The Development of Diphthongs in Vedic Sanskrit.

• Author: Ryan, Kevin M.

In Old Indie metrics, syllable weight is conventionally treated as a binary opposition between heavy (guru) and light (laghu). Nevertheless, as Hoenigswald (1989) and Kobayashi (2001: 94-96) observe, the Rgvedic (RV) poets preferentially avoid placing overlong (also known as superheavy or trimoraic) syllables in metrical cadences, making for a third grade of weight that is metrically relevant, albeit not categorically regulated. (1) A superheavy syllable has a rime of the shape VXC,, where VX indicates a long vowel, diphthong, or short vowel plus consonant, and C, indicates at least one consonant. The cadence here refers to the final four or five (see [section]1) positions of the line, excluding the ultima, which is metrically indifferent. In Vedic and elsewhere, the cadence is the strictest part of the line. To be sure, superheavies are not hard to find in cadences, but heavies are significantly less likely to be superheavies in the cadence than elsewhere in the line. Hoenigswald (1988, 1989, 1991) connects this avoidance to other evidence for superheavy markedness in Vedic and adduces comparative support. (2)

This article corroborates superheavy avoidance in cadences, going beyond Hoenigswald and Kobayashi by using statistical models that control for possible confounds such as word boundary distribution. Aside from drawing additional support for cadential avoidance in the general case, the article pursues three new directions. First, it examines whether different types of superheavies are treated differently, finding, for one, that ultraheavies (VX[C.sub.2]) are avoided even more stringently than VXC superheavies. Second, it exploits avoidance as a diagnostic to determine whether particular sound changes had occurred at the time of composition of the text. Finally, it considers implications of these results for the proposed stratifications of the RV.

Old Indie has four diphthongs, transcribed e, o, ai, and au. All are metrically heavy, including e and o, which are usually transcribed without macrons because they lack short counterparts. Uncontroversially, the original (say, early Vedic or pre-Vedic) pronunciations were e [ai], o [au], ai [ai], and au [ay] (Wackernagel 1896: 35-41; Whitney 1889: 12; Allen 1953: 62-64). This is clear from both external (comparative) and internal evidence, such as sandhi (e.g., a plus i yields e). Nevertheless, the received values of the vowels for both Sanskrit and Vedic are e [e], o [o], ai [ai], and au [au]. These latter realizations, which remain standard to date, were set forth in even the earliest phonetic descriptions, the Pratisakhyas. (3) These outcomes entail two sound changes, as in Figure 1. First, e and o flatten, eventually to monophthongs. Second, the cores (low vowel components) of ai and au shorten. Flattening of some degree must precede shortening, lest the pairs merge.

The question remains as to when the changes occurred. Their completion by the time of the Pratisakhyas, as well as their confirmation by Greek loans (ca. 300 BCE; Wackernagel 1896: 35), merely supply a terminus ante quem: the changes must have been complete by the end of the Vedic period. In principle, they could have occurred any time within that period, or even before it (after Proto-Indo-Iranian). (4) This article proposes that the changes took place at a relatively early date, but not so early as to predate the Vedic corpus, namely, midway through the compositional period of the RV. Based on superheavy avoidance in cadences, ai and au are shown to transition from trimoraic to bimoraic within the RV. Meanwhile, e and o must have at least partially flattened by the later strata of the RV. That is, e (and o, mutatis mutandis) might have flattened either fully, to the monophthong [e], or partially, to a minor diphthong such as [ei] or [aej. Either way, it would not have retained its original value as a major diphthong [ai].

1. SUPERHEAVY AVOIDANCE IN THE RV

   The sharp drop in the incidence of superheavies from the pre-cadence to the cadence is depicted for the RV in Figure 2. The figure is divided into three panels for three metrical line ipada) types, namely, dimeter (8-syllable) and trimeter (11- or 12-syllable), hereafter simply 8, 11, and 12. Lines of other lengths, which are infrequent in the RV, are excluded, as are lines in special meters, (5) lines from the Valakhilya, and repeated lines after the first instance. The remaining 31,766 lines compose the normal meter corpus used in this article.

   Figure 2 shows the percentage of heavies that are superheavy in each position. (6) Cadences are indicated by solid circles and pre-cadences by empty circles. To a First approximation, the cadence is the final four positions of the line, ignoring the weight of the ultima. In 12, however, the cadence extends to the fifth position from the end, that is, 12-8, as is clear from Figure 2 as well as from the general strictness of 12-8, which is filled by a heavy 97% of...