24.
962, Spring 2018 Class 21 — 5/3/18
Class 21
Allomorphy
5/3/18
1 Types of Allomorphy
A LLOMORPHY:
One “morpheme” ( ≈ same morphosyntactic/morphosemantic feature (bundle)) surfaces with different
phonological content (morph) in different contexts; i.e. one morpheme has multiple allomorphs.
? Many different phenomena fall under the broad umbrella of “allomorphy”, running the gamut from purely
phonological to purely morphological/lexical.
• Carstairs-McCarthy (2001) lays out four main logically possible types. (All types exist.)
(1) Logically possible types of allomorphy (Carstairs-McCarthy 2001:113) [order different here]
a. Morphs phonetically similar, distribution describable in purely phonological terms.
b. Morphs phonetically dissimilar, distribution describable in purely phonological terms.
c. Morphs phonetically similar, distribution not describable in purely phonological terms.
d. Morphs phonetically dissimilar, distribution not describable in purely phonological terms.
• Type (1a) is clearly all about phonology.
• Type (1d) clearly has nothing to do with phonology.
? Type (1b) and (slightly less so) Type (1c) is where the real interesting stuff lies.
⇒ Understanding how cases like these truly work will shed light on the division of labor between
phonology and morphology (see esp. Paster 2009, 2015, Nevins 2011).
1.1 Purely phonological allomorphy (Type (1a))
• Different allomorphs can arise because of the application of general phonological processes.
◦ Sometimes referred to as “phonologically driven allomorphy”.
• For example: regular allomorphs of /-z/ and /-d/ suffixes in English.
(2) Plural s in English
1. (epenthesis) /-z/ → [-Iz] / C[+strident] +__ /pæS+z/ → [pæSIz]
2. (voicing assim.) /-z/ → [-s] / C[-voice] +__ /kæt+z/ → [kæts]
3. (elsewhere) /-z/ → [-z] /dOg+z/ → [dOgz]
• English doesn’t allow strings of sibilants (1.) or obstruent clusters that disagree in voice (2.), so the
different allomorphs can be explained fully by the phonological context, by invoking general phonological
properties of the language.
→ So, no special morphological devices are required in order to explain the distribution, and parsimony
suggests they should not be employed here.
1
24.962, Spring 2018 Class 21 — 5/3/18
• To confirm that we can do everything in the phonology using a single UR, here’s a quick analysis:
(3) Regular plural allomorphy in English is phonological
a. Sibilant-final stems → [-Iz]
/pæs-z/ *[+strid][+strid] AGREE[voice] D EP V-IO I DENT[voice]-IO
a. pæsz *! *!
b. pæss *! *
c. + pæsIz *
d. pæsIs * *!
b. Voiceless-final stems → [-s]
/kæt-z/ *[+strid][+strid] AGREE[voice] D EP V-IO I DENT[voice]-IO
a. kætz *!
b. + kæts *
c. kætIz *!
d. kætIs *! *
c. Other stems → [-z]
/dOg-z/ *[+strid][+strid] AGREE[voice] D EP V-IO I DENT[voice]-IO
a. + dOgz
b. dOgs *! *
c. dOgIz *!
d. dOgIs *! *
1.2 Morphological allomorphy (Types (1d) and (1c))
• Sometimes, different allomorphs look (totally/mostly) unrelated phonologically (i.e. cannot be derived
from same or similar UR via anything resembling regular phonology) and are distributed arbitrarily (from
a phonological perspective) across particular morphosyntactic/morphosemantic contexts.
◦ Usually referred to as suppletion.
→ If phonology has nothing to do with it, these cases must reside (somehow) in the morphology.
• For example: irregular verbal allomorphy in English.
(4) Present ∼ Past suppletion in English
a. go [goU] ∼ went [wEn(-)t] (totally phonologically unrelated)
b. bring [brIN] ∼ brought [brO(-)t], stand [stænd] ∼ stood [stUd] (partially phonologically related)
c. take [teIk] ∼ took [tUk], drive [draIv] ∼ drove [droUv], lead [lid] ∼ led [lEd], etc.
