Review of Clitic and Affix Combinations
Date: Wed, 26 Oct 2005 11:45:51 +0200
From: Hana Skoumalová
Subject: Clitic and Affix Combinations: Theoretical perspectives
EDITORS: Heggie, Lorie; Ordóñez, Francisco
TITLE: Clitic and Affix Combinations
SUBTITLE: Theoretical perspectives
SERIES: Linguistik Aktuell/Linguistics Today 74
PUBLISHER: John Benjamins
Hana Skoumalová, Institute of Theoretical and Computational Linguistics,
This volume is a collection of papers which describe treatment of clitics
and affixes in various theoretical frameworks. The book consists of two
parts - "Clitic sequences" and "Clitics vs. Affixation", each of them
contains five articles. The book opens with an overview by the editors.
There are endnotes and references after each paper. The book also contains
table of contents, names and addresses of contributors, and three indexes.
In the introduction "Clitic ordering phenomena: The path to
generalization", the editors Lorie Heggie and Francisco Ordóñez list some
of the problems connected to clitics and affixes. First, they mention the
hypothesis that affixes are the endpoint of a grammaticalization process
(called 'cline') involving the reduction of words to clitics and then to
affixes. This hypothesis is interesting for the diachronic perspective but
in the synchronic view we need criteria for distinguishing clitics from
affixes. The authors list some counterexamples to criteria stated by
Zwicky and Pullum (1983). In the further text, the authors concentrate on
some problems with the ordering of clitic and affixes: second position
clitic phenomena, ACC-DAT vs. DAT-ACC order in Romance, prohibition of
certain combinations of clitics, which is shown on 'me-lui' constraint,
ordering dependent on the grammatical functions of the clitics, extra
clitics (ethical dative, clitic doubling), and change of the clitic form
under certain conditions (spurious 'se' in Spanish). The next part of the
paper deals with representative approaches to clitic combinations, where
two main directions occur: templatic approach and representational
approach such as Optimality Theory. Finally, the authors list several
problems for future research, as the combination of first and second
person dative object with a third person accusative object, correlation
between non-syncretic clitics for third person and the ordering of dative
and accusative forms, and replacement of dative clitic with a locative
In "Romance clitic clusters: The case connection", Louis H. Desouvrey
investigates ordering of object clitics in preverbal position in French
and spurious 'se' in Spanish.
Features of French clitics are presented in a table, where Case ([A]
and/or [O]) of every clitic and the animacy are shown. The author supposes
that every clitic is specified for Case, unlike the strong forms of
pronouns, which lack Case feature. The behavior of verbs and pronouns is
governed by the interaction of their features. The author also formulates
five constraints on the features and well-formedness conditions on the
representation. The placement of pronominal clitics is a result of clitic
climbing. The clitics are generated in the same position as the nominals
and then they are moved to the left, because they are specified for Case
and their original position violates Obligatory Contour Principle (OCP).
If there are two pronouns in the sentence, the question arises which of
the two clitics moves first (and occupies the first position in the clitic
cluster). The order depends on the original order of both arguments, on
animacy hierarchy, and number of Cases for which single clitics are
specified. Eliminating violations of OCP and other constraints we get the
right order for various clitic clusters.
In Spanish, the situation differs in that first and second person are
assumed to exist in two different but homophonic morphemes: accusative and
oblique. This is very important for the further explanations but
unfortunately, the author does not elucidate why it is so. The clitic
movement from the position where they are generated is motivated by the
same principles as in French: they move so that the violation of OCP was
eliminated. The spurious 'se' replaces third person clitic 'le' in
sentences where third person accusative clitic 'la' or 'lo' also occurs.
The explanation is as flows: clitic with two Cases ('le') is in pole
position (closer to the verb), hence it absorbs both Cases of the verb and
thus the second clitic is not licensed. The remedy of this situation is to
remove the Accusative feature from 'le', which yields the 'se' as the
closest element. The same mechanism is suggested for Italian, where the
strong form pronoun replaces the indirect object clitic in a clitic
The rest of the paper is devoted to co-occurrence restrictions on clitics
(me-lui constraint, Person Case Constraint). These restrictions are
explained as violations of OCP or other constraints which cannot be
eliminated. Author's conclusion is that only Case and animacy are relevant
to the syntax of clitics.
