Myc. tu-wo, Hom. θύος and the vocalism of s-stems in Proto-Indo-European
A neuter s-stem based on PIE *dʰu̯ eh₂- '(produce) smoke' is well-attested from the earliest Greek: Myc. tu-wo /tʰu(u̯ )os/ 'aromatic substance' (incense, perfume), pl. tu-we-a ~ tu-we-a₂ /tʰu(u̯ )eha/, Hom.+ θύος (pl. θύεα) 'burnt offering(s)'. But a “face-value” Transponat PIE nom./acc. *dʰúh₂-os, obl. *dʰuh₂-es- violates the standard Erlangen model of accent-ablaut patterns in the inflection of athematic formations, where an s-stem with zero-grade root is anomalous (cf. [Schindler 1975]): s-stems belong to the “proterokinetic” inflectional type, with ideal structure nom./acc. R(é)-os, obl. R(Ø)-és-. Greek has other s-stems with zero-grade root; but most are transparently secondary, e.g. βάρος 'weight' (Hdt., Aesch.+) ← βαρύς 'heavy' (Hom.+). For θύος, the only available secondary pattern is the deverbative one, i.e., θύος ← θύω 'offer, sacrifice (a victim)' (Hom.+). Nevertheless, Greek deverbative s-stems are otherwise mostly late. Given the early attestation of Myc. tu-wo and Hom. θύος, these forms may continue an inherited s-stem, for which the anomalous zero-grade root vocalism invites further comment. Such a structure is not isolated. Schindler [1975: 264–265] had already identified two probable PIE s-stems with pervasive zero-grade root, i.e. *sríHg-os 'cold, frost' (Gk. ῥῖγος, Lat. frīgus) and *púH-os 'pus' (Gk. πύος, Lat. pūs). And it may be possible to add another such item, i.e. Ved. dúvas- 'gift, offering', better assigned (as *dúH-es-) to 2. deu- '(religiös) verehren etc.' [IEW: 218–219] than to the questionable construct *deh₃u- 'give'.
Recent scholarship has leveled significant criticism against the Erlangen model (the “paradigmatic” approach), with proterokinetic inflection itself as a major target (cf. [Kümmel 2014]), in favor of a new theory of Indo-European accent (the “compositional” approach, cf. [Kiparsky 2010]). Yet this revisionist discussion has had nothing to say about aberrant s-stems with zero-grade root vocalism. What, then, can be said about s-stems like *sríHg-es-, *púH-es-, and (if the analyses here are correct) *dʰúh₂-es- and *dúH-es-?
Beside *sríHg-es-, no full grade is attested for any of the root's nominal or verbal forms. Despite Stüber [2002: 152], there is no reason to regard *sriHg- as a “secondary root” with unspecified origin. Much the same picture appears for *púH-es-; the idea [Stüber 2002: 200] that the pervasive zero grade reflects an early “generalization” based on a primordial ablauting paradigm is merely an article of faith. For the root of *dʰúh₂-es-: full-grade forms are virtually unattested, and the uncertainty about the root shape ([LIV1] *dʰeuh₂- vs. [LIV2] *dʰu̯ eh₂-) is symptomatic. All of this is strongly reminiscent of other PIE “non-apophonic zero-grade roots”, of which the bestknown is *bʰuH- 'be, become', in the classic analysis by Jasanoff [1997]. Cf. also PIE *suH-i̯ u- ~ *suH-nu- 'son', notorious precisely for its failure to conform to expectations about proterokinetic inflection, much as with the zero-grade s-stems. For the underlying verbal root ([LIV2] ?*seu̯ H- 'gebären'), no verbal full grades are attested, and there are no nominal forms with arguably archaic full grades. Additional cases probably exist, e.g. *ḱu̯ eh₁- 'swell' [LIV2: 339–340], with no verbal full grades, symptomatic Schwebeablaut material, and a zero-grade proterokinetic men-stem in Gk. κῦμα 'wave'.
In sum, neuter s-stems with pervasive zero-grade root represent a hitherto neglected contributor to the reevaluation of proterokinetic inflection and, more broadly, the reappraisal of the “Erlangen” (paradigmatic) approach to the reconstruction of accent and ablaut in athematic inflectional categories.