Studies in Chinese Linguistics
Towards a Theory of Morphology as Syntax
Chris Collins, Richard S. Kayne
New York University
Abstract
Phenomena traditionally thought of as morphological can be accounted for in
terms of syntactic operations and principles, hence bringing forth questions
that traditional morphology fails to ask (for instance, concerning the licensing
of empty morphemes). The language faculty contains no specific morphological
component, nor any post-syntactic morphological operations.
Keywords
morphology, allomorphy, syncretism, suppletion, late insertion
Studies in Chinese Linguistics, Volume 44, Number 1, 2023, 1–32 DOI: 10.2478/scl-2023-0001
© 2023 Chris Collins, Richard S. Kayne. Studies in Chinese Linguistics is published by Sciendo on
behalf of T.T. Ng Chinese Language Research Centre, Institute of Chinese Studies, The Chinese
University of Hong Kong. This work is licensed under the Creative Commons AttributionNonCommercial-NoDerivatives 4.0 International License.
2 Morphology as Syntax
1 Introduction
In this paper, we discuss the relationship between syntax and morphology. In
particular, we are interested in the question of the extent to which morphological
generalizations can be accounted for in terms of syntactic operations and
principles. The thesis to be defended is the following (see Caha 2020: 8 for a very
similar statement concerning Nanosyntax):
(1)
Morphology as Syntax (MaS)
Morphological generalizations are accounted for in terms of syntactic
operations and principles. There is no morphological component in Universal
Grammar (UG), nor are there post-syntactic morphological operations.
MaS is a program for research, which has as an underlying assumption that
separation of morphology and syntax cannot be achieved in any natural way. In
other words, it is impossible to analyze morphology in isolation from syntax.
For example, it is impossible to understand syncretism in a verbal paradigm
without an analysis of the syntax of the language (see section 7 on Spanish for
a case study).
MaS is a consequence of the Strong Minimalist Thesis (SMT) of Chomsky
2000, which states that “language is an optimal solution to legibility conditions”,
and therefore a theory lacking a morphological component is plausibly more
optimal. Furthermore, just like the operations and principles of syntactic theory can
be subject to scrutiny from the point of view of the SMT (for example, D-Structure,
S-Structure, traces, c-command, labels, etc.), proposed operations and principles of
morphology can be subject to scrutiny from the point of view of MaS.
For syntactic theory, we assume the broad outlines of minimalist syntax,
recently formalized by Collins and Stabler (2016). The basic assumptions
are sketched in section 2. Nevertheless, we are not making the claim that all
morphological generalizations can be reduced to syntactic principles as they are
understood today. Rather, new insights about syntax can be learned by trying to
give syntactic explanations of morphological generalizations.
As for the relationship between phonology and syntax, works such as
Dobashi 2020 argue for a particularly close connection between certain aspects
of phonology and syntax. Moreover, one can ask more generally if the kinds of
structures and constraints found in phonology fall under syntactic theory (cf.
Nevins’ (2010) account of vowel harmony). For example, can tone spreading in
autosegmental phonology be understood in the same terms as successive cyclic
movement in syntax? For the purposes of this paper, we put these important
issues aside and assume that there is a clear distinction between phonology and
syntax, on the grounds that the primitives manipulated in syntax (e.g., formal
features such as PL) are distinct from the primitives manipulated in phonology
(e.g., [+voiced]).
Chris Collins, Richard S. Kayne 3
2 Syntactic Framework
In order to defend (1), the theory of syntax to be employed needs to be discussed.
We are assuming the minimalist theory of syntax, as outlined and formalized
in Collins and Stabler 2016, the most important assumption being that syntactic
objects are combined by Merge:
(2)
For all syntactic objects X and Y, Merge(X,Y) = {X,Y}
In (2), Merge takes two objects and produces an unordered set (but see Kayne
2019b, 2022 which argue for incorporating linear/temporal order into the definition
of Merge). For the purposes of this paper, we adopt (2) and leave open the question
of how morphological phenomena bear on the issue of whether or not the definition
of Merge incorporates linear/temporal order.
There are two subcases of Merge: internal Merge (movement) and external
Merge. Internal Merge yields various patterns such as: successive cyclic movement,
roll-up movement, remnant movement, smuggling, etc. Movement is subject to
locality conditions such as the Minimal Link Condition, Relativized Minimality
(MLC/RM) and Phase Impenetrability Condition (PIC). All syntactic operations
are subject to economy conditions.
The set of syntactic objects includes the set of lexical items (morphemes). We
assume that lexical items can be defined as follows (for alternatives in the spirit of
MaS see Collins 2017: 61, Kayne 2016):
(3)
LI: {FF, PHON},
where FF is a set of formal features and PHON is a sequence of
phonological segments.
This definition excludes late insertion (since it assumes that lexical items pair up
FF and PHON pre-syntactically (in the lexicon)) and raises important questions
including: Can PHON = ∅ (the empty sequence)? What is the set of formal features
of UG? What kinds of sets of FF are allowed? Can they be of arbitrary complexity?
Kayne (2005: 212) (see also Bobaljik 2012: 212) formulates the following hypothesis
(see Collins 2022 for discussion):
(4)
Principle of Decompositionality
UG imposes a maximum of one interpretable syntactic feature per lexical item.
The question of formal features in the definition of lexical items is very closely
related to the question of hierarchies of functional projections in the clause (e.g., the
left periphery, the adverb hierarchy, extended projection of lexical categories, and
other hierarchies discussed in the cartography literature). The syntactic hierarchies
interact with syntactic constraints on movement to yield cross-linguistic variation.
A striking example of this kind of work is from Cinque (2005) who argues that
a version of Greenberg’s Universal 20 (Greenberg 1966) can be accounted for in
terms of a universal hierarchy of modifiers combined with restrictive theories of
linear order and movement. Greenberg’s Universal 20 states (see Cinque 2005: 315):
4 Morphology as Syntax
(5)
“(a) that in prenominal position the order of demonstrative, numeral, and adjective
(or any subset thereof) conforms to the order Dem > Num > A, and (b) that in
postnominal position the order of the same elements (or any subset thereof)
conforms either to the order Dem > Num > A or to the order A > Num > Dem.”
In MaS, one would expect such syntactic argumentation to carry over into the
domain of morpheme order. In fact, Cinque (2015) argues that the order of Mood,
Tense and Aspect morphemes can be accounted for by similar principles.
On syntactic approaches to morpheme order see (amongst others) Baker
1985, 1988, Buell et al. 2014, Julien 2007, Kayne 2010a, Koopman 2005,
2017, Muriungi 2014, and Zyman and Kalivoda 2020. On syntactic approaches
to clitic order see Kayne 1994, Ordóñez 2002, Săvescu-Ciucivara 2009,
Shlonsky, to appear, and Terzi 1999. On MaS compliant ways of analyzing agent
nominalizations see Collins 2006, Kayne 2008, Fábregas 2012, Ntelitheos 2012,
and Gotah and Lee 2022.
Once a syntactic object is formed, it must be spelled out at the phase level. So,
we assume that there is an operation Transfer that has two components TransferPF
and TransferLF.
(6)
For any syntactic object SO,
Transfer(SO) = {TransferPF(SO), TransferLF(SO)}
Transfer is often referred to as Spell-Out in the syntax literature. Spell-Out has the
function of establishing a linear/temporal order of the morphemes, which we will
assume conforms to the Linear Correspondence Axiom (Kayne 1994). Spell-Out
also determines which occurrences of a syntactic object are spelled out (e.g., in
remnant movement).
3 Empty Elements
Related to Spell-Out is the issue of when a morpheme can fail to be pronounced
(or alternatively, when a morpheme can be spelled out as zero). Any morpheme
that is unpronounced must either be lexically zero (specified in the lexicon as
having no phonological form) or syntactically licensed as unpronounced in some
way or the other. The issue of zero morphemes and how they are licensed is of
great importance in the MaS framework. Here, we briefly survey the relevant
syntactic principles.
First, if a constituent X undergoes movement (internal Merge), typically only
the highest occurrence is spelled out (although some care is needed to get SpellOut to work correctly for remnant movement, see Collins and Stabler 2016; also
Nunes 2004).
Second, if a constituent X has an identical antecedent, in many cases it may
undergo ellipsis (see Van Craenenbroeck and Merchant 2013 for a survey; see
section 4.1 below for an example).
Chris Collins, Richard S. Kayne 5
Third, there are many effects falling under the Doubly Filled Comp Filter
in syntax. These are situations where the head and the specifier of a maximal
projection cannot be filled overtly at the same time (see Collins 2007; Koopman
2000: 350–353; Koopman and Szabolcsi 2000). The version from Collins 2007 is
given below (Edge(X) includes both the head and the specifier of X):
(7)
No Crowding Condition
a. Edge(X) must be phonetically overt.
b. The condition in (a) applies in a minimal way so that either the head, or the
specifier, but not both, are spelled-out overtly.
Fourth, Kayne (2006) proposes that the specifier of a phase is in general not
spelled-out. He has used that principle to explain the presence of a null MUCH
in expressions like enough money (not *enough much money or *much enough
money as opposed to too much money or so much money).
Fifth, Emonds (1987) proposes the Invisible Category Principle which
regulates, amongst other things, the distribution of null prepositions in English.
