SYNCRETISM WITHOUT PARADIGMS : REMARKS ON WILLIAMS 1981, 1994

running head: Syncretism without paradigms
article type:
theoretical article (morphology-syntax)
full title:
Syncretism without Paradigms: Remarks on Williams 1981, 1994
author:
Jonathan David Bobaljik
affiliation:
McGill University
address:
McGill University
Department of Linguistics
1085 Dr. Penfield
Montréal, QC H3A 1A7
CANADA
tel: (514) 398-4224, fax: (514) 398-7088
email: jonathan.bobaljik@mcgill.ca
version info:
written:
November 2001
revisions:
February 2002
1

---

---

SYNCRETISM WITHOUT PARADIGMS:
REMARKS ON WILLIAMS 1981, 1994
Abstract:
Arguments from patterns of syncretism taken to show that the paradigm is “a real object,
and not the epiphenomenal product of various rules” (Williams 1994:22) are re-examined
from the perspective of vocabulary-based theories of morphology. Recurring patterns of
syncretism within a language may be captured equally well in theories without
paradigms, as in those with them; in neither type of theory do the patterns come for free,
and the devices employed in both types of theory are of equivalent complexity. While the
patterns themselves thus do not constitute an argument either way, a particular
requirement on relations among paradigms within a single language, proposed as a
universal in Williams 1994, is statable only in theories encompassing paradigms. If
upheld, this Instantiated Basic Paradigm requirement would constitute a powerful
argument in favour of paradigms. Data from Russian declension shows that the proposal
is untenable. Grammar may be paradigmatic in various ways, but grammars do not
require explicit reference to the structure of paradigms, and are limited to making
reference to the pieces that make up a paradigm (features, vocabulary items) and rules for
combining these.
Keywords:
Paradigms, Lexicalism, Distributed Morphology, Impoverishment (versus) Rules of
Referral
1

---

SYNCRETISM WITHOUT PARADIGMS
0. INTRODUCTION
Plank (1991) begins with the observation that “[t]he earliest extant grammatical texts are
paradigms” (p.161). The long linguistic and philological traditions have established a wealth of
knowledge about the properties of paradigms, notably regarding the issue of syncretism, but one
fundamental question has not been definitively answered, namely (1):
(1)
Does knowledge of language (grammar) include knowledge (memorization) of paradigms
themselves or just of the pieces that constitute paradigms and rules for generating them?
Consider, by way of a simple, illustrative example, the (partial) paradigm of a regular English
verb given in (2):
(2)
Present
Past
Singular
Plural
Singular
Plural
1 psn
play-ø
play-ø
play-[d]
play-[d]
2 psn
play-ø
play-ø
play-[d]
play-[d]
3 psn
play-[z]
play-ø
play-[d]
play-[d]
As is well known, the information contained in this paradigmatic representation can be
generated from the set of morpho-syntactic features indicated (in this case, two tenses, three
persons, two numbers), along with a disjunctively ordered list of morpheme realization rules, or
equivalently, competing VOCABULARY ITEMS in the terminology of Distributed Morphology
(Halle & Marantz 1993). That is, given the same feature set and in abstracting away from the
usual phonological considerations—necessary on either approach—(3) will derive (2).
(3)
Vocabulary Items
-d
?
past
-z
?
3 sg
-ø
=
default / elsewhere
2

---

SYNCRETISM WITHOUT PARADIGMS
The question in (1) thus asks whether an English speaker’s knowledge of their
language—their grammar—is more accurately represented by (2) or (3). This question
constitutes a major divide between classes of theories of morphology. On the one hand are
paradigm-based theories, which assume that paradigmatic structures such as (2), in addition to
their contents, are part of the grammar; Williams (1994) argues explicitly for this position. On
the other hand are vocabulary-(item)-based theories, which maintain that (3) is the best
representation of the grammar, and that paradigms are epiphenomenal, derived constructs.1
In this article, I seek to counter one set of arguments in favour of paradigm-based theories
over vocabulary-item-based alternatives. Specifically, I will examine the arguments given in
Williams (1994) regarding patterns of syncretism and his conclusion that the paradigm is “a real
object, and not the epiphenomenal product of various rules” (Williams 1994:22). I will argue
first that Williams provides a valid and important critique of naïve vocabulary-item-based
theories, but that this critique does not lead inescapably to the conclusion he draws. In particular,
the argument from syncretism ultimately does not distinguish between the two classes of
theories. The patterns of syncretism that Williams identifies may motivate an enrichment to
naïve vocabulary-item-based theories (specifically, the kind of feature-manipulation device
instantiated by Impoverishment rules in Distributed Morphology, cf. Bonet 1991, 1995), but the
patterns require an equivalent enrichment to naïve paradigm-based theories. The patterns of
syncretism are not a priori predicted by either class of theory, and can be accommodated with

