Elamite should not be confused with the proto-Elamite language or its writing system, which have yet to be proved identical or affiliated with Elamite.
Elamite has been classified into four periods based on its known history: Old Elamite, Middle Elamite, Neo-Elamite, and Achaemenid Elamite. For information on the text corpora from these periods, consult that listed section.
As with several other languages written in Mesopotamian cuneiform, the historical ambiguities and deficiencies of that script have made our understanding of Elamite phonology more difficult. In particular, voicing distinctions in stops are not always clear.
This information is tentative. Moreover, one must take into consideration diachronic phonological changes in the language through time. Based on Stolper 2008 pg. 57.
|Stops||p, b,k,g t,d|
|Fricatives||s, sz, z, h?, f/v?|
|Liquids and Nasals||n,m l,r|
There are vowels /a/, /i/, /u/, and /e/. Vowel length is not phonemic, nor are there diphthongs (Stolper pg. 59).
A number of ambiguities and variations in spelling practices have been sources of speculation about Elamite phonemes.
One is the possibility of tense and lax consonants. As noted by Erica Reiner (Reiner 1969 pg. 112 ff.), Old Persian loanwords into Elamite with voiced intervocalic stops are spelled as single consonants in the syllabary (i.e. (V)C-CV), as in OP Babiru = ba-pi-li. On the other hand, voiceless intervocalic stops are spelled as gemminated consonants (i.e. VC-CV) as in OP Gautama = ka-ma-ad-da. Thus in Elamite intervocalic voiceless stops are represented by gemmination, whereas intervocalic voiced stops are represented by single consonants. However in non-intervocalic position the voiced distinction is not maintained in spelling, suggesting that voicing is not the only distinguishing feature among stops. One solution is to say that Elamite maintains a tense/lax distinction in stops, where intervocalic tense stops are realized as voiceless gemminated consonants in the Akkadian syllabary while lax ones are realized as single voiced/non-voiced ones. This phenomenon has a parallel in Tamil, which is known to have a tense/lax distinction and exhibits the same behavior with Sanskrit loanwords (see Reiner pg. 115).
Another example is the existence of word-final consonantal clusters….