The principle states:
(8)
Invisible Category Principle
“A closed category B with positively specified features Ci may remain empty
throughout a syntactic derivation if the features Ci (save possibly B itself) are
all alternatively realized in a phrasal sister of B.”
On a possibly related note, Choi and Harley (2019: 1346, 1347, footnote 32)
observe that it is often the case when X triggers suppletion of some other head Y,
X is realized as a zero allomorph.
Sixth, there are null pronominals of various sorts, including PRO, pro and
implicit arguments, each with its own licensing conditions (on null pro in object
position, see Rizzi 1986; on the syntactic status of implicit arguments, see Bhatt
and Pancheva 2017 and Collins 2022; on null subjects, see Holmberg 2005, 2010).
See also Carstens 1991, 1993 on diminutive morphology in Bantu, and in particular
the Zero-Morpheme Licensing Principle.
There are many other areas where people have proposed empty elements
in syntax. Among other principles, the MaS proposal is that empty morphemes
in what is traditionally called morphology are licensed in the same way as other
empty categories in syntax.
4 Examples
In this section, we will go over a number of short case studies illustrating some
analyses that conform to MaS, and some that do not.
4.1 Ellipsis
Caha (2013: 1023) argues that Case Attraction in Classical Armenian should be
analyzed as ellipsis (see also Erschler 2018 for another relevant case study).
6 Morphology as Syntax
(9)
a. N-ABL
b. N-ABL
c. N-ABL
N-GEN
N-GEN AGR=ABL
N-GEN AGR=ABL
(underlying)
(case agreement)
(ellipsis)
In (9a) there is a head noun and a complement noun. Normally the nominal
complement is in the genitive. However, if the head noun has dative, ablative or
instrumental case, the complement can optionally show up in that case instead.
Caha argues that there is a process of case agreement between the complement
and head noun (see (9b)), followed by ellipsis of the case marker of the complement
noun (see (9c)). The reason why ABL can trigger ellipsis of GEN is that ABL
actually contains GEN structurally. See Caha 2013 for details.
Two aspects of this explanation deserve comment from the perspective
of MaS. First, ellipsis under identity with an antecedent is a mechanism that is
independently needed in syntax (in the spell-out of syntactic objects). Therefore,
Caha’s explanation of Case Attraction falls squarely in the MaS program. Second,
just like in current discussions of ellipsis in syntax, there is no need to assume that
any structure has been removed in (9c) (unlike impoverishment in DM). Rather,
GEN is simply unpronounced under identity with a subpart of ABL.
4.2 Infintival [e] in Italian
In Italian, the infinitive normally ends in the vowel [e]. Cardinaletti and Shlonsky
(2004: 529) note that “when a pronoun appears as an enclitic to the infinitive, [e] is
obligatorily absent.” The examples below are from pages 529, 537:
(10) a. Lo vorrei
vedere.
it
would.want
see
‘I would want to see it.’
b. Vorrei veder(*e)lo.
(11) a. Lo
sono andato a
him
am
gone
to
‘I went to visit him.’
b. Sono andato a trovarlo.
trovare.
visit
As shown in these examples, when the clitic is before the finite verb, the infinitive
ends in [e]. When the clitic follows the infinitive, [e] is absent. Cardinaletti and
Shlonsky (2004: 532) propose that:
(12) “[e] is an independent morpheme and not part of the nonfinite inflection,
which is [r].”
Adopting this assumption, we propose that the final [e] heads a functional projection
whose specifier may host a clitic moved from a lower position, if there is one (for
related proposals about clitics see Săvescu-Ciucivara 2009). For example, in (10b)
and (11b), the clitic lo has moved into Spec FP. By the No Crowding Condition in (7),
the head of F is not pronounced. In other words, the empty F is licensed by the NCC,
one of the principles in section 3. If the clitic precedes the finite verb, then Spec FP is
either empty or contains a trace, and so F is pronounced as shown in (10a) and (11a).
Chris Collins, Richard S. Kayne 7
How this account relates to clitic clusters (as in Săvescu-Ciucivara 2009), we
leave to further work.
4.3 Contextual Restrictions and Double Plurals
Kramer (2016a: 544) analyzes double plurals in Amharic. The possible plurals of
the word for “baby” are given below:
(13) a. hɨs’an-at
b. hɨs’an-otʃtʃ
c. hɨs’an-at-otʃtʃ
(p. 528)
(p. 544)
(p. 544)
In her framework, the irregular plural (13a) is a little n with an uninterpretable
[+pl] feature. The regular plural (13b) is formed by adding Num which has an
interpretable [+pl] feature. The double plural in (13c) combines the irregular and
regular plurals.
The structure of the double plural is given in (14):
(14) [NumP [nP [Root] n] Num]
To account for the data in (13), Kramer assumes the following vocabulary items:
(15) a. Num, [+pl] ←→ -otʃtʃ
b. n, [+pl] ←→ -at / {HƗS’AN,….}
(regular)
(irregular)
Kramer elaborates (p. 530) “Overall, since every nominal has a regular plural and
there are double plurals, I conclude that regular and irregular plural morphology do
not compete for insertion in Amharic; in other words, they do not occupy the same
syntactic head (Num). Instead, I propose a split analysis of number: the ‘regular’
plural suffix is the realization of Num[+pl] and irregular plural morphology is the
realization of n[+pl]…”.
How are such contextual restrictions as those in (15) handled in MaS?
Contextual restrictions on morphemes define where a morpheme can be merged
into a syntactic object. They are to be understood in terms of relations familiar
from syntax (c-selection, s-selection, l-selection, and the relation between the parts
of idiomatic expressions). In other words, there is no such thing as morphologyspecific contextual restrictions on morphemes (or on vocabulary insertion). Since
syntactic selectional restrictions obey locality conditions, by MaS, contextual
restrictions on morphemes will obey the same locality conditions (for a survey of
existing theories, see Choi and Harley 2019). However, the general goal should be
to reduce or eliminate the use of such restrictions as much as possible, since each
restriction is a stipulation.1
It is sometimes easy to translate DM style vocabulary items into MaS style
lexical items (morphemes). From the point of view of MaS, the regular plural
1
See Matthews and Yip 2011: 124 “...classifiers are not merely selected by individual nouns, but can
be applied meaningfully to classes of nouns and to new situations.”
8 Morphology as Syntax
in Amharic has the following lexical entry (on morphology and selection, see
Fabb 1988):
(16) LI: {-otʃtʃ, Num[+pl]} selects nP
If the positions of NumP and nP in the functional sequence were part of UG, the
selectional restriction (“selects nP”) would not need to be listed in the lexical item
in (16).
On the other hand, the irregular plural at- is more restricted and can be
defined as follows:
(17) LI: {-at, n[+pl]} selects {hɨs’an ‘baby’,…}
The restriction is that the irregular plural -at can only appear with one of a small
number of roots. Such a restriction is similar to the lexical restrictions found in the
following phrases in English:
(18) a. by dint of
b. to keep tabs on
In both cases, there is a lexical item that can appear in a very specific lexical
context. In (18a), the noun dint only appears in the expression by dint of, and in
no other expression. In other words, there is lexical selection, where dint lexically
selects the preposition by (if dint is a noun, the presence of the preposition of is
predictable and does not need to be accounted for by lexical selection). Similarly,
tabs (in the relevant sense) can only appear with keep in (18b). We assume in this
case that the idiom chunk tabs is lexically selecting the verb keep.
In MaS, (16) and (17) are lexical items, and are not inserted post-syntactically.
Rather, they are merged together with the root or the noun to form plurals in
Amharic. The syntactic derivation of the double plural in Amharic is given in
(19). As is customary, the lexical items in the derivation are represented by their
phonological forms:
(19) a. Merge(hɨs’an, -at)
=
{hɨs’an, -at}
(satisfying the selectional requirements of -at)
b. Merge({hɨs’an, -at}, -otʃtʃ)
=
{{hɨs’an, -at}, -otʃtʃ}
(satisfying the selectional requirements of -otʃtʃ)
c. Spell-Out({{hɨs’an, -at}, -otʃtʃ})
=
hɨs’an-at-otʃtʃ
Since Kramer assumes that the n and Num heads appear finally, the result of (19b)
would be spelled out as in (19c) (perhaps involving several instances of internal
Merge), with no late insertion.
4.4 Allomorphy
A case related to Amharic irregular plurals arises with English irregular plurals.
Consider the DM treatment of the irregular plural ox-en given by Embick (2015:
171–172):
Chris Collins, Richard S. Kayne 9
(20) a. [+pl]
b. [+pl]
c. [+pl]
←→
←→
←→
-en/{OX, CHILD, …}
∅/{FISH, MOOSE…}
/-z/
The vocabulary items in (20) represent a late insertion analysis for the English plural.
The late insertion analysis of the English irregular plural is considered a foundational
example for the analysis of allomorphy in the DM literature, see for example McGinnis
2017: 395, Bonet and Harbour 2012: 196, Bobaljik 2012: 7, Embick and Noyer 2007:
298–299. In section 5 below, we will argue that late insertion is incompatible with
MaS, but apart from such, there are other problems with (20).
