|Title:||Assimilation, markedness and inventory structure in tongue root harmony systems|
|Authors:||Roderic F. Casali|
|Abstract:||This paper is concerned with the related topics of assimilatory dominance and markedness relations in tongue root harmony systems. While a range of views has existed on these topics, much work in recent years adopts the view that either [+ATR] or [-ATR] can function as the dominant value in a language, and that markedness relations are determined in conjunction with tongue body height. That is, neither [+ATR] nor [-ATR] values are marked in absolute terms; instead, low [+ATR] vowels (or, in some work, all non-high [+ATR]) vowels and high [-ATR] vowels are assumed to be marked. These assumptions are reasonable ones that are supported not only by familiar functional considerations (gestural antagonism in the production of high retraced and non-high advanced) vowels but by the typological observations that languages (e.g., Yoruba) in which [-ATR] is dominant are strongly attested (as, of course, are those in which [+ATR] is dominant) and that it is the high [-ATR], low [+ATR] and mid [+ATR] vowels that are sometimes missing in languages with an [ATR] contrast.
I present typological evidence in this paper, however, that while such a view has elements of correctness as a kind of average picture of what occurs, it is incomplete in important respects. I argue for a fuller typological picture in which both dominance and markedness relations are strongly correlated with the contrastive vowel inventory of a language, behaving differently in systems with a tongue root contrast in high vowels ("/2IU/ systems") and systems ("/1IU/ systems") with an [ATR] contrast only in mid vowels. Not only do the two systems characteristically show different dominant [ATR] values ([+ATR] and [-ATR] respectively), as claimed in earlier studies (e.g., Casali 2003), but different (and essentially opposite) markedness relations in non-low vowels as well. /2IU/ systems characteristically treat both mid and (more surprisingly) high [-ATR] vowels as unmarked (e.g., in positional neutralization patterns) relative to their [+ATR] counterparts. /1IU/ systems display the opposite relations, treating [-ATR] as marked in both (as expected) high vowels and (more surprisingly) non-high vowels. (Low vowels exhibit some unexpected differences as well, though the problem of making full sense of these is complex.) Thus, in both systems, the marked value in non-low vowels is also the one that is characteristically dominant. It is also noteworthy that there is no single inventory type in which both mid [+ATR] and high [-ATR] vowels characteristically pattern as marked. The vowels [e], [o] behave as marked sounds in /2IU/ systems but unmarked in /1IU/ systems, while the vowels [ɪ], [ʊ] behave as unmarked in the former and marked in the latter.
I argue that the reversals of both markedness and dominance relations that exist in /1IU/ and /2IU/ systems defy easy explanation in functional terms, but that a formal explanation, based on representational underspecification, is more promising: the differences in patterning make reasonable initial sense under the assumption, proposed in various previous studies, that the two systems employ different specified [ATR] values, [-ATR] and [+ATR] respectively. A partial explanation is also speculatively proposed for why the two systems employ different values.
|Area/Keywords:||Phonology, Vowel Harmony, [ATR], Assimilation, Markedness, Contrast, Underspecification, Tongue Root Features|