Recently I’ve been gradually working on assembling (together with some other editors) an appendix of Proto-Finnic roots for the English Wiktionary. Today I ran into an interesting issue: the word for ‘voice’.
Finnish and Votic ääni, Karelian eäni ~ iäni, Ludian iäń, Veps äń allow straightforwardly putting together PF *ääni. Given Samic *jēnë ‘voice’ and Hungarian ének ‘song’, this is evidently inherited already from Proto-Uralic.
The reason for the long *ä in Finnic is not entirely clear. In wake of Ante Aikio’s recent re-defense of “Lehtinen’s Law”, i.e. the development *ä > *ee before single sonorant + *ə, we’d actually expect **eeni; his paper suggests that raising was in this word for some reason blocked in the word-initial position, given how Proto-Finnic has no *ee-. There is no counterevidence for this — but neither are there any parallels, and so the question appears to be impossible to settle.
Southern Finnic languages however at first glance seem to have a different root: Estonian hääl, Livonian ēļ, Votic ääl (in some other dialects; I don’t think I will check right now their distribution) suggest instead *hääli.
Yet this is similar enough that etymologists have long considered it possible that they in fact also derive from *ääni, with some kind of irregular deformation. According to SSA,[1] this observation is due to Ahlqvist already in the mid-1800s. If so, then it appears to be possible to presume an earlier alteration to *ääli, followed by secondary addition of h- in Estonian. There is no reason to assume that aspiration was ever present in Votic or Livonian (although, of course, if it were, it would have disappeared all the same).
But how back would this variant go, then? Estonian and Votic are fairly closely affiliated, but Livonian is not; arguably it is more distantly related to the two than Finnish and the rest of the Northern Finnic continuum are. Formally, the variants *ääni ~ *(h)ääli, as attested side-by-side in Votic, should therefore be both reconstructed already to late Proto-Finnic. (If it wasn’t for Samic and Hungarian supporting *n, we could even suspect that the direction of alteration was rather *ääli > *ääni, and that the “Northern” variant in Votic was an Ingrian loanword.)
Positing *ääli already in PF seems to allow / be further supported by a new etymology, for Finnish ääliö ‘idiot’. This is usually explained as “onomatopoetic”, but the morphology doesn’t support this idea; the ending -iO is just about nonexistent in onomatopoetic words. I suspect we rather have here an old example of stereotyping disabled people by their behavior, and that the word is a derivative of the *ääli variant of ‘voice’. Thus, *ääl-ijo ‘sound-maker’ > ‘person who makes weird noises’ > ‘idiot’.
It’s interesting to moreover compare Finnish hälistä ‘to be noisy, to chatter’, älistä ‘to cry, to moan’. These two are evidently onomatopoetic variants of each other — but if we allow the option that *ääli existed already in late Proto-Finnic, we could just as well also allow the existence of a variant *älə in early Proto-Finnic, prior to Lehtinen’s Law. And it appears to be the case that sufficiently early derivatives with an altered stem vowel were not subject to LL. Aikio has noted some examples such as Fi. säle ‘splinter’. Another interesting case I’ve noticed is *mälə > *määlə > LPF *meeli > Fi. mieli, Es. meel ‘mind’, contrasted with PF *mäl-o > Estonian mälu ‘memory’. If so, then a derivative *äl-ićə- could have regularly yielded älistä.
This scenario seems to additionally provide a tiny amount of extra support for the idea that *ää > *ee was blocked in word-initial position: we’d now have two examples of it, i.e. both *ääni and *ääli. (Effectively perhaps less, since they’re variants of one another.)
If I wanted to further proliferate stem variants, I could also posit *hälə coming into existence already this early, and derive hääl on one hand, hälistä on the other from this in a similar fashion. But a *h this early is anachronistic (LL precedes *ð > *t, which precedes *ti > *ci, which precedes *š > *h), and in light of words like hello, hallo it is clear that /h/ can have an onomatopoetic function in ‘voice’-related words. It seems thus reasonable to consider it to have come about independently in Estonian and Finnish.
[1] i.e. the etymological dictionary Suomen sanojen alkuperä (1992–2000), ed. Erkki Itkonen & Ulla-Maija Kulonen.
Freelance reconstruction 
Hi,
Quick comment on this:
“and in light of words like hello, hallo it is clear that /h/ can have an onomatopoetic function in ‘voice’-related words”
The etymology of hello/hallo/etc. seems to be unknown; it may originate in the term “holla” (“stop!”), from French holà, or it may be from the German word holen “summon, fetch”, which has relatives in many other IE branches (e.g. Latin calare, Greek kaleîn “to call (together), summon”).
More generally, though (and with all due respect), why would a word used for greeting people or getting people’s attention have an especially strong connection to voice-related words?
A possibly-better candidate for a sound that is associated with the act of speaking (though perhaps not “voice” in general) is [l]: witness Greek laleîn “to talk, chat”, English blab “to talk indiscreetly/thoughtlessly”, Irish labrur “I speak”, etc.
German lallen “make indistinct noises while drunk under the table”, labern “talk a lot (of tedious nonsense)”…