For example, there is a difference between irregular and regular plurals in
compounds (see also Pinker 1999: 179 who cited the examples of “draft horse,
mule and oxen power”):
(21) a. ox cart, oxen cart
b. dog sled, *dogs sled
While both forms in (21a) seem to be more or less acceptable, dogs sled in (21b) is
sharply degraded, which suggests that the regular plural and the irregular plural
-en do not occupy the same syntactic position, contrary to the claim of the analysis
in (20). Rather, -en occupies a lower position than -s. Following Collins (2018), we
propose that -en occupies an inner plural position (PL1) and -s occupies an outer
plural position (PL2). Only the inner plural is compatible with noun compounds.
In other words, English has an inner and outer plural, in a way similar to Amharic
(discussed in the previous section).
A similar contrast holds for derivational morphology:
(22) a. childrenize, oxenize
b. *kidsize, *oxesize, *cowsize
The examples in (22a) seem relatively acceptable, but the examples in (22b) are
sharply unacceptable. Just as in the case of compounds, such a contrast suggests
two different positions for the two different plural markers (-en versus the regular
plural -s). Only a plural noun with an inner plural can take a derivational suffix
like -ize.
Our split plural analysis of ox-en is similar to Punske and Jackson’s (2017:
268) analysis of English plurality: “In English, the higher position Num encodes
individuation while the lower position, n, encodes general plurality.” They show
convincingly that only lower plurals appear in compounds.2
2
We unsurprisingly take compounds in general to be put together by the syntax by Merge, with
non-transparent compounds being akin to phrasal idioms; certain compounds will contain silent
elements (cf. Lees 1963 and Levi 1978), as in the case of a New Yorker plausibly being “a New
York LIVE IN -er”; the extent to which Chinese compounds as discussed by Liao (2014) are similar
needs to be determined, with Huang 2015 being particularly relevant.
10 Morphology as Syntax
In a MaS style analysis, the relevant lexical items are given below:
(23) a. LI1:
b. LI2:
{PL2, -s}
{PL1, -en}
selects X
selects ox
In these lexical items, X stands for some projection in the DP that is as high as
PL1P (so that there is an inner and an outer plural position). If the positions of
PL1P and PL2P2 in the functional sequence were part of UG, there would be no
need for the selectional restrictions to be listed in the lexical items in (23).
What is the relationship between PL1 and PL2? There are three possibilities.
First, we could adopt Kramer’s (2016a) analysis that the outer plural is interpretable
while the inner plural is uninterpretable. In that case, PL2 = PL and PL1 = uPL.
Second, it could be that they each are interpretable with different semantic values,
as suggested by Punske and Jackson (2017) (recall, individuation versus general
plurality). Third, it could be that they each are interpretable, but with exactly the
same semantic values. In that case, PL1 = PL2 = PL. We will return to this issue
in a moment.
These morphemes do not compete with one another, in that many speakers
allow both kinds of plurals:
(24) a. oxes
b. oxen
As Pinker (1999: 52) notes: “Most Americans meet oxen mainly in writing, and
commonly say oxes instead.” On the DM analysis in (20), such a state of affairs
illustrated in (24) would be completely unexpected. Since the irregular plural -en
is more specific (having a more specific context for vocabulary insertion), it should
block the regular plural. However, under the split plural analysis, the data in (24)
is completely expected, just like in Amharic both the irregular and regular plurals
are possible (see (13)). See Collins 2018 for more discussion of the theoretical
significance of morphological doublets for the issue of late insertion.
A piece of evidence for the split plural analysis is the following contrast:
(25) a. ??oxens
b. *oxesen
While both forms seem unacceptable, the order of the plural markers in (25a) is
much better than (25b). This contrast is consistent with our analysis in that the
order in (25a) reflects the order of inner plural before outer plural. A future project
would be to investigate whether oxens appears in internet searches, and whether
the form exists in different English dialects. A related question is whether the
double plural childrens appears in child English.
The relation between outer plural -s and inner plural -en has much in common
with the relation between negative un- and negative in-, with ??oxens versus
Chris Collins, Richard S. Kayne 11
*oxesen parallel to ?unineffective versus *inuneffective. The parallel suggests that
un- is an outer negation and in- is an inner negation.3
Nonetheless, if there is an inner plural and an outer plural, then why is the
double plural (25a) not perfectly acceptable, and what accounts for the difference
between English and Amharic double plurals (see (13) above)? One possibility is
that the inner plural in Amharic is uninterpretable (following Kramer (2016a)),
but both the inner plural and outer plural are interpretable in English (they both
have plural semantics). If both the inner plural and outer plural are interpretable in
English, (25a) might be relatively unacceptable by Full Interpretation, in the sense
of Chomsky 1986: 98.
Although it is unclear that the above analysis can serve as a general model for
contextual allomorphy, one important general assumption is the following (see (3)
for the definition of morpheme):
(26) Contextual allomorphy involves two or more different morphemes.
In the example at hand, it is not that the irregular -en and the regular -s are
competing for insertion in a given position. Rather, each of the morphemes
heads its own functional projection. In general, on the approach to contextual
allomorphy outlined in this paper, in any given case there will be at least four
questions to answer. Consider the example of oxen again, which gives rise to these
relevant questions:
(27) a. Why can ox combine with -en?
b. Why can ox combine with -s?
c. Why can fox not combine with -en?
d. Why can fox combine with -s?
Here are the answers to those questions: (a) ox can combine with -en because ox
satisfies the contextual restrictions of -en (see (23b)). (b) ox can combine with
-s because ox satisfies the contextual restrictions of -s (see (23a)). (c) fox cannot
combine with -en because fox does not satisfy the contextual restrictions of -en.
(d) fox can combine with -s because fox satisfies the contextual restrictions of -s.
The mechanisms of late insertion, competition, blocking and the elsewhere
condition will play no role in the answers to these questions, since our proposal is
3
For a proposal to the effect that the prefixal (vs. suffixal) character of un- and in- is explainable, see
Kayne 2017.
The Cantonese negative prefix m- discussed by Matthews and Yip (2011: 44) may be less close to
English un- or in-, at least when prefixed to verbs, than it is to the silent NOT that Kayne and Moro
(to appear) associate with negative verbs like deny.
The fact that Cantonese and Mandarin have prefixal counterparts hó- and ke-, respectively, of
English suffixal -able, as Matthews and Yip (2011: 44) and Liao (2014: 7) point out, might reflect a
single irreducible parametric difference; alternatively, Cantonese and Mandarin have, via the use
of silent elements, something closer to able-to-be-V-ed than to V-able.
12 Morphology as Syntax
that such post-syntactic mechanisms are not part of UG and are irrelevant to the
analysis of contextual allomorphy.
A reviewer asks whether the account for oxen/oxes is applicable to plural
suppletion (e.g. mouse/mice)? Collins (2018) suggests that irregular plurals are
analyzed as man-F-PL (where F is a covert version of the -en one finds in oxen).
The covert morpheme F triggers a stem vowel change (man to men) and also
blocks the final regular plural. A difference between oxen/oxes and *childs, *mans,
*womans is that the latter are all unacceptable. Collins (2018) suggests that these
can be understood as child, man and woman lexically selecting the irregular plural
morpheme F (where ox would not select an irregular plural). Further analysis of
these cases will not be pursued here.
On an account of suppletion (e.g., *goed versus went) in the spirit of MaS see
Kayne 2018, 2019a, 2020, to appear.4
4.5 Insertion of dissociated morphemes
Distributed Morphology postulates the post-syntactic insertion of morphemes,
referred to as ornamental or dissociated morphemes in the DM literature (see
Embick 2015: 65). See also Choi and Harley’s (2019: 1331) discussion of insertion of
dissociated morphemes as node-sprouting. Embick and Noyer (2007: 309) add that:
(28) “As a working hypothesis, it has been suggested that only features irrelevant
to semantic interpretation, that is, features that are not interpretable, can be
introduced at PF…”
Morphemes that are introduced this way include agreement nodes (AGR), case
features, theme vowels (for verbs), class markers (for nominal declension classes)
(see also Halle and Marantz 1993: 135), and honorific morphemes (Choi and
Harley 2019).
An example of a dissociated morpheme in Italian is raised by Calabrese
(2015: 74) (see also Harris 1998: 44):
(29) “Thematic Vowels (TV) are special morphological elements adjoined to
certain functional heads in morphological structure by the rule in (11).”
That rule is given below:
(30) X0
→
X0
X0
4
TV
That kind of approach to suppletion may be extendable to Cantonese positive vs. negative
existential yáuh/móuh, as discussed by Matthews and Yip (2011, sect. 14.3), though with the latter
containing negative m-, in a way that recalls both (the n- of) Russian net and Mandarin méiyǒu, as
discussed by Li and Thompson (1981, chap. 12) and Yip and Rimmington (2004, sect. 16.1).)
Chris Collins, Richard S. Kayne 13
From the point of view of MaS, the problem with (30) is that it allows morphemes
to be combined with other morphemes in a way that is very different from Merge.
To put it another way, it posits that something would block such a morpheme from
being introduced by Merge in ordinary syntax.
