Subjunctive Vowel harmony in verb stems and suffixes

185

6.2 Subjunctive

The subjunctive suffix indicates subjunctive mood, and it occurs in the FV slot of the verbal inflection template. It does not distinguish tense but can refer to both the past and future. Despite the fact that both the subjunctive and the applicative suffixes have front mid-vowel variants that alternate between [e] and [ ], these variants have quite different distributions. Compare the patterns in 151 repeated from 148 above and 152 below. 151 Applicative vowel harmony - r following i, , a, , u -er following e, o 152 Subjunctive vowel harmony -e following i, e, , a, o, u root vowel: e - following With the applicative suffix, [i u a] roots pattern with the [-ATR] suffix [- ]. With the subjunctive, however, the high and low root vowels pattern with the [+ATR] suffix [-e]. I suggest that these differences likely indicate that the underlying [ATR] value of these suffixes is different; the applicative is - whereas the subjunctive is -e. Note also that in the subjunctive there is an asymmetry in the alternations which follow the two [-ATR] mid vowels. Following root vowels, the subjunctive is [-ATR], showing that the suffix harmonizes with the root. However, following , the subjunctive suffix remains [+ATR] and even seems to spread [+ATR] leftward to the root vowel. Examples of subjunctive verbs are presented in 153 below. The same seven verb roots which were used for the applicative examples are also used here and throughout the rest of this chapter to illustrate harmony patterns. 186 153 Subjunctive examples N N N N K K K K -; . -; . -; . -; . ST 2 ST 2 ST 2 ST 2 9 2 : V X V X V X V X ` . : 8 8 9 8 As the examples show, [- ] occurs only following . When the subjunctive suffix is combined with root vowels, as in [tu- e -e], the vowel of the underlyingly [-ATR] root is realized as [+ATR]. Also note that a does not induce any sort of harmony in either in the applicative or the subjunctive suffix, even though it patterns with and as a [-ATR] vowel by triggering prefix dissimilation see Chapter 5. Instead, under my analysis, the low and high root vowels always pattern with the underlying form of the suffix. Though it is not my goal to formally account for these patterns, I do highlight some potential means of explanation. The asymmetric alternations of the subjunctive suffix might be explained by some combination of a variety of factors, including: 1 a highly ranked constraint enforcing harmonic sequences of mid vowels, 2 a restriction on which vowel qualities can occur in the FV position, and 3, an asymmetry in the strength of the two [-ATR] mid vowels. Concerning 1, the vowel alternations in both applicative and subjunctive verbs, as well as in noun stems, point to the existence of this type of mid-vowel harmony constraint. 187 Concerning 2, it is a general tendency for marked segments in a language to have a more restricted distribution than unmarked segments Rice 2007; Steriade 1995, and this is well-attested in Bantu languages in particular. For example, Beckman 1997 notes that only a subset of vowels are allowed in later stem positions in Shona verbs. We have also already seen this phenomenon in Ikoma, especially in the restriction against [ ] in prefixes and as the second root vowel in some noun stem vowel combinations. In Ikoma verbs, only a subset of vowels occur in final vowel position. The high and low vowels [i u a] are most frequent final vowels, and the subjunctive vowel [e] is also common. However, [ ] are nearly non-existent. The vowel [ ] in fact never occurs in this position, and the only time [ ] occurs here is following the root vowel, as in the example [tu-k - ] ‘let’s do.’ Based on these distributional observations, we can consider this is a possible factor in the frontback asymmetry. If we proceed with the distributional argument as a possible explanation for the asymmetry, we must also consider the third point introduced above concerning an asymmetry in the strength of the vowels and . Preservation of is highly valued in Ikoma, which we see not only in this subjunctive pattern but also with patterns discussed later in this chapter. As I discuss in §6.4 below, there are other cases of apparent leftward [+ATR] spreading which triggers a root alternation from to [e], but not from to [o]. We begin to see a tendency in which always spreads [-ATR] rightward except to high vowels and is never a target of [+ATR] spreading. On the other hand, the front vowel is always a target of [+ATR] spreading, and in the case of the subjunctive, does not spread [-ATR] rightward. Archangeli and Pulleyblank 1994 have also noted a correlation between backness and [-ATR]. They cite the case of Lango, in which [-ATR] 188 spreads rightward to suffixes but only from back [-ATR] vowels, not from front vowels. We see, then, that there is at least some precedent for having a stronger connection to [-ATR] compared to . See the summary of this chapter §6.8 for more discussion of the asymmetry between and . Based on the three factors mentioned above, imagine an OT analysis in which there is an undominated constraint forcing harmonic sequences of mid-vowels. If this constraint exists, there are two problematic subjunctive mid-vowel sequences C C-e and C C-e. 74 The logical possibilities are for both vowels to be realized as [-ATR], or for both to be realized as [+ATR]. If it is true that there is a general avoidance of the marked vowels in peripheral positions such as the FV position, then the preferred repair might be for C C-e to be realized as [CeC-e]. However, if it is also true that is a “strong” [-ATR] vowel and is especially resistant to alternation, then the preferred repair for C C-e is for the subjunctive vowel to be [-ATR], resulting in [C C- ]. Therefore, we see how the combination of these three factors could result in two different strategies being used to repair the mid-vowel harmony violation. These ideas could be implemented by a particular ranking of constraints