Remerciez-le!

Remerciez @Admin pour avoir partagé cet document gratuitement, de la manière la plus simple, en partageant sur les réseaux sociaux.

INDO-EUROPEAN ACCENT AND ABLAUT Edited by Götz Keydana Paul Widmer and Thomas O

INDO-EUROPEAN ACCENT AND ABLAUT Edited by Götz Keydana Paul Widmer and Thomas Olander Museum Tusculanum Press University of Copenhagen 2013 © Museum Tusculanum Press and the author(s) 2013 Indo-European Accent and Ablaut © Museum Tusculanum Press and the authors 2013 Edited by Götz Keydana, Paul Widmer & Thomas Olander Cover design by Thora Fisker Set by Thomas Olander Printed in Denmark by Tarm Bogtryk A/S ISBN 978 87 635 4043 8 Copenhagen Studies in Indo-European, vol. 5 ISSN 1399 5308 Published with support from: Roots of Europe – Language, Culture, and Migrations Museum Tusculanum Press Birketinget 6 DK 2300 Copenhagen S www.mtp.dk © Museum Tusculanum Press and the author(s) 2013 Metrical grid theory, internal derivation, and the reconstruction of PIE nominal accent paradigms Ronald I. Kim Adam Mickiewicz University, Poznań This paper applies “metrical grid theory” as developed by Halle and Idsardi to the standardly reconstructed accent-ablaut paradigms of PIE. It is shown that root nouns can be straightforwardly mod- eled, as well as acrostatic and amphikinetic inflection: the strong and weak case endings are underlyingly unaccented and accented, respec- tively, and the root is unaccented in root nouns of the standard type (e.g. *h2ént- ~ *h2n"t-´ ‘front, face’) and amphikinetic paradigms (e.g. *d#ég$#-om- ~ *d#g$#-m-´ ‘earth’), but accented in acrostatic paradigms (e.g. *pód- ~ *péd- ‘foot’ , *wód-r" ~ *wéd-n"- ‘water’). Hysterokinetic and neuter proterokinetic paradigms can also be modeled without dif- ficulty, the latter by assuming a dominant unaccented nom./acc. sg. ending *-∅, while animate proterokinetic paradigms require the as- sumption of one or more ad hoc rules. These analyses are then con- sidered in the context of internal derivation, where it is argued that amphikinetic inflection represents the default type, derived by the addition of a dominant zero suffix. The remaining types of internal derivation are reviewed in turn, followed by a conclusion summariz- ing the remaining difficulties and issues for future research.1 The “kinetic revolution” in Indo-European studies during the 1960s and ’70s radically altered our understanding of nominal and, to a lesser extent, verbal inflection in Proto-Indo-European and the oldest Indo-European languages. 1 I thank Götz Keydana and Paul Widmer for kindly inviting me to take part in the Göttingen Workshop on Indo-European Accentology, and all the participants there for their comments on the oral version of this paper. Special thanks also go to Craig Melchert for reading a final draft and making numerous helpful sugges- tions. All errors and opinions remain entirely my responsibility. © Museum Tusculanum Press and the author(s) 2013 Ronald I. Kim 64 Following upon the pioneering insights of Pedersen (1926, 1933) and Kui- per (1942), scholars such as Hoffmann, Narten, and Schindler reconstructed several contrasting accent and ablaut classes for the protolanguage; their conclusions are now generally accepted in one form or another by all main- stream Indo-Europeanists. Only in recent years, however, have scholars seri- ously attempted to model these accentual paradigms in light of advances in contemporary phonological theory. In particular, the prosodic framework developed by Idsardi and Halle has been successfully applied to Russian and other modern Slavic and Baltic languages with lexically contrastive stress, whose stress systems are shown to operate according to certain universal rules and language-specific parametric settings. This “metrical grid theory” has encountered a number of difficulties in accounting for the accentual paradigms posited for PIE, calling into question the validity both of the theoretical approach as well as of the reconstructed paradigms themselves. The following paper does not seek to address all the problems raised by the reconstruction of PIE accent and ablaut classes, par- ticularly such fundamental issues as the prehistoric relation of ablaut and stress, or the relative strength or weakness of the evidence for individual paradigms. Instead, it aims to investigate to what extent metrical grid theory is able to model the main inflectional types, namely acrostatic, amphikinetic, proterokinetic, and hysterokinetic, and their relation to one other through processes of internal derivation. 