Xjyuk

I've read in various sources that the verb nosco 'know' had a long vowel in the first syllable in Classical Latin pronunciation: nōscō [noːskoː]. I'm wondering what the linguistic evidence is for the vowel being long before -sc- in this word.

The most direct type of evidence would probably be a statement from a Classical author: do any of them talk about the length of this vowel? The next most direct type of evidence that I can think of would be the quality of the vowel in Romance descendants, but I'm not sure that I know all of the relevant sound changes. Italian has close o in coˈn[o]scere, which seems to support the long vowel reconstruction (based on the correspondence Latin ō = Italian close [o], Latin ŏ = Italian open [ɔ] or [wɔ]), but I know that Italian vowel qualities sometimes went through more complicated changes. Old French conoistre > French connaȋtre is possibly also evidence for Latin ō, since if I'm reading this Wikipedia table right, the word should have developed to connuitre instead if it had had short ŏ in Latin. Is this an accurate summary of what these two forms tell us about the Latin word? Do these reflexes, or those in any other Romance language, suffice to establish the length of the vowel in Latin?

The comparative or etymological evidence is even harder for me to understand. De Vaan says that in Proto-Indo-European, the present would have been a zero-grade form *ǵnh₃-sḱé-, which would have developed in Latin to *gnāscō. The actual form gnōscō is explained by de Vaan 2008 as resulting from the introduction of a full-grade vocalism *ǵneh₃-sḱé-, taken from either the aorist or the perfect. That makes sense, but it's still a bit unclear to me when this substitution is supposed to have taken place (did it affect any other Indo-European languages, or just Latin?), and whether it is an isolated case of substitution or an example of some more generally applicable process of analogy that affected the present-tense forms of other verbs.

A note re: evidence from IE comparanda

PIE *nH > Sanskrit ā, Avestan ā, Latin nā, etc. but Greek nē/ā/ō (Beekes 2011: 151).

Some of the relevant IE cognates are Greek γιγνώσκω, OPers. xšnāsāhiy, and Sanskrit jānā́ti; however, only PIE *nh₃ > Greek nō.

Weiss 2009/2011: PIE *R̥H_iC > *RĒ_iC

In Greek: *CR̥h₃C > CRώC

cf. PIE *ǵnh₃-sk̂é- Greek γιγνώσκω

"A sequence of a syllabic liquid or nasal followed by a laryngeal becomes the non-syllabic variant of the liquid or nasal followed by a long vowel corresponding in quality to the coloring effects of the laryngeal" (p. 52; emphasis mine - Alex B.).

In Italic and Celtic: CR̥HC > CRāC

cf. PIE *ǵnh₃-ské - Latin *gnāsco (the predicted form that got replaced by gnōsco), which would be homonymous with Latin nāsco, nātus and this is how de Vaan - in our case - explains (g)nōsco, ad hoc - analogical leveling from either the perfect *ǵnéh₃- or the aorist *ǵe-ǵnóh₃-.

NB: PIE eh₃ > Latin ō

I believe this explanation is rather standard now, cf.

LIV, p. 170, endnote 14: "Mit R(e), die zur Differenzierung von nāscor 'werde geboren' (s. *ĝenh₁) dient, vgl. Klingenschmitt 68⁹; Rix, l.c."

The zero grade in the present: it's because of the suffix -sk-, which, as Beekes 2011 writes, "always had thematic inflection and the zero grade of the root" (p. 257)

Regardless of the specific details how/why the full grade was used instead of the expected zero-grade in the sk̂é-present in Latin for this particular verb, the fact stands - the root vowel was long, and that's what your OP is about, in its current form.

Since posting the question, I was able to consult Peter Schrijver's "The Reflexes of the Proto-Indo-European Laryngeals in Latin" (1991) (cited by de Vaan), which, along with Alex B.'s answer, has helped me to understand better the etymological arguments in favor of long ō in Latin nōscō.

Like de Vaan 2008 and Beekes 2011 (cited in Alex B.'s answer), Schrijver notes that the present tense would have been expected to be zero grade; various possible origins are proposed for the development of a full-grade vocalism in the present tense. Schrijver says that the Greek forms γνωτός and γιγνώσκω prove that this root should be reconstructed with the laryngeal *h₃ (p. 147). Greek γνω is the regular reflex of PIE zero-grade *ǵnh₃.

My understanding is that unlike in Greek, *h₃ by itself did not have a reflex of o in Latin. But it did "color" an adjacent *e, giving it the quality of o. A laryngeal also lengthened a preceding vowel. Since eh₃ or oh₃ became Latin long ō, a word from a root with h₃ could only wind up with short ŏ in Latin if it developed from a form with a vowel after rather than before the laryngeal: *h₃e/o.

It seems that *h₃e/o is not a plausible candidate for the source of the o in the first syllable of Latin <nosco>: I'm not entirely sure why, but here's what I've gathered from Schrijver.

Schrijver indicates that a *CnHV sequence (or any *CRHV sequence) would show vocalized (syllabic) n as a rule (p. 198). He explains the gnĭt- sequence in cognĭtus as developing from earlier -genot- (< *ǵnh₃et-), with vocalization that was later reversed by syncope of medial unstressed e (p. 202). But that kind of syncope would not be regular word-initially, and so couldn't be used to explain nosco (even if there were some explanation of where an -o/e- between the h₃ and s could have come from).

Schrijver does mention a few cases where word-initial *CRH appears to have yielded Latin CR- clusters, but he suggests that this was the result of early loss of laryngeals that only occurred in this context before a lengthened grade vowel: this condition is used to explain ǵlōs from *glh₂-ōu-s (p. 199) and perfects of the form gnōvī, plēvī, strāvī, flāvī, nēvī, if it is assumed that they derived from the forms *ǵnh₃-ēu-, *plh₁-ēu-, *strh₃-ēu-, *bʰlh₁-ēu-, *snh₁-ēu- (p. 131).

So my current understanding is that there is no vocalism of PIE gnh₃ that would yield Latin word-initial nŏ-, which implies that <nosco> had a long vowel.