[also irregular plurals: e.g. mouse ∼ mice, foot ∼ feet, goose ∼ geese, etc.]
(clearly phonologically related by semi-regular “ablaut”, but not by regular phonology)
2
24.962, Spring 2018 Class 21 — 5/3/18
• The distribution of go vs. went must be located purely in the morphology/lexicon:
◦ There are no phonological rules (no matter how ad hoc) which can relate the two allomorphs.
◦ Furthermore, the distribution appears to be conditioned by the morphosyntax:
(5) a. go / __PRES [really the elsewhere context; appears in infinitive, etc.]
b. went / __PAST
• The same probably goes for pairs like bring ∼ brought, stand ∼ stood.
◦ Even though the respective allomorphs clearly share some phonological material, the ways in which
they differ don’t look remotely like regular phonology.
• Cases like {take ∼ took, drive ∼ drove, lead ∼ led, etc.} are closer to the borderline (basically Type (1c)).
◦ They differ only in the vowel, sometimes by just one or two features.
◦ There are substantial classes of verbs that all pattern in similar ways (cf., e.g., Pinker & Prince 1988,
Albright & Hayes 2003).
→ The point is: it looks like phonological generalizations could be involved, so maybe it’s not purely morpho-
logical/lexical.
1.3 Phonologically Conditioned (Suppletive) Allomorphy (Type (1b))
• The types of cases which are likely to be most enlightening are what is sometimes called “Phonologically
Conditioned (Suppletive) Allomorphy” (PCSA; Carstairs-McCarthy 2001, Paster 2009, 2015):
→ Two allomorphs which are not transparently phonologically related, but whose distribution is clearly
phonologically governed.
• Classic example: English indefinite article a ∼ an
(6) English indefinite article
a. a ([eI, 2, @,...]) / __#C... e.g., a tiger (*an tiger)
b. an ([æn, 2n, @n, n,...]) / __#V... e.g., an elephant (*a elephant)
"
• The distribution seems to be structured so as to avoid hiatus:
◦ The consonant-final allomorph (an) appears just in case the vowel-final allomorph would have created
a hiatus.
• Historically, it’s actually the reverse: the /n/ is original, and was deleted pre-consonantally (basically to avoid a coda).
• All the same points would hold thinking of it in this way synchronically as well.
• However, we can’t simply view this as a phonological repair for hiatus:
◦ English does not use n as an epenthetic consonant
◦ English generally tolerates hiatus
→ The fact that this morpheme employs a strategy to avoid hiatus is a property just of the morpheme, not of
the language generally.
◦ (Though it looks like it operates in the cross-linguistically expected direction — the allomorphs are
distributed so as to avoid a marked structure.)
3
24.962, Spring 2018 Class 21 — 5/3/18
• This effect could alternatively be analyzed as a sort of (highly specific) derived environment effect, or with
a special cophonology.
→ However, other examples of PCSA, where the two morphs are fully phonetically/phonologically
distinct, cannot.
• Carstairs-McCarthy (2001:114–115) mentions a number of such examples (see also Paster 2009, 2015,
Nevins 2011, and references therein).
◦ One set comes from noun class markers in Bantu (Guthrie 1956), where phonologically distinct
morphs surface in phonologically complementary distribution.
• In FaN, one class has a simple V vs. C distinction.
◦ Any other distribution would result in phonotactic issues — either an initial cluster or hiatus.
(7) FaN Class 5: a∼dz (Guthrie 1956:551)
a. Consonant-initial stems have [a-]: a-fan (*dz-fan) ‘forest’
b. Vowel-initial stems have [dz-]: dz-al (*a-al) ‘village’
• In one noun class in TsOgO, consonant-initial and front-vowel–initial stems pattern together (showing a
default, presumably) against back-vowel–initial roots.