"Constraining Optimality: Clitic sequences and Feature Geometry" by David
Heap is a discussion with Grimshaw's (1997, 2001) OT accounts of clitic
selection and ordering. The material used in the paper are Romance
clitics, and especially their variations in Spanish.
Grimshaw gives the following Universal Markedness Hierarchies (UMHs) for
Person and Case in Italian and Spanish:
*2 >> *1 >> *3
*DAT >> *ACC
These constraints are stipulated without being motivated and they treat
all persons regardless of the number. In Spanish dialect, however, we can
observe asymmetry between SG and PL: in singular, both variants 'me se'
and 'se me', or 'te se' and 'se te' are allowed while in plural only the
variants 'se nos' and 'se os' are possible.
For Spanish clitics the author proposes a structure which is a hybrid of
Bonet's Feature Geometry for clitics and Harley and Ritter's Feature
Geometry for number and gender paradigms. In this proposal, 'se' has the
simplest structure with the node CL only, the third person clitics are the
most complex. This reflects the assumption that more complex structures
are are more marked and vice versa.
For the linear ordering of clitics in standard Spanish, the author
proposes a constraint called Least Leafy to the Left (LLL): Arrange
clitics from the morphologically least specified to most specified. If we
allow variably underspecified clitics we can get variations found in
Spanish dialect. The structure of 'se' can be enhanced by features which
are not required but which are compatible with it. Thus it can have the
same amount of specification as 'me' or 'te' but not the same as 'nos'
In "The syntax of clitic climbing in Czech" by Milan Rezac, the author
tries to formulate syntactic analysis of clitic climbing in Czech, within
the Principles and Parameters framework.
In Czech, clitic cluster occupies the second position in a sentence. The
cluster itself consists of several disjoint sets of clitics. The author
claims that there are ten such sets, but Franks and King (2000) never said
the exact number. The author also oversimplified the discussion on
conditionals and thus he created non-existent combination 'by jsem'
(cond+perf.aux). As the paper deals with climbing of dative and accusative
clitics, this mistake is not too important. There are more serious errors,
in the Czech examples, which suggests that Czech is not author's first
language and that he should consult his examples with native speakers.
The author presents Czech verbs with finite and non-finite clausal
complements and the division of verbs taking non-finite complements to
raising and control type. Clitic climbing is only allowed from non-finite
complements, no matter whether the matrix verb is of raising type, object-
control or subject-control type. Climbing is, however, limited to
complement infinitives. The next claim, that climbing is all-or-nothing
phenomenon should not be declared so categorically: the sentences (11c)
and (11d) sound awkward but they are not worse than (11a), where both
clitics stay at the embedded infinitive. Finally, clitics cannot climb
In the next part, impersonal constructions with reflexive 'se' are
examined. The author claims that in Czech, there is the option for the
internal argument to stay in accusative, but this is wrong. Such a
sentence can be said in a language play but not in a "normal" speech. The
matrix verb assigns nominative to the infinitival object and agrees with
it. The sentences in (14) and (15) are thus wrongly marked.
The next issue is binding domain. A clitic that has climbed cannot be
coreferential with the matrix subject, while one that has not can. Clitic
climbing should block a subject-oriented anaphor within the infinitive
from being bound by a matrix object, even though it is interpreted as the
subject of the infinitive. This is exemplified by sentence (18), which is
unfortunately again wrongly asterisked. All these observations were to
serve to the conclusion that there are two kinds of infinitival
complements. One is non-restructuring infinitive (NRI), which is CP and
blocks clitic climbing. The other is restructuring infinitive (RI), which
is a bare VP and requires clitic climbing.
There is also interaction between Case and climbing. If the matrix verbs
has no object or dative object, the climbing is allowed. However, if it
has accusative object, no climbing is allowed. There is an additional
restriction, which is double dative ban. If the matrix verb has a dative
object, a dative clitic cannot climb from the infinitival complement. And
the last issue is the Person-Case Constraint. This constraint is believed
to be universal and the author brings theoretical arguments why it is so.
Unfortunately, his examples in (43) are again wrongly marked. The sentence
glossed 'I will show him.dat you.acc tomorrow' is correct, as well as the
other examples. The prohibited combination of clitics can result from
clitic climbing as well, as in (44).
In "Romance clitic clusters: On diachronic changes and cross-linguistic
contrasts", Fabrice Nicol investigates Romance pronominal clitic cluster
and especially the diachronic changes in French. In Old Romance, the
accusative-dative ordering ('illum mihi') was standard. Most Modern
Romance varieties have shifted to dative-accusative order ('mihi illum').
