Tuesday, September 9, 2014


P. J. Barber, Sievers' Law and the History of Semivowel Syllabicity in Indo-European and Ancient Greek. Oxford classical monographs. Oxford; New York: Oxford University Press, 2013. Pp. xv, 437. ISBN 9780199680504. $185.00.

Reviewed by Götz Keydana, Georg-August-Universität Göttingen (gkeydan@gwdg.de)

Version at BMCR home site


Eduard Sievers proposed his eponymous law in 1878. It has been disputed ever since, although the core facts have never been questioned: In Vedic Sanskrit, forms containing an etymological semivowel /i̯/ show an epenthetic homorganic vowel if /i̯/ is preceded by what Sievers called a heavy syllable. Thus, we find martiya- 'human' besides avya- 'from sheep', both with a PIE suffix *-i̯o-. Similarly, in Germanic, forms like Gothic g.sg. harjis 'army' alternate with forms like g.sg. hairdeis 'herdsman' (with <ei> denoting [iː]), both from *-i̯e-so. However, the exact conditions of this alternation are a matter of debate. Is Sievers' Law a sound law, or rather a synchronic repair for illicit phonological structure? What is the trigger for the epenthesis, prosody or phonotactics? Does the law target */i̯/ or any resonant? And finally, is the law of PIE origin or rather an independent development in some subphyla? In his meticulous work on Sievers' Law in Indo-European and Ancient Greek, Peter Barber tries to give answers to all these questions, focussing specifically on the last one: If the law could be shown to have been active in Greek, this would strongly suggest its PIE origin.

In chapter 2 of his book, Barber discusses the evidence from Germanic and Vedic, viz. the two subphyla where the law was still active in historical times. Turning to Germanic first, Barber surveys the relevant data, and concludes that the law is best described "as a set of inherited synchronic rules or constraints operating on syllable structure, or higher metrical structures" (p.17). Consequently, he goes on to present various attempts to model the law as part of a synchronic grammar. This overview, however, lacks rigour. The author does not even try to compare the different approaches, let alone assess their descriptive adequacy. Rather, he contents himself with the vague conclusion that "they all seem to rely, to one degree or another, on the notion that a Cj-onset was somehow disfavoured and liable to become Ci-" (p.20). This leads him to assume a constraint blocking Cj-, whose importance for word-internal clusters the author ponders also for PIE (p.387). Barber tries to strengthen this assumption by surveying Germanic word-edges and Gothic writing conventions. However, he has to admit that "[t]he absence of words with initial *Cj- remains a puzzle" (p.22), and although "there is something special about the orthographical treatment of word-internal j" (p.21), the evidence is inconclusive. This further strengthens the impression that the constraint is completely ad hoc and that the author should have looked for alternatives. A promising one was offered by Byrd (2010), who in his formulation of the law relies on well-founded constraints only. Byrd's important contribution, however, is only mentioned cursorily in footnotes.

The second part of chapter 2 is devoted to Vedic. Again, the relevant data are presented in a very careful and illustrative way. The classical evidence for Sievers' Law comes from metrics, where a word like martyaḥ 'man' regularly counts as trisyllabic (martiyaḥ). Barber, however, argues that there is more evidence for the law in Vedic (p.26). Present stems of verbs of the so-called 5th class are formed by adding to the root a suffix -nu-, -nv-, -nuv- (in the zero grade). The variant -nuv- (instead of -nv-) occurs only when the preceding verbal root ends in a consonant. Thus, we get su-nv-ánti 'they press', but aś-nuv-ánti 'they obtain'. At first glance, this does indeed look like a Sievers' Law phenomenon. However, the reason for the epenthesis of /u/ in aś-nuv-ánti is most probably morphological transparency: Without the epenthesis, the /n/ would become syllabic and change to /a/, yielding an opaque form aśavánti. As with the -n-infix (cf. Keydana 2008, Byrd 2010), Vedic avoids this problem by keeping the nasal consonantal, in this case by epenthesis. Forms like aśnuvánti not being pertinent, the domain of the law narrows down substantially: Pace Barber there is no evidence that it targeted any resonant other than */i̯/ (the examples for Sievers' Law with liquids adduced by Praust 2000 (not quoted by Barber) are highly uncertain).

