/ #Phonology #Phonetics 

Colloquium 6 (28st May, 10AM): Donca Steriade (MIT)

Harmony and myopia in vocalic plateaux

The sixth colloquium held at the Department of Linguistics, SNU in 2021 is “Harmony and myopia in vocalic plateaux” by Donca Steriade.

The speaker currently works at MIT, USA. She acquired her PhD in Linguistics from MIT in 1982.


I analyze a phenomenon that looks like vowel harmony, but is triggered by a preference to preserve an identity relation (a plateau) between vowels in adjacent syllables. A schematic illustration is this: a rule raises word-final /a/ to [i], e.g. /uba/→[ubi]. When applied to underlying strings of contiguous /a/ syllables, like /abaka/, this rule causes the /a/’s to raise across-the-board to [i]: /abaka/→[ibiki]. Sequences of different heights, like /abi/, remain unchanged. This shows that non-final /a/’s change not because they assimilate to /i/ but to preserve the identity relations between members of the input [a]-plateau.

Across-the-board shifts of this sort are found in surprisingly many systems: Romanian, Javanese, Yakan, Rotuman, Welsh, possibly Telugu, and elsewhere. Comparable ATB tonal shifts are frequent, requiring constraints that protect tonal plateaus. The analysis of all these phenomena invokes constraints that preserve input identity relations. It can be shown that analyses which attribute the ATB behaviors to multiply linked vocalic or tonal autosegments don’t do justice to such patterns.

Having outlined the general analysis of this phenomenon, I take a closer look at the ATB-shift of Romanian. This reveals a pattern that contradicts the belief that harmony is myopic (Wilson 2006, McCarthy 2011), i.e. the idea that harmony applies in local steps, without anticipating non-local consequences of any one iteration. The Romanian process is initiated by local raising [a] to [ə] (e.g. CaCi→CəCi). Raising extends to full strings of [a]’s: e.g. [CaCaCaC-i]→[CəCəCəC-i]. If, however, a position in the chain of [a]’s disallows [ə], only the local raising applies: [aCaCaC-i]→[aCaCəC-i], not _[əCəCəC-i], as initial [ə] is banned, and not _[aCəCəC-i] either. A raised [ə]</a/ does not spread to a preceding [a] unless the entire string of [a]’s can raise to [ə].

The existence of this non-myopic process is a challenge to all analyses of harmony based on the assumptions of Harmonic Serialism (McCarthy 2011). A somewhat similar non-myopic harmony is documented by Stanton 2019, in Gurindji.

I will show that a minor change in our conception of harmony allows us to analyze both apparently myopic processes and clearly far-sighted ATB shifts like that of Romanian.

Due to cyber security concerns, we do not share ZOOM links in the public domain. However, if you want to partipate in this event, please leave 1) your full name, 2) email, 3) name of the event you want to participate, and 4) basic information about yourself via contact menu of this website.