DISTRIBUTED MORPHOLOGY*
Jonathan David Bobaljik
U Conn • Storrs
August 2015
Distributed Morphology (DM) is a theoretical framework that emerged in the early
1990s. The name is introduced in Halle and Marantz (1993, 1994), with important
precursors including Halle (1990), Bonet (1991), Noyer (1997), and Pesetsky (1995). At
a broad level, DM represents a set of hypotheses about the interaction among components
of grammar: Morphology, in DM, is (a part) of the mapping from the output of a
syntactic derivation to the (input to) the phonology, as sketched in (1):1
(1) Architecture
Syntactic derivation
Output (Spell-Out)
Morphology
Phonology Semantics
DM is centred around two key hypotheses; both common in various parts of the
literature—where DM claims novelty is in the synthesis of these two leading ideas under
the architecture in (1):
(2) Syntax-all-the-way-down: The primary mode of meaningful composition in the
grammar, both above and below the word-level, is the syntax. Syntax operates on
sub-word units, and thus (some) word-formation is syntactic.
(3) Late Insertion / Realization: The pieces manipulated by the syntax (functional
morphemes) are abstract, lacking phonological content. The pairing of
* This chapter attempts to elucidate some core ideas in the framework of Distributed Morphology, but does
not aim to be a thorough review of the relevant literature. For conversations that have been particularly
helpful in formulating the perspective presented here, I thank Mark Baker, Andrea Calabrese, David
Embick, Alec Marantz, Andrew Spencer, Susi Wurmbrand, Jochen Trommer and members of the Network
Grundmechanismen der morphologischen Exponenz, as well as the many class participants with
whom I have discussed the merits of competing morphological theories.
1 There are a variety of views on the relationship of Spell-Out to LF/Semantics that are consistent with the
overall DM architecture. On one view, there is ‘covert’ syntax mapping Spell-Out to LF, on another view,
overt and covert movement are interleaved, distinguished by whether a moved element is pronounced in a
high or low position, and thus Spell-Out takes place after ‘covert’ movement. On this latter view (see
Bobaljik 2002a and references therein), the representation that is the input to Morphology corresponds to
LF in GB/Minimalist usage. Related to these differences is the question of whether Spell-Out applies to
portions of the syntactic derivation in cyclic fashion (see Embick 2010 for discussion, and section 4.1
below).
2
phonological features with the terminals of the syntax (vocabulary insertion or
exponence) happens post-syntactically, in the mapping from syntax to
phonological form (PF).
The functions of morphology in other approaches, and of the Lexicon in particular, are in
DM distributed (hence the name) over multiple points in the architecture. In particular,
there is no single Lexicon, understood as at once a list of (i) the minimal meaningful units
of grammar or building blocks of words, (ii) the minimal pairings of form (sound) and
function (meaning), and (iii) non-compositional aspects of the meaning of
words/morphemes in particular contexts. Instead, there is on the one hand a list of the
syntactic atoms, manipulated by (and thus accessed by) the syntax, in the construction of
complex terminal nodes. Items on this list would include features that project to a
syntactic node (say [PLURAL]), and (possibly language-particular) bundles of features that
constitute a single node: for example English (plausibly) groups both tense and
agreement (person and number) under a single INFL node in the syntax. A second list, the
Vocabulary, associates morphosyntactic features and their phonological exponents.
Idiosyncratic meanings of morphemes in context (idiomaticity) is part of a third list—the
Encyclopaedia—discussed in section 5.
It is a matter of observation that there are prima facie mismatches between the abstract
nodes of the first list and the nodes that are spelled out by vocabulary items; in other
words, that the correspondence between the lists is not one:one. One such mismatch is
underspecification of vocabulary items relative to feature contrasts in the syntax. For
example, in Russian, predicate adjectives (and participles) show agreement for number
and gender with the subject of their clause, even, as in first and second person pronouns,
where gender is not contrastive:
(4) a. ja bol’n-a I sick-FEM ‘I am sick’
b. ja bolen-Ø I sick-MASC ‘I am sick’
In DM, as in other realizational frameworks, such examples may be modelled by positing
that the syntactic representation is fully specified, and thus that the subject position
includes a gender contrast, but that the Russian Vocabulary only has items such as (5),
which lacks a gender specification and thus realizes both feminine and masculine
subjects. This leaves a many:one relation between the morphosyntactic and
morphophonological representations, with fewer contrasts in the overt form than are
present underlyingly in the syntax.
(5) [1 SG NOM] ó ja
Allomorphy provides a paradigm example of the opposite situation, in which the relation
between syntax and phonological realization is one:many. In German, for example, there
is no reason to suspect that there is more than one nominal plural suffix in the syntax;
hence plural nouns may all have the abstract syntactic representation: [[NOUN]-PL].
However, the German vocabulary provides a variety of vocabulary items that express this
node, including: -Ø, -(e)n, -e, -er and -s, with or without concomitant stem changes
(readjustments, see below).
The examples of mismatches just considered exploit the realizational nature of the
framework. Other examples of mismatches arise between the syntactically motivated
hierarchy or constituency and the morphophonologically evident hierarchical structure. In
the development of the framework, various mechanisms have been posited to account for
3
these, chiefly as operations that manipulate the output of the syntax prior to vocabulary
insertion; see section 3.
For example, it is a mainstay of much syntactic theorizing building on Chomsky’s
influential (1957) analysis of do-support that the inflectional features in English head a
syntactic node (INFL) that is independent of the verb, as in (6a).2 Yet, evidently, in simple
declaratives, the main verb and inflectional features form a morphological unit, as in
(6b)—a verb, such as walk-ed.
(6) a. IP b. X˚
3 3
3 V˚ INFL˚
INFL˚ VP
3
3
V˚
If (6b) is derived syntactically, for example via head movement, then the levels align. But
word order contrasts, famously between French and English (Pollock 1989), have led
some researchers to conclude that English lacks a syntactic head movement operation
deriving (6b). Thus, main (non-auxiliary) verbs in French precede elements (adverbs,
negation) that English main verbs follow, and only in a verb-movement language like
French does the main (non-auxiliary) verb invert with the subject in inversion contexts. If
it is indeed correct that English lacks a syntactic operation deriving (6b) from (6a), then
we are faced with an instance of a mismatch: INFL˚ and V˚ do not form a constituent in
the syntax, but they do form such a unit in the morphology. An important line of study
within DM seeks to propose a theory of possible mismatches of this sort. For the case at
hand, an operation of Morphological Merger has been appealed to. Various formulations
have been offered, such as (7), based on Marantz (1989:261):
(7) Morphological Merger
A syntactic complementation relation: [ X˚ YP ]
may be realized in the morphology as an affixation relation:
X affixed to Y, the head of YP: [[ Y ] X ] or [[ X [ Y ]]
Merger in (7) reconstitutes Chomsky’s Affix-Hopping as a morphological operation.
Since English inflectional morphology is exclusively suffixing, while syntactic
complementation is head-initial, combining INFL as an affix to the verb will necessarily
result in a suffix, and the appearance of INFL ‘hopping’ across the verb.3
Cliticization provides related examples where again, syntactic constituency and morphophonological
constituency are imperfectly aligned, with a hopping effect. Marantz
(1989), drawing on Sadock (1985) gives the example of the Latin conjunctive clitic =que.
Syntactically, the coordinator is presumably a head that occurs between the conjuncts,
2 In many current syntactic analyses within the GB/Minimalist tradition, there is much additional structure
as well which we may ignore for the sake of exposition.
3 If the affixation relationship that is the output of Merger could also be Y affixed to X: [[ X ] Y ], then
Merger would be morphologically indistinguishable from Head Movement. This provides one way of
thinking about proposals to recast syntactic head movement in non-syntactic terms.
4
either in a flat structure as in (8a), or as the head of a more articulated coordination
phrase. Yet the morpho-phonological constituency does not reflect this—the coordinator
=que occurs after the first phonological word of the last conjunct in a coordination.
(8) a. NP
q|p
NP and NP
b. [ [ bon-i puer-i ] & [ pulchr-ae puell-ae ] ]
c. [ [ bon-i puer-i ] [ pulchr-ae –que puell-ae ] ]
good-M.PL boy-M.PL beautiful-F.PL-AND girl-F.PL
‘good boys and beautiful girls’
Embick and Noyer (2001) propose that this too involves a species of Morphological
Merger. They note in particular that this is unlikely to involve syntactic movement. Not
only is there no independently motivated syntactic analysis that would readily
accommodate (8c), but the ordering is in some cases demonstrably dependent on
phonology: if there is a preposition in the second conjunct, then =que attaches to the
preposition when the latter constitutes a phonological word, but when the preposition is
light and phonologically dependent on the word to its right, then =que follows both the
preposition and the noun:
(9) a. circum-que ea loca
around-AND those places
‘and around those places’
b. in rēbus-que
in things-AND
‘and in things’
Examples of this sort not only serve to illustrate the kinds of operations posited within
DM for analyzing mismatches between syntactic and morpho-phonological constituency,
they also call attention to the dependency of morphological analysis on assumptions
about the syntax. If English has verb movement to INFL after all, or if the Latin
conjunction facts do turn out to involve syntactic movement, then the morphological
operations needed would of course be different.4 It is worth noting in this context the
emerging perspective of Nanosyntax (Starke 2009, Caha 2009, Svenonius 2012), which
shares the core tenets of DM in (2)-(3), but holds, in effect, on conceptual grounds that a
more elegant theory would make no recourse to post-syntactic operations that alter the
syntactic representation prior to vocabulary insertion. Of course, simplification of the
morphological component by eliminating such operations is not an argument in and of
itself, to the extent that apparent simplification in one domain may come at the expense
of unwarranted complexity in the other. As theories of syntax:morphology mismatches,
DM accounts are answerable to both the syntax and the morphology in justifying
assumptions in both components.
To recap, then—as a framework for grammatical analysis, DM takes a derivational
stance. The derivation of a complex word begins in the syntax, which combines abstract
4 Similar effects have been noted in Slavic, where apparently phonologically-restricted cliticization feeds
syntactic movement. See Radkevich (2010) and for an analysis in which morphological cliticization is
cylically interwoven with syntactic movement, see Calabrese and Pescarini (2014).
5
(i.e., phonology-free) pieces according to general principles. It is the output of the syntax
that is then interpreted by the Morphology. A key part of the morphology, conceived of
as the mapping from syntax to phonological form, is vocabulary insertion, the pairing of
syntactic terminals with (possibly null) phonological underlying representations. Between
these two ends of the morphological derivation, a variety of operations may apply which
manipulate the representation in limited ways, including via Morphological Merger (of
terminals), illustrated above. With this general outline in mind, we turn to a more refined
presentation of some key elements of the theory.
1. REALIZATION AND UNDERSPECIFICATION
Late insertion, or realization, of morphosyntactic representations was discussed above in
connection with (4)-(5). As in other realizational frameworks, a central role in DM is
accorded to underspecification of vocabulary items. A well-worn example of
underspecification is provided by the English present tense inflection (of main verbs).
The English vocabulary contains the following two items, which are candidates for
realizing in INFL node in (6b).5
(10) Vocabulary of English (fragment)
a. [ 3SG, PRES ] ó -s
b. [ PRES ] ó Ø
The formal statements of vocabulary items are tantamount to rules of exponence (cf.
Matthews 1972), and as such, two general principles of rule interaction are operative in
their application.6
(11) Rules Apply
A rule applies wherever its structural description is met.
(12) Elsewhere Condition
Where more than one mutually exclusive rule may apply, (only) the most highly
specified rule applies.