In standard French, three constraints apply: (i) Person unicity - there is
no more than one first or second person in a cluster; (ii) Person first -
first or second person come first; (iii) No Direct Object person - there
is no first or second person as a Direct Object in a clitic cluster.
In the Minimalist theory, every clitic is represented as a bundle of
features (case, affix and person-gender-number). In the text, which is
rather technical, the author explains how the cluster with 'mihi illum'
order is derived. Further, he formulates Morphological Opacity (MO) as
follows: Let F be an Interpretable Feature that cannot be erased in the
syntax. If the maximal word-level projection X0max contains such a feature
F, then X0max=[F[...X0...]]. First and second person features cannot be
erased in the syntax and therefore there will not be any first or second
person clitic in second linear position.
In the next section, Italian and Spanish are investigated. These languages
tolerate some clitic combinations that violate MO when one of the clitics
is a reflexive. This requires parametrization of MO for the two
languages. Diachronic changes in Italian and Spanish are then described,
as well as raising and causative constructions in Standard French. The
result of this discussion is a final version of MO. This part is closed by
formulating Case Syncretism property (CSP): In languages with 'mihi illum'
clusters, at least one feature of a third person cluster pronoun is case-
The following three sections contain discussion of the single groups of
Romance languges: Case-syncretic group (Valencia Catalan, Portuguese and
Galician, Italian Basilicatese, Piedmontese, Sardinian, Veneto, Modern
Occitan, Modern Rumanian) Conservative group (with 'illum mihi' ordering -
Aragonese, Majorca Catalan, Modern Provençal: Niçois, Old
French/Occitan/Provençal, (conservative variety of) Modern Occitan,
Corsican), and Mixed ordering group (Aragonese, Barcelona and Minorca
Catalan, Modern Gascon, Modern French).
The last section discusses related theoretical issues and unsolved
The paper "Strong and Weak Person Restrictions: A feature checking
analysis" by Elena Anagnostopoulou investigates the constraint which
prohibits 1st and 2nd person weak direct object together with weak
indirect object ('me lui' or Person-Case constraint). There are two
versions of the constraint. In the strong version there is absolute
prohibition of 1st and 2nd person weak direct object in the presence of
weak indirect object regardless of the person. In the weak version, 1st
and 2nd person weak direct object cannot co-occur with 3rd person weak
There is an interesting correlations between Person-Case constraint and
agreement restriction in some languages, e.g. Icelandic: In the presence
of a dative subject, the agreeing nominative object has to be 3rd person.
The constraint applies only in the case when the nominative object agrees
with the verb. The strong version of PCC and the restriction on
nominative objects in Icelandic are compared, and these similarities are
found: (i) in both cases the restriction arises in environment involving
an argument with an indirect object role and another argument with a
direct object role or, in infinitivals, a lower subject; (ii) the indirect
object argument typically bears morphological dative or genitive case
while the other argument has structural Case; (iii) the argument with
structural Case has to be 3rd person; (iv) reflexive pattern with 1st and
2nd person pronoun cannot co-occur with the dative argument; (v) the two
constraint arise whenever both the dative and the argument with structural
Case relate to the same functional head via movement or agreement; (vi)
the constraints are relaxed in order to circumvent the prohibition of
In the next section, a theoretical analysis of the two constraints
follows. The author proposes use of split feature checking. Further, the
analysis of the weak version of PCC is discussed, and the author proposes
to to use Multiple Agree with the two objects. In the last part of the
paper, the author shows that Multiple Agree can be used also in analysis
of inverse language, which is shown on Passamaquoddy.
"Non-morphological determination of nominal affix order in Korean" by
James Hye Suk Yoon starts with a discussion on nature of nominal particles
in the agglutinative East Asian languages - whether they are affixes added
to the nominal root in morphology or whether they should be treated as
heads of functional projections in the syntax. The Korean nominal
particles behave as phrasal affixes, similar to English possessive: the
particle attaches to the right edge rather than to Head, the nominative
particle displays allomorphy and in turn causes stem allomorphy with
certain stems, it attaches after lexical affixes, it must attach within
the phrase it is associated with, and it can appear in each conjunct
separately or once in a coordinate structure.