◦ We can probably understand this as a restriction on adjacent vowels mismatching in backness.
(8) TsOgO Class 7: ge.∼s (Guthrie 1956:551)
a. Consonant-initial stems have [ge.-]: ge.-deku̧ ‘chin’
b. Front-vowel–initial stems also have [ge.-]: ge.-e.pa ‘bone’, ge.-Edu̧ ‘chin’[?]
c. Back-vowel–initial stems have [s-]: s-o.ma ‘thing’, s-OtO ‘fire’
• In Kongo, a number of noun classes show no overt marker with stems beginning in a non-nasal consonant,
but a consistent overt marker for other stems (i.e. vowel-initial and nasal(+consonant)-initial).
◦ Not completely clear what problem this distribution would be solving... (we’ll come back to this
question)
(9) Kongo (Guthrie 1956:551–552)
a. Class 5: Ø∼di
i. Non-nasal–consonant—initial stems [Ø-]: (Ø-)pata ‘village’
ii. Vowel-initial stems have [di-]: di-aku ‘egg’
iii. Nasal-initial stems have [di-]: di-nkondo ‘banana’
b. Class 7: Ø∼ki
i. Non-nasal–consonant—initial stems [Ø-]: (Ø-)sanu ‘comb’
ii. Vowel-initial stems have [ki-]: ki-ula ‘frog’
iii. Nasal-initial stems have [ki-]: ki-nzu ‘pot’
c. Class 8 (plural of Class 7?): Ø∼bi
i. Non-nasal–consonant—initial stems [Ø-]: (Ø-)sanu ‘combs’
ii. Vowel-initial stems have [bi-]: bi-ula ‘frogs’
iii. Nasal-initial stems have [bi-]: bi-nzu ‘pots’
d. Class 14: Ø∼u/w
i. Non-nasal–consonant—initial stems [Ø-]: (Ø-)fuku ‘night’
ii. Vowel-initial stems have [w-]: w-anda ‘net’
iii. Nasal-initial stems have [u-]: u-mfumu ‘chieftainship’
4
24.962, Spring 2018 Class 21 — 5/3/18
• In circumstances like these, it seems clear that there must be at least two URs available (under some
conditions) for each particular morpheme.
◦ i.e., no phonological transformation(s) could derive both morphs from a single UR.
• Yet it is phonological information which is conditioning the distribution of the two morphs.
1.4 Big Questions
• Along the lines of Nevins (2011:esp. §3), some of the most significant questions are:
? Which aspects of “allomorphy” reside in the phonology vs. the morphology?
,→ How do we decide for any given case?
? What exactly do the mechanisms of non–purely-phonological allomorphy look like?
,→ What is the proper input to the phonological evaluation?
2 Vocabulary Insertion in Distributed Morphology
• For purely morphological allomorphy, it is (more or less) clear that this must take place in the morpholog-
ical component, because phonological information plays no role in determining the choice.
? There are ways of sticking this into the phonological component nonetheless, if we allow the phono-
logical component to see substantial morphological information (see below).
• Consider suppletion in English be:
(10) Person/number suppletion in English be
Present Past
Singular Plural Singular Plural
1st am [æm] are [aô] was [w2z] were [wô]
"
2nd are [aô] are [aô] were [wô] were [wô]
" "
3rd is [Iz] are [aô] was [w2z] were [wô]
"
• Within the present and past tense paradigms, there are 5 distinct morphs: am, is, are, was, were
◦ There are some phonological similarities between some forms (esp. was∼were), but it seems unlikely
that any of these could be related by phonological transformations.
◦ Rather, it seems like it’s purely morphosyntactic information which is making the determination.
• Distributed Morphology (DM; Halle & Marantz 1993, et seq.) uses a system of “Vocabulary Insertion”
(VI) to relate morphosyntactic information to stored lexical items (Vocabulary Entries) which contain
arbitrary phonological information (i.e. URs).