(31) Merge(TV, X0)
However, when considering (31), it can be observed that (30) is redundant with the
syntactic operations of UG. In addition, the rule in (30) is highly specific to the TV
morpheme, which in turn raises the question of whether each dissociated morpheme
will need its own special rule, greatly complicating particular I-languages, and
raising serious questions of explanatory adequacy. Furthermore, (30) raises the
question of how such rules vary cross-linguistically. Therefore, it is very different
from Merge which is part of UG and able to combine any two syntactic objects.
However, Calabrese does not discuss why the more general formulation in (31)
would not work, nor how it would have affected his analysis.
A more challenging example of this kind of analysis is offered by Baker and
Kramer (2014). They claim that case markers in Amharic (‘from’, ‘to’, ‘by’, ‘for’)
are introduced post-syntactically. Some examples are given below.
(32) a. kä-bet
from-house
‘from a house’
b. kä-tɨllɨk’
bet
from-big
house
‘from a big house’
c. bä-zzih
bet
in-this
house
‘in this house’
d. lä-Gɨrma
wändɨmm
to-Girma
brother
‘to Girma’s brother’
e. kä-rädʒdʒɨm-u
näggade
suk’
from-tall-DEF
merchant
shop
‘from the shop of a tall merchant’
(p. 150)
(p. 150)
(p. 145)
(p. 151)
(p. 155)
In these examples, we adopt the word divisions given by Baker and Kramer. We
assume that their description of the distribution of the case markers is correct
(see (34) below), based on many complicated facts, concerning the distribution of
definite markers and the syntax of relative clauses. Due to space constraints, we
will refrain from recapitulating those arguments here.
To handle this kind of data, Baker and Kramer (2014) assume that in these
examples there is a null adposition following the noun, which assigns a case feature
to its complement.
14 Morphology as Syntax
(33)
PP
DP
Gɨrma
[+Ablative]
P
∅
=
kä- Gɨrma ‘from Girma’
The case marker itself is inserted post-syntactically by the following rule
(m-word = morphological word). A more complicated, recursive version of this
rule is needed to handle cases like that of (32e). For brevity’s sake, we put the
recursive version aside.
(34) Insertion Rule (Preliminary Version)
If a feature F is to be inserted within constituent X, then attach F to the m-word
Z such that Z asymmetricially c-commands all the other m-words in X.
Once again, from the point of view of MaS, the problem with (34) is that it allows
morphemes to be combined with other morphemes in a way that is very different
from Merge. Moreover, this rule is completely different from the one postulated by
Calabrese in (30), raising the question of how many such post-syntactic insertion
rules there are. Lastly, (34) is meant to be a rule of post-syntactic case insertion,
but relies on the notion of asymmetric c-command which is a purely syntactic
relation, again raising the issue of whether the insertion rule should be carried out
by syntactic mechanisms.
In formulating an alternative analysis, we adopt from Baker and Kramer
2014 the restriction of the case marker to m-words. We also adopt their idea that
there is a null post-position present. However, unlike them, we assume that the
case maker and its m-word are combined by Merge: Merge(K, m-word), which in
MaS, is the only way to introduce morphemes into a structure. Additionally, there
must be a local syntactic relation between the overt case marker K and the null P
(see (38) below).
Under these assumptions (and the LCA), the structure in (33) would be
replaced by the following:
(35)
PP
KP
K
kä
P’
DP
Gɨrma
P
∅
<KP>
In this example, the KP is the sister of P’. As for the other examples in (32b–e),
they resemble pied-piping:
(36) a. I wonder who she saw.
b. I wonder whose picture she saw.
c. I wonder whose brother’s picture she saw.
Chris Collins, Richard S. Kayne 15
In (36a), the embedded [+Q] complementizer requires a wh-phrase in its specifier,
but this requirement can be satisfied under pied-piping as shown in (36b).
Accordingly, Kayne (1994: 24) gives the following condition:
(37) The wh-phrase in interrogatives must asymmetrically c-command the
[+wh] head.
Adapting this condition to the case at hand, we have:
(38) The KP must asymmetrically c-command the corresponding null preposition P.
In (35), KP asymmetrically c-commands P. Now consider (32d), which has the
following structure:
(39) [PP [DP [KP lä- Gɨrma] wändɨmm] P’]
‘to Girma’s brother’
In the specifier (possessor) of the complement of the postposition, the KP
c-commands P, just as in pied-piping in (36b) where the possessor whose
c-commands the [+Q] complementizer. The Amharic example in (32e) is parallel
to the case of pied-piping in (36c), where whose is the specifier of a specifier.
Due to space restrictions, we are unable to analyze all the complicated
examples (some involving relative clauses) discussed by Baker and Kramer.
Whether or not our analysis is correct, it links the current set of facts to the general
theory of pied-piping.
In summary, MaS holds the following to be true:
(40) There is no post-syntactic insertion of morphemes.
5 Against Late Insertion
The preceding sections have given an overview of a syntactic approach to
morphology but the discussion so far leaves open the following question: What
is the status of late insertion? That is, is late insertion consistent with the general
MaS framework? In this section, we will argue that it is not, by the basic point:
Late insertion involves wide ranging formal mechanisms that operate on syntactic
structures but are independent of syntactic theory. By assuming early insertion,
those mechanisms may be rendered redundant.
Two theories that adopt the assumption of late insertion are Distributed
Morphology and Nanosyntax. As Baunaz and Lander (2018: 12) phrased it: “In
realizational, late-insertion theories like DM and Nanosyntax, however, sound and
meaning are not inherently linked but are separate entities, and it is only when the
syntactic derivation reaches a certain point that the meaning is paired with (for
some, replaced by) sound.” In the remainder of this section, we will focus on DM
and leave a discussion of Nanosyntax for future work.
The syntactic operation Merge takes two syntactic objects X and Y and
puts them together to form a third syntactic object Z = {X,Y}. This operation can
16 Morphology as Syntax
be looked at in two different ways, corresponding to the two interfaces (CI and
SM) that need to interpret the syntactic object formed. From one angle, Merge
is combining X and Y so that they can be composed semantically. From another
angle, Merge is combining X and Y so that when they are linearized (at Spell-Out),
the result is a phonological string determined by the phonological form of X and
of Y. In effect, Merge is introducing phonological forms into syntactic objects, so
there is no need for extra rules to govern the introduction of phonological forms
into terminals.
Late insertion theories do not adopt these assumptions about Merge.
In DM, morphemes have no phonological form. They are combined to form
syntactic objects (by Merge and a set of post-syntactic operations), and then postsyntactically the terminals (morphemes) are provided with phonological forms by
vocabulary insertion. Implementing late insertion theories requires wide ranging
and complicated additions to that part of syntactic theory dealing with the spell-out
of syntactic objects (the phonology-syntax interface).
As discussed in section 2, Spell-Out incorporates general principles of
morpheme order (e.g., the LCA) and principles dealing with the spell-out of
occurrences (e.g., in remnant movement). These general Spell-Out principles are
also applicable in late insertion theories.
First, there is an independent list of vocabulary items (the Vocabulary) that
are defined by the pairing of formal features and phonological forms, which in
functional vocabulary alone consists of hundreds such vocabulary items for each
I-language. In addition, there is another list of terminals (morphemes) that are
combined by Merge. In many cases, the vocabulary items and the terminals will
be defined by identical sets of formal features. For example, suppose the English
preposition to is defined by the formal feature DAT. Then in DM, there will be a
terminal DAT as well as a vocabulary item: DAT ←→ to. All formal features will
appear twice in the grammar: in the terminals and in the vocabulary items.
Each time vocabulary insertion applies to a particular syntactic terminal,
every one of those vocabulary items will have to be checked to see if it matches
the terminal. That is, the features of the vocabulary items need to be a subset of the
formal features of the terminal (the Subset Principle, see Embick and Noyer 2007:
298). If a syntactic object contains 10 terminals (a small tree) and there are 100
vocabulary items (a small Vocabulary) for a particular grammar, then there will be
1,000 operations of checking. Of course, in practice, there may be ways to optimize
this search, and to cut down on the total number of operations needed.
Second, once a small subset of matching vocabulary items is selected, they
will be ranked according to how well they match the terminal. If there are five
matching vocabulary items, the one with the most features wins (the Subset Principle
again). So, if V1 matches terminal T with three features and V2 matches T with four
features, V2 wins. This means that UG must be able to take two vocabulary items,
compare them in terms of size and choose the biggest one. Also, it is clear from
Chris Collins, Richard S. Kayne 17
the DM literature that the competition does not just refer to the formal features
of the vocabulary item, but also contextual restrictions on vocabulary items (Halle
and Marantz 1993: 123). As far as we know, the algorithm which carries out this
comparison (in DM) has not been formalized. Are the features and contextual
restrictions actually counted or is some other method used? In some cases, where
there is a tie, it might be necessary to refer to feature hierarchies (Embick and Noyer
2007: 298, fn. 14), adding another level of complexity and stipulation to the process.
It is important to note here that the mechanism of comparing the size of the
feature sets of two different vocabulary items (both of which match a terminal)
is completely foreign to syntactic theory. For example, the choice between
Merge(X,Y) and Merge(X,Z) is never dependent on how large the feature sets of
Y and Z are.
