loanwords depending on the voicing of the original word.) Another thing that helps with not gaining a systematic voicing distinction is that your language seems to be located in a rather peripheral area, so while its speakers would of course be exposed to some amount of Russian and other European languages, the pressure to pronounce words from these languages accurately won't be as strong as it would be in a more cosmopolitan place like, say, the city of Helsinki.

All important points. There isn't any voicing-related allophony in Kosi at the moment, except when consonants end up immediately adjacent to one another. In plosive-, affricate-, and fricative-initial clusters, a difference in voicing is resolved with progressive assimilation: /bs/ -> [bz], /sv/ -> [sf]. In nasal- and liquid-initial clusters, such a conflict is settled with regressive assimilation: /ms/ -> [m_0s], /rt/ -> [4_0t].

Fricative+plosive combinations have raised a dilemma for me, because (as I recall reading a long time ago) Russian has regressive voicing assimilation when obstruents collide. Kosi /avstrali/ <- Russian [afst4alija] 'Australia' would have to be pronounced [avzd4ali] and /vintovka/ <- [vintofka] 'rifle' would have to be pronounced [vintovga]. I'm not so fond of this idea.

You could make such a rule that obstruents which do not have its voicing counterpart (e.g. /v s/ without /f z/) are handled as sonorants when it comes to voicing assimilation, e.g. /vk/ -> [fk], /sd/ -> [zd] etc.

[Trebor]

You could make such a rule that obstruents which do not have its voicing counterpart (e.g. /v s/ without /f z/) are handled as sonorants when it comes to voicing assimilation, e.g. /vk/ -> [fk], /sd/ -> [zd] etc.

That sounds like an elegant solution, but Kosi /asvalt/ 'asphalt' <- Russian [asfal_jt], for example, would have to be pronounced with [zv] too, wouldn't it?

[Cedh]

You could also define a hierarchy that determines which sounds tend to retain their underlying voicing and which ones change it. For example, you could say that plosive-plosive clusters assimilate regressively (that's IMO more likely, both because Russian does this and because plosive voicing is easier to hear before a vowel than after one), whereas fricatives always assimilate in voicing to an adjacent plosive, and /v/ always assimilates in voicing to an adjacent /s S/.

With these rules, some of the Russian loanwords from this thread would come out something like the following:

/posredstvenn1j/ 'mediocre' > /bosredstvenni/ [bosretstfenni]
/grejpfrut/ 'grapefruit' > /krejbvrut/ [krejbvrut]
/izgnannik/ 'outcast/exile' > /isknannik/ [isknannik]
/diftong/ 'diphthong' > /divtonk/ [diftoN_0k]
/formulirovka/ 'wording/formulation' > /vormulirovka/ [vormulirofka]
/infrastruktura/ 'infrastructure' > /invrastruktura/ [invrastruktura]
/jazva/ 'ulcer' > /jasva/ [jasfa]
/povstanets/ 'rebel/insurgent' > /bovstanets/ [bofstanets]
/predskazanie/ 'prediction/forecast' > /bredskasanje/ [bretskasanje]
/sverkajuS_j:ij/ 'brilliant/sparkling' > /sverkajuSi/ [sferkajuSi]
/onemevSij/ 'speechless' > /onemevSi/ [onemefSi]
/asfal_jt/ 'asphalt' > /asvalt/ [asfalt]
/afst4alija/ 'Australia' > /avstralija/ [afstralija]

There were no words in this sample with obstruent clusters that would come out as voiced. I don't know any Russian, so I'm making up some hypothetical words now (notating only the Kosi side).

/abda/ [abda]
/abta/ [apta]
/adka/ [atka]
/akda/ [agda]
/atba/ [adba]
/ebsetve/ [ebzetfe]
/esbevte/ [ezbefte]
/edSekve/ [edZekfe]
/eSdevke/ [eZdefke]
/isvi/ [isfi]
/ivSi/ [ifSi]

[Click]

That sounds like an elegant solution, but Kosi /asvalt/ 'asphalt' <- Russian [asfal_jt], for example, would have to be pronounced with [zv] too, wouldn't it?

What's so bad with voicing assimilation? If you really don't want it, you can make /v/ realised as [P] after a voiceless consonant.

[Cedh]

Another option would be to define a voicing assimilation hierarchy on the phoneme level, while not specifying a direction of assimilation at all. Something like /d/ > /t/ > /b/ > /k/ > /s S/ > /v/ would seem to make sense for this language (with sounds appearing earlier in the list being less likely to change their voicing value), and it would give mostly the same results as above, except that /adka/ would come out with a voiced medial cluster as [adga], and /atba/ would come out with a voiceless cluster as [atpa].

Maybe it would also be useful to have a majority rule for clusters of three or more consonants, for instance if more than half of all consonants in a cluster are phonemically voiceless, the whole cluster will be phonetically voiceless.