enforcing mid-vowel harmony and faithfulness to , which both dominate a constraint against verb-finally. To summarize, one possible explanation for the subjunctive vowel alternations is to attribute the asymmetry to three different factors a mid-vowel harmony constraint, restrictions on the occurrence of the marked [-ATR] mid vowels in FV position, and an 74 Leitch 1997 discusses the very same subjunctive verb forms in his work on the Bantu C languages. In some of those languages, the subjunctive vowel does not alternate and instead remains disharmonic, which he takes as evidence that the subjunctive final vowel is outside of the domain of stem harmony. However, in other Bantu C languages, harmony extends all the way to the right edge of the word, thus including the subjunctive vowel. Returning to Ikoma, we see that harmony of the subjunctive vowel is enforced, which suggests that the harmonic domain includes the subjunctive vowel. 189 asymmetry in the strength of the [-ATR] vowels. Positing these different factors to account for the asymmetry might seem complicated, but each of these points are well- founded in Ikoma and in other languages. Furthermore, each of these well-motivated factors could be expressed in the form of competing constraints which could be ranked in a way that would result in these patterns. As with the applicative suffix in the previous section, the patterns might be difficult to analyze, but the data itself is quite clear. The quality of the subjunctive vowel alternations can be quantified with vowel formant measurements. Speaker B’s subjunctive vowel formant averages are in 154 below, and those averages are plotted in 155. In the table and plot below, averages for [- ] are from tokens of the subjunctive vowel following root vowels only. Averages for [-e] are from tokens following i e a o u. The symbol E is used only for measurements of the subjunctive vowel following roots with underlying . Dividing the measurements in this way allows for us to see the similarity of the suffix vowel following roots with that which follows other root vowels. 154 Subjunctive vowel formant averages Speaker B F1 F2 Tokens - 435 1811 following 6 -e 329 1954 following i e a o u 30 -E 340 2016 following [e] 6 155 Subjunc As the measurements s realized as [-ATR] to m In these cases, the ro The root vowel values of root vowels w values for all seven vow average root vowel me infinitives with only th nctive vowel formant averages plotted Speake show, when the subjunctive [-e] follows , th o match the [-ATR] value of the root, but instea root vowel is also realized as [+ATR] [e]. el change from to [e] is also clearly seen in when they are followed by the subjunctive suf vowels are shown in 156. For comparison, I al easurements for the same fourteen verbs when the FV [-a] suffix. The row for root vowels 190 ker B , the suffix is not ad it remains [+ATR]. in the average formant suffix. The average I also include the en they occur as basic els is highlighted. 191 156 Vowel formant averages of root vowels with the subjunctive Speaker B Root vowel Basic CVC-a Subjunctive CVC-e F1 F2 F1 F2 i 314 2413 282 2149 e 404 1968 385 1973 476 1769 393 1972 u 282 689 235 776 o 408 882 367 871 467 967 457 1096 a 637 1382 604 1540 The table above shows a drop in F1 for from 476 Hz to 393 Hz when the subjunctive suffix is added. A lower F1 measurement corresponds to an acoustically higher vowel. When F1 for underlying is compared to the F1 for e root vowels with the subjunctive, we see that they are very close, with only an 8 Hz difference. F2 is also virtually identical, with a difference of only 1 Hz. These measurements provide strong evidence that the root vowel is not only acoustically raised before the subjunctive suffix, but also that it appears to have completely merged with e root vowels in this same environment. In other words, the subjunctive suffix causes neutralization of the [ATR] contrast between e and . This type of neutralization suggests that the raising of is not simply an effect of anticipatory co-articulation that is, simply a phonetic phenomenon but instead that it is the result of complete i.e. categorical assimilation. This could be understood in terms of [+ATR] spreading from the suffix to the root, or in some underspecification models we might understand it in terms of [-ATR] delinking from the root. In order to visu the formant averages fo chart in 156 above. 157 Plotted a We see clear merging o back vowels, however, Individual token the measurements are t vowel. isualize this complete assimilation, in 157 belo s for root vowels with the subjunctive suffix. Th d averages of root vowels with the subjunctive g of e and here represented as E before th er, do not merge. kens of the root vowel measurements are in 15 taken from two verb roots per root vowel, tota 192 elow I include a plot of The values are from the ve suffix Speaker B the subjunctive. The 158 below. Once again, totaling 4-6 tokens per 158 Plotted t Note that tokens of e The subjunctive speech of Speaker B, b This evidence from mu neutralized before the s asymmetry, since an occurring with the high discussion there includ or not assimilation is g these five suffixes the same asymmetrical rais d tokens of root vowels with the subjunctive S e and are heavily overlapping, whereas o a ive harmony patterns described here have been , but also in the analysis of wordlists by Speake multiple speakers makes a good case that under e subjunctive. There is, however, an interesting and o remain distinct. We see this same front igh-vowel suffixes as well, which I discuss in § udes additional vowel formant analysis and t-te s gradient or categorical. The data in that section he subjunctive suffix and the four high-vowel s raising of but not . 193 Speaker B and are not. en verified not only the kers A and D as well. erlying e and are ing frontback ntback asymmetry §6.4 below. The tests to clarify whether tion show that each of l suffixes triggers the 194

6.3 Inversive and Inversive Stative