In lieu of an abstract, here is a brief excerpt of the content:

  • Nonmyopic Harmony and the Nature of Derivations
  • Rachel Walker

Unbounded harmony caused by feature spreading is typically myopic (Wilson 2003, 2006). This means that whether spreading proceeds to a neighboring element is not sensitive to whether a segment beyond the neighboring element can undergo harmony. To illustrate, suppose that a language has a regressive harmony for some feature and a sequence of segments [. . . α β γ. . .]. Whether spreading proceeds from γ to β is not sensitive to whether spreading can continue on to α. In other words, the operation of myopic harmony is determined by local factors, not global ones. Although myopia follows from the iterative application of an assimilation rule that spreads a feature to a single adjacent target, studies by Wilson (2003, 2006) and McCarthy (2003, 2004, 2009) have demonstrated that it presents difficulty for the classic framework of Optimality Theory (OT) (Prince and Smolensky 2004) under conventional assumptions about featural representations and the harmony-driving constraints.

This work focuses on a complementary issue. The first point to be established is that nonmyopic bounded harmony exists: patterns where adjacent segments undergo assimilation only when a nonlocal viable target is present for a bounded harmony process. The theoretical argument to be made about these systems is that while they are straightforwardly handled within classic OT, they pose a problem for a proposal to restrict the magnitude of change that can occur in a single step of a derivation, as in certain serialist approaches in OT, such as OT with candidate chains (OT-CC) (McCarthy 2007). [End Page 169]

1 Nonmyopic Harmony in Romance Metaphony

Minor Romance languages or "dialects" spoken in the central Veneto region and on the island of Grado show a type of vowel harmony referred to as metaphony in which a posttonic vowel triggers raising of a preceding mid vowel in a syllable that receives main stress. The particulars of the raising patterns in these regions are chiefly the same. I will demonstrate that metaphony is nonmyopic in these systems, and I have included examples from both regions to strengthen the basis of support for this claim. The data are drawn primarily from Walker 2005, with additional examples from Cortelazzo 1978, Tarlao 1983, and Brunelli 2000.

In stressed syllables, the dialects of central Veneto (CV) and Grado contrast [i, e, ε, a, ɔ, o, u]; in unstressed syllables, only [i, e, a, o, u] occur. In metaphony, a posttonic high vowel causes /e/ and /o/ to raise to [i] and [u], respectively, in a syllable assigned main stress. Examples are provided in (1).1 The trigger vowel in these examples is /i/. These languages lack /u/ in their inflectional system, and no data were found where /u/ occurred in a stem in a context to trigger harmony. Given, however, that /u/ can trigger metaphonic raising in certain other Romance languages (Hualde 1989, Maiden 1991), it is reasonable (but not crucial) to generalize that triggers are high vowels.

(1) Central Veneto
'sock (m sg/pl)'
'sing (1sg/1pl impf. subj.)'
'move (1sg/2sg)'
'button (m sg/pl)'
Grado
'believe (3sg/2sg)'
'blessed (m sg/pl)'
'break (1sg/2sg)'
'tree (m sg/pl dim.)'

In addition to alternations that occur under suffixation, noninflectional posttonic stem vowels can trigger metaphony. This is apparent in [gúm(b)i-o] 'elbow (m sg)' (CV), for which speakers who show variable metaphony have an alternate form [góm(b)i-o], and in [súri+-o] 'mouse (m sg)' (Grado), which is sórcio in Standard Italian.

A key observation for the issues under focus is that the harmony can be initiated by a high vowel in a syllable that is not adjacent to the stressed syllable. This can occur in words with antepenultimate [End Page 170] stress in which the penult and antepenult contain /e/ or /o/. For speakers who show this pattern, vowels in both syllables undergo raising.

(2) Central Veneto
'order (1sg/2sg)'
Grado
'shin (m sg/pl)'
'young man (sg/pl)'

Although a mid vowel in an unstressed penult can raise when the preceding stressed syllable undergoes raising, such vowels do not raise otherwise. Elsewhere (Walker 2005), I establish that metaphony does not affect /ε, a, ɔ/ (e.g., CV: [gát-o]/[gát-i] 'cat...

pdf

Share