A vexing problem in the context of Sievers' Law are data like Vedic matsya- 'fish' (never matsiya-). Barber treats the issue extensively in section 2.3.4. His conclusion that the evidence "is confined to Vedic and relies on examples which are unlikely to be older than Indo-Iranian" (p.37) is indisputable. One might even go further and claim that the etymologically isolated matsya- is indeed the only valid example. The compound virapśa- 'wealth', discussed by Baker on p.32, is not pertinent. The preform *vira-pk̑wo- instead of *vira-pk̑uwo- is probably due to the fact that each part of the compound was syllabified separately. *pk̑w-, therefore, formed an onset and was not in the domain of the law. This is even more true if the law only targeted /i̯/ (the same applies to ūrdhvá-, p.32-36). Surprisingly, Barber does not discuss synchronic phonological approaches to Sievers' Law in Vedic. The already mentioned Byrd (2010) would have been worth treating, as also Kobayashi (2004), which Barber does not mention at all.

In section 2.4 Barber extends the view to languages other than Vedic and Germanic and to phenomena related to but not identical with Sievers' Law. This section is quite superficial. Subphyla like Balto-Slavic are not treated at all.

The final part of the second chapter is devoted to Edgerton's extension of Sievers' Law to word-initial sequences and to Lindeman's Law, restricting the application of Edgerton's extension to monosyllabic words. He convincingly refutes Schindler's rule (reproduced incorrectly on p.49), an attempt to unify Sievers' Law and Lindeman's Law by confining the domain to the last syllable. He then goes on to examine the Early Vedic data pertinent to Lindeman's Law in greater detail. The conclusion from this important glimpse into the monosyllabicity criterion is sobering: "[T]here is a limit on what we can know, and we ought to exercise caution in projecting potentially superficial features of the Rigveda on to Indo-European phonological rules" (p.64).

The interim résumé of the first part is thus that there are "rule-governed word-internal syllabicity alternations in Germanic and Vedic" (p.65). We might add that his observations make it plausible that even in Vedic the law did not operate word-initially. Against Barber we might further add that in both branches of IE only /i̯/ was affected (see above). In other words, the difference boils down to the trigger: In both branches it was prosodic, based on foot structure in Germanic, syllable structure in Vedic. Adding the residues of the law found in other IE languages, we get a neat picture of a typical life cycle of a phonological rule (Bermúdez-Otero and Trousdale 2012). In PIE it started out being post- lexical (see Byrd 2010) – this is the state still observable in Vedic. In Germanic, the rule became lexical (leading to different stem types). In the other IE languages, it vanished altogether, leaving some fossilized traces in individual words. Barber's conclusion, that "substantial evidence from an independent source, such as Greek, becomes very desirable" (p.65), though still true, is thus less pressing than the author thinks.