5 By convention, fragments of the vocabulary relevant to a given point are presented as a disjunctively
ordered list, from most to least specific. Standard DM notation writes these as correspondence rules (ó)
relating a morphosyntactic representation to a phonological one. An alternative within DM treats these as
rewrite rules, replacing syntactic features with phonological ones (i.e., with à in place of ó; see Halle
1990, Trommer 1999, Bobaljik 2000). The two views may differ, for example, in whether features that
have been spelled out remain visible in the derivation to serve as a condition for later rules. See Bobaljik
(2000) and Bonet and Harbour (2012) for relevant discussion.
6 In Halle and Marantz (1994:276), these are grouped together under the rubric “Underspecification”, and
in Halle (1997), (11)-(12) are conflated into a single Subset Principle. Note that this is largely a matter of
exposition and nomenclature: formulations such as “only vocabulary items whose specified features are a
subset of the features in a given terminal node are able to [be inserted at that] node” (Harley, 2008:263) are
simply elaborations of (11), but there is no contentful “subset principle” postulated beyond (11). In DM, as
elsewhere, there are competing views as to the precise formulation of (12), in particular, whether “more
specified” should be limited to subset-superset relations (as in Kiparsky 1973), or should include other
cases, such as feature-counting (Halle 1997, Harley 2008:262).
6
Returning to English inflection—if the subject is, say, first person plural, then the
features [1PL, PRES] will appear at the INFL node in (6b) and constitute the input to
vocabulary insertion. The item (rule) in (10a) may not apply, as its structural description
is not met; only (10b) is compatible with this context. On the other hand, where the
subject is third person singular, both exponents in (10) are eligible for insertion, but as
(10a) is more specific (a proper subset of the environments characterized by (10b)), (10a)
must be inserted: She walk-s, and not *She walk-Ø.
German provides a further illustration. Consider the agreement morphology in the simple
past tense, illustrated in (13) (German shows separate terminal nodes for tense and
agreement, unlike English):
(13) German (weak) Past Tense: sagen ‘to say’
a. [2 PLURAL] ⇔ -t
b. [PLURAL] ⇔ -n
c. [2] ⇔ -st
d. [ ]7 ⇔ Ø
A possible analysis of the agreement exponents is given in (13a-d). Note in particular that
all four exponents are eligible to express the second person plural (their structural
descriptions are all met in the context [2 PLURAL]), but the Elsewhere Condition correctly
regulates this competition, determining a unique winner. Formally, the suffix -en marks
only the plural, and is not sensitive to person; the impossibility of *sag-t-en as a 2PL form
is attributed to the more highly specified item -t, which wins the competition in that
context.
The same (familiar) elsewhere logic regulates the competition among lexically
conditioned allomorphs, providing a formal description of the interaction of regular and
irregular affixes, as in (14):
(14) Vocabulary of English (fragment)
a. [ PAST ] ó -t / ]V__ ; where V ∈ {dream, dwell etc.}
b. [ PAST ] ó Ø / ]V__ ; where V ∈ {run, hit, fly etc.}
c. [ PAST ] ó -d / ]V__
In the usual manner, the (intrinsic) elsewhere ordering in (14) ensures that the irregular
exponents block the regular past tense exponent for verbs that are listed as irregular
(whether this is formalized as a list of restrictions on the rules, as in (14), or via a
diacritic). Two comments are worthy of note here. First, the vocabulary items of DM are
to be understood as phonological underlying representations, subject to further
phonological rules—(14c) has three phonologically predictable surface realizations: [d ~ t
~ əәd].8 Second, DM separates choice of affixal exponent from stem alternations, such as
vowel changes (if any)—the latter are formally the purview of readjustment rules
(equivalently: morphophonological rules, minor phonological rules, etc.). Readjustment
7 An empty feature set, the elsewhere case, is consistent with any environment and thus constitutes the
default realization of a given node.
8 Many morphologists thus reserve the term ‘allomorphy’ in this context to alternations that must be
handled by the morphology (as in (14)), and thus do not include phonologically-predictable changes, such
as voicing alternations in the English past tense and plural, under this term.
PERSON SINGULAR PLURAL
1st
sag-te-Ø sag-te-n
2nd
sag-te-st sag-te-t
3rd
sag-te-Ø sag-te-n
7
rules alter the form of an underlying representation, and thus they necessarily occur after
the rules of vocabulary insertion. Readjustment rules are fundamentally phonological in
nature, but are restricted to apply to some, but not all, morphemes that meet their
phonological description. The dissociation of readjustment rules from affixal exponents is
clearest in the case of verbs that make use of both: tell ~ tol-d.
Elsewhere-governed competition among vocabulary-items may also be invoked to
capture more abstract competitions. Adding (15) to the Vocabulary fragment is one way
of capturing the fact that in English, verbs which lack an -(e)n participle form take the
same affix (including -Ø) in the participle as they do in the simple past (I have dwel-t /
dream-t / hit-Ø / sung-Ø …), even though they may show a different stem alternation
(readjustment).9
(15) PAST, PARTICIPLE ó -(e)n / ]V__ ; where V ∈ {write, give, etc.}
Although the above examples are drawn from inflection, competition in DM is used to
model allomorphy in derivational morphology as well. Descriptions of many languages
may include an array of phonologically distinct affixes performing the same function, for
example, the nominalising suffixes of English include: -ness, -ity, -th, etc., which
combine with different ranges of adjectival stems to yield nouns. To the extent that these
affixes have systematically differing syntax or semantics, then they may indeed realize
distinct abstract morphemes. On the other hand, the mechanisms of the theory allow for
the possibility that these all realize a single abstract morpheme, and constitute instead
surface allomorphs (different exponents) of, say, a syntactic head n˚ which merely
contributes the category feature Noun (see, e.g., Marantz 1997, Harley 2009, Embick
2010; see Lowenstamm, 2010 and Van Craenenbroeck and De Belder, in press, for an
alternative conception.)
In sum, one of the pillars of DM, shared with a wide variety of competing frameworks, is
the property of realization, invoking underspecified rules of exponence to provide the
morphological/phonological realizations to logically prior, abstract morphosyntactic input
representations. Morphology interprets, rather than projecting, syntactic structure.
It should be noted that the sense of underspecification used here refers specifically to the
characterization of items in the vocabulary. Understanding vocabulary items as
constituting rules of exponence, underspecification in this sense is simply the general
property of rules that they may have differing levels of generality in their structural
description. A separate question is whether the morphosyntactic representations that are
the input to vocabulary insertion may be underspecified, that is, whether a node may lack
a value for a feature that it (otherwise) bears. The distinction between the two (distinct
but overlapping) senses of underspecification may be illustrated with reference to
Russian nominal gender.
9 Many approaches to morphology distinguish a root (the most deeply embedded morpheme) from a stem
(the possibly internally complex element to which inflection attaches); in the word destroying, the prefix
de- combines with the root √STROY, to form a stem, to which the suffix –ing is attached. This
terminological distinction may be useful in description, but the category of stem as such has no privileged
morphological status in DM. In particular, while DM recognizes domains for morphophonological
interactions (see section 4.1 below), to the extent there are effects that seem to pick out the ‘stem’, these
arise only to the extent that stems happen to coincide with other recognized domains.
8
Like many languages, Russian lacks a morphological gender contrast in the plural; as the
nominative, third person pronominal paradigm in (16) illustrates.10
(16) Russian gender inflection (short adjectives, nominative case)
a. [PLURAL] ⇔ -i
b. [FEM] ⇔ -a
c. [NEUT] ⇔ -o
d. [ ] ⇔ Ø
Within DM, Halle (1997) proposes to describe the Russian paradigm with the vocabulary
items in (16a-d). We may assume that nominal inflectional nodes may abstractly be
specified for both gender and number, hence a node may bear the specification
[FEM,PLURAL]. In Halle’s account, this node will be realized as -i, because this item is
ranked highest on the disjunctive list of competing vocabulary items in (16). In contrast
to the examples considered above, where the Elsewhere Condition established an intrinsic
ordering among competing exponents, in Halle’s analysis, the ordering of (16a) before
(16b-c) is extrinsic—in essence, an arbitrary fact of Russian that needs to be learned.
Alternatively, as Harley (1993), Noyer (1997) and others have suggested, the order
among the exponents in (16) may be established by a feature hierarchy. Noyer (1997)
posits the hierarchy PERSON > NUMBER > GENDER, which ensures that number wins out
over gender when no other considerations establish order, as in the case at hand.
With or without a hierarchy, Halle’s account treats the lack of gender contrasts in the
plural as a contingent property of Russian grammar, a consequence of the inventory of
vocabulary items in the language. Williams (1994) objects to accounts of this general
type on the grounds that they fail to capture what appears to be a systemic, rather than
accidental, property of Russian morphology (see also Bobaljik 2002b, Harley 2008). The
syncretism seen in (16)—the lack of a gender contrast in the plural—is not merely a
property of the short nominative inflectional endings, but is instead a meta-syncretism,
holding of all inflectional paradigms in the language—gender is never morphologically
contrastive in the plural (a pattern that is not uncommon cross-linguistically, cf. Corbett -
1991).
An alternative to using underspecified vocabulary items, as in (16), is to countenance
underspecification of the morphosyntactic representations of the nodes that are the input
to vocabulary insertion. Bonet (1991, 1995) argued that the morphological operations in
(1) include feature-deletion rules, which she termed Impoverishment rules. These rules
operate on fully-specified syntactic matrices, but delete features prior to vocabulary
insertion, thereby yielding systematic neutralizations in surface forms.11 Stating (17)
10 The inflectional suffixes of these pronouns are shared with other nominal classes, including interrogative
and demonstrative pronouns, nouns, short form adjectives, and the past tense (participle) inflection. An
influential analysis of the masculine singular ending (surface Ø) treats it as underlyingly a yer vowel, which
is subject to a rule of yer-deletion in (among other contexts) word-final position.
11 Bonet’s primary focus was Romance clitic clusters, in which combinations that are expected on the basis
of the clitics that appear in isolation are (in part) replaced by unexpected clitics, drawn from elsewhere in
the inventory. A famous case is the Spanish spurious se, in which, for example, the combination of third
person accusative lo and third person dative le, yields the opaque cluster se lo instead of expected, but
ungrammatical *le lo. Bonet argues that many such cases can be treated by deleting features of one or the
other clitic, leading to the insertion of less (featurally) marked exponents. For recent discussion, see, among
GENDER SINGULAR PLURAL
MASC on-Ø on-i
FEM on-a on-i
NEUT on-o on-i
9
once as a general Impoverishment rule of Russian will formally describe the metasyncretic
effect: at the point of vocabulary insertion, no node will ever bear both plural
and gender; underlying [FEM, PLURAL] will lose its feminine feature, and surface just as
[PLURAL]. For the same reason, this account will also eliminate the need for extrinsic
ordering in (16): (16a-c) will simply never compete and thus need not be ordered with
respect to one another.
(17) a. GENDER à Ø / [ ___ PLURAL]
b. * [ GENDER, PLURAL ]
Noyer (1997), extending ideas of Calabrese (1995), suggests that (many such)
impoverishment rules are the result of markedness constraints, such as (17b), which
underlie typological generalizations (see also Nevins 2011). For example, if (17b) is a
markedness constraint (which may or may not be active in a given language), but there is
no corresponding *[GENDER , SINGULAR] constraint, a good portion of Greenberg’s
Universal 37: ‘A language never has more gender categories in non-singular numbers
than in the singular’ (1963:112; cf. Corbett 1991:156) receives an account. Numerous
interesting questions arise at this point, and further discussion of deletion operations will
be taken up in section 3.3, below.