(11) Components of VI in DM
a. List of Vocabulary Entries (≈ lexicon)
b. Set of (intrinsically-ordered) Vocabulary Insertion rules, which essentially relate morphosyn-
tactic (morphosemantic, in the case of roots) information to particular Vocabulary Entries
c. For a particular derivation, the Vocabulary Entries which are selected by applying the VI rules
are sent to the phonological component, where they function as the URs/Input to the phono-
logical evaluation
5
24.962, Spring 2018 Class 21 — 5/3/18
• VI rules can contain contextual information (much like phonological rules, as it were).
→ For cases like English be allomorphy, the context is fairly clearly morphosyntactic in nature.
(12) VI rules for English be
a. BE ↔ /æm/ / PRES , 1, SG
b. BE ↔ /I(z)/ / PRES , 3, SG ([z] could be the 3SG . PRES agreement marker)
c. BE ↔ /aô/ / PRES
d. BE ↔ /w2z/ / PAST, 1, SG
e. BE ↔ /w2z/ / PAST, 3, SG
f. BE ↔ /wô/ / PAST
"
◦ The rules with person specifications are intrinsically ordered before those without via the “Elsewhere
Condition”, which states that more specific rules are ordered before less specific ones.
• The question we will be interested in is: can the context also include phonological information? (And
if so, is that a good idea?)
◦ Standard answer in DM: yes, but only if the context is contained in a morpheme which has already
been spelled-out — i.e. is lower in the tree, assuming bottom-up cyclic spell-out (see, e.g., Bobaljik
2000, 2012).
3 VI with phonological contexts, and Sub-categorization
• For cases of PCSA, we could extend a morphologically-oriented DM VI type analysis to phonological
information.
• Consider the example of genitive allomorphy in Dja:bugay (Pama-Nyungan) (Paster 2015:219–223; from
Patz 1991):
◦ With vowel-final stems, the genitive suffix surfaces as [-n]
◦ With consonant-final stems, the genitive suffix surfaces as [-Nun]
? Assume that -Nun cannot be related to -n in any phonologically regular way.
(13) Genitive allomorphy in Dja:bugay
a. Vowel-final stems → [-n]
guludu-n (*guludu-Nun) ‘dove-GEN’
gurra:-n (*gurra:-Nun) ‘dog-GEN’
djama-n (*djama-Nun) ‘snake-GEN’
b. Consonant-final stems → [-Nun]
girrgirr-Nun (*girrgirr-n) ‘bush canary-GEN’
gañal-Nun (*gañal-n) ‘goanna-GEN’
bibuy-Nun (*bibuy-n) ‘child-GEN’
6
24.962, Spring 2018 Class 21 — 5/3/18
• We can describe this distribution using VI rules in the morphological component of the grammar where
the context is comprised of phonological information.
(14) VI rules for Dja:bugay genitive
a. GEN ↔ [-Nun] / [STEM ...C]__
b. GEN ↔ [-n] (elsewhere)
• However, as Nevins (2011:esp. 14–16) points out for similar examples, this approach seems to be missing
a crucial phonological generalization:
◦ The [-Nun] allomorph seems to be used in order to avoid a complex coda (word-final CC cluster).
(15) a. /gañal+GEN/ 9 *gañal-n (violates *CC#) → gañal-Nun
b. /guludu+GEN/ → guludu-n (satisfies *CC#) 9 *guludu-Nun
• Nevins (2011:22) outlines a way to integrate the phonological explanation into a morphological insertion
account: mention constraint violation/satisfaction in the phonological context.
(16) VI rules for Dja:bugay genitive (revised)
a. GEN ↔ [-Nun] if it removes/avoids a *CC# violation
b. GEN ↔ [-n] (elsewhere)
→ This, though, would seem to require a sort of comparison/look-ahead which is odd in (standard) DM.
