AUTHOR: Milan Rezac
TITLE: Phi-features and the Modular Architecture of Language
SERIES TITLE: Studies in Natural Language and Linguistic Theory
Michael Barrie, Department of English, Sogang University
Rezac’s monograph presents a wealth of empirical data, the kind one finds tucked
away in a footnote of many a journal article with the caveat “left to future
research”. The data discussed largely fall under the umbrella of Person Case
Constraints (PCC), in which a first or second person direct object cannot appear
with an overt indirect object, or some variation thereof. While the study
focuses mainly on French and to a lesser extent Spanish (making it essential
reading for Romance linguists), it delves into a number of languages around the
world. This monograph is much more than an analysis of quirky facts in a number
of languages. Rezac ties a common thread among all of them to not only develop a
theory of repair mechanisms, but to argue for a modularized view of grammar,
showing that certain PCC phenomena have clear syntactic effects, while others
are purely morphological. Thus, the monograph has two clear goals:
(1) to argue for a special repair mechanism, R, that can detect impending
failures at the interfaces and insert an extra Probe in the Numeration to rescue
the derivation, and
(2) to argue for a modularized, cognitive view of grammar, in which syntactic,
morphological and phonological phenomena are encapsulated and sheltered from one
Chapter 1 introduces the conceptual basis for the book’s thesis, as well as the
empirical foundation for the study, starting with the following asymmetry in
French (simplified from his discussion).
(i) Elle vous le présentera.
she you.DAT him.ACC will.introduce
‘She will introduce him to you.’
(ii) *Elle le présentera à vous
she him.ACC will.introduce to you
(‘She will introduce him to you.’)
(iii) *Elle nous vous présentera.
She us.ACC you.DAT will.introduce
(‘She will introduce us to you.’)
(iv) Elle nous présentera à vous
she us.ACC will.introduce to you
‘She will introduce us to you.’
The direct and indirect objects appear as clitics attached to the verb, (i).
Normally, the indirect object cannot appear in a separate prepositional phrase
(PP), (ii). However, a 1st or 2nd person or reflexive cannot appear as a direct
object clitic if there is an indirect object in the clause, (iii). In this case,
the indirect object can exceptionally appear as a PP, (iv). Rezac is quick to
point out that the repair structure in (iv) is available only if the standard
structure, (iii) would give rise to ungrammaticality. This chapter goes on to
discuss modularity as a general strategy for cognitive organization, using
modular properties of the visual cortex as an example. He then discusses the
modularity of syntax as a set of operations distinct from morphophonology and
meaning -- roughly phonetic form (PF) and logical form (LF), respectively,
although Rezac is careful to remain theory neutral at this point. Rezac then
goes on to discuss the ubiquity of phi-features in the grammar. Specifically, he
introduces data to be discussed later in the book showing that phi-features are
not the sole province of syntax proper, but are also accessed by morphology and
LF. This chapter ends with a discussion of the empirical topic of this book --
repair strategies. He starts with some examples of irreparable failures such as
the impossibility to repair extraction out of a coordinate structure, even with
a resumptive pronoun, thereby dismissing Optimality Theoretic approaches to
syntax. His argument, however, tacitly assumes that alternative structures are
not made available by GEN in such a framework (his examples (34a-b)).
v. Who set out for Pohjola __ along with Niera’s son?
vi. *Who set out for Pohjola __/he and Niera’s son?
Specifically, v. could be in the evaluation set, and is ultimately selected by
the grammar. He states here that the repair paradigm in i.-vi. above is
cross-linguistically robust and begs a unified treatment. The core of his theory
of repairs is summed up in his formulation of R.
R: An uninterpretable feature may enter the numeration only if needed for Full
Interpretation of the syntactic structure built from it.
Chapter 2 discusses the role of phi-features in morphology. Here, Rezac is more
concrete regarding his view of the organization of grammar. He adopts a fairly
standard Y model that firmly puts morphology after syntax (i.e., no
pre-syntactic generative lexicon), and uses this discussion as a vehicle to
sharpen his view of modularity. And he now advances a fine-grained notion of
modularity and defines a “modular signature”. The modular signature spells out
the domain of interactions (or interfaces), the kinds of information available
to the module, and the computational processes permitted in the module (such as
Merge in the case of syntax). Rezac distinguishes between syncretic and opaque
cliticization. Syncretisms are better known and involve the loss of distinction
for a given feature or features. Thus, French ‘les’ is syncretic for gender,
although gender is distinguished in singular forms. Opaque cliticization
involves a lack of clear source of phi-features as opposed to a true
neutralization. For instance, in Old and Middle French a plural dative could
combine with a singular accusative; however, plural marking could shift from the
dative to the accusative clitic. Thus, the source of plurality is opaque in the
overt realization. These two phenomena are clearly morphological, following
Rezac’s discussion. He presents data from a number of languages showing that
they do not have any effect on the syntax. Additionally, he discusses gaps, such
as the lack of a participial form for ‘stride’ in English, and attributes this
to a morphological module.
Chapter 3 is devoted to person hierarchies in the syntax module. Person
hierarchies are well known from Algonquian languages, where the verb
preferentially agrees with a 2nd person subject over a 1st person subject. He
shows here that person hierarchy effects in numerous languages (notably Ojibwa
and Mapudungun) are syntactic rather than morphological. Unlike the phenomena in
chapter 2, person hierarchy effects have visible consequences for the syntax.
