Intro to LinguisticsAutumn 2009
[ʀ]=uvular trill[q] = voiceless uvular stop
Inuktitut
[iglumut]‘to a house’[pinna]‘that one up there’
[ukiaq]‘late fall’[ani]‘female’s brother’
[aiviq]‘walrus’[iglu]‘(snow) house’
[aniguvit]‘if you leave’[panna]‘that place up there’
[aglu]‘seal’s breathing hole’[aivuq]‘she goes home’
[iglumit]‘from a house’[ini]‘place, spot’
[anigavit]‘because you leave[ukiuq]‘winter’
Inuktitut has 3 phonemic vowels: [i], [a], [u]. We know that [i] and [a] are phonemically different because of the minimal pair [ini] ~ [ani], [i] and [u] are phonemically different because of [iglumit] ~ [iglumut], and finally [u] and [a] are phonemically different because of [ukiaq] ~ [ukiuq].
Greenlandic Eskimo
[ivnaq]‘bluff’[qasaloq]‘bark’[ipeʀaq]‘harpoon strap’
[ikusik]‘elbow’[imaq]‘sea’[qilaluvaq]‘white whale’
[tuluvaq]‘raven’[qatigak]‘back’[itumaq]‘palm of hand’
[sakiak]‘rib’[sava]‘sheep’[ugsik]‘cow’
[nuna]‘land’[oʀpik]‘tree’[ine]‘room’
[neʀdloq]‘goose’[nanoq]‘bear’[maʀʀaq]‘clay’
[iseʀaq]‘ankle’[iga]‘pot’[isse]‘eye’
[igdlo]‘house’[seʀmeq]‘glacier’[sako]‘tool’
This is trickier; there don’t appear to be any minimal pairs here. The real reason minimal pairs indicate that 2 sounds are meaningfully different is because they show that there’s no way to always consistently predict the distribution of the two sounds. (In Inuktitut, [ani] ~ [ini] illustrate that you can’t predict where [a] vs. [i] occur.) So we can also investigate whether sounds are phonemically or allophonically different by coming at it from the other direction: is their distribution predictable? If the distribution of any pair of sounds is predictable, that pair is allophonically different; if the distribution between any pair is not predictable, that pair is phonemically different.
Here are all the sounds that can precede and follow each of these 5 vowels in the data above:
__a / a__ / __i / i__ / __u / u__ / __e / e__ / __o / o__i
g
q
ʀ
v
s
m
n
l / t
k
q
ʀ
v
s
l
# / p
t
k
q
s
# / a
p
t
k
g
v
s
m
n
l / t
k
n
l
# / g
v
s
m
n
l / p
s
m
n / q
ʀ
# / k
n
l
# / q
ʀ
#
These lists illustrate that, based on immediate neighboring sounds (which is the relevant conditioning environment here), the difference between [a] and each other vowel is never predictable. [a] and [i] can both follow [q] and [s], and they can both precede [t], [k], [v], [s], and [l]. While there are some sounds that only precede [a] (e.g. [g], [n]), or only follow [i] (e.g. [p], [m]), the fact that there are some environments (e.g. after [s], or before [t]) where it is impossible to predict whether [a] or [i] should occur means that there’s no way to always consistently predict which sound appears where, so the difference between these 2 sounds must be meaningful, so phonemic. By similar logic, the differences between [a] and [u], [a] and [e], and [a] and [o] must also be phonemic; make sure you understand why.
Okay, so [a] is a phonemically unique vowel. What about the rest? Well, [i] and [u] occur in lots of the same contexts: after [t], [k], and # (word edge), and before [g], [v], [s], [m], [n], [l]. Since you can’t ever predict the occurrence of [i] vs. [u] – since [i] and [u] aren’t in perfectly complementary distribution – they must also be meaningfully different.
The difference between [e] and [o] is meaningful, too – both show up after [n], and before exactly the same set of sounds – [q] and [ʀ], and at the ends of words. These contexts are, finally, predictably different from those where [i] and [u] show up: [e] and [o] occur only before [q], [ʀ], and #, while [i] and [u] never occur in these contexts. This suggests that [e]/[o] are allophones of [i]/[u], or vice versa.
Now we have to sort out the complicated relationship between these four sounds. In the diagram below, solid lines represent meaningful differences, while dashed lines represent predictable differences. It looks at this point like there are two phonemic vowels here, and the other two are predictable allophones of these variants. But which are the underlying representations, [i]-[u] or [e]-[o]? And are the allophonic pairs [i]-[e] and [u]-[o] or [i]-[o] and [u]-[e]?
[i][u]
[e][o]
The underlying representations are [i] and [u]. These high vowels show up in lots and lots of contexts, while [e] and [o] show up in extremely specific contexts: before uvulars and word-finally. Rule (a) below is much simpler than (b); it’s much easier to imagine [i]-[u] becoming [e]-[o] in these few contexts than [e]-[o] almost always changing their forms.
(a) The real rule (almost):[i] and [u] become [e] and [o] when they occur before uvulars or word-finally.
(b) Too complex to be the right rule:[e] and [o] become [i] and [u] when they occur before labials, labiodentals, alveolars, velars, and vowels.
Finally, does the underlying representation [i] become the allophone [e] or [o]? Well, think about the features:
[i]:high, front, unrounded[u]:high, back, unrounded
[e]:mid, front, unrounded[o]: mid, back, unrounded
Again towards the goal of writing a simple rule (as simple rules are most often phonologically correct), if [i] changes to [e] before uvulars and word-finally, it only needs to change its height feature; if [i] [o], it needs to change both height and backness. So the real, complete rule looks like this:
The whole real rule:[i] becomes [e], and [u] becomes [o], when they occur before uvulars or word-finally.
1