◦ In order to know to use the more specific allomorph, you need to first know that the default allomorph
would yield a violation.
◦ Yet the intrinsic ordering inherent to VI seems to suggest that you discharge the earlier rule without
reference to the later rule.
3.1 Sub-categorization
• Paster (2006, 2009, 2015) advocates for a “sub-categorization” approach to these sorts of facts.
◦ See also Yu (2007) on using this approach for infixation.
◦ It falls broadly within the Sign-Based Morphology / Construction Grammar approach developed by
Sharon Inkelas and other Berkeley people.
• In this approach, each allomorph is assigned its own sub-categorization frame.
◦ This sub-categorization frame may include phonological restrictions about which types of stems the
allomorph can attach to.
◦ It may also specify its own cophonology (depending on how far into that theory you want to go).
◦ (This is also where (morpho)syntactic and semantic conditions on affixation live.)
• (Putting aside that last point,) if we adopt Nevins’ suggestion about what sort of phonological information
can be contained in the context of VI rules, then two approaches become very similar, so I’m not going to
go into further.
7
24.962, Spring 2018 Class 21 — 5/3/18
4 Allomorph selection in the phonology, and the question of optimization
• In most cases, we can do these same things (and maybe better) in the phonological component itself.
◦ Several specific proposals exist, but they mostly have the property of allowing multiple morphs to
co-exist in the input to the phonological component, and allowing constraints in CON to adjudicate
between them.
◦ The upshot of this is that standard phonological constraints be employed to choose between the
different morphs in the part of the grammar where they’re supposed to be, i.e. in the phonology.
• This general approach is attributed to McCarthy & Prince’s early work in OT (McCarthy & Prince
1993a,b), and often referred to as the “P[honology] M[orphology]” approach.
→ If/when “phonological” constraints outrank morphological “constraints”, the phonology can cause the
morphology to do things it doesn’t want to do.
◦ e.g. use a dispreferred allomorph, displace morphemes within the word, etc.
◦ (There usually isn’t a clear definition given for what counts as a P constraint and what counts as an M
constraint...)
Important Prediction:
• These approaches predict that all phonologically conditioned allomorphy must be phonologically
optimizing; i.e. whenever a non-default allomorph surfaces, it must improve the phonological struc-
ture in that case relative to the default allomorph.
• This is because the only thing which can divert a derivation away from the default allomorph is
high-ranking phonological constraints.
• I’ll exemplify this approach with “U SE:X” constraints (loosely based on MacBride 2004), which works as
follows:
◦ The phonological input consists of all possible (allo)morphs associated with a given morpheme (i.e.
morphosyntactic feature (bundle))
≈ The output of Vocabulary Insertion is a set of morphs.
◦ Each morph is associated with a U SE constraint, which is ranked in CON.
◦ Definitionally, U SE :D EFAULT U SE :A LTERNATIVE
→ If P U SE :D EFAULT, and P is violated by the output form with the default morph, then the alternative
morph will be selected.
? Mascaró (2007) and Bonet, Lloret, & Mascaró (2007) use a constraint they call “P RIORITY”, which does
exactly this but with a single constraint. The relevant lesson and mechanics are essentially the same.
4.1 Dja:bugay genitive allomorphy with U SE :X constraints
• We can easily implement the above insights of Dja:bugay genitive allomorphy with this approach (see
Kager 1996):
(17) VI rule for the Dja:bugay genitive: GEN ↔ /{-n, -Nun}/ [no context needed]
• We take /-n/ to be the default morph, and /-Nun/ to be the alternative morph; therefore:
(18) Ranking: U SE :/-n/ U SE :/-Nun/
8
24.962, Spring 2018 Class 21 — 5/3/18
• The default fails to surface just in case it would create a final CC cluster; therefore:
(19) Ranking: *CC# U SE :/-n/
• These pieces put together derive the full distribution:
(20) Dja:bugay genitive allomorphy with U SE :X constraints
a. Vowel-final stems → [-n] (default)
/guludu+{-n, -Nun}GEN / *CC# U SE :/-n/ U SE :/-Nun/
a. + guludu-n *
b. guludu-Nun *!
