Locality in Vowel Harmony

3. Contrastiveness, Markedness, and Feature-Based Locality

3 Contrastiveness, Markedness, and Feature-Based Locality The Search procedure introduced in chapter 2 finds and copies needed feature-values from the closest element in the search domain. An important source of crosslinguistic variation in delimiting the search domain results from language-specific reference to two paradigmatic properties defined by the inventory of sounds in the language. Once these two properties are defined, di¤erences between languages reduce to use of the same Search algorithm over di¤erent sets of elements; these di¤erences can be predicted by inspecting given languages’ featural alphabets. This chapter will explore crosslinguistic variation in what is excluded from the search domain and will conclude by summarizing how the range of ‘‘transparency’’ e¤ects in harmony is limited by the grammatical options made available within a relativized-search-domain approach to feature-copying rules. 3.1 Contrastiveness Cuts Potential Donors in Finnish Harmony We begin by examining Finnish, which has a [Gback] harmony system similar to that observed for Turkish in chapter 2. The essive case su‰x alternates between [-na] (with [þback] vowel) and [-nä] (with [
back] vowel) depending on the value of the vowel to its left. However, not all vowels count: in particular, the high front vowel /i/ and the mid front vowel /e/ act as if they were not there (Ringen 1975). To illustrate with /i/: (1) The high front vowel /i/ is invisible in Finnish harmony a. pöütä-nä ‘table-ess’ b. pouta-na ‘fine.weather-ess’ c. koti-na ‘home-ess’ d. pappi-na ‘priest-ess’ e. väkkärä-nä ‘pinwheel-ess’ f. makkara-na ‘sausage-ess’ g. tühmä-nä ‘stupid-ess’ h. tuhma-na ‘naughty-ess’ As examples (1a–b) show, the [
back] su‰x [-nä] follows a root that is all [
back], and the [þback] su‰x [-na] follows a root that is all [þback]. However, (1c–d) show that when the closest leftward vowel to the su‰x is /i/, it is skipped, and the search continues leftward to the next vowel, which in these cases is [þback], eventually yielding the [þback] variant of the su‰x. What’s going on here? Shouldn’t the search halt immediately with the closest vowel in the domain? To answer this puzzle, we return to one of the central points raised in chapter 1: that humans often do not compute distance ‘‘as the crow flies,’’ but rather within a system. Let’s examine the properties of the Finnish vowel inventory. In the table in (2), I indicate ‘‘missing’’ vowels with 7.1 (2) Finnish vowel inventory [
back,
rd] [
back, þrd] [þback, þrd] [þback,
rd] i ü u 7 [þhigh,
low] e ö o 7 [
high,
low] ä a [
high, þlow] As (2) shows, there are no [
low, þback,
round] vowels in the inventory , and thus, while /ü,ö,ä/ and /u,o,a/ di¤er only in [Gback], the [
low,
back,
round] vowels /i,e/ have no harmonic counterpart. The notion of having or not having a harmonic counterpart can be formalized once we refer to features.2 (3) Definition of contrastive A segment S with specification aF in position P is contrastive for F if there is another segment S0 in the inventory that can occur in P and is featurally identical to S, except that it is
aF. This definition determines whether an element has a ‘‘twin’’ with respect to a certain feature-value. We are observing that a paradigmatic property of a vowel (whether it is contrastive or not) determines its syntagmatic behavior (whether it participates in vowel harmony or not). An easy mnemonic for this pattern might be ‘‘If you don’t have a twin, you can’t be a donor.’’ Since the vowels /i,e/ are noncontrastive for [Gback] in Finnish, they pattern as invisible in [Gback] harmony, not acting as donors and not 70 Chapter 3 being included in the search. To put it di¤erently, whether or not a vowel is contrastive for a harmonic feature determines whether or not it is included at all in what the search looks at. The set of items that are potentially included in the search (based on, for example, whether they are contrastive or not) constitutes the domain of search. The modification of the Search algorithm to exclude noncontrastive items as potential donors is shown in figure 3.1. Relativization prunes the...