Third, once the winning vocabulary item is determined, its phonological form
will be inserted into the terminal, so when the syntactic object is linearized, the
phonological features of the vocabulary item are incorporated into the output. Such
insertion changes the syntactic object formed (by replacing one of its morphemes
with a morpheme specified for phonological form). So, it is important to stipulate
that this process lies outside of syntax, which is constrained by the No Tampering
Condition (no altering syntactic objects already formed).
Similar remarks hold for other post-syntactic operations like fission, fusion,
impoverishment, and post-syntactic insertion of dissociated morphemes. All of these
would violate the No Tampering Condition, if they were syntactic operations. In other
words, these other operations are not syntactic operations, even though they have
syntactic structures as both input and output, which adds a level of complication.
The above remarks demonstrate that late insertion requires wide ranging and
powerful formal mechanisms that do not find any basis in syntactic theory. On
the contrary, the MaS program adopts early insertion, so that Merge itself (not
vocabulary insertion) introduces phonological material into syntactic structures.
The hypothesis is that by adopting early insertion, it will be possible to dispense
with these non-syntactic formal additions needed to implement late insertion.
This section can be summarized as:
(41) There is no late insertion.
6 Syncretism
In both DM and Nanosyntax, the primary motivation for late insertion is
accounting for morphological syncretisms (see Embick and Noyer 2007: 299).
Here, we propose a theory of syncretism that does not rely on late insertion:
(42) Syncretism involves a single morpheme being used in two or more different
syntactic contexts.
In this definition, morpheme means a pair {FF, PHON} (see (3)).
18 Morphology as Syntax
The assumption in (42) generalizes the approach of Kayne 2010b, in which
he analyzes syncretism between dative clitics, locative clitics and others (chapter 6)
and syncretism between 1PL object clitics and locative clitics and others (chapter 7).
In the cases Kayne discusses, the contexts are defined in part by empty categories.
Here is an example to illustrate how (42) works. Greenberg (1966) states the
following two generalizations concerning syncretism:
(43) Universal 37. A language never has more gender categories in nonsingular
numbers than in the singular.
(44) Universal 45. If there are any gender distinctions in the plural of the pronoun,
there are some gender distinctions in the singular also.
In English, these generalizations capture the behavior of the third person plural
pronouns (they, them, their) which do not show the gender distinctions found in the
singular (e.g., he versus she versus it). The question is why there are no masculine
and feminine forms parallel to 3PL they:
(45)
Masc
Fem
SG
he
she
PL
PL
they
*hey
they
*shey
(* means “does not exist”)
In this paper, we outline some preliminary steps towards an analysis of Greenberg’s
generalization for English in the framework of MaS.
First, we assume they is specified for gender (not merely underspecified for
gender). One piece of evidence for this assertion is that English does show gender
contrasts in the singular: he, she, it. This means that gender is a feature of the
English pronominal system, including the plural pronouns (even if it is not realized
overtly). A second piece of evidence is that even fairly closely related IndoEuropean languages have gender contrasts in the plural pronouns (3PL subject
clitics in French: elles versus ils). Assuming a universal functional hierarchy,
English would also have to have these features represented in its plural pronouns.
However, there is also some syntactic evidence internal to English:
(46) Every one of them (said about a group of boys) hates himself.
It is well known that a reflexive pronoun in English must match the gender of its
antecedent (see for example Collins and Postal 2012 for extensive discussion).
Since the reflexive pronoun in (46) is masculine, it implies that its antecedent must
also be masculine, and hence that them must be specified for gender.
In MaS, we can invoke independently needed syntactic hierarchies of
functional projections. In this case, we assume that the syntactic structure of the
pronoun they is as follows (on the internal structure of pronouns, see Koopman
Chris Collins, Richard S. Kayne 19
2000: chapter 3, and also Vanden Wyngaerd 2018, for relevant discussion of gender
see Bernstein 1993, Picallo 2008, Kramer 2016b).
(47) [DP D [#P # [GenP Gen NP]]]
For example, according to Vanden Wyngaerd (2018: 280) (see also Liao 2014: 4),
Mandarin plural pronouns transparently involve a plural morpheme: “Some languages
form the plural of pronouns with the same morpheme that is used with nouns (or
certain noun classes).” The Mandarin forms are given below in (48). Similarly,
Matthews and Yip (2011: 45) note that Cantonese plural -deih can be added only to
pronouns and to yàhn ‘person’. We take these facts as cross-linguistic evidence for
the structure in (47) involving a plural head.
(48)
1
2
3
SG
wo
ni
ta
PL
wo-men
ni-men
ta-men
In (47), since we are dealing with a pronoun, NP is null and D is specified as
definite. # is specified as plural and Gen is specified as masculine (see (46)). Now
clearly, they is bimorphemic: th-ey. For example, it is parallel to th-em and th-eir
which are third person pronouns, and also parallel to th-ese and th-ose, which are
demonstratives. We analyze th- as a definite determiner because of its presence
in the definite determiner the. But if th- is a definite determiner, then what is -ey?
Considering that it cannot be gender since it is overt and does not vary between
masculine and feminine, we suggest that it is an irregular form of the plural
morpheme used in pronouns.
What about gender? We argued in (46) that it is syntactically present. In
this case, it must be syntactically present, but phonologically null. Putting these
assumptions together, we have the following structures (MASC and FEM are not
pronounced, as indicated by the capital letters):
(49) a. [DP th- [#P -ey [GenP MASC NP]]]
b. [DP th- [#P -ey [GenP FEM NP]]]
On this analysis, the 3MPL and 3FPL are syncretic in English, because the two
morphemes th- and -ey are being used in two different contexts defined by the null
MASC and FEM morphemes (see (42)). In MaS, this pattern of explanation will
hold for all cases of syncretism.
Unlike the singular (e.g., he/him/his versus she/her/her versus it), no gender
is realized overtly in 3PL pronouns in English. Why is gender unpronounced
here? There are two possible analyses in MaS: (1) Gen is a lexical zero (specified
in the lexicon as zero in the context of the plural), (2) Gen has a phonological
form (perhaps he) that is unpronounced in this particular context. However, the
assumption that Gen is lexically specified as zero does not seem sufficiently strong.
For example, it would allow an English dialect to show gender distinctions in the
20 Morphology as Syntax
third plural, but as far as we know, there are no such English dialects. And in fact,
the syncretism holds more widely across Germanic languages. Therefore, what
remains to be understood in this analysis is how the empty GEN morphemes in (49)
are licensed syntactically.
We propose that in this case the higher feature (plural number) is licensing
the null lower feature (gender). The prediction is that no specific overt gender (e.g.,
feminine) will license null plural number, which would be the case of pronouns in
some language that show a plural distinction in the masculine, but where singular
and plural are syncretic in the feminine. For example, in such a language, she
would be either feminine singular or plural, but he would only be masculine
singular. As far as we know, this generalization is accurate, but we will leave the
exact mechanisms for future work.
7 Metasyncretism
Harris (1998) and Embick (2015: 154) analyze syncretism between 2PL and 3PL
in the Latin American Spanish verbal paradigm. As can be seen in the paradigm
below, the 2PL and 3PL of the present tense of hablar ‘to speak’ are identical.
(50) hablar ‘to speak’
present
1SG hablo
2SG habla-s
3SG habla-∅
1PL habla-mos
2PL habla-n
3PL habla-n
As Embick (2015: 26) notes: “Latin American Spanish shows only five distinct
phonological exponents, with an -n appearing in both the second person plural and
third person plural contexts.”
Embick (2015: 27) proposes to analyze this syncretism in terms of the
following vocabulary item:
(51) [-1, +PL] ←→ -n
Because of the underspecified nature of the vocabulary item, it can be inserted
post-syntactically in both a second person plural terminal and a third person
plural terminal.
However, the identity of 2PL and 3PL is not limited to this one paradigm.
Rather, as Harris (1998: 31) notes, the syncretism between 2PL and 3PL is pervasive
in Latin American Spanish:
(52) “Unlike standard Iberian, Latin American dialects systematically lack second
person plural morphology: every semantic/syntactic second person plural item
Chris Collins, Richard S. Kayne 21
is realized overtly with third person plural morphology. This generalization
is all-inclusive, covering not only all verb inflection but also nominative and
object-of-proposition pronouns; long and short possessive adjectives and
pronouns; accusative, dative, and reflexive clitic pronouns; etc.”
Since the syncretism extends to various paradigms (verbal, clitic and pronominal),
it will miss a generalization to analyze it in terms of underspecified vocabulary
items. For example, the vocabulary item for -n in (51), will not apply to the
syncretism between 2PL and 3PL object clitics (e.g., los and las, for masculine and
feminine respectively).
To account for this pervasive syncretism, Harris postulates an impoverishment
rule. We present Embick’s (2015: 154) version below:
(53) [+/- 2] → ∅ [__,+pl]
As Harris (1998) comments, this rule “…removes the feature [2pers] in the context
of [+plural], leaving no person feature at all. This is the formal counterpart of
‘third person’, the default person in Spanish (and perhaps universally). All other
features are unaffected; in particular, the features of case, gender and number
necessary for realization of overt phonological distinctions in these properties
remain intact. It is important to bear in mind that impoverishment is a purely
morphological operation; syntactic and semantic representations are not affected
by it at all.”
