Week 5: Morphology and Alternation

Linguistics 107
Winter 2014
Week 5: Morphology and Alternation
I. Rule ordering
1. /aI/ Raising (Northeastern US and Canada)
/aI/ → øI / __ -syllabic
-voice
‘tripe’
‘right’
‘hiker’
‘life’
[t®øIp]
[®øIt]
[høIk®`]
[løIf]
‘tribe’
‘ride’
‘tiger’
‘live’
[t®aIb]
[®aId]
[taIg®`]
[laIv]
2. Flapping (revised to apply to both /t/ and /d/)
+coronal → +continuant
+anterior
+sonorant
-continuant +voice
/ [+syllabic] __ +syllabic
-stress
3. Flapping is a neutralizing rule
‘heat’
‘heed’
/"hit/
"hit+IN
"hi|IN
["hi|IN]
/"hid/
"hid+IN
"hi|IN
["hi|IN]
Phonemic
Morphology
Flapping
Phonetic
4. Other examples of neutralization due to flapping
‘betting’ vs. ‘bedding’
["bE|IN]
["bE|IN]
‘wetting’ vs. ‘wedding’
["wE|IN]
["wE|IN]
‘butted’ vs. ‘budded’
["bø|´d]
["bø|´d]
1
5. Examining the interaction between /aI/ Raising and Flapping
The crucial words are those with underlying /aI/ before a voiceless alveolar stop
‘write’ vs. ‘writing’
/"®aIt/
/"®aItIN/
‘ride’ vs. ‘riding’
/"®aId/
/"®aIdIN/
(Phonemic forms)
/aI/ Raising depends on voicing of following consonant, but Flapping changes voicing
7. Predictions
• If /aI/ Raising applies to phonemic form we would except it to apply in words like ‘writing’
• If, on the other hand, /aI/ Raising applies to phonetic form, we would not expect it to apply
in ‘writing’ because we have a flap instead of /t/.
8. In fact, for some speakers, raising does apply, while for others it does not
Group A: Raising does not apply
‘write’ vs. ‘writing’
["®øIt]
["®aI|IN]
‘ride’ vs. ‘riding’
["®aId]
["®aI|IN]
Group B: Raising does apply
‘write’ vs. ‘writing’
["®øIt]
["®øI|IN]
‘ride’ vs. ‘riding’
["®aId]
["®aI|IN]
(Note that there is also a Group C (not relevant for present purposes) for which raising does not
apply at all)
9. In Group A, Flapping precedes Raising
‘writing’
‘riding’
/"®aItIN/
"®aI|IN
--["®aI|IN]
/"®aIdIN /
"®aI|IN
---["®aI|IN]
Phonemic
Flapping
Raising
Phonetic
In Group B, Raising precedes Flapping
/"®aItIN/
"®øItIN
"®øI|IN
["®øI|IN]
/"®aIdIN /
---"®aI|IN
[""®aI|IN]
Phonemic
Raising
Flapping
Phonetic
2
10. Vowel Length in Chimwi:ni (Bantu; Somalia)
11. Some morphology
-el
-an
-o…w
-a
applicative (action on behalf of someone)
reciprocal (action to each other)
passive suffix
filler vowel so that verbs don’t end in a consonant
12. Vowel length is contrastive
[ku-ba…ram-a] ‘to talk’ vs. [ku-balam-a] ‘to promise’
[x-pe…lek-a] ‘to be able to be swept’ vs. [ku-pelek-a] ‘to send’
13. But vowel length not contrastive in syllables before the third from the end, i.e. before the
antepenult
[ku-re…b-a] ‘to stop’
[ku-re…b-el-a] ‘to stop for’
but [ku-reb-el-an-a] ‘to stop for one another’ and not *[ku-re…b-el-an-a]
14. Pre-Antepenultimate Shortening (better formalizing later)
V → [-long] / when > 3 syllables from end of utterance
/ku-re…b-el-an-a/
ku-reb-el-an-a
[ku-reb-el-an-a]
Output of morphology
Phonology: Pre-Antepenultimate Shortening
Phonetic output
The root for ‘stop’ thus has two allomorphs:[reb] and [re…b]
15. Pre-Antepenultimate Shortening is a neutralizing process, since phonemic contrast in vowel
length is neutralized in syllables before the antepenult
16. Some more examples of Pre-Antepenultimate Shortening
bo…z-el-e
d°Zo…hari
x-fa…n-a
‘he stole
‘jewel’
‘to do’
vs.
vs.
vs.
boz-el-en-i
d°Zohari-j-e
x-fan-iliz-a
‘what did he steal?’