While the discussion above merely scratches the surface in many ways, the examples
given touch on the major motivations for, and applications of, a realizational approach to
morphology, one in which phonological exponents are associated with morphosyntactic
representations after the syntactic derivation. A central role is played here by the
investigation of mismatches between syntactically-motivated representations, and those
observed in the morphophonological string: a variety of devices, including underspecified
vocabulary items, as well as morphological operations such as Merger and
Impoverishment, serve together to constitute a theory of possible mismatches.
2. SPELLING OUT SYNTAX
While many theoretical frameworks incorporate a realizational architecture, in which
rules of exponence spell out the features of a morphosyntactic representation, frameworks
differ significantly in the properties they attribute to that representation. A central tenet of
DM, noted above, is that the starting point of the morphological (component of the)
derivation is the representation that is the output of the syntax, including, potentially,
internally complex X˚ nodes created in the syntax. This view was discussed with
reference to (6) above—even a simple case such as a finite verb in English is assumed to
have an internally complex hierarchical structure that is determined (in part) by the
syntax: (6b).12 Other realizational frameworks, in particular Word and Paradigm theories,
explicitly deny that the morphosyntactic representation has internal syntactic structure
(this view is laid out with particular clarity in Anderson 1992: Chapter 10).
others, Nevins (2007) and Pescarini (2010).
12 The qualification “in part” is relevant here because DM recognizes operations that manipulate the
syntactic representation prior to vocabulary insertion, such as Morphological Merger. Thus, the abstract
complex terminal [ [ verb ] INFL ] in (6b) may not (in English) be formed in the syntax per se (Chomsky’s
narrow syntax), but it nevertheless has an internal hierarchical structure that is a function of the syntactic
representation with distinct INFL and V nodes—(6a).
10
Central arguments for recognizing word-internal syntactic structure (and thus
composition of words in the syntax) come from two domains.
One domain, alluded to already, is the syntactic evidence for such structure, for example,
the evidence from English do-support for an INFL node in the syntax, distinct from the
verb. Similarly, many schools of thought recognize evidence for syntactic decomposition
of words in examples such as (18), with a transitive verb open meaning (something like)
CAUSE-TO-BE-OPEN. Sentences of this form are famously ambiguous (Dowty 1979, von
Stechow 1996), allowing (at least) a repetitive reading, as in (18a), and a restitutive
reading (18b), under which (18) is true so long as the window had been open before.13
(18) Leo opened the window again.
a. Leo opened the window, and he had [ opened the window ] before.
b. Leo opened the window, and [the window was open ] before.
If adverbs like again modify syntactic constituents, then the restitutive reading points to a
syntactic constituent consisting of the NP (the) window along with the stative portion of
the verb (i.e., BE.OPEN) but excluding the eventive component of the verb meaning, the
lower VP node in a structure like [VP CAUSE [VP BE.OPEN window ]]. We return briefly to
some related issues in section 5.
The other important domain of evidence for internally hierarchically structured
morphosyntactic representations comes from the morphology proper, and especially from
the domain of contextual allomorphy, and other (apparent) morpheme-morpheme
interactions within a word. For this reason, a major object of inquiry within DM is the
question in (19), continuing a research agenda laid out in one way or another in works
including Williams (1981), Siegel (1978), Lieber (1982), and Carstairs (1987):
(19) Under what conditions may one morpheme (M1) condition allomorphy for
(including the appearance or absence of) another morpheme (M2)?
Since syntactic features in DM may constitute (abstract) morphemes, DM would be
consistent with the discovery of conditions on allomorphy that are crucially defined over
hierarchical structures. Schematically, in DM, a root X associated with multiple
inflectional features may have an articulated morphosyntactic representation, such as
(20a), where the corresponding MSR in a Word-and-Paradigm theory would be flat, the
features associated as an unstructured bundle with the lexeme X (20b):
(20) a. 3 b. X [F1, F2]
3 [F2]
X [F1]
In DM, then, conditions such as (structural and/or linear) adjacency, and relative
closeness among features may turn out to play a key role in answering (19), where such
relationships are undefined if features are unstructured. Work in DM arguing for the
relevance of hierarchical structure in contextual allomorphy and other morphological
13 In the limiting case, the window is installed open, is closed once, and then opened by Leo. In such a
context, there is no repletion of an event of opening at all, merely the restoration of the window to the open
state. The current literature debates whether there are additional readings, suggesting even further
decomposition.
11
interactions includes: Bobaljik (2000, 2012), Embick (2003, 2010), Harbour (2007), and
Arregi and Nevins (2012); see also Bonet and Harbour (2012) for a review. Two
illustrative examples are presented here.
We begin with a fairly simple example. It is generally held (e.g., Lieber 1980, Kiparsky
1982) that morphological irregularity does not survive category changing derivation. The
basic verb fly in English has an irregular (strong) past tense, as in (21b). But the
denominal verb fly permits only a regular past tense flied (for example, in its baseball
sense - ‘to hit a fly ball’, (21c), as well as other senses identified in the OED).
(21) a. Superman will fly out.
b. Superman flew out. [ [ fly ]V INFL ]
c. Superman flied out. [ [ [ fly ]N ]V INFL ]
One account of this generalization is crucially structural: in (21b), the verb and INFL are
local to one another—the root conditions the zero allomorph of past tense INFL (see
(14b)), while the feature [PAST] in INFL triggers a readjustment rule on the verb root. By
contrast, in the denominal form (21c), additional (though unpronounced) structure
intervenes between the root and the INFL node, disrupting the local relation and
prohibiting morphological interactions. Only the default past tense INFL is possible, since
the lexical identity of the root (equivalently, a diacritic feature of the root) is not visible
to the INFL node.14
Keeping to well-discussed examples, another manifestation of the same concept is the
difference between nominalizations and gerunds. There are many vocabulary items that
express “nominalization” in English, and the choice of affix is in part lexically
determined by the root/stem: marri-age, refus-al, destruc-tion, break-Ø, etc. On the other
hand, gerunds (a species of nominalization) are always expressed by the item -ing:
marrying, refusing, breaking, and never show the idiosyncratic stem changes
characteristic of nominalization: note destroy-ing, rather than *destruct-ing. Here again,
one may propose (as Embick 2010 does) that the nominalising suffix is local to the root
in the root nominalizations (hence may interact in allomorph selection), but that there is
additional structure (for which there is syntactic evidence) in the gerunds, and this
additional structure renders the nominalising suffix too remote from the root to show
lexically-conditioned interactions.15
A similar invocation of structural locality conditions is used to explain a range of
generalizations in the domain of adjectival suppletion in Bobaljik (2012). That work
consists in part of a large, cross-linguistic survey of comparative and superlative
morphology. The account makes crucial reference to the internal structure of words, prior
to the application of rules of exponence, a structure that is denied under Word-and-
14 There is a certain (but not complete) overlap between theories, such as Embick (2010), which appeal to
structural locality conditions on allomorphy, in this case, some species of intervention, and theories such as
Lieber (1980) and Williams (1981) in which allomorphy is extremely local, with long-distance effects
achieved by feature percolation, with the intermediate node blocking percolation of features.
15 Embick argues for a more complex theory than hinted at here, in particular, with a role to play for cyclic
versus non-cyclic nodes in the structure. Embick also posits further structure beyond that in (21d-e), in
particular, a category-neutral ROOT, which is syntactically the complement of category-determining nodes
(denoted n, v etc). Only cyclic nodes intervene to disrupt locality, permitting morphological dependencies
to cross some nodes.
12
Paradigm theories. Specifically, I argue there that (relative)16 superlative grades of
adjectives are always derived from (structurally include) the comparative grade, as in
(22):
(22) 3
3 SPRL
ADJ CMPR
This hierarchical relationship is morphologically transparent in a great many languages as
shown here, although the expression of the affixes as prefixes or suffixes is a point of
cross-linguistic variation:17
(23) POSITIVE COMPARATIVE SUPERLATIVE
a. kam kam-tar kam-tar-in ‘little’ (Persian)
b. šüa šüan-ar šüan-ar-ste ‘pretty’ (Cimbrian German)
c. mlad-ý mlad-ší nej-mlad-ší ‘young’ (Czech)
d. nagy nagy-obb leg-nagy-obb ‘big’ (Hungarian)
e. nüs˚əә ç’a-nüs˚əә a-ç’a-nüs˚əә ‘pretty’ (Ubykh)
Even in languages like English where (22) is no longer synchronically transparent,
evidence for the presence of a comparative node embedded in superlatives comes from
patterns of root (suppletion). The overwhelmingly most common pattern (as noted for a
smaller sample by Ultan 1972) has a single suppletive root shared by both the
comparative and superlative degree, regardless of whether the structure in (22) is
transparent or not, thus:
(24) POSITIVE COMPARATIVE SUPERLATIVE
a. god bed-re bed-st ‘good’ (Danish)
b. špatn-ý hor-ší nej-hor-ší ‘bad’ (Czech)
c. k’argi-i u-mǰob-es-`i sa-u-mǰob-es-o ‘good’ (Georgian)
d. šig’ pɛr’-am pɛr’-mus ‘good’ (Kildin Saami)
e. kwad nax nax-deda ‘many’ (Kabardian)
The presence of the comparative node in the representation of superlatives in (22) serves
not only to provide a formal account of the sharing of a suppletive root allomorph in both
grades, but also to exclude a pattern like *good – better – goodest, a pattern which is
essentially unattested in the Bobaljik (2012) sample (with a few debatable counterexamples
discussed there). Key here is the Elsewhere Condition. Given (22), a root
allomorph specified for the context: “/ ___ ] CMPR” will of necessity win out over a
context-free allomorph (the positive) in both the comparative and superlative structures.
Although rare, root suppletion in the superlative grade, with a root allomorph distinct
from the comparative, is attested, as in Latin bonus – melior – optimus ‘good – better –
best’, with distinct root allomorphs in all three grades. Regular adjectives in Latin take a
16 Relative superlatives are those meaning ‘more A than all others’, for some adjective A. This term
contrasts with what are sometimes called absolute superlatives, with a meaning like ‘A to the highest
degree’ or ‘extremely A’. In Italian, for example, the suffix -issim-o/a marks absolute superlatives
(bellissima ‘very beautiful’), while relative superlatives are periphrastic (la più bella ‘the most [lit: more]
beautiful). The discussion below refers only to relative superlatives.
17 For sources and qualifications, as well as additional data, see Bobaljik (2012).
13
superlative in -iss-imus, which includes (a reflex or allomorph of) the comparative,
spelling out each of the pieces in (22): beat-us – beat-ior – beat-iss-imus ‘happy –
happier – happiest’). But the reflex of the comparative is missing in the one case where
the superlative and root appear to interact: opt-imus, *opt-issi-imus. This is as predicted,
since the only way for ADJ and SPRL to interact is when CMPR is not a distinct head, either
removed via a deletion operation, or combined with the adjectival root into a single locus
of insertion (via fusion, or non-terminal spell-out, see below).18 Smith et al. (to appear)
argue that the general schema laid out in Bobaljik (2012) for adjectival grades is
replicated in other morphological domains, notably suppletion in pronouns for case and
number, suggesting a wider role for structural locality conditions within words.
