The older consonant cluster sn- loses its s in Latin:

nix "snow" vs. English snow

na "supper" vs. older Latin cesna

Two questions:

1) Since word-medial -sn- was clearly lost within the history of Latin, is the same thought to be true of word-initial sn-?

(I'm pretty sure that sn > n is not pan-Italic, as I think there are attested forms with initial sn- from some of the non-Latin Italic languages, but I don't recall any examples of initial sn- within Latin.)

2) Is there thought to be any link between the changes sn > n and gn > n in Latin?

I ask because the latter change also seems to have happened within the history of Latin (gnōtus ~ nōtus "known, recognized"), although it does not seem to have extended into word-medial environments at all (ignis "fire", lignum "wood", etc.).


  • 1
    Word initial sn might go via hn, since "hnow" is a child speech substitute for "snow".
    – Greg Lee
    Sep 14, 2015 at 23:42
  • @GregLee What's strange is that there is no general change of s > h in Latin, but in Greek, which did undergo a s > h shift, there apparently wasn't a complete shift of sN > N (where N = any nasal). For example, there is a sporadic or dialectal alternation in Greek between smikrós and mikrós ("short"), the first of these being the older form (according to what I've heard).
    – user8017
    Sep 15, 2015 at 0:00
  • 1
    Well, the change I mentioned in child speech is not the result of a change s > h. It's from a progressive voicing assimilation of n to preceding s, then loss of the s conditioning factor. "hnow" is just a way of writing that.
    – Greg Lee
    Sep 15, 2015 at 2:24
  • One possible explanation that comes to mind for variants like smikros / mikros is that, since sN > N intervocalically is regular in Greek, the s was lost in cases where a preceding word began with a vowel, e.g. ho smikros (treated as one phonological word), but not elsewhere, and the alternants were then lexicalized as two variants of the same word.
    – TKR
    Sep 15, 2015 at 4:24
  • 2
    @czypsu Not that I know of. All the branches in which you would expect sn- to be retained have a cognate beginning with this cluster: Germanic (English snow), Slavic (Slovene sneg), Goidelic (Modern Irish sneachta), etc.
    – user8017
    Sep 15, 2015 at 15:58

1 Answer 1


As Sihler puts it, [in Latin] "an s is lost before most voiced consonants" (para 225), cf. Meiser "(Konsonant+) s schwindet im Anlaut vor Nasal." Thus, in Anlaut (word-initial) clusters *sn, *sm, *sl the *s is lost. He admits that for *sm and *sl the evidence is "meagre." There is no written evidence coming from Latin; this is a reconstruction (i.e. educated guess) based on evidence coming from other IE languages and from the development of Latin Inlaut sn, sm, sl clusters - there is written evidence for the latter. Presumably, the change was s => z=> h (?) => 0

Latin nix 'snow' <= Proto-Italic *sn(e)iw- (de Vaan)

Latin nido 'nest' <= Proto-Italic *nizdo- (de Vaan)

Latin cena 'meal' <= Proto-Italic *kert(e)snā- (de Vaan), cf. Oscan kersnu, Umbrian śesna, Old Latin cesnas

Leumann 1977 adds that Anlaut *sn, *sl, *sm were retained in the Sabellian (or Sabellic) group (para 194), i.e not in Latin:

Latin no nare - Umbrian SNATU (Meiser), Proto-Italic *(s)nāje/o- (de Vaan)

With *gn, there is actual written evidence coming from Old Latin, e.g. gnosco (nosco) or gnatus (natus) (para 220); cf. Meiser "Anlautender Velar schwindet vor n." This process must have happened by the second century BCE (Meiser 1998, Weiss 2009), cf. still gn- in GNATEIS (122 BCE), as opposed to the very first n- NATVM (117 BCE). Presumably, the change was gn => ŋn => nn

[will add more later]

  • Perhaps _sn_ > _n_ belongs together with changes such as *_-zd- > _-d-_, as in _nīdus_ "nest" < * *nizdo-?
    – user8017
    Sep 15, 2015 at 3:02
  • @user8017 of course.
    – Alex B.
    Sep 15, 2015 at 3:41

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.