‘her jewel’
‘to do with’
17. Some other rules in Chimwi:ni
18. Prepausal Shortening (applies to utterance-final vowels)
V → [-long] / __ ]Utterance
3
19. Word-final Lengthening (applies to all word-final vowels)
V → [+long] /__ ]Word
na
kolko
kama
hud°Zo
‘by’
‘than’
‘like’
‘the one who eats’
na… noka
kolko… mi
kama… mpÓaka
hud°Zo… mbele
‘by a snake’
‘than me’
‘like a cat’
‘the one who eats first’
20. Prepausal Shortening and Word-final Lengthening conflict with each other in Utterance-final
position; we thus need to see which one wins out
21. Word-final Lengthening applies before Prepausal Shortening
/kama mpÓaka/
kama… mpÓaka…
kama… mpÓaka
[kama… mpÓaka]
Output of morphology
Phonology: Word-final Lengthening
Phonology: Prepausal Shortening
Phonetic output
22. Next question: Is there a crucial ordering between Word-final Lengthening and PreAntepenultimate Shortening?
23. Check the outputs under both orderings
Relevant case = a word preceding a trisyllabic word at the end of a phrase: /kuna kahawa/
Derivation A: Pre-Antepenultimate Shortening precedes Word-Final Lengthening
/kuna kahawa/
---kuna… kahawa…
kahawa
[kuna… kahawa]
Output of morphology
Phonology: Pre-Antepenultimate Shortening
Phonology: Word-Final Lengthening
Prepausal Shortening
Phonetic output
Derivation B: Word-Final Lengthening precedes Pre-Antepenultimate Shortening
/kuna kahawa/
kuna… kahawa…
kuna
kahawa
[kuna kahawa]
Output of morphology
Phonology: Word-Final Lengthening
Phonology: Pre-Antepenultimate Shortening
Prepausal Shortening
Phonetic output
24. The actual form is [kuna kahawa]
4
25. Orderings thus far:
Word-Final Lengthening
Pre-Antepenultimate Shortening
Prepausal Shortening
26. Of course, there is no reason to rank Pre-Antepenultimate Shortening and Prepausal
Shortening relative to each other since they never apply in the same environment
27. One more rule:
Pre-Long Shortening
[+syllabic] → [-long] / __ C0 +syllabic
+long
C0 means any number of consonants
x-so…ma-a
x-Si…k-a
x-pe…l-a
‘to read’
‘to seize’
‘to sweep’
x-som-o…w-a ‘to be read’
x-Sik-o…w-a ‘to be seized’
x-pel-o…w-a ‘to be swept’
28. We need to establish relative ordering of Pre-Long Shortening and Word-Final Lengthening
since they compete with each other in the case of a word-final vowel appearing before a word
whose first vowel is long
29. Forms like the following establish the necessary ranking
kubiga pa…si
‘to iron clothes’
kudara le…mbe
‘to sharpen a razor’
x-t°Simbil-a si…mba ‘to run from a lion’
30. Because the final vowel of the first word in each phrase is short, we know that Pre-Long
Shortening has to apply after Word-Final Lengthening.
/kubiga pa…si/
kubiga…
kubiga
[kubiga pa…si]
Output of morphology
Phonology: Word-Final Lengthening
Phonology: Pre-Long Shortening
Phonetic output
5
31. If the opposite order were true, we would expect a long vowel and not a short vowel
Incorrect result:
/kubiga pa…si/
----kubiga…
[kubiga… pa…si]
Output of morphology
Phonology: Pre-Long Shortening
Phonology: Word-Final Lengthening
Phonetic output
32. Note that it doesn’t matter for rule ordering purposes whether the final vowel in /kubiga/ is
short or long: Pre-Long Shortening must apply after Word-Final Lengthening to ensure that we
get a short vowel at the end of [kubiga]
33. There is also no need to worry about ordering Pre-Long Shortening relative to Prepausal
Shortening or to Pre-Antepenultimate Shortening, since they all have the same effect of
shortening vowels
34. Final Rule Ordering
Word-Final Lengthening
Pre-Long Shortening
Pre-Antepenultimate Shortening
6
Prepausal Shortening
Week 5 (Continued): More on Phonological Analysis: Lardil
1. Lardil vowels
[+high]
[-high]
[+front]
i
e
[-front]
u
a
2. Lardil consonants
Bilabial
Stops
Nasals
Rhotics
Laterals
Glides
p
m
Laminodental
t1
n1
Apicoalveolar
t
n
r
l
Laminopalatal
tj
nj
w
lj
j
Retroflex
Velar
Ê
=
«
l¢
k
N
3. Noun morphology
Uninflected
1. kentapal
2. ket1ar
3. mija«
Acc. Non-Fut
kentapal-in
ket1ar-in
mija«-in
Acc.Fut.
