Greek -σι- Abstracts and the Reconstruction of Proterokinetic *-tí- in PIE
Although the Greek primary –σι- stems, forming feminine nomina actionis and abstracta e.g. βάσις, βάσεως f. ‘stepping’, generally appear with zero grade vocalism of the root and with recessive accentuation, they are regularly treated as being in their remote past ‘proterokinetic’ (schematically *Ré-ti- ~ Rø-téy- where R = root, e.g. *gwém-ti-s ~ gen.sg. *gwṃ-téy-s). This view is canonized in the most recent historical grammars of Greek and has become standard fare in the Indo-Europeanist literature (cf. Rix 1992:146, more hesitantly Sihler 1995:279). However, the comparative evidence on which this reconstruction is primarily based has been seriously challenged (Kümmel fthcm.). This rethinking of the comparative data urges us to reconsider the prehistory of the Greek -σι- abstracts. I propose that we begin instead from the reconstruction of PIE Rø-tí- (*gwṃtís, *mṇtís etc.), since this accounts best for the descriptive data and I argue that from here we can develop a better account of the Greek (and Vedic) evidence: the change from oxytone –tí- to barytone stems that we see before our eyes in Vedic must have occurred in the prehistory of Greek. To account for this change, I argue that recessive accent developed due to the suffix’s derivational properties.
To explain the loss of inherent accentuation in the -tí- suffix we look to the suffix’s derivational behavior, especially in compounds, where accent was lost in derivation, e.g. PGk. *gwṃtís but *ékgwṃtis > βάσις, ἔκβασις. I argue that such patterning led to a loss of inherent accent of the suffix in Proto-Greek; when this occurred, -ti- abstracts used instead default, recessive accent. Note that the *–tí- suffix was commonly used in compounds (Vine 2004:371), and as demonstrated by Meillet (1925) this strong compositional behavior is reconstructible for Proto-Gk. and PIE based on numerous pairs like βάσις and ἔκβασις, φάτις f. ‘report’ (Hom.+) and πάρφασις (Hom.) etc. The change posited here is interestingly paralleled by the Vedic evidence which, usually held to show synchronic fluctuation in accent, is better taken as a shallow, diachronic change within the Vedas: oxytone type matís f. ‘thought’ (RV) yields to a late Vedic barytone mátis ‘id.’ (ŚB; cf. AiGr. 2.2: pp.631-2; evidence assessed in Lundquist 2014).
As a parallel for the loss of accent in a nominal class due to derivational behavior, Probert (2006:ch.12) demonstrates that beside the inherently accented Caland suffixes –μό-, -νό- we have recessively accented secondary suffixes –ιμο-, -υνο- which lost inherent accent due to their secondary derivation. In a few cases, certain formations show recessive accent due to decompositional derivation (Probert 2006:264n.15 on φῶνος, χείμερος). These cases of decompositional accent illustrate exactly what I am positing for the prehistory of -σι- abstracts in Greek.
Greek and Latin Linguistics