b. Consonant-final stems → [-Nun] (alternative, driven by *CC#)
/gañal+{-n, -Nun}GEN / *CC# U SE :/-n/ U SE :/-Nun/
a. gañal-n *! *
b. + gañal-Nun *
• The point is: the (accidental?) existence of an alternative allomorph allows the phonology to solve a phono-
tactic problem by morphological means rather than purely phonological means.
◦ Why not solve it by phonological means, e.g. epenthesis? Ranking.
• Whether the language otherwise allows final clusters (D EP V-IO *CC#) or otherwise repairs them via
epenthesis (*CC# D EP V-IO) [I don’t know which is true], as long as D EP V-IO U SE :/-n/, selecting
the alternative allomorph will be less costly than performing epenthesis.
(21) Allomorphy not epenthesis
/gañal+{-n, -Nun}GEN / D EP V-IO *CC# U SE :/-n/ U SE :/-Nun/
a. gañal-n *! *
b. + gañal-Nun *
c. gañal-in *! *
4.2 Is PCSA always optimizing?
• When the selection between different morphs is phonologically optimizing (i.e. improving on some high-
ranked phonological constraint), the P M approach works very nicely.
• The problem is, there are at least some cases where it doesn’t look optimizing.
4.2.1 Apparently phonologically arbitrary distributions
• There are a number of PCSA patterns where the allomorphs don’t seem to serve any different phonological
function.
9
24.962, Spring 2018 Class 21 — 5/3/18
(22) Arbitrary PCSA patterns (Nevins 2011:15)
• In both of these cases, you get an alternation in the featural composition of a consonant, but this alternation
serves no discernible purpose.
→ At present, such cases look like they do require arbitrary specification in the morphology.
• However, if some phonological motivation could be conjured up, then we could do without it.
◦ It seems like most of the apparently arbitrary patterns are “syllable-counting allomorphy”, i.e. the
distribution of different morphs appears to be governed by the syllable count of the stem.
◦ I don’t think we have a good handle on how these work to begin with.
4.2.2 Apparently(/allegedly) phonologically “perverse” distributions
• Some PCSA distributions seem not only arbitrary, but actually counter to expected phonological patterns
— “perverse” in Paster‘s (2015) terms.
• The banner case is definite suffix allomorphy in Haitian Creole (see Klein 2003, Paster 2015:229, and
other references therein).
◦ Consonant-final stems take -la, but vowel-final stems take -a, yielding a hiatus which could have been
avoided by using -la.
(23) Haitian Creole definite suffix (data taken from Paster 2015:229)
a. Consonant-final stems → [-la]
pitit-la ‘the child’
ãj-la ‘the angel’
kay-la ‘the house’
madãm-la ‘the lady’
b. Vowel-final stems → [-a]
panié-a ‘the basket’
trou-a ‘the hole’
figi-a ‘the face’
chẽ-a ‘the dog’
10
24.962, Spring 2018 Class 21 — 5/3/18
• If we were to assume that the relevant phonological factor were O NSET, this would be impossible to
explain in the phonology.
→ Paster (2009, 2015) and others thus use this case to argue against the P M approach.
• But that’s probably not what it’s actually about:
◦ Following Klein (2003) (Nevins 2011:8), the relevant fact is morphophonological alignment (or some-
thing similar...).
◦ Using -la with consonant-final stems allows them to avoid re-syllabification across the morpheme
boundary, which would create a mismatch between syllable boundary and morpheme boundary.
(24) A LIGN (S TEM , R; σ, R)
Assign a violation * if the right edge of the stem does not coincide with the right edge of a syllable.
◦ This constraint will also be satisfied by suffixing -a to vowel-final stems.