1 Many theories lie between these extremes. Some theories for example treat paradigms as emergent constructs (thus
not memorized properties of a language, in contrast to Williams) but nevertheless admit of grammatical principles
which refer directly to these structures; Wunderlich (1995) outlines one such theory, Stump (2001) another. While
there are perhaps few morphological theories that espouse the strong position taken by Williams, such a position
does seem to be implicit in work on morphosyntax (for example Rohrbacher 1999) and morphophonology (for
example Kenstowicz 2000, McCarthy 2001) which refers directly to paradigm structure. For reviews and critiques of
3

---

SYNCRETISM WITHOUT PARADIGMS
directly comparable formal devices in either class of theory. In passing, I will briefly compare
Impoverishment to a similar device, namely Rules of Referral as proposed by Zwicky (1985) and
developed by Stump (1993, 2001), noting that of the two, only impoverishment allows for a
restrictive theory of syncretism, one which in fact captures nicely much of the data used by
Stump to argue for rules of referral, and goes beyond this to provide a reason for the general
directional nature of these rules.
After having shown where the two classes of theories do not differ, I will argue that one part
of Williams’s theory of syncretism does constitute a real point of difference between the two
theories. Specifically, Williams proposes a universal requirement of an Instantiated Basic
Paradigm (explained below). I will show that such a requirement crucially refers to implicational
relations among paradigms, and thus must be stated over paradigms and can not be stated in a
theory such as DM which treats paradigms as epiphenomenal constructs, arising from the
combination of vocabulary items and impoverishment rules in a given language. If the
Instantiated Basic Paradigm requirement is a part of Universal Grammar, then paradigms are a
part of grammars, and the more restrictive vocabulary-item-based theories are inadequate. As it
turns out, the Instantiated Basic Paradigm requirement is empirically untenable, a fact noted by
Baerman (2000). The conclusion we must draw is that on this one point where the two classes of
theories are in principle distinguishable, and on which Williams’s theory includes a superset of
the apparatus in the vocabulary based theory, exactly the additional expressive power which
allows a paradigm-based theory to state the Instantiated Basic Paradigm requirement is in fact
not made use of by UG. Thus, considerations of restrictiveness point towards the vocabulary-
item-based theories, enriched with impoverishment, such as Distributed Morphology.

Rohrbacher’s proposals, see Lardiere 2001 and Bobaljik to appear.
4

---

SYNCRETISM WITHOUT PARADIGMS
It should of course be noted from the outset that I neither claim (nor aim) in this paper to
address all arguments that have been put forward in favour of the extra expressive power of
paradigms within UG. I am simply targeting one set of what appear to be particularly good
arguments for paradigms, and showing that they do not go through. I leave for future work the
extension of this investigation to other arguments for paradigms, such as those put forward in
work by A. Carstairs-McCarthy (for recent criticisms of which, see, e.g., Noyer 1997 and Halle
& Marantz 2001), and Stump (2001).
1. UNDERSPECIFICATION AS A THEORY OF SYNCRETISM
A review of the role of underspecification in explaining certain kinds of syncretism will serve as
a useful point of departure for this article. A paradigm, as presented in (2), is nothing more than a
structured list of forms, a convenient descriptive device. Much of the interest in going beyond
lists of forms and developing theories of paradigm structure comes from the cross-linguistic
prevalence of syncretism, that is, recurrence of a single form in multiple cells of the paradigm. In
English (2), the -d form is syncretic, occurring throughout the past tense, and the -Ø forms occur
everywhere in the present tense except the 3sg. The presentation in (3) constitutes a theory of this
syncretism. The set of morphosyntactic features (person, with three values, and number and tense
with two values each), effectively defines the range of possible exponents (the paradigm space),
and then the list of vocabulary items, consulted disjunctively from top to bottom, yields the form
for any given combination of features. In this presentation, there are not five homophonous zero
affixes, specified for the different contexts of insertion, rather the zero affix is treated as
unspecified, having the distribution it does by virtue of the fact that there is a more highly
specified vocabulary item, namely -z, specified to occur only in the context of third person
singular. In turn, the 3sg -z need not be positively specified to occur only in the present tense, it
5