In sum, although the details of the theory, and correctness of particular analyses, are a
matter of current debate, the study of structural locality conditions, formulated over
abstract, word-internal hierarchical structures, prior to, and constraining, vocabulary
insertion, provides important support for the general thesis that the morphosyntactic
representation is indeed internally (hierarchically) structured.19
3. MORPHOLOGICAL OPERATIONS
The preceding sections laid out the two key components of DM, elaborating on the
assumptions in (2)-(3). In practice, of course, there are a number of other assumptions
beyond (2)-(3) associated with DM. The most important of these is the assumption that
there exist morphological operations which may manipulate the representation after
syntax, but prior to vocabulary insertion. Two such operations, Morphological Merger
and Impoverishment, were mentioned in passing above, and the literature contains
proposals for a variety of other operations, which are surveyed in superficial terms here.
Although there is substantial debate within DM as to what operations should be
recognized, it is worth stressing that it is a leitmotif of DM that there is a fundamental
systematicity to morphology—that there is order to be discovered in the apparent chaos
of morphological data, and that this order is indicative of the role of universal
grammatical constraints, restricting the space of possible mismatches in observed
syntactic and morphophonological structure. The wealth of operations put forward within
the general umbrella of DM is, from this perspective, somewhat of an embarrassment of
riches. With this in mind, work within DM has sought limits to the descriptive power of
the various devices, and not all authors recognize all of the operations that are surveyed
below.
18 See also Radkevich (2010) and Moskal (2015b) for further exploration of this idea, and for evidence that
suppletion quite generally affects only nodes that are contiguous with or local to the triggering feature.
19 Evidence from word-internal locality conditions on morphological operations thus provides a more
direct range of evidence for structure in morphology than does affix order, as for example in Baker’s
(1985) Mirror Principle effects. On the one hand, parallels between syntactic (or semantic) constituency
and morphological constituency may be described on theories without internal structure by appeal to a
common basis for both sorts of ordering, and hence are not a priori decisive. On the other hand, affix order
often does diverge from a direct reflection of syntactically and semantically motivated constituency (see,
e.g., Muysken 1981), and structure-based theories must either posit ad hoc complications of the syntax, or
acknowledge mechanisms for manipulating affix order post-syntactically. In DM, the syntactic structure is
the input to morphology, hence parallelism is the default, with additional operations such as templatic
restrictions (Bonet 1991, 1995, Noyer 1998a), metathesis (Arregi and Nevins, 2012), and the like requiring
special pleading.
14
3.1.Regrouping operations
One family of morphological operations alters the constituency after the syntax, but prior
to vocabulary insertion. An operation of this sort is Morphological Merger, discussed
above. Note that Merger creates a morphological word out of elements that do not,
together, form a complex X˚ in the syntax. Morphology may be dependent on syntactic
structure, but this does not require that all word formation happens “in the syntax”.
Cliticization, mentioned above as well, has a similar effect, but rather than combining
two heads (X˚ nodes), cliticization typically combines an X˚ element with a linearly
adjacent word or phrase, yielding a less tightly phonologically integrated combination. A
general question, not specific to DM, is the issue of the demarcation between clitics and
affixes, as well as the degree of syntax involved in cliticization. Rebracketing under
adjacency (an operation posited in other theories, see Sproat 1985, Williams 2003) is also
invoked within DM (see Radkevich 2010) to alter constituency relations, but not linear
order, internal to a complex X˚, for example: [ [ X Y ] Z ] à [ [ X ] Y Z ]. The effects of
rebracketing come into play in their interaction with assumptions about locality, and in
particular, about possible contexts for portmanteau morphology (see below). For one
view of a typology of operations that combine distinct heads, see Embick and Noyer
(2001); for a debate of some of the specific analyses presented there, see Hankamer and
Mikkelsen (2005) and Katzir (2011).
3.2.Fusion and portmanteaus
Merger and the related operations just mentioned rearrange the (terminal) nodes of the
syntax, but do not alter their number. Given the syntactic structure (25a), Merger may
apply, yielding (25b)—a synthetic expression: smart-er—or Merger may fail to apply,
yielding a periphrastic expression: more intelligent. Either way, the adjectival root and
the comparative node each constitute a distinct locus of vocabulary insertion.
(25) a. CMPRP b. X˚
3 3
3 ADJ˚ CMPR˚
CMPR˚ ADJP
3
3
ADJ˚
There are, however, cases in which two nodes in the syntax appear to correspond to a
single overt exponent with no internal morphological segmentation possible. Thus,
alongside bi-morphemic smart-er, bett-er, we find worse as the comparative of bad. It is
of course possible that worse is really worse-Ø, with a lexically conditioned Ø
comparative exponent alongside the suppletive root (parallel to the analysis of fly – flew
in (21b)). On the other hand, DM provides for the analysis of portmanteau morphology
(Matthews’s 1972 cumulative exponence) via the operation of fusion (Halle and Marantz
1993:116). Fusion combines two sister nodes into a single X˚, with the features of both
input nodes, but no internal structure. A fused node reduces the number of terminals,
providing for but a single locus of vocabulary insertion.
Embedded within a theory that involves strict locality conditions on morphological
operations, the two treatments of worse—reciprocally conditioned allomorphy and
fusion—make potentially distinct predictions about morpheme-morpheme interactions.
This was mentioned above, in connection with the Latin superlative: after fusion, the
15
adjectival root is local to the superlative morpheme in (22), in a way that it would not be
without fusion; treating worse without fusion, but with a zero comparative allomorph
would not alter locality relations.20
3.3.Deletion
Nodes and features may also be deleted by impoverishment rules, exemplified in (17)
above. Impoverishment rules delete features from a morphosyntactic representation, prior
to vocabulary insertion, with the result that impoverishment yields surface neutralization
of underlying contrasts, for example, suppressing gender contrasts in the plural or for
local (first and second) persons. Noyer (1997) suggests that impoverishment rules may be
motivated as repairs, affecting representations that violate markedness conditions (such
as (17b)), potentially drawn from a universal markedness hierarchy. Noyer also suggests
that feature hierarchies determine the particular repair to a given markedness statement.
Thus, if the combination *[FEM, PL] is excluded, it is FEM that will delete, due to the
hierarchy NUMBER > GENDER. Others argue that such deterministic approaches are too
constrained, and the there is variation, even among closely related languages, that is best
seen as alternative patterns of deletion in response to a single, shared markedness
constraint. Arregi and Nevins (2012), for example, note that the combination of markers
for first and second person arguments in the Basque auxiliary is systematically avoided,
but that different dialects of Basque use different strategies in such contexts. The
variation may be modelled as differences in what feature or features are deleted as repairs
in response to the same markedness condition. Arregi and Nevins argue moreover that
deletion operations may target either individual features or an entire node (see also Halle
and Marantz 1993). In the former case, the result is the insertion of a less marked
exponent, possibly, though not necessarily, phonologically null; in the case of deletion of
an entire node (which they term obliteration), the result is the failure of any exponent to
be inserted.
In the standard formulation, impoverishment rules are deletion rules, and hence the
output is a representation that is underspecified for feature values (in (17a), lacking
gender). Noyer (1998b), drawing on evidence from Nimboran, contends that this view is
too restrictive, and that persistent redundancy rules insert the unmarked value of a feature
when impoverishment deletes a marked value. If masculine is the unmarked gender, then
deleting feminine yields masculine. Calabrese (2011) extends this further still, claiming
that the input to vocabulary insertion is never formally underspecified, and that the only
underspecification that should be countenanced is in rules of exponence, such as (5).
Impoverishment is a powerful device, but it has less formal descriptive power than Rules
of Referral (Zwicky 1985, Stump 1993) which are in essence rules that may change any
combination of features into any other combination—an ‘anything goes’ approach. In
other words, impoverishment allows for the characterization of only a proper subset of
the syncretic patterns allowed by rules of referral. DM thus typically starts from the more
restrictive theoretical position, in which observed typological generalizations are
explained within the theory (certain patterns are formally unstatable), where theories such
20 See Chung (2007, 2009) for additional elaboration of the Fusion operation, and arguments that Fusion
rules may be cyclically interleaved with vocabulary insertion. See Radkevich (2010) for proposals to recast
Fusion as insertion directly at the higher node in a structure like (25b). Insertion at complex terminals is
also proposed in Neeleman and Szendrői (2007) and Caha (2009), among others. The combination of
rebracketing and fusion (or rebracketing and non-terminal insertion) is in many respects equivalent to the
operation of Spanning proposed in Williams (2003) and more recently in Svenonius (2012) and Merchant
(2015).
16
as Paradigm Function Morphology (Stump 2001) must seek explanations for observed
typological patterns outside the morphological system.
As an empirical matter, there are numerous challenges to the claim that syncretism is
fundamentally modelable as underspecification or neutralization by impoverishment. See
Baerman et al. 2005 for a large survey, including numerous cases which they contend
cannot be modelled in this manner. On the other hand, many apparent counter-examples
have turned out to be susceptible to alternative analyses (including accidental homophony
in the limiting case). Two illustrative cases are considered here.
Person-marking (pronouns, agreement) constitutes a well-defined, closed class, amenable
to cross-linguistic comparison. In a large survey, Cysouw (2003 and related work) argues
that there are statistical trends (some patterns are significantly more common than
others), but that no pattern is impossible. Others have looked at the same domain and
found that the empirical domain is best described by a theory that does draw on a
universal inventory of features, with neutralization (underspecification) as the primary
source of syncretism, but with recognition of a minor role for (synchronically) accidental
homophony (see, among others, Bobaljik 2008, Pertsova 2011, Sauerland and Bobaljik
2013, Harbour to appear). To some extent, what matters is being able to distinguish
between syncretism as a property of the grammar of a language, and accidental
homophony. Syncretism (1PL=3PL) in the German (weak) past tense was considered in
(13) above, and an underspecification account was considered. The corresponding person
marking in the present tense differs from the past in having an additional instance of
identity, here between the affial markers of 3SG and 2PL:
(26) German (weak) Past Tense: sagen ‘to say’
Should the theory accommodate this identity of form as an instance of syncretism? It is
difficult to see how to model both the 3SG=2PL identity and the 1PL=3PL identity
simultaneously using the device of impoverishment. An alternative is to treat the
exponents of 3SG and of 2PL as distinct instances of –t, with no formal relation between
them—homophony, but not syncretism. In this particular case, a range of evidence has
accumulated pointing to a qualitative difference here: the identity that is expressible as
impoverishment (1PL=3PL) is represented in German speakers’ grammars (it is metasyncretic
and participates in identity effects in ellipsis and resolution of disjuncts, for
example), while the surface identity that is not representable as ellipsis (3SG=2PL) fails to
show these properties and is accidental (see Albright and Fuß 2013 for an overview of
approaches to syncretism, including discussion of the German facts).
Peculiarities of Chukchi (ckt, Chukotko-Kamchatkan) inflection provide a different kind
of challenge to the claim that impoverishment is the primary driver of syncretism.
Simplifying: transitive verbs in Chukchi are typically marked by a prefix agreeing with
the (ergative) subject and a suffix agreeing with the (absolutive) object (see (27a-b)).
Certain combinations of subject and object (in certain moods) depart from this pattern.
