“The Mechanisms of Tone Assignment in Ancient Greek: A New Solution”

I propose a solution to an old problem: how was high tone assigned in Ancient Greek (Attic) to words conventionally described as being “recessive,” the default and most common accentual pattern in the language (Probert 2006)? Linguists have come tantalizingly close, but we still lack a single explanatory system that works perfectly. The contributions of Steriade and Sauzet established that tone assignment is based on trochaic feet constructed right-to-left (Itō & Mester, who posit a purely moraic mechanism, disagree, followed by Revithiadou). Golston made the crucial obvservation that these feet must be moraic rather syllabic. In general, this work collectively relies on making the rightmost trochee prominent and then hypothesizing processes by which tone is shifted to the left to produce the expressions familiar to us from the descriptions of ancient grammarians and medieval markings of accentuation. However, all proposals require at least one one-off rule to cover exceptions, most commonly the accentuation of words such as ἄνθρωπος.

Noting the regularity of rightward tone shift in world languages and inspired by an observation of De Boer about explanations of tone shift in certain dialects of Japanese that are puzzling because they posit mechanisms “leaving no traces of the earlier situation in the form of dialect islands, or irregularities within the dialects,” I hypothesize that the exceptional status of words like ἄνθρωπος constitute precisely such surviving irregularities in Greek. In other words, this situation is the result of processes that have shifted tone rightward in most cases but left a few instances of the original mechanisms of tone assignment before the shift occurred. Rather than try to explain why tone shifts leftward in a few metrical-tonal shapes, I explain why tone shifts rightward in most of them. Such conditioned shift can be found in Japanese dialects of the Izumo area, where they are induced by vowel height, including some dialects in which the high tone shifts when the first syllable has a close vowel and the second an open one, and some others where a close vowel in the second no longer has the ability to block the shift (De Boer and Loukareas).

Cross-linguistically, rightward shift is common in tone and pitch languages. Compare Bantu languages; see the brief overview in Odden, who observes, “a number of languages have rules of tone shifting. The most common form of shift is rightward shift by one syllable.” In Greek a general rightward shift blocked in some environments seems more economical and likely than the current proposals.

There is no room here to detail the complete argumentation, but under certain perfectly consistent conditions that I will outline, we can assume a generalized rightward tone shift in Greek that is blocked by a single condition—a following footed or heavy syllable—but not blocked elsewhere. This explains how ἄνθρωπος, with the heavy, footed syllable ‑θρω‑ following a tone assigned to the fourth mora, does not undergo tone shift and express as *ἀνθρῶπος. Modifications to earlier assumptions allow all accentual expressions to be accounted for with similar consistency.