From the point of view of MaS, the use of impoverishment rules to capture
pervasive syncretisms, like that illustrated above for Latin American Spanish, is
problematic in a number of ways.
First, as Harris (1998: 40) points out: “Spell-out transfers syntactic arboreal
structures into the Morphology model, where they continue to be subject to
syntactic-type operations, among others. Thus, morphological representations and
their operations do not necessarily differ radically from their syntactic counterparts.”
This quote recognizes the syntactic nature of impoverishment rules. They
take syntactic structures (those specified for +2 or -2) and produce new syntactic
structures, by removing features. However, such impoverishment rules are not
true syntactic rules. Rather they apply post-syntactically. They are not syntactic
rules because they do not obey constraints on syntactic computation (e.g., the No
Tampering Condition). In effect, adding impoverishment rules introduces a new
kind of unconstrained second syntactic component, operating outside of the core
syntactic system. This is the general problem of second syntax.
Second, there is an issue of restrictiveness (on which, see Chomsky 1981: 5). What
are the constraints on impoverishment rules? Can any combination of formal features
be deleted? If not, why not? Are there languages with the following impoverishment
rules for verbal paradigms?
22 Morphology as Syntax
(54) a.
b.
c.
d.
e.
f.
g.
h.
i.
j.
k.
l.
m.
n.
Delete +1
Delete -1
Delete +2
Delete -2.
Delete +PL
Delete -PL
Delete +1, -PL
Delete +1, +PL
Delete -1, -PL
Delete -1, +PL
Delete +2, -PL
Delete +2, +PL
Delete -2, -PL
Delete -2, +PL
The above rules do not even take into account gender or other proposed features
which would add another layer of possible deletions.
In general, if there are n features, there will be 2n -1 impoverishment rules.
For example, if there are five features, there will be 31 impoverishment rules. It thus
appears that the operation of impoverishment is unrestrictive unless accompanied
by a strong theory of constraints, which was not given in the above sources.
Third, since impoverishment operations have contextual restrictions, there
is an issue with ordering. Harris (1998: 42) makes use of ordering between two
impoverishment rules to obtain the correct subjunctive verb forms in Spanish. The
possibility of ordering impoverishment rules adds further to the unrestrictiveness
of the system, since each choice of ordering leads to a new I-language.
Fourth, when there is a pervasive syncretism involved, as in the case of Latin
American Spanish, it might reflect a deeper syntactic property of the language.
All dialects of Spanish use the third person forms usted and ustedes to refer to the
addressee. In Iberian Spanish, the use of usted and ustedes correlates with formality.
In Latin American Spanish dialects, the use of ustedes to express the second person
plural is the only option. In those dialects, there are no morphologically second
person plural forms (such as the subject pronoun vosotros).
Collins and Ordóñez (2021) analyze usted and ustedes as imposters, in the
sense of Collins and Postal 2012. In particular, they analyze usted as follows:
(55) Structure of usted
[DP D [TU usted]]
In this structure, there is a null 2SG pronoun TU that is embedded in a 3SG DP
whose head noun is usted (see Collins and Postal 2012 for a precise structural
proposal). Crucially, an overt 2SG pronoun tu (or sometimes vos) is found in all
dialects. As often with proper names, the determiner is null. On this analysis, the
Chris Collins, Richard S. Kayne 23
reason why usted refers to the addressee is because of the presence of the null 2SG
pronoun TU. The reason why usted shows 3SG verb agreement is because of the
3SG head noun.
The structure of ustedes is parallel, except there is an additional plural morpheme:
(56) Structure of ustedes
[DP D [TU usted]-s]
The plural morpheme forms the semantic plural of TU, but syntactically it merges
with the 3SG [TU usted]. In this case, ustedes refers to a plural addressee, but
agrees in 3PL with the verb.
On the above account, the pervasive syncretism between 2PL and 3PL in
Latin American Spanish is given a purely syntactic analysis. Here are the steps in
the argument:
(57) Syntactic account of meta-syncretism between 2PL and 3PL in Latin
American Spanish:
a. Latin American Spanish dialects lack the 2PL pronoun vosotros, the 2PL
clitic os, and 2PL possessive forms vuestro/a/os/as.
b. It is not necessary to assume that there is a constraint of the form *2PL
ruling out these forms, rather the relevant forms simply do not exist.
c. 2PL and 3PL are syncretic in those dialects because reference to a plural
addressee is only expressed with the imposter ustedes.
d. There is no need for an impoverishment operation.
See Collins and Ordóñez 2021 for further detail.
Harley (2008: 255) defines meta-syncretism as “…a syncretism that holds for
a particular set of features in a language, regardless of the particular affixes used
in any particular instance of the syncretism”. The 2PL and 3PL syncretism in Latin
American Spanish is a meta-syncretism because it runs across pronouns, clitics,
and subject-verb agreement. If the above account of Latin American Spanish can
be generalized, we have the following:
(58) Meta-syncretism is the result of syntactic properties of a language
(not morphological impoverishment).
8 Comparison of Theories
In this section, we will give a brief comparison of MaS (Morphology as Syntax)
with DM (Distributed Morphology) and NS (Nanosyntax) along several
dimensions (for an introduction to Nanosyntax, see Baunaz and Lander 2018).
Before examining the differences, it is important to note that all three theories
adopt minimalist syntactic assumptions. For example, they all recognize Merge as
the syntactic structure building operation. See Stump 2001: chapter 1 for a broader
overview of morphological theories.
24 Morphology as Syntax
First, MaS does not have late insertion, contrary to both DM and NS. For
example, in DM a terminal (morpheme) will get a phonological form by vocabulary
insertion post-syntactically. In MaS, lexical items are defined as a pair of a set of
formal features and a phonological form, so the phonological features piggy-back
into the syntactic structure when the lexical items are merged. Another related
difference is that MaS has no notion of competition, unlike DM (which has the
subset principle) or NS (which has the superset principle). As a reviewer points
out, it is possible to define a theory with early insertion and competition between
morphemes. However, it is also clear that for DM and NS, competition is crucially
defined in terms of late insertion.
It is possible to make points about late insertion at a more abstract level.
Consider the relation of subset (not proper subset). There are three relations that
can hold between two sets based on that relation. A is a subset of B, B is a subset of
A, and A and B are equal. These three configurations define three morphological
theories: DM is based on the subset relation (for late insertion); NS is based on
the superset relation (for late insertion); finally, MaS is based on the equality
relation: the formal features of a lexical item are equal to the formal features of a
syntactic terminal, because syntactic trees are built by merging lexical items. So,
in this sense, DM, NS and MaS constitute a partition of the logical possibilities
of Merge based syntax. The different possibilities are illustrated below (A and B
are sets of features):
(59) a. A
b. A
c. A
⊆
⊇
=
B
B
B
(subset, Distributed Morphology)
(superset, Nanosyntax)
(equality, Morphology as Syntax)
Second, MaS has no post-syntactic operations. For example, there is no postsyntactic movement of morphemes that one finds in the DM literature. There is
also no post-syntactic insertion of dissociated or ornamental morphemes (e.g.,
theme vowels). Nor are there operations such as impoverishment (or fusion and
fission) that operate on syntactic structures and produce syntactic structures,
but are not syntactic operations. In this regard, MaS is similar to NS which also
eschews such post-syntactic operations (with the exception of late insertion).
Another way of putting the same point is in terms of the expression “syntactic
hierarchical structure all the way down” (see Halle and Marantz 1994: 276).
Syntactic hierarchical structures are those that are created by syntactic operations
(such as Merge) and that obey syntactic principles (such as binary branching).
Because of the widespread use of post-syntactic operations (such as fusion,
fission, impoverishment and post-syntactic insertion of dissociated morphemes),
DM definitely does not fall under this description. For example, suppose we have
the structure [X Y], and that an ornamental morpheme Z (e.g., a theme vowel)
is inserted post-syntactically to yield [X [Y Z]]. Since [Y Z] is not formed by a
syntactic operation, it is not syntax (on the reasonable assumption that only those
structures formed by syntactic operations can be characterized as syntactic).
Chris Collins, Richard S. Kayne 25
Similarly, only MaS and NS fall under the rubric of “single generative
engine”. While structure is only formed through Merge in MaS and NS, in DM
there is the possibility of forming structure through the post-syntactic insertion
of dissociated or ornamental morphemes (sprouting). Therefore, DM cannot be
characterized by the expression “single generative engine”.
Third, both MaS and DM claim that morphemes bear phonological features.
For MaS, the phonological features are specified in the lexicon. For DM, the
phonological features are inserted post-syntactically by vocabulary insertion. For
NS, late insertion is defined in terms of syntactic phrases. A phrase gets paired up
with a phonological form post-syntactically.
The differences are summarized in the following chart:
(60)
late insertion
competition
post-syntactic operations
morpheme based
MaS
no
no
no
yes
DM
yes
yes
yes
yes
NS
yes
yes
no (only late insertion)
no (phrase based)
Actually, from the summary, it can be seen that these four properties define a
space of sixteen different Merge-based theories. However, we will leave it to
future work to explore the other possible theories, and the relationships between
the four properties that define them.
9 Conclusion
In this paper, we have achieved the following. First, we sketched the basic assumptions
of MaS, a syntactic approach to analyzing putative morphological generalizations.