kentapal-u«
ket1ar-u«
mija«-u«
Gloss
‘dugong’
‘river’
‘spear’
Acc.Fut.
mela-«
wanka-«
kuNka-«
Gloss
‘sea’
‘arm’
‘groin’
4. Preliminary Hypothesis
• Accusative Non-future = -in
• Accusative Future = -u«
5. Some more data
Uninflected
4. mela
5. wanka
6. kuNka
Acc. Non-Fut
mela-n
wanka-n
kuNka-n
6. We now have two allomorphs for Accusative Non-future and Accusative Future
• Accusative Non-future = -in ~ -n
• Accusative Future = -u« ~ -«
7. Two hypotheses to test
Hypothesis 1: Underlying suffixes are -in, -u« and vowels are deleted in some cases
7
Hypothesis 2: Underlying suffixes are -n, -« and vowels are added in some cases (epenthesis)
8. Testing hypothesis 1
/kentapal-in/
---[kentapalin]
/kentapal-u«/
---[kentapalu«]
/mela-in/
melan
[melan]
/mela-u«/
mela«
[mela«]
Underlying
V-Deletion
Surface
9. V-deletion formalized
[+syllabic] → Ø / [+syllabic] __
10. This rule is natural: many languages avoid adjacent vowels (hiatus)
11. Testing hypothesis 2
/kentapal-n/
kentapalin
[kentapalin]
/kentapal-«/
kentapalu«
[kentapalu«]
/mela-n/
---[melan]
/mela-«/
---[mala«]
Underlying
Epenthesis
Surface
12. Problem with this hypothesis: Although epenthesis is natural to break up consonant clusters,
how do we know which vowel to insert? Why /i/ in [kentapalin] but /u/ in [kentapalu«]?
Epenthesis is unable to predict which vowel to insert; we must assume that the /i/ in kentapalin
and the /u/ in kentapalu« are part of the stem since this information is not predictable by our rule.
13. Summing up the analysis
• Accusative Non-future = -in
• Accusative Future = -u«
V-deletion
[+syllabic] → Ø / [+syllabic] __
14. Some more data
Uninflected
7. Nuka
8. kaÊa
Acc.Non-fut.
Nuku-n
kaÊu-n
Acc.Fut
Nuku-«
kaÊu-«
Gloss
‘water’
‘child’
15. Why do we get an alternation between /a/ and /u/?
16. These data present problems, since we already saw other data in which /a/ stays the same
throughout the paradigm (repeated here)
8
Uninflected
4. mela
5. wanka
6. kuNka
Acc. Non-Fut
mela-n
wanka-n
kuNka-n
Acc.Fut.
mela-«
wanka-«
kuNka-«
Gloss
‘sea’
‘arm’
‘groin’
17. If we make a chart to see environments in which /a/ is kept and environments in which /a/
changes to /u/ we see that there is no plausible environment which could predict this.
/a/ is kept
4) l _ n
5) k _ n
6) k _ n
/a/ changes to /u/
7) k _ n
8) Ê _n
18. We thus can’t write a rule changing /a/ to /u/ since there is no environment which this rule
applies in consistently
19. Another hypothesis: The /u/ in words like (7, 8) is part of the root and lowers to /a/ in wordfinal position
20. We are pushed in this direction because we have one set of forms in which /a/ alternates with
/u/ and another set in which /a/ remains throughout the paradigm
[Nuka], [Nuku-n], [Nuku-«] vs. [wanka], [wanka-n], [wanka-«]
21. Final Lowering (preliminary)
u → a / __ ]word
22. Testing our rule on our data
/Nuku/
---Nuka
[Nuka]
/Nuku-in/
Nukun
---[Nukun]
/wanka/
------[wanka]
/wanka-in/
wankan
---[wankan]
Underlying
V-Deletion
Final Lowering
Surface
23. Prediction: There are no final /u/ in the language; this is correct (for the most part; more
later)
24. Final Lowering generalized to apply to front vowels
Uninflected
9. tjempe
10. wiÊe
11. ke=Êe
12. Ni=e
Acc. Non-Fut
tjempe-n
wiÊe-n
ke=Êi-n
Ni=i-n
Acc.Fut.
tjempe-«
wiÊe-«
ke=Êi-wu«
Ni=i-wu«
9
Gloss
‘mother’s father’
‘interior’
‘wife’
‘skin’
25. Final Lowering formalized
[+syllabic] → [-high] / __ ]word
26. Clean-up Rule:
/a/ Adjustment (turns non-high non-front vowels into /a/)
+syllabic
-front
→
-high
+low
-round
-back
27. /a/ Adjustment must apply after Final Lowering, since it applies to the output of Final
Lowering
Otherwise we would get forms like *[Nuko]
28. Another alternation: In forms like [ke=Êiwu«], [Ni=iwu«] we have a [w] showing up before
the accusative future suffix [-u«].
