When did Old Latin develop initial stress?

Question

The stress system of Classical Latin is thought to have been preceded by a period of fixed initial stress. When did that earlier system arise, replacing the inherited Proto-Indo-European mobile accent?

The evidence for initial stress is the widespread pattern of vowel reductions in non-initial syllables. These reductions do not occur in the very earliest Latin inscriptions: they seem to have arisen sometime around the middle of the first millennium BC. At the same period, other languages of Italy (Oscan, Umbrian, Etruscan) show similar vowel reductions, so this phenomenon appears to be an example of areal diffusion.

It seems natural to assume that initial stress itself may therefore also have been an areally diffused innovation of the same period. However, initial stress is often reconstructed for Proto-Italic, as for example in Fortson, Indo-European Language and Culture. Is there any actual reason to think so? Do we have any evidence for whether Old Latin initial stress was inherited from Proto-Italic or arose later?

Answer 1

-very brief and disorganized notes (not a full answer), maybe someone else will be willing to write a more coherent answer-

Weiss 2020: 527

"Primary stress on the initial syllable is inferred on the basis of the syncope that affects medial syllables"

also see Lindsay 1894 pp. 157-160 (e.g. word-initial stress in facilius and mulierem in the plays of Plautus and Terrence, cf. Sommer and Pfister 1977: 74 fắcĭlĭŭs, sĕ́quĭmĭnī, cĕ́cĭdĕrō, mŭ́lĭĕrĕm; alliteration, e.g Meiser 1998/2010: 53 Cato Agr. 141,2f. "fruges frumenta vineta virgultaque ... pastores pecuaque salva servassis"); also see Leumann 1977 §243 (pp. 246-248) and esp. Sommer and Pfister 1977 pp. 73-74 for a very informative review.

Clackson and Horrocks 2007/2011 (pp. 47-48 and esp. 92-94)

Latin: The Lapis Satricanus (around 500 BCE) shows unweakened forms, their example is Mamartei, which they claim is supported by the other Latin inscriptions of the 7-6th centuries (they don’t mention which inscriptions though);

Etruscan (not an Italic language!): the first half of the 5th century (cf. Weiss 2020: 118, ft. 17 "around 470 BCE", "beginning at the end of the 7th century", p. 527), their example is Avile (6th century, Weiss 2020: 527) or Avele (6th/5th century, Weiss 2020: 527) in the earlier texts, whereas Avle in the later texts (for Aulus);

Sabellian languages: they write that “In texts written in the Oscan script long vowels are sometimes written with a doubled vowel sign […] found (with one exception) in word-initial syllables, suggesting a maintenance of vowel length under the word accent, but loss elsewhere” (p. 47)

Syncope of short medial vowels – between 6th-4th centuries, their example is Peracis (ca. 500 BC) vs. perkium (from a later Oscan text)

They also discuss an interesting case, destr- in Oscan and Umbrian vs. dextr- in Latin (cf. Greek δεξῐτερός). So they write that “we can know that the change which led to the similarity between Latin dextr- and Sabellian destr- took place when they were separate languages” (p. 48).

also cf. Umbrian mersto-'iustus' vs. Latin modestus (Sommer and Pfister 1977: 73)

Meiser 2017:

“The chronology is unclear; to be sure, it precedes the syncope of word-internal short vowels” (p. 748);

‘An argument for at least a Proto-Sabellic date is the fact that in Oscan as well as in Umbrian the outcome of PI *ẽ (< PIE *n̥) is different, depending on its position within a word” (p. 748),

PI *ẽ (< PIE *n̥) > an (word-initial), en (elsewhere)

Oscan anter, Umbrian ANDER (Latin inter) vs. Umbrian iveka IUENGUA /iwengaf/ acc.pl. ‘heifers’, Latin iuvenca (p. 750).

Weiss 2020:

C2, "probably a diffused trait" (p. 207);

either a development of Proto-Italic or an innovation that "arose in one of the languages of the Italic cultural Koine" which "spread across pre-existing linguistic boundaries" (p. 118, footnote 17).

also see Kurylowicz 1958 (L'accentuation des langues indo-européennes), pp. 381-384, for some interesting observations

also see Nishimura 2008 (a Ph.D. dissertation, UCLA) or Nishimura 2014 and Brent Vine (e.g. Vine 2012); Pultrova