An example is given in (27c). Although the clausal morphosyntax (e.g. case-marking) is
transitive, the verb shows an intransitive morphological form: the agreement morphology
PERSON SINGULAR PLURAL
1st
sag-(e) sag-en
2nd
sag-st sag-t
3rd
sag-t sag-en
17
(both prefix, here null, and suffix) are those for a 3SG intransitive subject, and in addition,
the verb bears a prefix that normally marks the antipassive (a form of derived
intransitive).21
(27) a. ɣəm-nan ɣət təә-ɬʔu-ɣət
I-ERG you.SG(ABS) 1SG.SUB-see-2SG.OBJ
‘I saw you.’
b. ərɣə-nan ɣəm ne-ɬʔu-ɣəm
they-ERG me(ABS) 3.SUB(TR)-see-1SG.OBJ
‘They saw me.’
c. ə-nan ɣəm Ø-ine-ɬʔu-ɣʔi
he-ERG I (ABS) 3SG.SUB(INTR)-AP-see-3SG.SUB
‘He saw me.’ (Skorik 1977: 44-45)
In the indicative mood, “spurious” antipassive forms such as (27c) (the name is from
Halle and Hale 1997) are obligatory for a subset of inverse contexts, in which the object
outranks the subject on the person hierarchy. Spencer (2000) contends that the Chukchi
data shows the insufficiency of Impoverishment and requires instead more powerful,
arbitrary feature-changing rules. Specifically, it is easy to see how an impoverishment
rule, such as (28) (perhaps a repair for a filter banning certain inverse configurations)
would yield an apparently intransitive form in the morphology, from an underlying
transitive morphosyntactic representation, consistent with DM. But, Spencer argues, it is
hard to see how deletion of features would yield the spurious addition of an extra
morphological piece, namely, the antipassive.
(28)
€
AGR
Subj: 3sg
Obj:1sg
"
#
$
%
&
'
à
€
AGR
Subj: 3sg[ ]
Bobaljik and Branigan (2006) take up Spencer’s challenge, and suggest that the facts are
consistent with Impoverishment after all. Keeping in mind that DM is a realizational
theory, the appearance of antipassive morphology (whether spurious, or a true
antipassive) must be treated as a realization of a particular terminal in a specific
morphosyntactic configuration. Bobaljik and Branigan argue that the spurious antipassive
(after the application of impoverishment) shares with the true antipassive the property of
an underlyingly transitive syntactic configuration (the verb has a logical object) but with
only the features of a single argument in the agreement node. Taking this configuration to
trigger the occurrence of the -ine- vocabulary item, Bobaljik and Branigan demonstrate
how deletion of features may be consistent with the appearance of seemingly ‘extra’
exponents, once properties of the morphosyntax as a whole are taken into consideration.
21 Despite the surface similarity between the third person transitive subject prefix ne- and the
antipassive -ine-, the two prefixes are clearly distinct: they are different elements in other moods, for
example, and occupy different linear positions in the verb relative to other morphology, such as the future
prefix r-. In some varieties, there is a second, suffixal, antipassive, which is also spuriously used with some
subject-object combinations. See the literature cited for further details.
18
3.4.Epenthesis? Autonomous morphological conditions
Much work in DM stresses parallels across grammatical modules—for example, the role
of syntactic locality conditions in explaining morphological patterns, or the interaction of
morphology and phonology (for recent monograph-length studies, see Embick 2010,
Bobaljik 2011, Arregi and Nevins 2012). Just as deletion operations are common in
phonology, another common process is epenthesis—the addition of structure not in the
underlying representation in order to meet surface (universal or language-particular) wellformedness
conditions. Various studies within DM have argued for epenthesis-like
operations in morphology as well: the addition of terminal nodes post-syntactically, prior
to vocabulary insertion. Halle and Marantz (1993) suggested that agreement morphemes
were to be treated in this way—not present in the syntax, but appended, in the
morphology, to designated functional nodes from the syntax, such as T[ENSE] (the
modern name for INFL in (6)); Noyer (1997) also makes use of morphological epenthesis,
arguing for language particular autonomous morphological structure conditions, which
must be met. A special case of the addition of nodes at morphological structure, in
Noyer’s presentation, is feature fission, in which a single node in the syntax is split into
two nodes in the morphological representation, in some instances leading to apparent
cases of extended exponence (multiple expression of a single feature).
One use of morphological epenthesis that has been suggested within DM is for the theme
vowels characteristic of many Indo-European languages. These have been a particular
focus of inquiry in Slavic and Romance, and recent work has identified analogous
elements in Bantu (Monich 2015). For example, Oltra-Massuet and Arregi (2005) argue
that stress in the Spanish verbal system can largely be reduced to the fairly simple
generalization that stress falls on the vowel most closely preceding the T˚ node, once
certain assumptions about the internal morphosyntactic structure of the verb are
recognized. Their analysis includes theme vowels (possibly null), added in the
morphology to every syntactic functional head; in the simple case of the infinitive, the
theme vowels (TV) are those indicative of conjugation class, as illustrated in (29):
(29) a. cant-a-r sing-TV-INF ‘to sing’
b. tem-e-r fear-TV-INF ‘to fear’
c. part-i-r leave-TV-INF ‘to leave’
In analyzing the conjugation of Spanish (and related languages) the theme vowels are
treated as elements distinct from both the root (which governs their selection) and the
more peripheral morphemes with which they combine (such as infinitive -r). In more
complex forms, involving additional internal syntactic complexity, there will be multiple
theme vowels, one per syntactic head. The present imperfective indicative has the
(partial) morpho-syntactic structure (30a), to which theme vowels and an agreement node
are epenthesized (30b), yielding the form (30c) after vocabulary insertion and stress
assignment:
(30) a. [ [ √SING v ]v T ]T
b. [ [ √SING v-TV ]v T-TV ]T -AGR
c. cant- Ø á b a -mos
The conditional, they argue (on semantic and cross-linguistic grounds) is even more
complex morphosyntactically, essentially composed of a (modal-like) future and a (past)
Tense node (31a), with accordingly three theme vowels (31b-c):
19
(31) a. [ [ [ √SING v ]v F ]F T ]T
b. [ [ [ √SING v-TV ]v F-TV ]F T-TV ]T -AGR
c. cant- Ø a r í Ø a -mos
There is no evidence for a Theme Vowel head or Theme vowel phrase (let alone three of
them in the conditional) in the syntax, yet they are evidently present in the surface
morphological form. The presence and distribution of theme vowels thus appears to be a
purely morphological requirement of these languages, and is a clear candidate for the
kind of operation of morphological epenthesis discussed here.
Note that like Halle and Marantz (1993), Oltra-Massuet and Arregi (2005) suggest that
Agreement nodes are also epenthesized in the morphology, and do not constitute heads in
the syntax, with language-particular conditions playing a role in their distribution.22 For
example, the conditional in (31) has a series of clausal functional nodes (here F and T),
but only the higher of these is associated with an agreement node in the morphology. The
morphologically complex future and conditional structures such as (31) show parallels
with compound tenses, such as the compound future of French—see the alternation in
(32), with the -r- of the conditional possibly the same vocabulary item as the -r of the
infinitive, and in both cases, only a single agreement node.
(32) a. je chant-er-ai.
1SG sing-R-1SG ‘I will sing’
b. je v-ais chant-er.
1SG go-1SG sing-INF ‘I will sing’
In other languages, compound tense constructions involve agreement on multiple verbal
heads. Ibibio (ibb, Niger Congo) shows multiple agreement both in compound tenses
(33a) and where the tense/aspect morphology is part of the morphological word with the
verb (33b-c), examples from Baker and Willie (2010:100-101, 109):
(33) a. N-sʌk n-yem ebot odo.
1SGS-AUX 1SGS-seek goat the
‘I am looking for the goat.’
b. Ɲɲin i-k-i-yem Emem.
1PL 1PLS-PAST2-1PLS-seek Emem
‘We were looking for Emem.’
c. utom se ɲɲin i-ma[a]-i-k[e]-i-nam.
work that 1PL 1PLS-PAST1-1PLS-PERF-1PLS-do
‘work that we had already done’
These examples, like others discussed above (see (6), (25)), reinforce the parallels
between syntactic (analytic) and morphological (synthetic) structure that DM calls to the
fore. Here, we note that agreement may (as in Ibibio) but need not (as in Romance) be
attached to multiple functional heads. It is orthogonal to the characterization of
22 This is not a core assumption of DM—the overall DM architecture does not depend on agreement nodes
being added in the morphology, and is also compatible with agreement nodes (or features) being present in
the syntactic representation.
20
agreement whether these functional heads are part of the same morpho-phonological
word as the verb root.23
In sum, the DM toolkit contains a range of operations that describe failures of a one:one
correspondence—between the syntactic output representation and the phonological
representation. Analyzing such mismatches is the meat-and-potatoes of DM
morphological analysis, although there are debates within DM about the proper
formulation of many of these operations. Standing above the individual debates is the
question of how much of this powerful machinery is actually needed, and which
operations may be reduced to which others, without loss of empirical coverage, with the
goal on the horizon being, of course, a theory of possible morphosyntactic mismatches.
4. MORPHOLOGY, SYNTAX, AND THE DIVISION OF LABOUR
DM focuses attention on the nature of the relationship between syntax and morphophonology.
As such, numerous questions of the division of labour between the various
components come under scrutiny in this approach, including questions that go beyond the
study of word internal form (the traditional purview of Morphology). Obviously, the
questions to be asked will differ depending on the theory of syntax that is assumed. In
practice, much of DM assumes a GB/Minimalist syntax, and issues of the following sort
are thus considered within the sphere of DM:
4.1.Spell-Out Domains
In most versions of DM, the mapping from syntax to morphology is conceived of in
derivational terms, with a significant role for cycles. In practice, cyclicity is manifest in
two ways. One understanding of cyclicity is the idea that vocabulary insertion
(exponence) proceeds within a complex X˚ from the root (or most deeply embedded X˚
node) outwards. Bobaljik (2000) argues that this assumption makes predictions about
how information at one node may be accessed in the course of operations applying at
other nodes. In essence: if vocabulary insertion is a process that converts
morphosyntactic representations into morphophonological representations node-by-node,
then at any given point in the derivation, nodes that have already undergone exponence
will have only morphophonological information while nodes that have yet to undergo
exponence will lack such information, containing only morphosyntactic information. In
this way, cyclicity enforces conditions on possible rule contexts: rules operating at a
given node may be sensitive outwards only to morphosyntactic information and inwards
only to morphophonological information, all else being equal. Bobaljik (2000) provides
empirical evidence from the Chukotko-Kamchatkan languages that support this
perspective.
23 Harris (2009), analyzing a particularly rich array of examples like (33b-c) from Batsbi (bbl, NakhDagestanian),
claims that such “exuberant exponence” of agreement is claimed “not to exist” (p.268-272)
in DM and is deeply problematic for the framework. This appears to rest on a misreading of the assertion in
Halle and Marantz (1993:138) that (in the general case) each “syntactic or morphological node” receives a
single exponent. In the cases at hand, we are dealing with multiple agreement nodes in the morphosyntactic
representation (whether added via epenthesis, or underlyingly present), each of which receives a single
exponent, just as in (33a); compare Halle and Marantz’s (1993:145) analysis of Potawotami (pot,
Algonquian), which Harris is reacting to, in which the features of a single argument are indeed shared
among multiple agreement nodes, added post-syntactically.
21
The other notion of cycle is as designated domain, reminiscent of phases in current
syntactic theories. The approaches to locality mentioned in section 2 largely incorporate
cycles in this latter sense: on this understanding, it is not the case that each node
constitutes a cycle (i.e., of rule application). Instead, certain nodes are specified as
defining privileged domains that encapsulate information in a way that material in one
domain has only limited effects on material in another domain (cf. D’alessandro and
Scheer, to appear).