29. Hypothesis 1: [w] is underlyingly part of the Accusative future suffix
30. This would mean that we would need a rule deleting [w] after all vowels except [i]; after this
rule, Vowel-deletion would have to apply
/wanka-wu«/
wankau«
wanka«
------[wanka«]
Underlying
W-deletion
V-Deletion
Final Lowering
/a/ Adjustment
Surface
31. While this could work for this form, there are /w/ elsewhere in the language occurring
intervocalically after vowels other than /a/: [NawuNa] ‘termite’
32. Forms like [NawuNa] show that /w/ deletion would not work
33. Hypothesis 2: [w] is part of the root in words like [ke=Êi-wu«]
34. This would mean that we need a rule which deletes [w] word-finally and before vowels other
than /u/; after this rule, Vowel deletion and Final Lowering would have to apply
/ke=Êiw/
ke=Êi
----
/ke=Êiw-in/
ke=Êiin
ke=Êin
Underlying
W-deletion
V-Deletion
10
ke=Êe
---[ke=Êe]
------[ke=Êin]
Final Lowering
/a/ Adjustment
Surface
35. Although this would work for these forms, we do get [w] occurring elsewhere in the
language in words like [wiÊe] ‘interior’
36. We would thus need [w]-Deletion to apply word-finally and intervocalically before vowels
other than [u]; this would be pretty unnatural particularly in a context like [iwi]
(I also wouldn’t be surprised if we did find forms with [w] in exactly this environment)
37. Hypothesis 3: [w] is inserted between [i] and [u]
/w/ Epenthesis
Ø → w / i __ u
38. /w/ Epenthesis must be ordered before Vowel deletion, otherwise it would predict forms like
*[ke=Êi«]
39. Here’s where we are so far
/w/ Epenthesis
↓
Vowel Deletion
Final Lowering
↓
/a/ Adjustment
40. Some derivations
/papi/
------pape
---[pape]
/papi-in/
---papin
------[papin]
/papi-u«/
papiwu«
---------[papiwu«]
Underlying
/w/ Epenthesis
Vowel Deletion
Final Lowering
/a/ Adjustment
Surface
/Nuku/
------Nuko
Nuka
[Nuka]
/Nuku-in/
---Nukun
------[Nukun]
/Nuku-u«/
---Nuku«
------[Nuku«]
Underlying
/w/ Epenthesis
Vowel Deletion
Final Lowering
/a/ Adjustment
Surface
11
41. Some new alternations
Uninflected
13. jalul
14. majar
15. wiwal
16. jilijil
Acc. Non-Fut
jalulu-n
majara-n
wiwala-n
jilijili-n
Acc.Fut.
jalulu-«
majara-«
wiwala-«
jilijili-wu«
Gloss
‘flame’
‘rainbow’
‘bush mango’
‘species of oyster’
42. Two Hypotheses
43. Hypothesis 1: Vowel insertion
Problem: There does not seem to be any way to predict which vowel to insert; in some words it
would be /u/, in others /a/, in still others /i/
44. Hypothesis 2: Final vowel deletion
45. This looks more promising, as long as we can explain why we get deletion in words (13-16)
but not in earlier words
46. There actually is an explanation for the difference between the two classes of words
47. Final vowels only drop if this will not create words that are shorter than two syllables
48. If final vowels dropped in words like [Nuka], [pape], etc. we would wind up with
monosyllabic words.