◦ If -a is the default, and U SE :/a/ O NSET, this all works out in the phonology.
(25) Haitian Creole definite suffix allomorphy
a. Vowel-final stems → [-a] (default)
/trou+{-a, -la}DEF / A LIGN -R U SE :/-a/ U SE :/-la/ O NSET
a. + trou.-a * *
b. trou.-la *!
b. Consonant-final stems → [-la] (alternative, driven by A LIGN)
/trou+{-a, -la}DEF / A LIGN -R U SE :/-a/ U SE :/-la/ O NSET
a. piti.t-a *! *
b. + pitit.-la *
• This looks good, but there’s actually a problem.
◦ Consider candidate (c) below, where you simply don’t resyllabify.
◦ This should win if the relevant constraint is A LIGN -R, because O NSET has to be ranked low to explain
the -a forms.
(26) Consonant-final stems → [-la] (alternative, driven by A LIGN)
/trou+{-a, -la}DEF / A LIGN -R U SE :/-a/ U SE :/-la/ O NSET
a. piti.t-a *! *
b. § pitit.-la *!
c. L pitit.-a * *
• I think this may be saveable if we replace the notion of syllabification with the notion of CV transitions.
◦ Assuming that transitions are automatic (unlike syllabification), there should be no substantial differ-
ence (a) and (c).
◦ Something like D EP[CV transitions] would rule out (a) and (c), in favor of (b). It would also not mess
with the vowel-final forms.
→ Maybe some of the other weird patterns can be explained if we think more creatively about the phonetics.
11
24.962, Spring 2018 Class 21 — 5/3/18
4.3 Opacity in allomorph selection
• One of the other main points of contention between the “allomorphy in the morphology” people and the
“allomorphy in the phonology” people is the status of opacity in allomorph selection.
• There are a number of patterns where allomorph selection appears to be based on the underlying represen-
tation not the ultimate output; for example, Turkish, Japanese, Galician (see, e.g., Nevins 2011:17–18).
◦ Most of these are cases of allomorphy determined by adjacent C vs. V, but then the relevant C is later
deleted.
• Paster uses such cases to argue for selection in the morphology, where only the UR is visible.
• Nevins points out, however, that this could simply be garden-variety opacity.
◦ Allomorph selection could precede the opacifying phonological process.
◦ This is doable in phonological frameworks which permit intermediate levels of representation, e.g.
Lexical Phonology / Stratal OT, or “Optimal Interleaving” in OT with Candidate Chains (OT-CC)
(Wolf 2008, 2015).
→ Opacity in allomorph selection may be amenable to general approaches to opacity, and thus does not on
its own necessarily decide between the different views of allomorphy.
References
Albright, Adam & Bruce Hayes. 2003. Rules vs. Analogy in English Past Tenses: A Computational/Experimental Study. Cognition
90(2):119–161.
Bobaljik, Jonathan David. 2000. The Ins and Outs of Contextual Allomorphy. In University of Maryland Working Papers in
Linguistics 10, 35–71. http://home.uni-leipzig.de/muellerg/dm4.pdf.
———. 2012. Universals in Comparative Morphology: Suppletion, Superlatives, and the Structure of Words. Cambridge, MA:
MIT Press.
Bonet, Eulàlia, Maria-Rosa Lloret & Joan Mascaró. 2007. Allomorph Selection and Lexical Preferences: Two Case Studies. Lingua
117:903–927.
Bye, Patrik. 2008. Allomorphy: Selection, not Optimization. In Sylvia Blaho, Patrik Bye & Martin Krämer (eds.), Freedom of
Analysis?, 63–92. Berlin & New York: Mouton de Gruyter.
Carstairs-McCarthy, Andrew. 2001. Some Implications of Phonologically Conditioned Suppletion. In Charles W. Kreidler (ed.),
Phonology: Critical Concepts, vol. V (The Interface with Morphology and Syntax), 111–139. London & New York: Rout-
ledge. [Originally in Yearbook of Morphology 1988.] Available on Google Books.