Moskal (2015a,b) provides a straightforward illustration of this type of reasoning. A
prominent idea within the DM literature (see section 5, and especially Harley 2014) is
that open-class lexical items (nouns, verbs) consist of at least a root and a categorydefining
functional morpheme. Thus the noun dog is at least bi-morphemic, with the
structure: [ [ √DOG ] n ], where n is a nominalizing morpheme that establishes the
category of the word. A common hypothesis is that such category-defining nodes
establish locality domains (see especially Embick 2010). This assumption in large part
derives the effect that inflectional allomorphy does not survive category-change (see (21),
above). Now, Moskal suggests that the closed class, functional vocabulary lacks the
root+category structure that characterizes the open-class vocabulary. An immediate
consequence is that domain-sensitive locality restrictions should then play out differently
between, say, pronouns and nouns, or between auxiliaries and main verbs. And this is
exactly what Moskal finds. In a large survey of suppletion, she finds a striking
asymmetry between pronouns, which commonly supplete for case, and nouns, which
strikingly fail to do so. Her explanation is that the n node in nouns establishes a wordinternal
domain boundary, and that root allomorphy can therefore not be conditioned by
information (such as case) that is too far outside this domain. Since pronouns lack the
√ROOT+n structure, they have no such domain and thus no such restriction.
The idea of privileged nodes defining cyclic domains thus seems to have currency in
syntax (phases), morphology, and phonology. A currently active research question is the
inter-relatedness among the domains in the various components. Clearly the cleanest
hypothesis would be that the same domains are (potentially) relevant in all three
grammatical components—in that sense we would only need to provide one answer to the
question of why certain nodes, and not others, establish domains. An intriguing line of
research, illustrated by Newell and Piggott (2014), has returned to phonological
alternations at morpheme juncture, looking for evidence for the alignment of
morphosyntactic and phonological domains within the complex word. For example, they
find that hiatus avoidance (a pressure to avoid heteromorphemic VV sequences) may be
resolved differently depending on whether the two morphemes are within a single
syntactic phase, or span a phase boundary. Other researchers have found a less perfect
mapping among domains in the various components – for a recent survey and proposal
for a weaker connection between syntactic and phonological domains, see D’Alessandro
and Scheer (in press).
4.2.Linearization and copies
Since DM trades in operations that manipulate the syntactic representation as part of the
mapping to phonological form (see above), DM’s morphological operations have the
potential to interact with other operations that are held to be a part of Spell Out in the
GB/Minimalist architecture. For example, one prominent line of thought holds that the
syntactic representation properly represents only hierarchical, i.e., constituent, structures,
with precedence relations among the constituents not a part of this “narrow” syntax (see,
e.g., Marantz 1983). On such approaches, post-syntactic algorithms determine linear
ordering relations among the syntactic nodes. See Kayne (1994) and Fox and Pesetsky
(2005) for two influential proposals. Relatedly, one view of syntactic “movement” is that
22
a single syntactic element may be represented in more than one location in the syntactic
structure, either via “copies” in a chain (Chomsky 1995 et seq) or via Multidominance
representations (e.g., Gärtner 1999, among others). In the typical case, an element that is
so represented is nevertheless pronounced only a single time in a given domain, but the
determination of which copy to pronounce (or which position to pronounce the element
in) may be made post-syntactically (see, e.g., Nunes 2004). In Bobaljik (2002a), I gave
arguments that linearization at the phrase level (syntactic headedness), the
pronunciation/copy-choice algorithms, and Morphological Merger, do indeed interact in
non-trivial ways, providing an account of Holmberg’s Generalization effects in Germanic
Object Shift that relates an apparent syntactic condition on word order alternations to the
morphological conditions on the post-syntactic creation of complex words (inflected
verbs) that are also at work in the analytic/periphrastic alternation of English do-support
contexts discussed above.
Other work investigating the interaction of DM’s morphological operations with the postspell
out operations posited elsewhere in the literature includes Embick and Noyer
(2001), Embick (2010), and Arregi and Nevins (2012). With reference to the Latin
cliticization example mentioned in (8), Embick and Noyer (2001) focus on the idea that
grouping operations like Merger may apply before or after linearization, but with a
different range of effects. Cliticization examples, like Latin =que must apply after
Linearization, since their effects depend on knowing what the “first word” in a particular
domain is, whereas purely structural regroupings may apply before Linearization. These
operations in turn may interact with cyclic spell-out (see fn. 4).
Relatedly, within the realm of Linearization of syntactic structure, Noyer (1997) argues
that linear order among terminal nodes need not be exhaustively determined even at the
point of vocabulary insertion, and that vocabulary items may carry idiosyncratic
specifications as either prefixes or suffixes. That is, a single node in the morphosyntactic
representation of a single language may be sometimes realized as a prefix and sometimes
realized as a suffix, as a function of that element’s feature content. Noyer applies this to
cases of apparent discontinuous bleeding—competition effects between prefixes and
suffixes in the Afroasiatic conjugation. Thus, in Tamazight (Berber), the first person
plural is marked by a prefix n-, but the first person singular is marked by a suffix -ɣ,
which Noyer takes to be competing exponents of a single agreement affix.24
4.3.Case and Agreement
Another facet of the spelling out of syntactic structure covered under the broad umbrella
of the DM literature is the signalling of syntactic relations, such as in the morphological
categories of case and agreement. It is well established that the distribution of case and
agreement is in part a function of syntactic configuration. Under the DM architecture in
(1), there are at least two ways in which morphology (the surface forms of words) may be
influenced by syntactic context. On the one hand, the features that are expressed as case,
agreement etc. may be distributed (e.g., assigned/checked) in the syntax itself, subject to
realization in the morphology. For example, the syntax may be responsible for assigning
an [ACCUSATIVE] feature to a direct object, which then may or may not receive a
language-particular realization. A syntactic Case Theory of this sort dominated much of
GB and (early) Minimalist theorizing for two decades, starting with Chomsky (1981). On
the other hand, since Morphology governs the relation between syntactic and
phonological representations, the DM architecture allows for the possibility that an
operation may be sensitive to syntactic structure, perhaps even defined over such a
24 See Banksira (2000) for an analysis of Chaha and for some discussion of competing proposals.
23
structure (c-command, government—AGREE in Minimalist terms), yet that operation may
not be a part of the syntax per se, but rather a part of the algorithms that map the output
of the syntax onto a phonological representation. A useful analogy is perhaps to rules of
phrasal phonology or sandhi phenomena, which are phonological in nature, but defined
with reference to syntactic structure.
Marantz (1991) exploits this aspect of the architecture in (1), arguing that case and
agreement—the morphological categories we have evidence of—should be seen in these
terms as well. Marantz’s emphasis is on the failure of the morphological case categories
(nominative, accusative, etc.) to align with the categories of a GB/Minimalist-style Case
Theory. A striking example of such a misalignment is drawn from the seminal work of
Zaenen, Maling and Thráinsson (1985) on “quirky” case Icelandic. Icelandic nominals
show all of the hallmark distributional properties and alternations that are accounted for
by Case Theory in Chomsky (1981), including the “Exceptional Case Marking”
configuration (where the subject of a non-finite clause, otherwise obligatorily
unpronounced, may be an overt nominal when the infinitive is the complement of a verb
like believe). Yet, in Icelandic, the distribution attributed to Case Theory can only be
appreciated by abstracting away from the actual case borne by the nominals. GB’s
abstract nominative (essentially, finite subject) can be realized as any of the four
morphological cases in Icelandic (nominative, accusative, dative or genitive), while
conversely, the distribution of morphologically nominative nominals (or those with any
other case) required them to be analysed as abstractly nominative in some contexts, but
abstract accusatives in others. Pushing further, Marantz argues that when the
morphological distribution of case (and agreement; see also Bobaljik 2008) is factored
out of the syntax, little is left of the Case Theory of GB as a syntactic theory of argument
licensing, and thus, that recognizing the distinct roles of syntax—in argument licensing,
and morphology—in the surface realization of syntactic representations, allows for a
simplification of (aspects of) both components of grammar.25 In a related vein, Bhatt and
Walkow (2013) exploit the possibility that agreement may be part of the post-syntactic
component to make sense of asymmetries in sensitivity to linear order in conjunct
agreement in Hindi.
As illustrated in the preceding paragraphs, the scope of the DM literature encompasses
not only topics that sit comfortably within the common understanding of Morphology
(inflection, derivation, syncretism, exponence). In light of DM’s primary nature as a
theory (or family of theories) of the overall architecture of grammar, proposals within
DM do not only have effects within morphology proper, but potentially interact with the
syntax and the phonology. As regards syntax in particular, DM’s architecture opens
questions about the division of labour among the components, leaving much of
Morphosyntax up for grabs.
5. ROOTS AND THE ENCYCLOPAEDIA
Before closing, there is one further question of the division of labour among components
deserving of special mention in light of the significant attention it has received in the
(broadly) DM-related literature. This is the scope of lexical decomposition in the syntax,
and associated questions about the representation of argument structure and of roots—the
most deeply embedded morphemes in the open class, content vocabulary such as nouns
and verbs. It is impossible in the space remaining to do any sort of justice to this large
25 For arguments regarding aspects of case licensing that appear to be syntactic, see Wurmbrand (2006)
and Legate (2008), and on agreement, Béjar (2003) and Baker (2010).
24
topic, but the following remarks will, I hope, serve to illustrate why this topic is of
particular interest within DM.
Word-formation in DM is (primarily) syntactic, and syntax is subject to compositional
interpretation. If one structure properly embeds another, then (all else being equal) the
larger structure should inherit (syntactic, semantic, and morphological) properties of the
embedded structure. For example, if the causative member of a causative-inchoative
alternation (as in (34)) embeds (the relevant portion of) the corresponding inchoative,
then properties such as the entailment relation holding from (34a) to (34b) are part of the
model.
(34) a. The farmer grew tomatoes. [vP NP v [VP V NP ] ]
b. Tomatoes grew. [VP V NP ]
Nominalizations have been widely discussed from this perspective in the DM literature,
returning to themes originally raised in Chomsky (1970) (see Marantz 1997, Harley and
Noyer 2000, Harley 2009). The basic question is why some nominalizations (in
particular, some -ing gerunds, such as (35a)) appear to inherit verbal properties (allowing
accusative complements and adverbial modification), while others (such as (35b)) lack
any discernable verbal syntax. In DM, the fact that marri-age contains marry is a
syntactic fact, hence the question of why marry’s syntactic (e.g., verbal) properties are
not inherited wherever it appears.
(35) a. [Kate(’s) quickly marrying William] was prompted by …
b. [Kate’s quick marriage to William] was prompted by …
Chomsky (1970) argued essentially that differences of this sort were structural (see
section 2, above; see Marantz 1997 for a discussion of this reading of Chomsky 1970).
While gerundive constructions are derived from the verb (and thus contain verbal
morphosyntax), “nominalizations” such as (35b) are derived directly, on Chomsky’s
proposal, from a root, such as √MARRY, which is neutral as to syntactic category, but
which is associated with lexical meaning, including argument structure. Pesetsky (1995)
and especially Marantz (1997) update Chomsky’s proposal, introducing the notation of
category-neutral root √ROOT, along with category-defining heads (terminal nodes) in the
syntax. The verb marry is thus [[√MARRY] v], while marriage is the spell-out of
[[√MARRY] n]. At the limit, as mentioned above, even the simplest of words (from the
lexical classes) have an internal syntactic structure: dog is structurally [[√DOG] n], where
n is a syntactic terminal that established the category feature: Noun.
