Proto-Indo-European had consonants, resonants, laryngeals, and vowels (see Lehmann's Proto-IE Phonology and other manuals on Indo-European phonology).

Douglas Simms (email dsimms at siue.edu) designed and implemented chart prototypes, which since then have collapsed *palatal and velar obstruents into one column on the Phonemes page leaving root sub-pages to mark distinctions.

Two obstruent sets here are the traditionally reconstructed /p t k kw/, /b d g gw/, and /bh dh gh ghw/ or the "glottalic" /p^(h) t^(h) k^(h) kw^(h)/, /[b] t' k' k'w/, and /b^(h) d^(h) g^(h) gw^(h)/. Equated sets reflect two hypothetical views of Proto-Indo-European phonology. Each system is based on the same attested forms, but has different implications for the way sounds may have changed in prehistory. Most handbooks reflect the traditional view. For more on the Glottalic Theory, see the section on phonology in Thomas V. Gamkrelidze & Vyacheslav V. Ivanov. 1995. Indo-European and the Indo- Europeans (1984 Russian original translated by Johanna Nichols, Berlin: de Gruyter) with references.

Consonants combined to form PIE root structures of the shape, CVC- (beginning with a Consonant (C) plus a middle Vowel (V), and ending in a Consonant). So the root for 'foot' began with the consonant /p/ or /p^h/ and ended with /d/ or glottalic /t'/. Traditional *p corresponds to English /f/ and *d to English /t/. Alternatively glottalic *p^(h) corresponds to English /f/, glottalic *t' to English /t/.

Vowels, as the nucleus of a root, might undergo Vowel Alternation (or Ablaut). PIE was hypothesized to have one vowel, /e/, often alternating with [o] or lack of a vowel ("a zero vowel"). Such vowel alternation might vary the grammar as it still does in English 'sing, sang, sung' ([in] of 'sing' reflects older [en], [an] of 'sang' older [on], and [un] of 'sung' older [n]).

Both obstruents and the fricative /S/ always behaved as consonants in the CVC- root, but resonants could behave as either consonants or vowels depending on their environment in a syllable. Like vowels, resonants could function as a syllable nucleus in the absence of a vowel. When vowel alternation varied the grammar of a word as in 'sing, sang, sung', the lack of a root vowel put the resonant in a vocalic position ('sung' reflects *sn.-g^wh as vocalic *n. regularly becomes /un/ in Germanic. Vocalic resonants are usually transcribed with a circle below them to distinguish them from consonantal resonants, but pending resolution of the font problem, a period follows vocalic resonants here).

The resonsont (glide) *w in *wed- / wet' 'water, wet' behaved as a consonant, but it too could have a vocalic counterpart. Not only did processes of vowel alternation ('wet, water') create new words, but as enlargements or suffixes created new word forms, resonants might become a syllable nucleus. The root *wed-(r) / wet'-(r) alternated with *ud-(r), ut'-(r) when the root vowel of *wed/t'-(r.) was reduced to zero. The resonant *w, as syllable nucleus, then takes its vocalic form *u, a form reflected in Greek (h)ud-or 'water' whence English borrowed 'hydro-phobia' (water-phobia).

A theoretical class of sounds, Laryngeals (H₁, H₂, H₃), "colored" vowels on either side of them. H₂, for example, colored an adjacent vowel /e/ with the value /a/. Similar to resonants, laryngeals also had consonantal or vocalic function, depending on their position in a syllable.