1 Preliminaries 1.1 The PIE accentual system The basic principles of the PIE accentual system have been known since the late 19th century. With the exception of pro- and enclitics, all PIE word forms were stressable and could carry the stress on at most one syllable. In contrast to most modern IE languages, PIE had lexically contrastive accent, i.e. the accentual properties of morphemes were not predictable on the basis of their phonetic shape alone (e.g. vowel length or syllable weight), but had to be learned individually. Lexical accent is reflected in Hittite, Vedic, an- cient (and modern) Greek, and many Baltic and Slavic languages, including © Museum Tusculanum Press and the author(s) 2013 Metrical grid theory, internal derivation, and PIE nominal accent 65 Lithuanian, Russian, and Serbo-Croatian; in addition, the effects of Verner’s Law presuppose lexical accent for an earlier stage of Proto-Germanic.2 The phonetic properties of PIE stress and the prehistory of the PIE accent and ablaut system are not strictly speaking recoverable by the comparative method, but must be inferred through internal reconstruction and typologi- cal arguments. According to the traditional view (going back at least to Ped- ersen 1926), (pre-)PIE was marked by dynamic accent or “stress accent” , and the distribution of full and zero-grade vocalism (i.e. *e vs. *∅) was condi- tioned by the presence vs. absence of surface stress. This stress accent was lat- er transformed into pitch accent, which based on the descriptions of ancient grammarians was present in Vedic (as udātta) and ancient Greek; this pitch accent then gave way to a stress accent in later stages, e.g. modern Greek and the majority of present-day IE languages, including English. These assumptions have been called into question by modern phonet- ics, which has shown that so-called “pitch accent languages” also exhibit differential levels of stress intensity (loudness, length, etc.) on accented vs. unaccented syllables, and conversely that “stress accent languages” do show systematic differences in pitch frequency on syllables with primary stress, secondary stress, or no stress. Furthermore, the almost universally held be- lief that the ablaut alternation between full-grade *e and zero-grade *∅ was conditioned by stress in pre-PIE, i.e. that underlying *e was syncopated in unstressed syllables, has recently been called into question;3 and the prehis- tory of qualitative ablaut alternations (*e ~ *o) remains for the most part un- clear.4 In any case, the secure reconstruction of forms such as *sept' ‘seven’ or pres. act. 2pl. *bhér-e-te ‘you carry’ for the earliest recoverable stage of PIE, with stressed zero grade viz. unstressed full grades, makes it clear that ab- laut was no longer purely allophonic. The ablaut alternations reconstructible for the protolanguage had probably long since become morphologized, i.e. reinterpreted as associated with specific morphological categories and para- digms, such as the ones to be examined below in §3. 2 The Iranian languages and Tocharian B have also arguably preserved traces of PIE lexical accent: see Mayrhofer 1989: 13 with refs., Ringe 1987. 3 See e.g. Keydana 2005: 22–5, 39–43 and passim; Kiparsky 2010: 158–9. 4 For competing views on the origin of o-grade vocalism in general, see Beekes 1985: 157–8, 1995: 166–7, Rasmussen 1989: 123–262, Szemerényi 1990: 124–7, 1996: 119–21 with refs. © Museum Tusculanum Press and the author(s) 2013 Ronald I. Kim 66 1.2 The development of metrical theory In the earliest version of generative phonology (Chomsky and Halle 1968), stress was treated as a distinctive feature of particular segments, analogous to e.g. [long], [back], or [nasal]. However, Liberman (1975) pointed out that stress alternations such as English phótogràph vs. photógrapher vs. phò- tográphic or sixtéen, Japanése, bambóo vs. síxteen Jápanese bámboo tábles would require a complex series of otherwise unmotivated rule derivations, and proposed instead that stress is the phonetic realisation of underlying phonological structure, specifically of groupings of segments into larger units, or “feet” . This view is now universally accepted and has given rise to various theoretical models to account for the wide variety of prosodic sys- tems in the world’s languages. Probably the most influential model of stress computation in the pho- nological literature has been that of Hayes (1985, 1995). In line with the core assumption of generative grammar, he argued that stress patterns reflect dif- ferent parsings of phonological structure into (usually binary) feet, either syllabic or moraic, and a finite number of parametric settings, e.g. to assign prominence to the left or right foot edge (i.e. build left- or right-headed feet), iterate foot building (from left to right, or right to left), or exceptionally skip segments at the edge of a prosodic domain (extrametricality). This frame- work successfully generates most of the stress patterns commonly found in the world’s languages, including some of great surface complexity (e.g. the famous Axininca Campa; see Hayes 1995: 288–96, McCarthy and Prince 1993: 159–70). However, it does not directly deal with lexical accent, which is found in many ancient and modern IE languages and reconstructed for PIE (§1.1), and also occurs in several non-IE languages.5 5 For alternative analyses to the one adopted below, both emerging from the hey- day of uploads/Litterature/ metrical-grid-theory-internal-derivation-pdf 1 .pdf