A key argument for this view, presented in Chomsky (1970) and drawn out in Pesetsky
(1995) and Marantz (1997) runs on two premises. First, the morphologically irregular
nominalizations are limited to root nominalizations (section 2 discusses this from the
perspective of structural locality conditions on allomorphy). Second, lexical
decomposition is syntactic, and the causative member of the causative-inchoative
alternation involves a phonologically null CAUS head (v in (34a)) in the syntax. Together,
these assumptions entail that only the inchoative member of a pair such as (34) can be
subject to a root nominalization. Famously, The farmer’s growth of tomatoes therefore
lacks a reading corresponding to a nominalization of (34a). The irregular nominalising
affix -th can only be conditioned by the root when the root and affix are syntactically
local to one another; this in turn is possible only if the nominalizing affix selects the
structure in (34b), but is not possible when the nominalizer is separated from the root by
the head that introduces the external argument, as in (34b), which head is nevertheless
25
needed for the relevant interpretation. The only nominalization of (34a) that is possible is
a gerund: the farmer(’s) growing tomatoes…26
A question often asked in the context of the proposal for category-neutral roots is why we
cannot simply freely use any root in any context, for example, why the root arrive unlike
grow or melt, fails to participate in the causative-inchoative alternation in (34), or why
there is, at least to a first approximation, no verb to cat. A closely related question is why
root-derivations often have apparently idiosyncratic (restrictions on) meaning:
transmission in the automotive sense, seems only tenuously connected to the meanings of
the root √TRANSMIT evidenced in that root’s verbal occurrence: (to) transmit.
One aspect of this question has been addressed in part already, and is independent of the
particular formalisms of DM. Since the causative-inchoative alternation involves a
syntactic decomposition with a (possibly null) head v-CAUS, then some restrictions on the
distribution of roots in this frame may be attributable to the selectional (i.e.,
combinatorial) restrictions borne by this (abstract) syntactic head (whether these
restrictions are syntactic or semantic in nature). If there are fine-grained differences
among types of V (and thus types of root), that are syntactically represented (for example
distinguishing internally-caused change of state from others, see Levin and RappaportHovav
1995), then the failure of arrive to be used transitively (*The courier arrived the
letter) is a combination of the syntactic fact that v-CAUS only occurs with certain subcategories
of V, and the root arrive is not of the right (sub)-category to combine with vCAUS
(or more accurately, if roots, like other morphemes, are subject to Late-insertion,
then the exponent arrive does not spell out the type of V that combines with v-CAUS).27
Other aspects of this question may turn on the nature of conventionalized aspects of
meaning. There is likely no grammatical (syntactic, semantic) explanation for the absence
of a verb to cat in English (compare: to ape, to dog). DM would predict it to be possible
in principle, but it is apparently absent (at least for many speakers).28 DM treats
knowledge about the lexicon, and knowledge of actual use, in a grammatical repository—
the Encyclopaedia—separate from the vocabulary and the list of syntactic atoms. The
Encyclopaedia includes knowledge of idioms, arguably represented as knowledge of
individual roots in context. In the Encyclopaedia, one could posit that √KICK takes on the
special meaning ‘die’ in the context of √BUCKET (which in turn, is devoid of meaning in
the context of √KICK). From this perspective, as Marantz (1995) argued, all (content)
words are in effect phrasal idioms, but with idiosyncratic definitions restricted to a
syntactically defined context: the first (syntactic) phase in which the root occurs (cf.
Ramchand 2008). But just as roots do not have special idiosyncratic meanings in all
lexical contexts (the root √PUNCH takes on no special meaning in the context of
√BUCKET), so too may individual roots fail to lack a specified meaning in a particular
grammatical context: the root √CAT happens to find a conventional definition in the
26 Subsequent work in this vein has argued that the simple distinction between root nominalizations and
larger structures is too blunt (see Harley 2009, responding to challenges from Borer 2003, for example), but
the general tack of the DM revival of Chomsky (1970) remains. Refinements lie in the amount of syntactic
structure posited within words, and in the understanding of the locality conditions on the interpretation of
such structure: in particular, restrictions on the domain of idiosyncratic meaning and irregular
pronunciation. For a sceptical evaluation of the argument from growth, see Müller and Wechsler (2014).
27 See Kramer (2015), among others, for related questions about associating roots with gender-bearing n
heads in the nominal domain.
28 Indeed, meanings for a verb to cat are listed in the OED. For relevant discussion of category-neutral
roots from a psycholinguistic perspective, see Barner and Bale (2002, 2005).
26
context [[ __ ] n] but lacks a conventionalized meaning in the context [[ __ ] v ].
Grammatically, [[ √CAT ] v ] is well-formed, but speakers of English must use cues from
context to interpret the nonce coinage.
6. SUMMARY
In sum, DM at its core comprises a framework of assumptions for thinking about the
grammar as a whole and the interaction of its parts. The Morphology part of the name
stresses an emphasis on facts traditionally seem as the purview of morphology—the
shape of words. The Distributed part of DM highlights the contention that the
grammatical knowledge of (pieces of) words is not monolithic, but enters the
grammatical computation at various points in various ways. The traditionally conceived
Lexicon is replaced by three lists:
(36) i. a list of the abstract morphemes that are the building blocks of syntax,
ii. a list of vocabulary items that spell out (morpho)-syntactic structures, and
iii. a list of the idiosyncratic meanings of individual pieces in particular
contexts.
Key to the theory are the assumptions that syntax is the primary engine of composition—
i.e., that word-formation is part of the syntax—and that morphology is realizational,
specifically post-syntactic. Beyond these broad assumptions, the overall architecture of
the theory opens various possibilities for the analysis of particular arrays of data. The
success or failure of the theory as a whole requires embedding the specific analyses
within a larger theoretical framework encompassing assumptions about the workings of
other components of the grammar.
REFERENCES
Albright, Adam and Eric Fuß 2013 Syncretism. In Jochen Trommer, ed., The morphology
and phonology of exponence. Oxford University Press, pp. 236-288
Anderson, Stephen R. 1992. A-Morphous Morphology. Cambridge: Cambridge
University Press.
Arregi, Karlos, and Nevins, Andrew. 2012. Morphotactics: Basque Auxiliaries and the
Structure of Spellout. New York: Springer.
Baerman, Matthew, Dunstan Brown, and Greville Corbett. 2005. The syntax-morphology
interface: a study of syncretism. Cambridge University Press.
Baker, Mark. 1985. The Mirror Principle and Morphosyntactic Explanation. Linguistic
Inquiry 16:373-415.
Baker, Mark, and Willie, Willie Udo. 2010. Agreement in Ibibio: From Every Head, To
Every Head. Syntax 13:99-132.
Baker, Mark. 2010. The syntax of agreement and concord. Cambridge: Cambridge
University Press.
Banksira, Degif Petros. 2000. Sound Mutations: The Morphophonology of Chaha.
Amsterdam: John Benjamins.
Barner, David, and Bale, Alan C. 2002. No nouns, no verbs: psycholinguistic arguments
in favor of lexical underspecification. Lingua 112:771-791.
Barner, David, and Bale, Alan C. 2005. No nouns, no verbs? A rejoinder to Panagiotidis.
Lingua 115:1169-1179.
Béjar, Susana. 2003. Phi-syntax: a theory of agreement, University of Toronto: Ph.D.
Dissertation.
27
Bhatt, Rajesh, and Martin Walkow. 2013. Locating agreement in grammar: an argument
from agreement in conjunctions. Natural Language and Linguistic Theory,
31:951-1013.
Bobaljik, Jonathan David. 2000. The ins and outs of contextual allomorphy, eds.
Kleanthes K. Grohmann and Caro Struijke, 35-71. College Park, MD: University
of Maryland.
Bobaljik, Jonathan David. 2002a. Realizing Germanic Inflection: Why Morphology Does
not Drive Syntax. Journal of Comparative Germanic Linguistics 6:129-167.
Bobaljik, Jonathan David. 2002b. Syncretism without paradigms: Remarks on Williams
1981, 1994. In Yearbook of Morphology 2001, eds. Geert Booij and Jaap van
Marle, 53-85. Dordrecht: Kluwer.
Bobaljik, Jonathan David. 2008. Where's phi? Agreement as a post-syntactic operation.
In Phi-Theory: Phi features across interfaces and modules, eds. Daniel Harbour,
David Adger and Susana Béjar, 295-328. Oxford: Oxford University Press.
Bobaljik, Jonathan David. 2012. Universals in Comparative Morphology: Suppletion,
Superlatives, and the Structure of Words. MIT Press.
Bobaljik, Jonathan David, and Branigan, Phil. 2006. Eccentric Agreement and Multiple
Case Checking. In Ergativity: Emerging Issues, eds. Alana Johns, Diane Massam
and Juvenal Ndayiragije, 47-77. Dordrecht: Springer.
Bonet, Eulalia. 1991. Morphology after syntax, MIT: Ph.D. Dissertation.
Bonet, Eulàlia. 1995. Feature structure of Romance clitics. Natural Language and
Linguistic Theory 13:607-647.
Bonet, Eulàlia, and Harbour, Daniel. 2012. Contextual allomorphy. In The Morphology
and Phonology of Exponence, ed. Jochen Trommer, 195-235. Oxford University
Press.
Borer, Hagit. 2003. Exo-skeletal vs. Endo-skeletal explanations: Syntactic Projections
and the Lexicon. In The nature of explanation in linguistic theory, eds. John
Moore and Maria Polinsky. Stanford: CSLI.
Caha, Pavel. 2009. The Nanosyntax of Case, University of Tromsø: Doctoral dissertation.
Calabrese, Andrea. 1988. Towards a theory of phonological alphabets, MIT: Ph.D.
Dissertation.
Calabrese, Andrea. 1995. A constraint-based theory of phonological markedness and
simplification procedures. Linguistic Inquiry 26:373-463.
Calabrese, Andrea. 2011. Investigations on markedness, syncretism and zero exponence
in morphology. Morphology 21:283-325.
Calabrese, Andrea and Diego Pescarini. 2014. Clitic metathesis in the Friulian dialect of
Forni. Probus 26:275-308.
Carstairs, Andrew. 1987. Allomorphy in Inflexion: Croom Helm Linguistics Series.
London: Croom Helm.
Chomsky, Noam. 1957. Syntactic Structures. The Hague: Mouton.
Chomsky, Noam. 1970. Remarks on Nominalizations. In Readings in English
Transformational Grammar, eds. Roderick Jacobs and Peter Rosenbaum, 184-
221. Waltham, Mass: Ginn & Company.
Chomsky, Noam. 1981. Lectures on Government and Binding. Dordrecht: Foris.
Chomsky, Noam. 1995. The Minimalist Program. Cambridge, Mass: MIT Press.
Chung, Inkie. 2007. Suppletive negation in Korean and Distributed Morphology. Lingua
117:95-148.
Chung, Inkie. 2009. Suppletive verbal morphology in Korean and the mechanism of
vocabulary insertion. Journal of Linguistics 45:533-567.
Corbett, Greville G. 1991. Gender. Cambridge: Cambridge University Press.
Van Craenenbroeck, Jeroen, and Marijke De Belder. In press. How to merge a root.
Linguistic Inquiry.
Cysouw, Michael 2003 The paradigmatic structure of person marking. Oxford University
Press.
28
D’Alessandro, Roberta and Tobias Scheer. In Press. Modular PIC. Linguistic Inquiry.
Dowty, David. 1979. Word Meaning and Montague Grammar. Dordrecht: Reidel.
Embick, David. 2003. Locality, Listedness, and morphological identity. Studia
Linguistica 57:143-169.
Embick, David. 2010. Localism versus Globalism in Morphology and Phonology.
Cambridge: MIT Press.
Embick, David, and Noyer, Rolf. 2001. Movement operations after syntax. Linguistic
Inquiry 32:555-595.
Fox, Danny, and Pesetsky, David. 2005. Cyclic Linearization of Syntactic Structure.