The next section contains lexicalist critique of syntactic analyses,
namely Sells' (1995) arguments that show difficulties with syntactically
oriented analyses. The main arguments are non-local c-selection, problem
of underlying structure, and paradox of movement and selection. In the
lexicalist approach, morphological templates are used, but this solution
brings some problems. First, all slots are optional, which is quite
unusual. Secondly, the interpretation of null particle is context-
dependent. Thirdly, there are no discontinuous dependencies holding among
the slots. And finally, the ordering of nominal particles is not
The author proposes a non-morphological analysis of nominal particle
ordering. First, he argues that the honorific Nominative marker '-kkeyse'
is in fact a Postposition and then he analyses constructions with Copula.
The question is whether Copula occupies the last slot in the template. The
answer is that there are several different types of Copula construction in
Korean: Canonical CC, Inverse CC and Cleft construction, and other
particles may intervene between the predicate nominal and the Copula.
Author's conclusion after the analysis is that the Korean nominal
particles can be treated in syntax.
"Clitic positions within the left periphery: Evidence for a phonological
buffer" by Adam Szczegielniak discusses auxiliary clitics in Polish.
Polish has pronominal clitics (which are weak pronouns) and auxiliary
clitics. The auxiliary clitics must be differentiated from auxiliary
affixes which attach to verbs. The auxiliary clitics occur in past tense
formation and serve as subject-verb agreement marker. They attach to the
elements preceding the verb. They can be stressed and coordinated. They
can break up a constituent but only when the constituent can be broken up
by other non-clitic elements (possessive NP, coordination or multi-
The author adopts Rizzi's (1997) model of Left Periphery phrase structure
and he argues that auxiliary clitics are generated in Fin (head of the
Left Periphery that can carry tense/agreement features). The last issue is
why the auxiliary clitics cannot be clause initial. The author proposes
that syntax over-generates and the structures with initial auxiliary
clitics are then ruled out by a phonological filter.
In "The Wh/Clitic-Connection" by Cedric Boeckx and Sandra Stjepanovic, the
authors observe the parallel between wh-phrases and clitics in Bulgarian
and Serbo-Croatian. First, they present the facts. In BG, wh-phrases are
not separable in syntax (they form a unit) and clitics form a cluster in
syntax, as well. In SC, wh-phrases do not need to form a unit and nor do
the clitics. Though multiple wh-fronting is possible in both languages,
they differ in that wh-phrases can be split in SC but not in BG. In
BG, "Attract All Wh" property takes effect, while in SC we can observe
selective attraction. The clitic 'li' can intervene in BG wh cluster,
which is explained by that 'li' is a focus affix in BG, unlike SC, where
this clitic is a complementizer.
The conclusion drawn from the above observations is that BG wh-phrases
form a cluster targeting a unique projection, while in SC the possibility
exists of targeting distinct projections. In both languages, wh-phrases
and clitics occupy the same type of projection in syntax. The difference
between SC and BG is that the former allows multiple such phrases, while
the latter does not.
In "Morphosyntax of two Turkish subject pronominal paradigms" by Jeff Good
and Alan C. L. Yu, the authors describe behavior of two Turkish subject
pronominal paradigm. One of the paradigms (k-paradigm) can be attached to
past and conditional suffix, while the other (z-paradigm) can be attached
to any other predicate, verbal and non-verbal with the exception of
optative and imperative predicates. While the k-paradigm suffixes can
occur in the end of a verb or between two tense, mood and aspect (TMA)
markers, z-paradigm suffixes can be only attached to the end of a verb.
Further, k-ending can be stressed (Turkish stress is word-final), while z-
endings cannot. z-endings can have wide scope over more than one conjunct
in coordination. Authors conclude that k-paradigm endings are affixes
while z-paradigm endings are clitics. This hypothesis is supported also by
the historical development of the two paradigms: z-endings are cliticized
pronouns but k-paradigm endings are the result of a reanalysis of
In the rest of the paper, hierarchy of lexical types for Turkish is
sketched, as well as signs of the single types. The verbs with k-ending
thus is of lexical type finite, while verb with z-ending is of type non-
finite. The finite verbs can have an element of type subj-suffix as a
daughter. The non-finite structures, on the other hand, combine with
elements of the type clitic-pro(noun) in the syntax.