Thus, for example, the agreed with element can participate in certain syntactic
constructions such as cross-clausal agreement. If agreement were purely
morphological, the syntax would not have access to this information to ban
syntactic operations based on it. Considering data from Arizona Tewa, Rezac
rejects a post-syntactic filtering mechanism to explain person hierarchy effects
and concludes that they are genuinely syntactic, thereby adding support to his
Having established clear morphological and syntactic effects of phi feature
interactions, Rezac discusses in chapter 4 how PCC violations are repaired in
spoken French, noting speaker variation where present. Rezac lays out the
fundamental aspects on the syntax of French clitics, including dative and
applicative constructions. Rezac covers an impressive range of empirical facts,
including naturally occurring examples. In this chapter Rezac discusses in
detail the various kinds of repairs for PCC violations available in French. He
continues with a discussion of the syntactic nature of the PCC repairs, offering
evidence from floating quantifiers, Condition B effects, and right dislocation,
thus firmly placing the PCC repair mechanism in the syntactic module he argues
for above. After a comprehensive exposition on various kinds of applicative
clitic constructions, Rezac ends the chapter with a discussion of irreparable
violations, employing these facts to bolster his modularity hypothesis. He shows
that in some cases the failures are not visible to syntax. Since the repair
mechanism, R, is firmly implanted in the syntax module, it does not see the
violation, so does not enact a repair.
Chapter 5 lays out the specifics of R -- how it operates and how it affects the
syntax. Given the potentially strong generative capacity of R, Rezac is careful
to delineate clearly its mode of operation and its limitations. In a nutshell,
the PCC problem and its solution thanks to R boil down to the following. Only
1st, 2nd, and reflexive pronouns have a person feature. Other 3rd person
pronouns/clitics do not. When such a clitic (an accusative clitic) intervenes
between the little v probe and its goal, the dative clitic, a blocking effect
ensues and the derivation crashes. The repair mechanism, R, detects this crash
and adds an additional uninterpretable feature to the Numeration, thus salvaging
it. In the case of the French PCC repair, the additional uninterpretable feature
comes in the form of a preposition to assign Case to the dative clitic, as Case
was swallowed up by the person-bearing accusative clitic. The global character
or R, according to Rezac, rests in the fact that it is part of the interfaces
between syntax and the lexicon, PF and LF. Rezac runs through the long list of
PCC violations in the numerous languages he discussed in the previous chapters,
showing how R accommodates the facts. He then makes the interesting connection
between PCC and transitivity, arguing that R can account for accusative objects
and ergative subjects. He leaves open in this chapter the parameterization of R.
Specifically, he does not answer the question of what kind of additional probe
would be made available in a given language. He ends the chapter with a
discussion on the limits of R. Those structures which cannot be repaired by R
Chapter 6 discusses phi-features themselves in more detail. Specifically, Rezac
delves into the issues of phi-features as a common alphabet for morphology,
syntax and interpretation, probing various kinds of mismatches. One such
mismatch he discusses in detail is French ‘on’. This pronoun is morphologically
3rd person singular, but can be interpreted as first person plural. Another
mismatch Rezac discusses is imposters in the sense of Collins and Postal (2012),
in particular, 3rd person markers that act as 2nd person (What would madam
like?). It turns out that the clitic forms of these do not induce PCC effects.
Finally, Rezac presents a brief conclusion that ties the work together.
Rezac certainly presents an impressive contribution to syntactic theorizing
using what are often seen as problematic data. Instead, Rezac gleans an
important cross-linguistic generalization and furnishes a novel and elegant
mechanism to account for it. In terms of content, Rezac’s analysis makes strong
predictions about the nature of such repairs. For instance, they are limited in
scope to the phase in which R detects the crash. Also as mentioned, they occur
only when needed. Both of these predictions are supported. Rezac clearly lays
out his goals in the introduction and regularly revisits them throughout the
monograph. Furthermore, this monograph presents a novel mechanism for PCC and
related effects in a number of languages. The majority of recent work on PCC is
restricted to a particular language or language family, Rezac ties together such
phenomena in a wide number of languages.
Although Rezac indicates that the kind of repair made available by R differs
cross-linguistically and that cross-linguistic variation is left to future
research, I would still like to see an indication of how this parameterization
looks. Does it correlate with other aspects of the grammar (like postpositions
and OV order)? I would also like to see how R is constrained. Can it happen
twice in a phase (as perhaps is the case with ditransitives)? Can it happen
three times? If not, why not? Also, what counts as having a [person] feature?
For Rezac, in French, 1st, 2nd, and 3rd person reflexives have a [person]
feature, but other 3rd persons do not. There has, of course, been much
discussion on this topic (Benveniste, 1966 ; Kayne, 2000 ; Nevins, 2007). Is
this parameterized across language? If so, how is it acquired by the child?
These unanswered questions, of course, indicate a clear avenue for future
research and are not to be taken as faults.
As a monograph, the work is well organized and contains substantial references.
Indeed, Rezac’s coverage of previous and related literature is thorough. Readers
familiar with Rezac’s style will quickly recall the density of his writing.
This, in fact, is one drawback to the volume, as the reader sometime has to work
his way through difficult prose rather than concentrate on the content and the
data, as such it is recommended only for very advanced graduate students and
other members of linguistic academia. This should not be grounds for avoiding
the book, as the argumentation is sound and presented in a logical manner. In
sum, Rezac’s book is important for anyone interested in the architecture of
grammar, minimalism, or Romance syntax.
Benveniste, Émile. 1966. La Nature des Pronoms. Problèmes de Linguistique
Générale, 251-57. Paris: Gallimard.
Collins, Chris & Paul Postal. 2012. Imposters Cambridge, MA: MIT Press.
Kayne, Richard. 2000. Person Morphemes and Reflexives. Parameters and
Universals, ed. by R. Kayne. Oxford: Oxford University Press.
Nevins, Andrew. 2007. The representation of third person and its consequences
for the person-case constraint. Natural Language and Linguistic Theory
ABOUT THE REVIEWER