132 Another point can now be noted. If it is indeed true that [+ATR] spreads leftward
from high vowels, then the logical question to ask is whether or not it spreads from [+ATR] mid vowels as well. In principle, this type of evidence would confirm that the
raising is indeed related to [ATR] and not height. However, clear evidence suggesting spreading from e o has not been found in noun stems. In order to find convincing
evidence, we would ideally need pairs similar to those in 96 above, in which there is a clear relationship between a [-ATR] mid vowel verb stem and a corresponding [+ATR]
stem with [e] or [o] as the second stem vowel. We would expect the prefix of both words to be high as well, resulting in a noun such as the hypothetical [omu-keme], from the
hypothetical underlying root -k m-. Until evidence like this is found, we cannot confirm that spreading from mid vowels exists. Note, however, that we do see spreading from
[+ATR] mid vowels in verbs, triggered by the subjunctive suffix -e see §6.2 for examples. This type of evidence provides strong support for [+ATR] spreading, because
it shows that the vowels i e u all trigger the same assimilatory process, and they form a natural class of [+ATR] vowels.
4.2.2 Gradient raising of long vowels
In the previous section I present evidence of categorical raising of short vowels. In this section, we see that long vowels are more resistant to raising. There are no clear
examples of any sort of categorical [+ATR] spreading to long vowels. Instead, [-ATR] long vowels are only partially affected by high vowels in V2 position. Examples of mid
vowels which show gradient raising are in 103 below, with front vowels in a and back vowels in b. Each of these examples are in fact long vowels. The initial stem vowels of
the three front-vowel examples are all lengthened compensatorily as a result of
133 prenasalization and labialization respectively. It seems that it is by chance that the front-
vowel examples in the data are all compensatorily lengthened and that all of the back- vowel examples have phonemic length.
103 Long vowels showing gradient raising
7 6
. 2
8 1
2 - 9
-7 ́
T ́
2 2 .: ́
́ 8 - :
́ 8 - .
Note that by the term “gradient” raising I mean two things. In some cases, the amount of raising is different from one token of a word to the next. But perhaps more
often, there are similar amounts of raising in all tokens of the word, but the raising is only slight, resulting in a quality in between what is clearly either [+ATR] or [-ATR]. The
degree of raising can be seen in the average values of both raised and unraised long mid vowels, shown in 104 below.
104 Long vowel V1 averages Speaker A
M MM
M M
MM M
G 8 G 8
G 8 G 8
M MM
M M
MM M
G 8 G 8
G 8 G 8
ZZZZ 397
1907 ZZZZ
379 949
ZZZZ 509
1840 ZZZZ
491 934
ZZZZW W
W W
512 1919
ZZZZW W
W W
480 946
For both front and back vowels, the averages for long vowels before i u indicated with the + symbol are only marginally different from their normal long-vowel counterparts.
These averages are plo since the entire plot rep
105 Plot sho
When individual token vowels are in a broad b
no tokens of raised [ Once again, for the sak
the entire plot represen lotted in 105 below. Note that I have removed
represents only long vowels. howing long vowel averages
ens are plotted, we see that the F1 values of the d but steady continuum throughout the entire m
] or [ ] reach unraised [ee] or [oo]. This is sh ake of clarity in the plot, I have removed the le
ents only long vowels. 134
ved the length symbols,
the “raised” [-ATR] mid-vowel space, but
s shown in 106 below. length symbols, since
106 Plot of i
In conclusion, i vowels are realized as [
u. Though this is argua gradient for long vowe
undergo slight raisin f individual tokens of raised and unraised long
, in this chapter we have seen that underlyingly as [+ATR] mid vowels when they are followed
guably a categorical change for the front vowels wels. Both phonemically long and compensatori
ising before i u, but it is clearly not a categoric 135
ng mid vowels
gly short [-ATR] mid ed by the high vowels i
els, the process is only torily lengthened and
rical change.
136
Chapter 5: Vowel alternations in prefixes