Publishing Partner: Cambridge University Press CUP Extra Publisher Login
amazon logo
More Info

New from Oxford University Press!


Style, Mediation, and Change

Edited by Janus Mortensen, Nikolas Coupland, and Jacob Thogersen

Style, Mediation, and Change "Offers a coherent view of style as a unifying concept for the sociolinguistics of talking media."

New from Cambridge University Press!


Intonation and Prosodic Structure

By Caroline Féry

Intonation and Prosodic Structure "provides a state-of-the-art survey of intonation and prosodic structure."

Review of  Construction Grammar in a Cross-Language Perspective

Reviewer: Lea Cyrus
Book Title: Construction Grammar in a Cross-Language Perspective
Book Author: Mirjam Fried Jan-Ola Östman
Publisher: John Benjamins
Linguistic Field(s): Linguistic Theories
Cognitive Science
Subject Language(s): Czech
Issue Number: 16.1913

Discuss this Review
Help on Posting

Date: Mon, 20 Jun 2005 17:15:53 +0200 (CEST)
From: Lea Cyrus
Subject: Construction Grammar in a Cross-Language Perspective

EDITORS: Fried, Mirjam; Östman, Jan-Ola
TITLE: Construction Grammar in a Cross-Language Perspective
SERIES: Constructional Approaches to Language 2
PUBLISHER: John Benjamins
YEAR: 2004

Lea Cyrus, Arbeitsbereich Linguistik, Westfälische
Wilhelms-Universität Münster


The volume under review consists of five chapters which can be
divided into two major parts: the first two chapters are an introduction
to the construction grammar (CxG) framework, while the last three are
in-depth studies dealing with various constructions in Czech,
Japanese, and French.

In the first chapter (pp. 1-10), which is also an introduction to the
collection, the two editors begin by putting CxG into a wider
perspective and describing the way it relates to other frameworks,
such as Case Grammar and Relational Grammar, Gestalt Grammar,
Head Driven Phrase Structure Grammar (HPSG), and Frame
Semantics. They go on to explain how the "cross-language
perspective" of the title is addressed in this volume. Their main
objective is to refute the accusation that CxG is suitable only for the
grammar of English, for which it was originally devised. This will be
achieved by presenting three studies (chapters three to five) that
successfully apply the CxG framework to languages other than English.

The second chapter (pp. 11-86), by far the longest in the volume, is
devoted to a detailed description of the workings of construction
grammar in the Fillmore/Kay tradition, in which a conceptual closeness
to frame semantics is combined with a formal apparatus borrowed from
HPSG. The two editors, who are also the authors of this paper, felt the
need to include this type of overview, since to date there exists no
published introduction to this framework (the publication date of
Fillmore et al. (to appear) has repeatedly been postponed and is now
scheduled for end of July 2005, and the various prepublished versions
of the book (e.g. Fillmore and Kay, 1995) are not available to a wide
audience). After an overview of the main ideas and key concepts of
CxG, Fried and Östman place particular emphasis on a detailed and
example-rich step-by-step introduction to the mechanisms of this
framework. They introduce the characteristic box notation and explain
how attribute-value matrices and co-indexation are used. They also
show how CxG takes care of a wide range of phenomena, such as
determination in noun phrases, valence representation, various cases
of linking, as well as raising and control. The major construction types
(lexical, phrasal, linking and ordering constructions) are introduced in
more or less detail and, as the chapter proceeds, the examples - and
boxes - become increasingly complex. In many ways, this introduction
follows Fillmore and Kay's textbook manuscripts (e.g. Fillmore and
Kay, 1995). Of course, some things have been left out, some have
been added, and the overall organisation is different, too, but all in all
the closeness is apparent. The main difference between the
manuscript version and this paper is that, while the former is mostly
based on English, the latter also shows how phenomena that occur in
other languages can be accommodated within the CxG framework.
Examples of this include experiential verbs in Russian, which are
obligatorily expressed without a nominative, a family of dative
constructions in Czech, a Czech pronoun that serves both to refer to
the second person plural and as a polite form of address, and a
Hungarian ordering construcion.

In Chapter 3 (pp.87-119), Mirjam Fried examines two Czech
experiential constructions which, at first sight, differ only in the case
they assign to the experiencer, which is dative in the more productive
version (-> D-construction) and accusative, sometimes alternating with
dative, in the other (-> A-construction). In what follows, Fried
convincingly challenges the view held by more traditional grammarians
that the accusative version is an exception to the more basic dative
version. A close examination of both constructions leads her to the
conclusion that what they have in common is the semantic frame which
she calls "localized experience" and which, through its valence,
determines the two participants that must be realized: one with the
semantic role 'interest' (the experiencer), the other with the
role 'locative' (the body part where the experience is located).
Conversely, it also determines which participants must not be realized
and thus suppresses potential stimulus or agent arguments
contributed by the predicate. The two constructions differ in the type
of head predicate they contain. The D-construction mostly integrates
intransitive verbs of emitting light or sound, whereas the A-
construction integrates transitive verbs of direct physical contact. The
actual morphosyntactic realization of participants then comes about
through an interplay of regular linking patterns and constructional
requirements: after the only argument contributed by the intransitive
predicate in the D-construction has been suppressed, there is no
predicate-specific argument left that could take on the experiencer
role, so a dative of interest is introduced - a common and regular
pheonomenon in Czech grammar. In the A-construction, however, the
experiencer need not be added, because it is realized by a
reinterpretation of the patient argument of the transitive verb. Like all
Czech transitive patient arguments, this is in the accusative. The
occasional dative-alternation can occur on the analogy of the more
productive dative-experiencer in the D-construction. The overall
conclusion of Fried's paper is that the regular mechanisms linking
semantic roles and morphological cases on the predicate level can be
overriden by constructional requirements, which leads to apparently
exceptional case marking.

