Distributed Morphology Today: Morphemes for Morris Halle

2. Phonological and Morphological Interaction in Proto-Indo-European Accentuation

2 The reconstruction of the grammar of word accentuation in Proto-Indo-European (PIE) has long been a central topic in historical linguistics, as well as the focus of a number of studies within generative phonology of the daughter languages that preserve relicts of the anterior system (Halle and Kiparsky 1977; Halle and Vergnaud 1987). Within traditional historical linguistics a particular reconstruction of PIE accentuation, based principally on the work of Warren Cowgill, Jochem Schindler, and Helmut Rix from the mid-1970s onward (e.g., Schindler 1972, 1975a, 1975b; Rix et al. 2001), has been accepted as a standard working hypothesis in much current research. Evidence for this reconstruction comes partly from the accentual systems of certain attested Indo-European languages (Vedic Sanskrit, Russian, Lithuanian, Greek); reflexes of the pervasive pattern of vowel alternations (traditionally known as grades), which was partly correlated with accent position, especially in Anatolian , Greek, and Indo-Iranian; and the effects of accent position on obstruent voicing in Proto-Germanic (Verner’s law) and on pitch-accent types in the Baltic languages. The Schindler-Rix reconstruction is the basis of the analysis to be presented here; in particular I will follow the recent detailed presentation of Ringe 2006, from which the majority of the data has been taken; see also Fortson 2010 for a basic overview of the reconstruction of PIE and its accentuation. Halle (1997b) presents a preliminary analysis of the Schindler-Rix reconstruction within Simplified Bracketed Grid Theory (SBGT, Idsardi 1992). The present chapter is intended to supplement this line of research and to widen the scope of discussion to include the role of cyclic and noncyclic phonology in producing the accent classes of derived stems in PIE. In addition, I depart from earlier studies in a number of details. First, although “fixed” accent (“acrostatic” and “mesostatic”) stems emerge as a pervasive class in later languages such as Russian and Lithuanian, they are understood here to be relatively marked in comparison with the older stem Phonological and Morphological Interaction in Proto-Indo-European Accentuation Rolf Noyer 22 Chapter 2 type in which the position of accent vacillates between two adjacent syllables (“hysterokinetic” and “proterokinetic”). Second, the underlyingly prespecified metrical structure of PIE morphemes includes not only left boundaries of metrical constituents (“feet”), as in Halle 1997b, but also right-constituent boundaries, which produce a new type of “back-accenting” behavior. Third, following ideas introduced in Halle 1998 in the analysis of English stress, I make crucial use of the premise that certain morphemes may fail to project metrical grid marks, and thus be invisible to metrical computations; this invisibility is distinct from ordinary “extrametricality” in which peripheral elements remain external to metrical constituents, although still metrically visible. 2.1 Halle and Vergnaud’s Basic Accentuation Principle In their surface form, words in Proto-Indo-European (PIE) are hypothesized to have contained a single accented syllable, where accent was phonetically implemented as a high tone—for example, *dh ugh2te ˉ´r ‘daughter’, *ék̂u ˆ os ‘horse’, *snusós ‘daughter-in-law’, *sóh2u ˆ l ˚ ‘sun’. Although the position of accent in a word is often described as “free”—as if entirely an arbitrary property of a word’s inflectional “paradigm”—it is in fact derivable from a combination of the underlying properties of the word’s root and affixes and a system of general phonological rules. Halle and Vergnaud 1987, and later Halle 1997b, propose a Basic Accentuation Principle (BAP) for PIE, based on an analysis of Russian, Vedic Sanskrit, and Lithuanian, which are argued to preserve the essential character of the historically anterior system. (1) Basic Accentuation Principle a. In underlying representation, stems and affixes are either accented or unaccented. b. Leftmost: Stress falls on the leftmost accented syllable if any. c. Default initial: Otherwise stress falls on the initial syllable. In the modern Russian words in (2), for example, an underlyingly unaccented stem such as skovorod- ‘frying pan’ contrasts with an underlyingly accented stem such as kómnat- ‘room’ when an accented suffix such as nom.sg -á is added (underlying accent is indicated by underlining): (2) a. skovorod-á / skóvorod-y ‘frying pan’ nom.sg/gen.sg b. kómnat-a / kómnat-y ‘room’ nom.sg/gen.sg In skovorodá the suffix is the leftmost underlying accent (it is the only one), and thus has surface accent. In the gen.sg skóvorody, however, neither the Phonological and Morphological Interaction in...