Similarly to the situation for
Proto-Indo-European, reconstructions of Proto-Uralic are traditionally not written in
IPA but in
UPA.
Vowels Proto-Uralic had
vowel harmony and a rather large inventory of vowels in initial syllables, much like the modern Finnish or Estonian system: Sometimes a mid vowel *
e̮ is reconstructed in place of *
i̮, or a low back rounded *
å in place of *
a. There were no monophonemic long vowels nor diphthongs, though sequences of vowel and semivowel within a single syllable (such as *äj) could exist.
Unstressed vowels Vowel inventory in non-initial syllables was restricted: only a two-way contrast of open and non-open vowels is incontestably reconstructible. The actual realization of this contrast is a question of debate: one view considers this two
archiphonemic vowels and , realized as four allophones , as per
vowel harmony. However, other scholars such as Zhivlov posit the existence of disharmonic *i-a stems in Proto-Uralic, which would imply that vowel harmony was not allophonic. For the non-open vowel(s), most branches reflect a reduced vowel , however: • The
Finnic languages show or depending on harmony, word-finally . • The
Samic languages show a variety of reflexes, but these reflexes can be traced back to a
Proto-Samic phoneme
*e̮, which is also the reflex of Proto-Uralic
*i and
*ü in stressed syllables. While
vowel reduction is a common sound change, Finnic is known to have
adstrate influence from language groups that would not have known reduced vowels (namely the
Baltic languages and the early
Germanic languages), so a value of already in Proto-Uralic remains a possibility. Although these three or four stem types were certainly the most prominent ones in Proto-Uralic, it is possible that other, rarer types may have existed as well. These include for example kinship terms such as "sister-in-law", found as *kälü in both Proto-Finnic and Proto-Samoyedic. Janhunen (1981) and Sammallahti (1988) reconstruct here instead a word-final labial glide: *käliw. A general difficulty in reconstructing unstressed vowels for Proto-Uralic lies in their heavy reduction and loss in many of the Uralic languages. Especially in the
Ugric and
Permic languages, almost no trace of unstressed vowels appears in basic word roots. The original bisyllabic root structure has been well preserved in only the more peripheral groups: Samic and Finnic in the northwest, Samoyedic in the east. The main correspondences of unstressed vowels between these are as follows: Developments in Mordvinic and Mari are rather more complicated. In the former, Proto-Uralic *-a and *-ä are usually reduced to *-ə; *-a is however regularly retained whenever the first syllable of the word contained *u. Proto-Uralic *-i is regularly lost after open syllables, as well as in some other positions.
Conditional vowel shifts A number of roots appear to diverge from the main picture of unstressed syllables in a different way: while Finnic, Samic and Samoyedic languages all have one of the "typical" stem shapes, they may not quite match. Words in these classes often feature discrepancies in the vowels of the first syllable as well, e.g. Finnic *a or *oo (suggesting Proto-Uralic *a or *i̮) against Samic *ā (suggesting Proto-Uralic *ä) or *oa (suggesting Proto-Uralic *o). The change is, however, masked by the shift of *i̮ to *a (which later develops to Proto-Samic *uo) in words such as: In a second group, a change *ä-ä > *a-e appears to have taken place in Finnic in words such as:
Consonants In the consonant system,
palatalization, or palatal-laminal instead of apical articulation, was a phonemic feature, as it is in many modern Uralic languages. Only one series of stops (unvoiced unaspirated) existed: The segments symbolized by č and š were likely retroflex. The phonetic nature of the segment symbolized by *x is uncertain, though it is usually considered a back consonant; *x is also reconstructed word-medially, and in this position it also develops to a Finnic long vowel, but has clear consonantal reflexes elsewhere: *k in Samic, *j in Mordvinic and *ɣ in Ugric. If a consonant, it probably derives from lenition of *k at a pre-Uralic stage; it is only found in words ending in a non-open vowel, while *k is infrequent or nonexistent in similar positions. The phonetic identity of the consonant is also subject to some doubt. It is traditionally analyzed as the palatalized counterpart of the
voiced dental fricative , that is, as ; however, this a
typologically rare sound value for which no direct evidence is found in any Uralic language, and a pure palatal fricative is another option; a third option is a palatal liquid like, e. g., Czech
ř. while
Frederik Kortlandt reconstructs palatalized and , alleging that they pattern like resonants.
Dubious segments The phonemes in parentheses—*ć, *š, *ĺ—are supported by only limited evidence, and are not assumed by all scholars. Sammallahti (1988) notes that while instances of *ć are found in all three of Permic, Hungarian and Ob-Ugric, there are "very few satisfactory etymologies" showing any correlation between the branches in whether *ć or *ś appears. In the other languages, no consistent distinction between these consonants is found. The evidence for the postalveolar sibilant *š however is "scarce but probably conclusive" (ibid): it is treated distinctly from *s only in the more western (
Finno-Permic) languages, but certain loans from as far back as the
Proto-Indo-European language have reflexes traceable to a postalveolar fricative (including *piši- or *peši- "to cook"). The possibility of *ĺ is not considered by him at all. In contrast, Janhunen,
Phonotactics No final
consonant clusters were allowed, so words could end with a maximum of one consonant only. The single consonants also could not occur word-initially, though at least for the first of these, this may be a coincidental omission in the data. A reconstruction "spleen" exists but is not found in Samoyedic and the most stringent criteria for a Proto-Uralic root thus excludes it. A similar case is "fox", a loanword from Indo-Iranian. Inside word roots, only clusters of two consonants were permitted. Since
*j and
*w were consonants even between a vowel and another consonant, there were no sequences of a "diphthong" followed by two consonants, like in Finnish
veitsi. While
voicing was not a phonemic feature, double (i.e.
geminate) stops probably existed ( "father-in-law", "five", "to push"). The singleton–geminate contrast in several descendant languages developed into a voiced–voiceless distinction, although Finnic is a notable exception, e.g. Finnish
appi,
lykkää. When, due to suffixation, consonant clusters arose that were not permitted, the non-low vowel was inserted as a prop vowel. This process was obscured in the Finnic languages by an opposing process which syncopated unstressed
*e in many cases.
Prosody Proto-Uralic did not have contrastive tone. The majority view considers stress to have been fixed on the first syllable, although this is not universally accepted.
Phonological processes Consonant gradation may have occurred already in Proto-Uralic: if it did, it was probably an allophonic alternation involving voicing of the stop consonants: , , . ==Grammar==