> _porsas_ is an interesting data point, but are there any other widely-accepted loans of this > kind (i.e., loans that appear to show unmerged short *o)? If not, then maybe we should be
> cautious in deciding exactly what to conclude from it: the *o could reflect dissimilation
> (*a-a > *o-a), the (sporadic) influence of the *p-r environment, or both.
Some other cases include (in addition to the three just brought up by Douglas):
_orja_ "slave" < *worg^-
_orpo_ "orphan"
_moni_ "many"
_uni_ (< *on@) "sleep" (~ Greek _onar_)
_rohto_ "medicinal plant" < *g^hroH-to-
The first four have cognates in other Uralic branches and seem fairly old, esp. _uni_ which would have to have been loaned early enough to participate in the widespread early Uralic sound change *o > *u / _C@. "Slave" has an alternate etymology from _Arya_ which would put it in the category described below. "Medicinal plant" also has an alternate etymology relating it to _ruoho_ "grass" (< Gmc) within Finnic, but this requires ad hoc assumptions on the morphology.
NB _orpo_ and _rohto_ do not seem to have a Proto-Finnic *-o representing an original 2nd syllable *o; cf. Karelian _orboi_ and _rohtu_, which indicate *orpoi < *orpai; and *rohtV with later independent suffixation in Fi. and Ka.
One proposed case would even date to Proto-Uralic: _koke-_ "to perceive, to check" (cognate to Samoyedic *ko- "to find") ~ *h3okw- "eye", but I find this to be weak comparision.
Then there is also an unexplained group of so-called "o-loans" where Finnic /o/ comes from Late PIE or Indo-Iranian *a, not *o: e.g.
_ohra_ "barley" < Baltic < *h2ak^tro-
_olut_ "beer" < *h2alut
_onki_ "fishing rod" < Gmc < *h2ank-
_ora_ "thorn, awl" < II
_otsa_ "forehead" < Gmc < *h2antio-
possibly _orsi_ "perch" < Baltic *ardis
(I suspect this is an issue on the Finnic side & somehow related to how these words mostly have zero initials.)
> Incidentally, did the o-vocalism of the derivational suffix -os (as in Finnish _tulos_ "result",
> etc.) arise through the same process as the derivational suffix -o (as in Finnish _tulo_
> "arrival", etc.), which as I recall is from earlier *-aj?
Strictly speaking, we have no direct evidence. The _tulos_ type is certainly from a suffix *-wksV (cf. e.g. Fi. _punos_ ~ Erzya _ponavks_ "plait", from *puna "hair"). Modern Finnish -o is however a confluence of at least four different suffixes:
1) "applicative" /-o-/ in verbs such as _puno-_ "to plait" is from a stem vowel *a plus *-j. This is historically identical to "applicative" /-i-/ in some other derived verbs such as _kukka_ "flower" > _kukki-_ "to bloom", but both were later applied analogically in some positions where the other would have been the regular outcome (including *punoi- here; **puni- would be expected).
2) /-o-/ in compounding forms of nouns such as _jalka_ "foot" > _jalko-_ is similarly from *a + *-j, and similarly historically identical to /-i-/ in compounding forms of nouns such as _lehmä_ "cow" > _lehmi-_.
3) the origin of "nomenverbal" /-o/ in cases like _tulo_ is unclear, but this was probably originally in complementary distribution to the similar -u/y in words like _pese-_ "to wash" > _pesu_ "washing"; _kylpe-_ "to bathe" > _kylpy_ "bath". This suggests *-w, as /-os/ < *-awksV mentioned above has similarly a variant /-us/, as in _kalasta-_ "to fish" > _kalastus_ "fishing".
4) an "in-law" /-o/ appears with certainty only in the obsolete _nato_ "sister-in-law", but this can be dated all the way to Proto-Uralic (cf. Proto-Samoyedic *nåto "brother-in-law"), and whatever its source, the same original also seems to have occurred in _vävy_ "son-in-law", _käly_ "sister-in-law", and possibly _anoppi_ "mother-in-law". The o/y alternation suggests origin from *-w, though there is no explicit evidence for this.
>> Some other direct examples of "vowel flattening" in loanwords:
>> _ankerias_ "eel" < Baltic *angurjas
>> _kirves_ "ax" < Baltic *kirvis
>> _sisar_ "sister" < Baltic *s(w)esoor
>> _mesi_ (_mete-_) "honey" < PIE *medhu-
>
> Is Finnish _mete-_ a case of vowel flattening, or simply the "deletion" of -u
> because *metu would have looked like a derivative of an otherwise non-existent root?
That kind of morphologically motivated deletion _is_ pretty much exactly what I mean by "vowel flattening". There is no evidence of an actual sound change *o > *a etc. having been involved at any point here.
_j.