Next, we illustrated the approach with some examples involving ellipsis, contextual
restrictions, allomorphy and post-syntactic insertion of dissociated (or ornamental)
morphemes. Then, we argued that MaS does not allow late insertion of phonological
material and discussed the implications for theories of syncretism. Afterwards, we
argued that the operation of impoverishment is unrestrictive, and at least in one
case, not necessary to capture the empirical facts. Lastly, we showed that MaS is
one of three logically possible kinds of Merge-based theories of morphology, with
the other two logical possibilities being represented by DM and NS.
To help the reader keep all the various assumptions of MaS in mind, we have
written up the following cheat sheet:
(61) a.
b.
c.
d.
There is no morphological component in UG.
Definition of a morpheme: {FF, PHON}
For all syntactic objects X and Y, Merge(X,Y) = {X,Y}
Contextual restrictions on morphemes are to be understood in terms of
relations familiar from syntax (e.g., c-selection, s-selection, l-selection,
idioms etc.).
e. There is no late insertion.
26 Morphology as Syntax
f.
g.
h.
i.
j.
k.
l.
m.
There is no competition (or blocking) between morphemes.
There is no Elsewhere Principle.
Null morphemes are licensed by syntactic principles.
Contextual allomorphy involves two or more different morphemes.
(e.g., -en is an inner plural morpheme, -s is an outer plural morpheme).
Syncretism involves a single morpheme being used in two or more
different syntactic contexts.
Meta-syncretism is the result of syntactic properties of a language.
There is no post-syntactic insertion of morphemes.
There are no other post-syntactic morphological operations.
Acknowledgments
For useful discussion, we thank Peter Svenonius, Selikem Gotah and Ruth Kramer.
For translating the title, abstract and keywords, we thank Zhuoye Zhao. We also thank
two anonymous reviewers and the editorial board for their feedback.
References
Baker, Mark. 1985. The mirror principle and morphosyntactic explanation. Linguistic
Inquiry 16. 373–415.
Baker, Mark. 1988. Incorporation: A theory of grammatical function changing.
Chicago: University of Chicago Press.
Baker, Mark & Ruth Kramer. 2014. Rethinking Amharic prepositions as case markers
inserted at PF. Lingua 145. 141–172.
Baunaz, Lena & Eric Lander. 2018. Nanosyntax: The basics. In Lena Baunaz, Karen
De Clerq, Liliane Haegeman & Eric Lander (eds.), Exploring Nanosyntax 4–56.
Oxford: Oxford University Press.
Bernstein, Judy. 1993. The syntactic role of word markers in null nominal constructions.
Probus 5. 5–38.
Bhatt, Rajesh & Rouyana Pancheva. 2017. Implicit arguments. In Martin Everaert &
Henk van Riemsdijk (eds.), The Blackwell companion to syntax, 1–35. Malden,
MA: Blackwell.
Bobaljik, Jonathan. 2012. Universals in comparative morphology: Suppletion,
superlatives and the structure of words. Cambridge: MIT Press.
Bonet, Eulàlia & Daniel Harbour. 2012. Contextual allomorphy. In Jochen Trommer
(ed.), The morphology and phonology of exponence, 195–235. Oxford: Oxford
University Press.
Buell, Leston, Mariame Sy & Harold Torrence. 2014. The syntax of mirror
principle violations in Wolof. In Carson T. Schütze & Linnaea Stockall (eds.),
Connectedness papers by and for Sarah VanWagenen, UCLA working papers
in linguistics 18, 71–101. http://phonetics.linguistics.ucla.edu/wpl/issues/wpl18/
papers/buellsytorr.pdf.
Caha, Pavel. 2013. Explaining the structure of case paradigms by the mechanisms of
Nanosyntax: The classical Armenian nominal declension. Natural Language and
Linguistic Theory 31. 1015–1066.
Chris Collins, Richard S. Kayne 27
Caha, Pavel. 2020. Nanosyntax: Some key features. Ms., Masaryk University, Brno.
Calabrese, Andrea. 2015. Irregular morphology and athematic verbs in Italo-Romance.
Isogloss [Special issue]. 69–102.
Cardinaletti, Anna & Ur Shlonsky. 2004. Clitic positions and restructuring in Italian.
Linguistic Inquiry 35. 519–557.
Carstens, Vicki. 1991. The morphology and syntax of determiner phrases in Kiswahili.
Los Angeles: University of California, Los Angeles dissertation.
Carstens, Vicki. 1993. On nominal morphology and DP structure. In Sam A.
Mchombo (ed.), Theoretical aspects of Bantu grammar. Stanford, CA: CSLI,
Stanford University.
Choi, Jaehoon & Heidi Harley. 2019. Locality domains and morphological rules.
Language and Linguistic Theory 37. 1319–1365.
Chomsky, Noam. 1981. Lectures on government and binding. Dordrecht: Foris.
Chomsky, Noam. 1986. Knowledge of language. New York: Prager.
Chomsky, Noam. 2000. Minimalist inquiries. In Roger Martin, David Michaels &
Juan Uriagereka (eds), Step by step, 89–155. Cambridge, MA: MIT Press.
Cinque, Guglielmo. 2005. Deriving Greenberg’s Universal 20 and its exceptions.
Linguistic Inquiry 36(3). 315–332.
Cinque, Guglielmo. 2015. Again on tense, aspect, mood morpheme order and the
“mirror principle”. In Peter Svenonius (ed.), Functional structure from top to toe,
232–265. Oxford: Oxford University Press.
Collins, Chris. 2006. A Note on derivational morphology. Ms., New York University.
Collins, Chris. 2007. Home sweet home. In Lisa Levinson & Oana Savescu-Ciucivara
(eds.), NYU working papers in linguistics 1. http://ling.auf.net/lingbuzz/001967.
Collins, Chris. 2017. Merge(X,Y) = {X,Y}. In Leah Bauke, Andreas Blümel & Erich
Groat (eds.), Labels and roots, 47–68. Boston: De Gruyter Mouton.
Collins, Chris. 2018. The logic of contextual allomorphy. Ms., New York University.
Collins, Chris. 2022. A merge-based approach to argument structure. Ms., New
York University.
Collins, Chris & Francisco Ordóñez. 2021. Spanish usted as an imposter. Probus 33.
249–269.
Collins, Chris & Paul Postal. 2012. Imposters. Cambridge, MA: MIT Press.
Collins, Chris & Edward Stabler. 2016. A formalization of minimalist syntax. Syntax
19. 43–78.
Craenenbroeck, Jereon van & Jason Merchant. 2013. Ellipsis phenomena. In Marcel
den Dikken (ed.), The Cambridge handbook of generative syntax, 701–745.
Cambridge: Cambridge University Press.
Dobashi, Yoshihito. 2020. Externalization: Phonological interpretations of syntactic
objects. New York: Routledge.
Embick, David. 2015. The morpheme: A theoretical introduction. Berlin: De Gruyter
Mouton.
Embick, David & Rolf Noyer. 2007. Distributed morphology and the syntaxmorphology interface. In Gillian Ramchand & Charles Reiss (eds.), The Oxford
handbook of linguistic interfaces, 289–324. Oxford: Oxford University Press.
28 Morphology as Syntax
Emonds, Joseph E. 1987. The invisible category principle. Linguistic Inquiry 18.
613–632.
Erschler, David. 2018. Suspended affixation as morpheme ellipsis: Evidence for
Ossetic alternative questions. Glossa 3(1). 12, 1–41.
Fabb, Nigel. 1988. English suffixation is constrained only by selectional restrictions.
Natural Language and Linguistic Theory 6. 527–539.
Fábregas, Antonio. 2012. Evidence for multidominance in Spanish agentive
nominalizations. In Myriam Uribe-Etxebarria & Vidal Valmala (eds.), Ways of
structure building, 66–92. Oxford: Oxford University Press.
Gotah, Selikem & Soo-Hwan Lee. 2022. Negation in Ewe (Tongugbe) agent
nominalization. Ms., New York University.
Greenberg, Joseph. 1966. Some universals of grammar with particular reference to the
order of meaningful elements. In Joseph Greenberg (ed.), Universals of language,
73–113. Cambridge, MA: MIT Press.
Halle, Morris & Alec Marantz. 1993. Distributed morphology and the pieces of
inflection. In Kenneth Hale & Samuel J. Keyser (eds.), The view from Building
20, 111–176. Cambridge, MA: MIT Press.
Halle, Morris & Alec Marantz. 1994. Some key features of distributed morphology.
In Andrew Carnie, Heidi Harley & Tony Bures (eds.), Papers on phonology and
morphology, MIT working papers in linguistics 21, 275–288. Cambridge, MA:
MIT Working Papers in Linguistics.
Harley, Heidi. 2008. When is syncretism more than a syncretism? In Daniel Harbour,
David Adger & Susan Béjar (eds.), Phi-theory: Phi-features across modules and
interfaces, 251–294. Oxford: Oxford University Press.
Harris, James. 1998. Spanish imperatives: Syntax meets morphology. Journal of
Linguistics 34. 27–52.