---

SYNCRETISM WITHOUT PARADIGMS
does not occur in the past tense because the past tense -d occurs more highly in the list and will
therefore take precedence in realizing the inflectional affix in the context of the features [3 sg
past]. Leaving aside familiar questions of order in the list (see section 2.1, below), the structure
of the theory in (3) derives the information in (2), but the theory does not contain a paradigm per
se. In such a theory, no grammatical principle or rule may appeal to properties of paradigm
structure (as opposed to properties of features or of vocabulary items), since the paradigm
structure is epiphenomenal.
Note that one property of a theory of this sort is that underspecification entails competition
among vocabulary items. Given the context [3 singular past], all three vocabulary items are in
principle compatible with this context, but it is the most highly ranked item in the list that is
obligatorily inserted (thus: She played, *She plays, *She play). Any theory that invokes
underspecification (and thus competition of this sort) is necessarily REALIZATIONAL.
Underspecification or competition for vocabulary insertion (or rule application) is always
determined relative to some context, and the context must therefore be determined first. In the
case of inflectional morphology, this means that the morpho-syntactic representation (called the
morphemic representation in Matthews 1972) which the vocabulary items are competing to
express must be determined prior to the actual choice among exponents. Realizational theories
contrast with strongly lexicalist theories such as Lieber (1980, 1982) and DiSciullo & Williams
(1987) in which the (syntactic) properties of a word are uniquely determined by the properties
(i.e., features) of that word’s constituent morphemes, where ‘morphemes’ are identified by their
phonological instantiations. On strongly lexicalist theories, the verb sing-s is third person
singular because the features [3,sing] are contributed by the “-s” suffix (DiSciullo & Williams
1987:27). The distinction between realizational and strongly lexicalist theories is related to the
6

---

SYNCRETISM WITHOUT PARADIGMS
general issue of paradigm structure (strongly lexicalist theories as the term is understood here are
typically incompatible with underspecification, and thus with the approach to paradigms in (3),
but see Wunderlich 1995 for a hybrid approach). Nevertheless, the issue will not be taken up
here, in part since the theory under investigation (Williams 1994) explicitly adopts
underspecification and thus realization. In what follows, then, the entire discussion will be cast in
a realizational perspective. In particular, I will cast the discussion of the paradigm-free theory in
terms of the framework of Distributed Morphology (DM, Halle & Marantz 1993). This
framework is not only realizational in the broad sense but also distinct from other realizational
theories such as Matthews (1972), Anderson (1992) Williams (1994) and Stump (2001) in that
DM claims that the morpho-syntactic representation relative to which the rules of exponence
(VOCABULARY INSERTION) apply is in fact none other than the syntactic representation—the
result of the concatenative rules of syntax having applied to abstract morphemes (bundles of
syntactic features). This choice —while I believe ultimately strongly justified—is for present
concerns an issue of expository convenience; it has no bearing on the main point, but is useful to
keep in mind in considering the specific proposals below.
2. META-PARADIGMS, OR RECURRENT PATTERNS OF SYNCRETISM
Though Williams (1981, 1994) accepts underspecification in morphology, he criticizes the kind
of vocabulary-item driven approach to syncretism which (3) instantiates. His major criticism is
that, in any given language, it is often true that “the pattern of syncretism is a quite abstract
structure, standing above particular words, particular rules, particular suppletive relationships”
(Williams 1994:26). We may illustrate Williams’s point with respect to the English verbal
system discussed above. The notation in (3) initially suggests that the reason the regular past
tense suffix wins out over the third person agreement (i.e., in 3 sg past contexts) is precisely
7

---

SYNCRETISM WITHOUT PARADIGMS
because the individual vocabulary item -d is ranked higher in the list of competitors than is -z.
But it is not an idiosyncratic property of this exponent of past tense that it blocks 3sg agreement.
Rather, it is a general property of English (ignoring be) that regardless of the specific past tense
or past participial affix (of which there are four: -n, “ed” ={-?d, -t, -d}, -t, and Ø) associated
with a given verb, that affix will always preclude agreement and in particular the perfectly
regular third person agreement.2 For example, the past tense of dwell is (for many speakers) the
phonologically unexpected dwel-t /dwel-t/ (cf., yelled /yel-d/). Even though this form takes an
affix distinct from the vocabulary item -d, the past tense in a third person singular context is the
same as the past tense in all other contexts (I dwelt, She dwelt, *Last year she dwells). As
Williams puts it: “even suppletive verbs, the limiting case or irregularity, respects the pattern of
syncretism; the verb go has went as its past tense form. Things could have been different: went
could have been the third past plural form, with goed (or something else) for all the other forms;
but then, go-went would have violated the language-wide pattern of syncretism” (p.25). On a
vocabulary-item driven approach, this means that all past tense formatives must be listed above
the third person singular in the list of vocabulary items in (3).

2 All four affixes occur with and without triggering vowel (or other stem) changes, hence the two are logically
distinct, see Halle & Marantz (1993), Noyer (1997), though see Carstairs-McCarthy 1994 and Burzio 2002 for
qualifications.
-Ø
-t
-ed
-en
+ Stem
bind, see
buy, send
tell, flee
break, drive
Change
bound-Ø, saw-Ø
bough-t, sen-t
tol-d, fle-d
broke-n, drive-n
- Stem
beat, put
dwell, spell
mind, ski
beat, see
Change
beat-Ø, put-Ø
dwell-t, %spell-t
mind-ed, ski-ed
beat-en, see-n
8

---