Guthrie, Malcolm. 1956. Observations on Nominal Classes in Bantu Languages. Bulletin of the School of Oriental and African
Studies, University of London 18(3):545–555. http://www.jstor.org/stable/610117.
Halle, Morris & Alec Marantz. 1993. Distributed Morphology and the Pieces of Inflection. In Ken Hale & Samuel Jay Keyser
(eds.), The View from Building 20: Essays in Honor of Sylvain Bromberger, 111–176. Cambridge, MA: MIT Press.
Kager, René. 1996. On Affix Allomorphy and Syllable Counting. In Ursula Kleinhenz (ed.), Interfaces in Phonology (Studia
Grammatica 41), 155–171. Berlin: Akademie Verlag.
Klein, Thomas B. 2003. Syllable Structure and Lexical Markedness in Creole Morphophonology: Determiner Allomorphy in
Haitian and Elsewhere. In Ingo Plag (ed.), The Phonology and Morphology of Creole Languages, 209–228. Tübingen: Max
Niemeyer.
MacBride, Alexander Ian. 2004. A Constraint-Based Approach to Morphology. PhD Dissertation, UCLA.
Mascaró, Joan. 2007. External Allomorphy and Lexical Representation. Linguistic Inquiry 38(4):715–735.
McCarthy, John J. & Alan Prince. 1993a. Generalized Alignment. In Geert Booij & Jaap van Marle (eds.), Yearbook of Morphology
1993, 79–153. Kluwer.
———. 1993b. Prosodic Morphology I: Constraint Interaction and Satisfaction. Linguistics Department Faculty Publication Series
14 (2001 version). http://scholarworks.umass.edu/linguist_faculty_pubs/14.
Nevins, Andrew. 2011. Phonologically Conditioned Allomorph Selection. In Marc van Oostendorp, Colin J. Ewen, Elizabeth Hume
& Keren Rice (eds.), The Blackwell Companion to Phonology, vol. IV (Phonological Interfaces), 1–26. John Wiley & Sons,
Ltd. https://onlinelibrary.wiley.com/doi/abs/10.1002/9781444335262.wbctp0099.
12
24.962, Spring 2018 Class 21 — 5/3/18
Paster, Mary. 2006. Phonological Conditions on Affixation. PhD Dissertation, University of California, Berkeley. http://
escholarship.org/uc/item/7tc6m7jw.pdf.
———. 2009. Explaining Phonological Conditions on Affixation: Evidence from Suppletive Allomorphy and Affix Ordering. Word
Structure 2(1):18–37.
———. 2015. Phonologically Conditioned Suppletive Allomorphy: Cross-linguistic Results and Theoretical Consequences. In
Eulàlia Bonet, Maria-Rosa Lloret & Joan Mascaró (eds.), Understanding Allomorphy: Perspectives from Optimality Theory,
218–253. UK: Equinox.
Patz, Elisabeth. 1991. Djabugay. In R. M. W. Dixon & Barry J. Blake (eds.), The Handbook of Australian Languages 4, 245–347.
Melbourne: Oxford University Press.
Pinker, Steven & Alan Prince. 1988. On Language and Connectionism: Analysis of a Parallel Distributed Processing Model of
Language Acquisition. Cognition 28(1–2):73–193.
Wolf, Matthew. 2008. Optimal Interleaving: Serial Phonology-Morphology Interaction in a Constraint-Based Model. PhD Disser-
tation, University of Massachusetts, Amherst.
———. 2015. Lexical Insertion Occurs in the Phonological Component. In Eulàlia Bonet, Maria-Rosa Lloret & Joan Mascaró
(eds.), Understanding Allomorphy: Perspectives from Optimality Theory, 361–407. UK: Equinox.
Yu, Alan C. L. 2007. A Natural History of Infixation. Oxford: Oxford University Press.
13