Huang, C.-T. James. 2015. On syntactic analyticity and parametric theory. In Audrey
Li, Andrew Simpson & Wei-Tien Dylan Tsai (eds.), Chinese syntax in a crosslinguistic perspective. New York: Oxford University Press.
Holmberg, Anders. 2005. Is there a little pro? Evidence from Finnish. Linguistic
Inquiry 36. 533–564.
Holmberg, Anders. 2010. Null subject parameters. In Theresa Biberauer, Anders
Holmberg, Ian Roberts & Michelle Sheehan (eds.), Parametric variation: Null
subjects in minimalist theory, 88–124. Cambridge: Cambridge University Press.
Julien, Marit. 2007. On the relation between morphology and syntax. In Gillian
Ramchand & Charles Reiss (eds.), The Oxford handbook of linguistic interfaces,
209–238. Oxford: Oxford University Press.
Kayne, Richard. 1994. The antisymmetry of syntax. Cambridge, MA: MIT Press.
Kayne, Richard. 2005. Movement and silence. Oxford: Oxford University Press.
Kayne, Richard. 2006. On parameters and on principles of pronunciation. In Hans
Broekhuis, Norbert Corver, Riny Huybregts, Ursula Kleinhenz & Jan Koster
(eds.), Organizing grammar: Linguistic studies in honor of Henk van Riemsdijk,
289–299. Berlin: Mouton de Gruyter (reprinted in Kayne 2010b).
Kayne, Richard. 2008. Antisymmetry and the lexicon. In Jeroen van Craenenbroeck
& Johan Rooryck (eds.), Linguistic Variation Yearbook 8, 1–31. Amsterdam:
Chris Collins, Richard S. Kayne 29
John Benjamins (also in Anna Maria di Sciullo & Cedric Boeckx (eds.). 2011.
The biolinguistic enterprise: New perspectives on the evolution and nature of the
human language faculty, 329–353. Oxford: Oxford University Press.) (reprinted
in Kayne 2010b).
Kayne, Richard. 2010a. Toward a syntactic reinterpretation of Harris and Halle (2005).
In Reineke Bok-Bennema, Brigitte Kampers-Manhe & Bart Hollebrandse (eds.),
Romance languages and linguistic theory 2008: Selected papers from ‘Going
Romance’ Groningen 2008, 145–170. Amsterdam: John Benjamins (reprinted in
Kayne 2010b).
Kayne, Richard. 2010b. Comparisons and contrasts. Oxford: Oxford University Press.
Kayne, Richard. 2016. The silence of heads. Studies in Chinese Linguistics 37. 1–37.
Kayne, Richard. 2017. Antisymmetry and morphology: Prefixes vs. suffixes. In
Clemens Mayr & Edwin Williams (eds.), Festschrift für Martin Prinzhorn,
Wiener Linguistische Gazette 82, 145–161. https://wlg.univie.ac.at/fileadmin/
user_upload/p_wlg/822017/Kayne_wlg.pdf.
Kayne, Richard. 2018. Toward an understanding of apparent suppletion in Romance
pronominal possessives. Probus 30. 251–276.
Kayne, Richard. 2019a. What is suppletion? On *goed and on went in Modern English.
Transactions of the Philological Society 117. 434–454.
Kayne, Richard. 2019b. The place of linear order in the language faculty. Linguistic
Analysis 42. 441–472.
Kayne, Richard. 2020. A note on the silent GO that underlies an instance of apparent
suppletion in Spanish. Isogloss 6/4. 1–12.
Kayne, Richard. 2022. Antisymmetry and externalization. Studies in Chinese
Linguistics 43. 1–20.
Kayne, Richard. To appear. On a more demanding approach to suppletion. In Giuliano
Bocci, Daniele Botteri, Claudia Manetti & Vincenzo Moscati (eds.), Issues
in comparative morpho-syntax and language acquisition. Oxford: Oxford
University Press.
Kayne, Richard & Andrea Moro. To appear. A note on zero and silent negation.
Koopman, Hilda. 2000. The syntax of specifiers and heads. London: Routledge.
Koopman, Hilda. 2005. Korean (and Japanese) morphology from a syntactic
perspective. Linguistic Inquiry 36. 601–633.
Koopman, Hilda. 2017. A note on Huave morpheme ordering: Local dislocation or
generalized U20? In Gautam Sengupta, Shruti Sircar, Madhavi Gayathri Raman
& Rahul Balusu(eds.), Perspectives on the architecture and the acquisition of
syntax, 23–47. Singapore: Springer Nature.
Koopman, Hilda & Anna Szabolcsi. 2000. Verbal complexes. Cambridge, MA: MIT Press.
Kramer, Ruth. 2016a. A split analysis of plurality: Number in Amharic. Linguistic
Inquiry 47. 527–559.
Kramer, Ruth. 2016b. The location of gender features in syntax. Language and
Linguistics Compass 10. 661–677.
Lees, Robert B. 1963. The grammar of English nominalizations. Bloomington:
Indiana University.
30 Morphology as Syntax
Levi, Judith N. 1978. The syntax and semantics of complex nominals. New York:
Academic Press.
Li, Charles N. & Sandra A. Thompson. 1981. Mandarin Chinese: A functional
reference grammar. Berkeley: University of California Press.
Liao, Wei-Wen Roger. 2014. Morphology. In C.-T. James Huang, Y.-H. Audrey Li &
Andrew Simpson (eds.), The handbook of Chinese linguistics, 3–25. Oxford:
Wiley Blackwell.
Matthews, Stephen & Virginia Yip. 2011. Cantonese: A comprehensive grammar, 2nd
edn. New York: Routledge.
McGinnis, Martha. 2017. Distributed morphology. In Andrew Hippisley & Gregory
Stump (eds.), The Cambridge handbook of morphology. Cambridge: Cambridge
University Press.
Muriungi, Peter Kinyua. 2014. Phrasal movement in the Bantu verbal complex:
Deriving affix scope and order in Kı ı̂ ̂tharaka. Syntax 17. 21–39.
Nevins, Andrew. 2010. Locality in vowel harmony. Cambridge, MA: MIT Press.
Ntelitheos, Dimitrios. 2012. Deriving nominals: A syntactic account of Malagasy
nominalizations, vol. 3. Leiden: Brill.
Nunes, Jairo. 2004. Linearization of chains and sideward movement. Cambridge, MA:
MIT Press.
Ordóñez, Francisco. 2002. Some clitic combinations in the syntax of Romance.
Catalan Journal of Linguistics 1. 201–224.
Picallo, M. Carme. 2008. Gender and number in Romance. Lingue e Linguaggio 1. 47–66.
Pinker, Steven. 1999. Words and rules: The ingredients of language. New York:
Basic Books.
Punske, Jeffrey & Scott R. Jackson. 2017. The bifurcated nature of plural: Reconsidering
evidence from English compounds. In Julia Nee, Margaret Cychosz, Dmetri
Hayes, Tyler Lau & Emily Remirez (eds.), Proceedings of the 43rd Meeting of
the BLS, 261–283. Berkeley, CA: Berkeley Linguistics Society
Rizzi, Luigi. 1986. Null objects in Italian and the theory of pro. Linguistic Inquiry 17.
501–557.
Săvescu-Ciucivara, Oana. 2009. A syntactic analysis of pronominal clitic clusters in
Romance: The view from Romanian. New York: New York University dissertation.
Shlonsky, Ur. To appear. Rescaffolding the bundle in Afroasiatic inflection: Tamazight
and Hebrew. Brill’s Journal of Afroasiatic Languages and Linguistics.
Stump, Gregory T. 2001. Inflectional morphology: A theory of paradigm structure.
Cambridge: Cambridge University Press.
Terzi, Arhonto. 1999. Clitic combinations, their hosts, and their ordering. Natural
Language and Linguistic Theory 17. 85–121.
Vanden Wyngaerd, Guido. 2018. The feature structure of pronouns: A probe into
multidimensional paradigms. In Lena Baunaz, Karen De Clerq, Liliane
Haegeman & Eric Lander (eds.), Exploring Nanosyntax, 277–304. Oxford:
Oxford University Press.
Yip, Po-Ching & Don Rimmington. 2004. Chinese: A comprehensive grammar.
London: Routledge.
Chris Collins, Richard S. Kayne 31
Zyman, Erik & Nick Kalivoda. 2020. XP- and X0-movement in the Latin verb: Evidence
from mirroring and anti-mirroring. Glossa 5(1): 20. https://doi.org/10.5334/
gjgl.1049.
Mailing address:
Email:
Mailing address:
Email:
Received:
Accepted:
Chris Collins
Department of Linguistics, New York University, 10 Washington
Place, New York, NY 10003
cc116@nyu.edu
Richard S. Kayne
Department of Linguistics, New York University, 10 Washington
Place, New York, NY 10003
richard.kayne@nyu.edu
March 15, 2022
January 27, 2023
32 Morphology as Syntax
形態即句法理論初探
Chris Collins, Richard S. Kayne
紐約大學
提要
傳統上被認爲是形態學的現象可以用句法操作和原則來解釋,由此也可以提出傳統
形態學所無法提出的問題(例如有關空語素的允准)。語言官能不包含形態學特有
的成分,也不包含任何後句法形態學操作。
關鍵詞
形態學,語素變體,融合,異幹互補,遲後填音