As an ordinary term of phonetics, the word 'sonant' is obsolete. The Webster definition below is equivalent to what we now call voiced.

It is still used (with a different meaning) to describe a class of sounds that have been reconstructed in Proto-Indo-European (PIE), the ancestor of most European and Indian languages. (Sometimes the word resonant is used instead, but this also has an ordinary phonetic meaning.) They are six sounds that can be used as both consonant and vowel, depending on their position in the word, and the word accent. The sounds are m n l r w y.

Traditionally w y are called semivowels, but I can see no valid linguistic use for this common term, and am partly noding sonants now so that I can refer to it when I node semivowel. The key point is that all six of the PIE sonants behave in essentially the same way.

A key feature of PIE grammar is that roots of verbs don't have any distinction of vowels. Each root is (or is based on) a monosyllable, beginning and ending with a consonant or two. The vowel in the middle is always e. (Of course, going this far back, it's open to question exactly how they were pronounced, so this is partly conventional.)

The normal range of vowels i e a o u is in part created by grammatical processes akin to modern English sing - sang - sung and song:see ablaut for details. In English these alternations are a minority, but in PIE they were the rule. Originally they are thought to have come about by phonetic processes, somehow related to the accent, but in time they were fixed as part of the grammar.

The other processes that created the normal range of vowels were phonetic. One involves the sonants, the other involves so-called laryngeals and can be ignored in this write-up.

The basic grammatical alternation was of three grades, as they are called: the e-grade, the o-grade, and the zero grade. For example, the root 'to fly' appears as pet- and pot- and pt- in different grammatical situations. (I omit the complicated details of what these situations actually are.)

If the closing consonant, or the first of two, is a sonant, then the same grades occur, but with phonetic changes. So the root 'persuade' (or 'bid' someone do something) appears as bheydh- and bhoydh- and bhydh-. But because y is a sonant, it changes from a consonant to a vowel when it's squeezed into the zero grade, which therefore appears as bhidh-.

Likewise w behaves as the vowel u in zero grade: the root 'pour' has grades ghew- and ghow- and ghu-.

In English some consonants can function as syllables by themselves: especially n as in button, deaden and l as in cattle, muddle. But these are unstressed. We only have stressed consonants in a few interjections like hmm and mmm and grrr; and in many American accents the vowel in words like hurt is actually a syllabic R.

But in Proto-Indo-European m n l r could be syllabic much more often. These four were sonants like w y and like them had a sound change from consonant to vowel in the zero grade. So 'cut' had grades tem- and tom- and tM-: where I'm putting the M in capitals to indicate it's syllabic. In books this is indicated by putting a small circle under the sonant.

Likewise 'grieve' had grades pendh- and pondh- and pNdh-.

Mostly these syllabic consonants have disappeared in the descendant languages, usually by the addition of vowels, sometimes by replacement. In Greek M and N became a, while R and L became ar or ra, and al or la. In Sanskrit M and N became a, but syllabic L and R survived. They also survive to this day in some Slavonic languages, such as Serbo-Croat and Czech, and most notably Slovak which has long syllabic L and R as well as short.