The main part of the book is devoted to Greek evidence for Sievers' Law. As it turns out, however, Greek cannot fulfill Barber's expectations. Undisputed evidence for Sievers' Law in Greek comes from comparatives like κρέσσων (ion.) 'stronger' besides κερδίων 'more profitable'. Other data are hard to come by, since both */i̯/ and */u̯/ underwent massive change in the prehistory of Greek. In his book, Barber concentrates on */i̯/, the reason being that "as *y was lost at such an early stage in Greek […] the partially fossilized reflexes of *y are more likely to preserve traces of the inherited situation than w" (p.68). He gives a detailed survey of sound changes related to */i̯/, concluding that the law ceased being operative at an early stage of Proto-Greek. The only (possibly) telling data are thus morphological categories of PIE age containing morpheme-initial */i̯/. These are comparatives, nominals in *-i̯e/o- and verbs in *-i̯e/o-. Barber's treatment of these categories is impressive. His survey of the data is as exhaustive as it is detailed. The author distinguishes carefully between sure and doubtful tokens and between old formations (which typically fit the Sievers pattern) and newer ones (which do not). However, with respect to the question of Sievers' Law in Greek the results of this effort are quite disappointing. Barber shows clearly that there are no traces of Sievers' alternations word-initially, the few attestations of *Cii̯V- in Greek belonging to roots where /i/ is attested elsewhere in the paradigm. Analogy, thus, is a viable scenario (p.140), e.g. in the root aorist ἐβίων < *e-gwiyeh3-m, which may have been influenced by the 3.pl. *ἔβιον < *e-gwih3-ont (p.133). Likewise, "there is no meaningful Greek evidence for polysyllabic stems preceding postconsonantal prevocalic *y" (p.385). Also, the matsya-type of Vedic finds no equivalent in Greek. As the author shows, it is unlikely that Ionic διξός continues *dikhth-i̯os. Rather, it is probably a recent formation with the suffix *-so- (p.201). αὔξομαι (p.288) and ἀνάσσω (p.319), too, cannot be shown to contain a suffix *-i̯e/o-. The data confirm Barber's assumption that Sievers' Law ceased to be active at an early stage of the development of Greek. Thus, Greek more or less confirms the picture gained from Germanic and Vedic, but it cannot contribute anything new – certainly a meagre result for such an impressive and involved study. This strange imbalance is particularly obvious in the chapter on verbs in *-i̯e/o-. After a long detour on Vedic verbs in -ya-, it culminates in the observation that these verbs cannot contribute to the problem of Sievers' Law at all since all ancient formations are based on zero grade roots, which, due to their phonotactics, block Sievers alternations in principle.

The book ends with a useful summary of the main findings. Barber concludes that Sievers' Law phenomena are undoubtedly attested in Germanic, Indo-Iranian, and Greek. As similar patterns do not occur outside of IE, independent developments in these subphyla are improbable. Thus, "it seems reasonable to take the evidence for Sievers' Law in Greek, Germanic, and Indo-Iranian as a basis for attributing such a rule to their immediate common ancestor" (388). Barber, however, gives no hint as to the exact formulation of that "rule".

Barber's book is an impressive achievement. His etymological analyses of the Greek data are highly valuable and well-balanced. They are indispensable for everybody working on Greek etymology. However, concerning its main goal, the book is slightly disappointing, as it basically confirms the long-standing assessment that Greek cannot contribute anything substantial to our understanding of Sievers' Law in PIE.1


1.   Bibliography:

Bermúdez-Otero, Ricardo and Graeme Trousdale (2012). Cycles and continua: on unidirectionality and gradualness in language change. In: Nevalainen, Terttu (Hrsg.) and Elisabeth C. Traugott (eds.): Handbook of the History of English: rethinking and extending approaches and methods. Oxford : Oxford UP, 619-720
Byrd, Andrew Miles (2010). Motivating Sievers' Law. In Jamison, Stephanie W.; H. Craig Melchert and Brent Vine (eds.), Proceedings of the 21st Annual UCLA Indo-European Conference, Los Angeles, October 30th and 31st, 2009. Bremen: Hempen. 45-67.
Keydana, Götz (2008). Hohe Sonorität in der Koda: Indogermanische Lösungen für ein phonetisches Problem. Historische Sprachforschung 121:54-64. Kobayashi, Masato (2004). Historical Phonology of Old Indo-Aryan Consonants. Tokyo: Research Institute for Languages and Cultures of Asia and Africa, Tokyo University of Foreign Studies.
Praust, Karl (2000). Altindisch dr̥-/dr̥̄-: seṭ oder aniṭ? In Forssman, Bernhard and Robert Plath (eds.), Indoarisch, Iranisch und die Indogermanistik: Arbeitstagung der Indogermanischen Gesellschaft vom 2. bis 5. Oktober 1997 in Erlangen. Wiesbaden: Reichert. 425-441.

No comments:

Post a Comment