Potential indicators of [+ATR] markedness
259 why a [+ATR] dominance analysis would be difficult. Though I do not attempt to solve
this problem, I do make some general observations and suggestions to point us in a direction which seems promising for a comprehensive formal account of these patterns.
In the following subsections, I summarize the patterns which seem to indicate [+ATR] markedness and then the patterns which are indicative of [-ATR] markedness. In
§1.4.1 of the introduction, I discussed a number of the most well-accepted criteria for determining which features and segments in a language are marked, such as asymmetric
assimilations, resistence to spreading, distributional restrictions, etc. I have also discussed the correlation between markedness and dominance. Recall that the marked features in a
language are the ones which are also generally dominant in the language. In other words, segments with the most restricted distribution are usually the dominant, active, spreading
ones, and these are also usually resistant to alteration. In the following discussion, some indicators are more often associated with “markedness” while others are more associated
with “dominance,” but it is helpful to remember that the marked value is usually the same as the dominant one.
7.2.1 Potential indicators of [+ATR] markedness
I begin by summarizing all of the patterns in Ikoma which are suggestive of [+ATR] markedness, shown in 212 below. Arguments and extended discussion of each of these
points have been presented in the previous chapters, so I only briefly summarize some of the central arguments here.
212 Potential indicators of [+ATR] as the dominantmarked value
a. Leftward [+ATR] spreading to from [i u] in noun stems
b. Leftward [+ATR] spreading to roots from high-vowel verbal suffixes
c. Leftward [+ATR] spreading to roots from subjunctive -e suffix
d. Rightward [+ATR] spreading from e o verb roots to the applicative suffix
260 Each of the points above refers to [+ATR] spreading, which is discussed in detail in §4.2
above noun stems and Chapter 6 above verb stems, in which I make arguments for categorical assimilation of in nouns but only in verbs.
For the sake of simplicity in the current discussion, I disregard [+ATR] spreading in nominal stems. As I discuss in §7.3 below, it is likely that Ikoma previously had a
7VH inventory with [+ATR] dominance, which could be the historical origin of the patterns of [+ATR] dominance still found in present-day Ikoma. It is of course difficult to
know for sure, but it is certainly a possibility that the [+ATR] spreading in noun stems is the result of a historical process which has been lexicalized, especially since the patterns
are symmetric in nouns but asymmetric in verbs. Points b and c above both refer to cases of apparent leftward [+ATR]
spreading in verbs. In Chapter 6 I described five different verbal suffixes which cause the [-ATR] root vowel to be realized as [e]. Though the process is regular and widely
occurring, an intriguing point to notice is that all instances of leftward spreading are triggered by a vowel at the right edge of the word.
88
Though the spreading seems to be very broad and general, it is actually quite easy to define the specific environment in
which it occurs. The subjunctive is a final vowel and thus always at the end of the word. Similarly, all of the high vowels which cause root vowel alternations are always final in
the word.
89
The identification of this word-final environment could have some real implications for our understanding of which [ATR] value is most systematically,
88
I am grateful to Myles Leitch for noticing this pattern.
89
Recall from §6.3 that there is one potential example of the inversive suffix with an root, but it is very hard to know for sure about the underlying vowel of the root. I choose to disregard this potential example
for now for lack of data. If we find in the future that it does in fact spread [+ATR] leftward, then the condition for [+ATR] spreading must be broadened, since the inversive does not occur at the end of the
word.
261 generally dominant. For example, we could argue that the existence of this well-defined
environment is suggestive that [+ATR] could be the more limited and therefore not systematic dominant [ATR] value, and the apparent [+ATR] dominance could be the
result of some sort of edge effect. Concerning the [+ATR] spreading to the applicative suffix, recall that this is the
only instance of rightward [+ATR] spreading in the language. Also, [+ATR] spreads only from e o in the root, not from i u, which suggests that the process might not necessarily
be motivated by preservation of [+ATR], but instead by a need for adjacent mid-vowels to be harmonic. In the discussion in §7.1.3 above, I outline a dominance reversal account
of this pattern, based on an analysis in which [-ATR] is the systematically marked value in Ikoma. I argue that the applicative is [-ATR] underlyingly and that it is subject to
harmony constraints which require mid-vowel sequences to be harmonic. It is very unusual for dominant [-ATR] suffixes to spread leftward to [+ATR] root vowels, so
harmony is repaired by spreading [+ATR] from the root to the suffix. Alternatively, the apparent [+ATR] spreading could be understood as a delinking of [-ATR] from the
suffix. In summary, in Ikoma there are clear processes in which underlyingly [-ATR]
vowels are realized as [+ATR] vowels. Though these patterns are frequent, we have identified a well-defined environment in which they occur. We have also seen that even
though the applicative does not occur word-finally, there are other possible explanations for the apparent [+ATR] spreading. Similar dominance reversal accounts might be found
for the word-final cases as well. This leaves open the possibility that [+ATR] dominance is not necessarily the only option for a unified analysis of Ikoma vowel harmony. Before
262 further evaluating this idea, let us move on to review the potential indicators of [-ATR]
markedness and see whether or not there could be an alternative explanation for any of those patterns.
7.2.2 Potential indicators of [-ATR] markedness