49. Many languages impose minimal word requirements;
e.g. English
Words may consist of a long vowel (usually but not always [+tense]) or diphthong: say, see,
cow
Words may consist of a closed syllable: cat, lip, put
Words may consist of two syllables: Minna, Emma
But, no monosyllabic words consisting of a short vowel in an open syllable: *kœ, *lI, *pU
50. Vowel apocope in Lardil
V → Ø / __ ]word
if output is minimally disyllabic
51. Some more data
Uninflected
17. jukar
18. wulun
Acc. Non-Fut
jukarpa-n
wulunka-n
Acc.Fut.
jukarpa-«
wulunka-«
12
Gloss
‘husband’
‘species of fruit’
wuÊaltji-n
karwakarwa-n
19. wuÊal
20. karwakar
wuÊaltji-wu«
karwakarwa-«
‘meat’
‘species of wattle’
52. Two hypotheses
53. Hypothesis 1: Consonant insertion
Problem: There does not seem to be any way to predict which consonant gets inserted
54. Hypothesis 2: Consonant deletion
This looks better, especially since there are no consonant clusters at the end of words
55. Cluster reduction (deletes second in a sequence of two consonants at right edge of word)
C → Ø / C __ ]word
56.Cluster reduction must apply after Apocope, since Apocope sets up context for rule to apply
in
/jukarpa/
------------jukarp
jukar
[jukar]
/jukarpa-in/
---jukarpan
------------[jukarpan]
/jukarpa-u«/
---jukarpa«
------------[jukarpa«]
Underlying
/w/ Epenthesis
Vowel Deletion
Final Lowering
/a/ Adjustment
Apocope
Cluster Reduction
Surface
57. Some more data
Uninflected
21. putu
22. murkuni
23. NawuNa
24. tipiti
25. t1apu
Acc. Non-Fut
putuka-n
murkunima-n
NawuNawu-n
tipitipi-n
t1aputji-n
Acc.Fut.
putuka-«
murkunima-«
NawuNawu-«
tipitipi-wu«
t1aputji-wu«
Gloss
‘short’
‘nullah’
‘termite’
‘species of rock cod’
‘older brother’
58. These forms illustrate loss of certain final consonants: in particular [k, m, w, p, tj]
59. In fact, the only consonants which Lardil words can end in are [n, l, r, «]
60. The set of sounds which can occur in final position are the apical sonorants ([+coronal,
-distributed, +sonorant]
61. Non-Apical deletion
13
C → Ø / __ ]word unless C = [+coronal, -distributed, +sonorant]
62. Non-Apical deletion illustrated
/putuka/
/jilijili/
---------------jilijile
------putuk
jilijil
------putu
---[putu]
[jilijil]
Underlying
/w/ Epenthesis
Vowel Deletion
Final Lowering
/a/ Adjustment
Apocope
Cluster Reduction
Non-Apical Deletion
Surface
63. Non-Apical deletion must apply after Apocope, since it is Apocope which sets up the
environment for Non-Apical deletion
64. Other evidence for this ordering comes from forms like [murkuni]
/murkunima/
Underlying
---/w/ Epenthesis
---Vowel Deletion
---Final Lowering
---/a/ Adjustment
murkunim
Apocope
---Cluster Reduction
murkuni
Non-Apical Deletion
[murkuni]
Surface
65. If Apocope applied after Non-Apical deletion, then we would expect the final vowel to be
lost and wind up with *[murkunim]
/murkunima/
---------------murkunim
--[murkunim]
Underlying
/w/ Epenthesis
Vowel Deletion
Final Lowering
/a/ Adjustment
Non-Apical Deletion
Apocope
Cluster Reduction
Surface
66. More rule orderings
Uninflected
26. muNkumu
27. tjumputju
Acc. Non-Fut
muNkumuNku-n
tjumputjumpu-n
Acc.Fut.
muNkumuNku-«
tjumputjumpu-«
14
Gloss
‘wooden axe’
‘dragonfly’
67. These forms show that Cluster Reduction must precede Non-Apical deletion, since Cluster
Reduction places Non-apical consonants into position for deletion
68. Final Lowering must precede Non-Apical deletion, since high vowels which come to be in
final position because of Non-Apical deletion do not lower, e.g. we get [Nalu] from underlying
/Naluk/ and we don’t get *[Nala]
69. Final Ordering
/w/ Epenthesis
↓
Vowel Deletion
Final Lowering
↓
/a/ Adjustment
Apocope
↓
Cluster Reduction
↓
Non-Apical deletion
70. Conclusion:
• Lardil is a case in which the uninflected isolation form is not the underlying form: Rather, the
closest forms to the underlying forms are those occurring in a protected environment before
suffixes
• The uninflected forms are subject to a number of rules which erode at and alter the material at
the right edge (e.g. Final Lowering, Vowel Deletion, Apocope, Cluster Reduction, NonApical deletion)
• Though superficially complex, the alternations are the result of fairly simple rules; it’s just that
many rules may apply to a single form drastically altering the underlying form
15