Dissertation Oral Presentation

Long-distance Compensatory Lengthening

Speaker
Scott Borgeson
Date
Wed May 18th 2022, 9:30 - 10:45am
Location
Greenberg Room, Margaret Jacks Hall (Room 126)
Zoom: https://stanford.zoom.us/j/92254753804?pwd=ZFVHY00zc1ZpM0xSZ1VKQzkydE01dz09

Abstract

In many languages, the deletion or shortening of one sound in a word is accompanied by the lengthening of another; this is known as compensatory lengthening (CL). Traditionally, CL has been observed to occur only between adjacent sounds, or sounds that are in adjacent syllables. In this thesis, I show that it can also occur between sounds that are separated by theoretically any distance. Evidence comes from two unrelated languages: Slovak and Estonian. In Slovak, the deletion of a vowel anywhere in the word can cause another vowel two syllables away to lengthen. In Estonian, the deletion of vowels from the third or fourth syllable of a word can cause the first syllable to lengthen. Previous accounts of compensatory lengthening assumed this to be impossible; once this assumption is done away with, we are left with no means of differentiating between long-distance CL of this sort and the more widely-attested local CL. As a result, we expect the two to be treated equally in any language. This is contrary to fact: many languages allow local CL but prohibit long-distance CL, but no language displays the inverse.

 

This thesis seeks to fill this gap and provide a means of differentiating between the two types of CL. To do this, I propose an expanded definition of the constraint Linearity (namely MT-Lin for multi-tier linearity), which punishes the inversion of precedence relations between nodes in the prosodic hierarchy, no matter what tier those nodes are on. I assume that CL amounts to the movement of a mora from one segment to another; since MT-Lin punishes the movement of a mora across intervening segments, it serves to enforce locality in CL.

 

When MT-Lin is high-ranking, CL may only be local. When it is low-ranking, other constraints may intervene and force CL to be longer-distance. In Estonian, for example, long vowels and geminates are generally prohibited outside of the initial, primary-stressed syllable. Hence, when a vowel in the fourth syllable of a word is deleted, its mora may not land on any segment in the second or third syllable, but must travel further into the word, to the first syllable. In other words, local CL results in highly marked structures that long-distance CL avoids, and as a result, the latter is preferred.

 

The constraints used in this analysis generate a typology of eight possible grammars. The most restrictive grammar forbids CL entirely. The next allows it but requires it to take place within a single syllable. More permissive grammars allow morae to enter into new syllables, but require them to stay within their original feet. Finally, the most permissive grammars (to which Estonian and Slovak belong) allow CL to take place at any distance. The latter require a very specific ranking of constraints, and hence are correctly predicted to be comparatively rare cross-linguistically. These grammars can also select local CL under certain circumstances; this correctly predicts that any language that tolerates long-distance CL should also tolerate local CL, but not vice-versa.