Theoretical Linguistics 31:1-45.
Gärtner, Hans-Martin. 1999. Phrase-Linking meets Minimalist syntax. In Proceedings of
the18th West Coast Conference on Formal Linguistics (WCCFL 18), eds. Sonya
Bird, Andrew Carnie, Jason D. Haugen and Peter Norquest, 159-169. Somerville,
MA: Cascadilla Press.
Greenberg, Joseph H. 1963. Some universals of grammar with particular reference to the
order of meaningful elements. In Universals of Language, ed. Joseph H.
Greenberg, 73-113. Cambridge: MIT Press.
Halle, Morris. 1990. An approach to morphology. In Proceedings of the North Eastern
Linguistics Society Annual Meeting (NELS 20), eds. J. Carter, R.-M. Dechaine,
B. Philip and T. Sherer, 150-184. Amherst: GLSA.
Halle, Morris. 1997. Distributed Morphology: Impoverishment and Fission. In MIT
Working Papers in Linguistics, eds. Benjamin Bruening, Yoonjung Kang and
Martha McGinnis, 425-449. Cambridge: MIT.
Halle, Morris, and Hale, Ken. 1997. Chukchi transitive and antipassive constructions.
Cambridge, Mass: MIT.
Halle, Morris, and Marantz, Alec. 1993. Distributed Morphology and the pieces of
inflection. In The view from Building 20: Essays in Linguistics in honour of
Sylvain Bromberger, eds. Ken Hale and Samuel Jay Keyser, 111-176.
Cambridge, Mass: MIT Press.
Halle, Morris, and Marantz, Alec. 1994. Some Key Features of Distributed Morphology.
In Papers on Phonology and Morphology, eds. Andrew Carnie, Heidi Harley and
Tony Bures, 275-288. Cambridge, Mass: MIT.
Hankamer, Jorge, and Mikkelsen, Line. 2005. When movement must be blocked: A
response to Embick and Noyer. Linguistic Inquiry 36:85-125.
Harbour, Daniel. 2007. Morphosemantic Number: From Kiowa Noun Classes to UG
Number Features. New York: Springer.
Harbour, Daniel. to appear Impossible Persons. MIT Press.
Harley, Heidi. 1993. Hug a tree. In Papers on phonology and morphology, eds. Andrew
Carnie, Heidi Harley and Tony Bures, 289-320. Cambridge, Mass: MIT.
Harley, Heidi. 2008. When is a syncretism more than a syncretism? Impoverishment,
metasyncretism, and underspecification. In Phi Theory: Phi-features across
Modules and Interfaces, eds. Daniel Harbour, David Adger and Susana Béjar,
251-294. Oxford: Oxford University Press.
Harley, Heidi. 2009. The morphology of nominalizations and the syntax of vP. In
Quantification, Definiteness, and Nominalization, eds. Anastasia Giannakidou
and Monika Rathert. Oxford: Oxford University Press.
Harley, Heidi. 2014. On the identity of roots. Theoretical Linguistics 40:225–276.
Harley, Heidi, and Noyer, Rolf. 2000. Licensing in the non-lexicalist lexicon. In The
Lexicon/Encyclopaedia Interface, ed. Bert Peeters, 349-374. Amsterdam:
Elsevier.
Harris, Alice C. 2009. Exuberant Exponence in Batsbi. Natural Language and Linguistic
Theory 27:267-303.
Katzir, Roni. 2011. Morphosemantic mismatches, structural economy, and licensing.
Linguistic Inquiry 42:45-82.
29
Kayne, Richard. 1994. The Antisymmetry of Syntax. Cambridge, Mass: MIT Press.
Kiparsky, Paul. 1973. 'Elsewhere' in Phonology. In A festschrift for Morris Halle, eds.
Stephen R. Anderson and Paul Kiparsky, 93-106. New York: Holt, Reinhart and
Winston.
Kiparsky, Paul. 1982. Lexical morphology and phonology. In Linguistics in the Morning
Calm, I. S. Yang, ed., 3-91. Seoul: Hanshin Publishing Compnay.
Kramer, Ruth.2015. The morphosyntax of gender. Oxford University Press.
Legate, Julie Anne. 2008. Morphological and abstract case. Linguistic Inquiry 39:55-101.
Levin, Beth, and Rappaport-Hovav, Malka. 1995. Unaccusativity at the syntax-lexical
semantics interface. Cambridge, Mass: MIT Press.
Lieber, Rochelle. 1980. On the organization of the lexicon. Bloomington, IN: Indiana
University Linguistics Club.
Lieber, Rochelle. 1982. Allomorphy. Linguistic Analysis 10:27-52.
Lowenstamm, Jean. 2010. Derivational Affixes as Roots (Phasal Spellout meets English
Stress Shift). Ms. Université Paris Diderot-Paris 7.
Marantz, Alec. 1989. Clitics and phrase structure. In Alternative Conceptions of Phrase
Structure, eds. Mark Baltin and Anthony Kroch, 99-116. Chicago: The
University of Chicago Press.
Marantz, Alec. 1991. Case and Licensing. In Proceedings of ESCOL '91, eds. German
Westphal, Benjamin Ao and Hee-Rahk Chae, 234-253: Cornell Linguistics Club.
Marantz, Alec. 1995. 'Cat' as a phrasal idiom. Cambridge, Mass: MIT.
Marantz, Alec. 1997. No Escape from Syntax: Don't Try Morphological Analysis in the
Privacy of Your Own Lexicon. UPenn Working Papers in Linguistics 4:201-225.
Matthews, P.H. 1972. Inflectional Morphology: A theoretical study based on aspects of
Latin verb conjugation. Cambridge: Cambridge University Press.
Merchant, Jason. 2015. How much context is enough? Two cases of span-conditioned
stem allomorphy. Linguistic Inquiry 46:273-303.
Monich, Irina. 2015. Morphosyntax and Tonology of Sotho Languages. PhD Dissertation,
UConn.
Moskal, Beata. 2015a. A case study in nominal suppletion. Linguistic Inquiry 46:376-
387.
Moskal, Beata. 2015b. Domains on the Border: Between Morphology and Phonology.
PhD. Dissertation, UConn.
Müller, Stefan and Stephen Wechsler. 2014. Lexical Approaches to Argument Structure.
Theoretical Linguistics 40:1–76.
Muysken, Pieter. 1981. Quechua Causatives and Logical Form: a Case Study in
Markedness. In Proceedings from the 1979 GLOW Conference, eds. Adriana
Belletti, Luciana Brandi and Luigi Rizzi.
Neeleman, Ad, and Szendrői, Kriszta. 2007. Radical pro-drop and the morphology of
pronouns. Linguistic Inquiry 38:671-714.
Nevins, Andrew. 2007. The representation of third person and its consequences for
Person-Case effects. Natural Language and Linguistic Theory 25:273-313.
Nevins, Andrew. 2011. Marked Targets vs. Marked Targets and Impoverishment of the
Dual. Linguistic Inquiry 42:413-444.
Newell, Heather and Glyne Piggott. 2014. Interactions at the syntax-phonology interface:
evidence from Ojibwe. Lingua 150:332-362.
Noyer, Rolf. 1997. Features, positions and affixes in Autonomous Morphological
Structure. New York: Garland Press.
Noyer, Rolf. 1998a. Clitic sequences in Nunggubuyu and PF convergence. Natural
Language and Linguistic Theory 19:751-826.
Noyer, Rolf. 1998b. Impoverishment theory and morphosyntactic markedness. In
Morphology and its relation to phonology and syntax, eds. Steven G. Lapointe,
Diane K. Brentari and Patrick M. Farrell, 264-285. Stanford: CSLI.
30
Nunes, Jairo. 2004. Linearization of Chains and Sideward Movement. Cambridge: MIT
Press.
Oltra-Massuet, Isabel, and Arregi, Karlos. 2005. Stress-by-structure in Spanish.
Linguistic Inquiry 36:43-84.
Pescarini, Diego. 2010. Elsewhere in Romance: Evidence from Clitic Clusters. Linguistic
Inquiry 41:427-444.
Pertsova, Katya 2011. Grounding systematic syncretism in learning. Linguistic Inquiry
42:225-266.
Pesetsky, David. 1995. Zero Syntax: Experiencers and Cascades. Cambridge, Mass: MIT
Press.
Pollock, Jean-Yves. 1989. Verb Movement, Universal Grammar and the Structure of IP.
Linguistic Inquiry 20:365-424.
Radkevich, Nina. 2010. On Location: the Structure of Case and Adpositions, University
of Connecticut: Ph.D. Dissertation.
Ramchand, Gillian Catriona. 2008. Verb Meaning and the Lexicon: A First Phase Syntax.
Cambridge: Cambridge University Press.
Sadock, Jerrold M. 1985. Autolexical Syntax: A proposal for the treatment of noun
incorporation and similar phenomena. Natural Language and Linguistic Theory
3:379-439.
Sauerland, Uli and Jonathan David Bobaljik. 2013. Syncretism Distribution Modeling:
Accidental Homophony as a Random Event. In Nobu Goto, Koichi Otaki,
Atsushi Sato, and Kensuke Takita, eds, Proceedings of GLOW Asia IX 2012, 31-
53.
Siegel, Dorothy. 1978. The Adjacency Constraint and the theory of Morphology. In
Proceedings of the, ed. M. Stein, 189-197. Amherst, MA: University of Amherst.
Skorik, Piotr Ja. 1977. Грамматика чукотского языка, часть II: глагол,
наречие, служебные слова [Grammar of Chukchi, vol II: verb, adverb and
auxiliary words]. Leningrad: Nauka.
Smith, Peter W., Beata Moskal, Jungmin Kang, Ting Xu, and Jonathan Bobaljik. 2015.
Pronominal Supletion: Case and Number. To appear in Proceedings of NELS 45.
Spencer, Andrew. 2000. Agreement morphology in Chukotkan. In Morphological
analysis in comparison, eds. Wolfgang U. Dressler, Oskar E. Pfeiffer and
Markus A. Pochtrager, 191-222. Amsterdam: John Benjamins.
Sproat, Richard. 1985. On deriving the lexicon, M.I.T.: Doctoral dissertation.
Starke, Michal. 2009. Nanosyntax: a short primer to a new approach to language.
Nordlyd 36:1-6.
von Stechow, Arnim. 1996. The different readings of wieder 'again': a structural account.
Journal of Semantics 13:87-138.
Stump, Gregory T. 1993. On rules of referral. Language 69:449-479.
Svenonius, Peter. 2012. Spanning. Ms. CASTL/Tromsø.
Trommer, Jochen. 1999. Morphology consuming syntax's resources: Generation and
parsing in a Minimalist version of Distributed Morphology. Potsdam: Universität
Potsdam.
Ultan, Russell. 1972. Some features of basic comparative constructions. In Working
papers on language universals, 117-162.
Williams, Edwin. 1981. On the notions "Lexically related" and "Head of a word".
Linguistic Inquiry 12:245-274.
Williams, Edwin. 1994. Remarks on lexical knowledge. Lingua 92:7-34.
Williams, Edwin. 2003. Representation Theory. Cambridge: MIT Press.
Wurmbrand, Susi. 2006. Licensing Case. Journal of Germanic Linguistics 18:174-234.
Zaenen, Annie, Maling, Joan, and Thráinsson, Höskuldur. 1985. Case and grammatical
functions: The Icelandic passive. Natural Language and Linguistic Theory
3:441-483.
31
Zwicky, Arnold M. 1985. How to describe inflection. In Proceedings of the Berkeley
Linguistics Society, 372-386.