In the paper "On the syntax of doubling", Juan Uriagereka explores the
relation between clitic doubling and inalienable possession. The author
first examines sentences like 'le vi el cordón de ella' (lit. her I-saw
the cord to her). This sentence has only one reading - with inalienable
possession. The inalienable reading, however is not caused by mere
presence of the clitic, as the sentence 'vi su cordón de ella' (lit. I-saw
her cord of her) has the inalienable reading, too. The inalienable
relation need not involve cliticization. However, when a possessive clitic
is used, the relevant interpretation must be inalienable. A sort of
inalienable possession is implied in any instance of clitic doubling, not
just possessive examples. This idea is called Inalienable Double
Hypothesis (IDH): the denotation of a double stands in an inalienable
relation with respect to the denotation of its clitic.
The next section answers syntactic questions raised by IDH. The passage is
rather technical and the conclusion is that doubling construction is
headed by a complete determiner introducing the relational structure,
whereas the possessive construction is not.
This book collects papers directed at a quite narrow field of linguistic
research. In addition, some problems ('me lui' constraint, spurious 'se'
in Spanish, 'mihi illum' order) are explored in more than one paper, which
brings the chance for the reader to compare the competing hypotheses.
Most papers are based on Minimalist theory, only one uses HPSG as the
theoretical framework. Some of the papers contain many technicalities
which makes them hard to read for someone who is not familiar with all the
details of the theory, but on the other hand, some other papers
(Desouvrey, Heap, Good and Yu, above all) are very clear and intelligible.
Bonet, E. 1991. Morphology after Syntax: Pronominal clitics in Romance.
Doctoral Dissertation, MIT.
Bonet, E. 1994. "The person-case constraint: A morphological approach."
MIT Working Papers in Linguistics 22. The Morphology-Syntax Connection: 33-
Bonet, E. 1995. "Feature structure of Romance clitics." Natural Language
and Linguistic Theory 13:607-647.
Boškovic, Ž. 2001. On the Nature of the Syntax-Phonology Interface.
Cho, Y-M. Yu and P. Sells. 1995. "A lexical account of inflectional
suffixes in Korean." Journal of East Asian Linguistics 4(2):149-174.
Chomsky, N. 1995. The Minimalist Program. Cambridge MA: MIT Press.
Chomsky, N. 2000. "Minimalist inquiries: The framework." In Step by Step,
R. Martin, D. Michaels and J. Uriagereka (eds), 89-155. Cambridge MA: MIT
Chomsky, N. 2001. "Derivation by phase." In Ken Hale: A life in language,
M. Kenstowicz (ed.), 1-52. Cambridge MA: MIT Press.
Franks, S. and T. H. King. 2000. A Handbook of Slavic Clitics. Oxford:
Oxford University Press.
Grimshaw, J. 1997. "The best clitic: Constraint conflict in morphosyntax."
In Elements of Grammar, L. Haegeman (ed.), 169-196. Dordrecht: Kluwer
Grimshaw, J. 2001. "Optimal clitic positions and the lexicon in Romance
clitic system." In Optimality-Theoretic Syntax, G. Legendre, J. Grimshaw
and S. Vikner (eds), 205-240. Cambridge MA: MIT Press.
Harley, H. and E. Ritter. 1998. "Meaning in Morphology: Motivating a
Feature-Geometric Analysis of Person and Number." ms. University of
Calgary and University of Pennsylvania.
Kayne, R. 2000. Parameters and Universals. Oxford: Oxford University Press.
Perlmutter, D. 1971. Deep and Surface Constraints in Syntax. New York NY:
Holt, Reinhart and Winston.
Rizzi, L. 1997. The fine structure of the left periphery. In Elements of
Grammar, L. Haegeman (ed.), 281-337. Dordrecht: Kluwer Academic.
Sag, I. and T. Wasow. 2000. Syntactic Theory: A formal introduction.
Stanford CA: CSLI.
Sells, P. 1995. "Korean and Japanese morphology from a lexical
perspective." Linguistic Inquiry 7:119-174.
Sigurðsson, H. Á. 1996. "Icelandic finite verb agreement." Working Papers
in Scandinavian Syntax 57:1-46. Zwicky, A. and G. Pullum.
1983. "Cliticization vs. inflection: English n't." Language 59:502-513.
| ABOUT THE REVIEWER:
ABOUT THE REVIEWER
Hana Skoumalová works at the Institute of Theoretical and Computational
Linguistics at Faculty of Arts, Charles University, Prague, as a research
worker. She also teaches courses on constraint-based grammars. Her
interests are syntax and morphology, and formal methods in linguistics.