Seiko Fujii's study (Chapter 4, pp. 121-155), which is based on 15
hours of transcribed speech, looks at the way deontic meaning, in
particular obligation, is encoded in Japanese conditional clauses. She
distinguishes three types of obligation-expressing conditionals. What
the three types have in common is the form of their protasis: it consists
of a full clause which is usually negated (something like 'If I don't do
this...') and the conditional linker "to". They differ both in the form of
the apodosis and in the way the obligation is encoded. In the first
case, the apodosis consists of another full clause. Here, the obligation
is encoded only by conversational implicatures and depends on the
speaker and hearer's subjective negative evaluation of the situation
referred to by the apodosis: the obligation arises from the need to
avoid this negative situation. In the second case, the apodosis consists
of a negative evaluative predicate only. Consequently, the negative
evaluation of the antecedent clause and the resulting obligation is
coded explicitly in this construction type and need not rely on
conversational implicatures. In principle, any negative evaluative
predicate can fill the predicate slot in this construction, but the data
show that two predicates ("ikenai", "dame", both approximately
mean 'bad') account for fifty percent of the cases, so this construction
type is to a large extent conventionalized. The third construction type
finally contains only the bare protasis, i.e. there is no apodosis. Since
there is no second clause to help establish the deontic meaning, it
must be the construction itself that does so. Fujii introduces the notion
of "constructional scheme" to group together those construction types
that share a common meaning (e.g. 'obligation'). Towards the end of
her paper, she briefly touches upon constructional schemes
expressing other kinds of deontic meaning (e.g. prohibition). While the
various schemes can all be expressed by the same construction types,
they differ in the form of the protasis and, more importantly, in their
choice of clause linker (in Japanese, there are several ways of
expressing condition). Since regularities can be observed both across
construction types and across constructional schemes, Fujii sees
these notions as well suited for structuring families of constructions.

Knud Lambrecht, in Chapter 5 (pp.157--199), examines a French
construction which is very common in spoken language but has neither
been discussed in major grammars of French nor included in major
dictionaries, possibly due to its confinement to spoken discourse. The
construction under investigation is a variant of copular subject-
predicate structures like "C'est un livre intéressant" in which the noun
is right-detached, leaving its modifier isolated in the predicate position,
as exemplified by "C'est intéressant, comme livre". Lambrecht refers to
this construction as the Right-Detached "comme"-N construction
(RDCN). A close comparison of this construction with the related right-
topic construction (R-TOP) in which a topical argument is right-
detached but coindexed with a preverbal pronominal ("Il(i) est
intéressant, ce livre(i)"), reveals a number of similarities, so that RDCN
can justly be called a variety of R-TOP. However, there are also quite
a few syntactic and semantic differences, which proves that RDCN is
indeed a construction in its own right. Similarly, the detached
subcomponent "comme"-N deviates semantically from otherwise
comparable uses of "comme"-N in that it is neither role- nor domain-
specifying. One important result is that the detached constituent
denotes "the category which is modified by the intra-clausal predicate
adjective and of which the subject denotatum is an instance" (p. 178).
Lambrecht then sets out to determine why and in what contexts RDCS
is favoured over its canonical counterpart and reaches the conclusion
that they differ in their information structure. The canonical
construction is pragmatically ambiguous in that it does not determine
whether it is the whole predicate NP or only the modifying adjective
that is in focus. The RDCS construction is different in this respect:
here, the information structure unambiguously specifies that only the
adjective is in focus, while the denotatum of the detached noun must
be both known and discourse-active.


My overall impression of this book is very favourable: all contributions
are well-crafted, provide interesting insights, and are certainly state-of-
the-art in CxG research. What is also very important: they read well.
Even those not familiar with Czech, Japanese, or French will have no
difficulty following the discussions. My evaluation is structured as
follows: First, I will say a few words about the overall organization of
the volume. I will then comment on the framework as represented in
this collection, particularly in Chapter 2. After that, I take a look at the
cross-linguistic perspective. Finally, I will point out a few minor
mistakes I noticed while reading the book.

I found the overall organization of the book slightly unusual, a kind of
hybrid between a textbook/monograph and a collection: it is an edited
volume, but apart from the editors, there are only two contributors:
more than half of the pages have been written by the editors
themselves. Also, the combination of papers made me wonder what
kind of audience the editors had in mind. While Chapter 2, the
introduction to CxG, is written in textbook style and obviously intended
for newcomers to the field who still need to learn the basics, the
subsequent papers seem to be directed at readers familiar with the
framework. As mentioned above, this chapter was included in the
collection because there is no introductory textbook to CxG available
yet - and it certainly is a good introduction. However, I doubt that those
in need of such an introduction would look for it in an edited volume
like this one, and I suspect that those who buy or borrow this book will
have enough background to be able to do without it. Furthermore,
once Fillmore et al.'s textbook is out, which will be very soon (unless
there is another postponement), this chapter will lose some of its
raison d'être and will then seem even more out of place in this type of

I will now turn to a few matters regarding the CxG framework itself.
Before doing so, I would like to stress that I took an immediate liking to
this approach the moment it first came to my attention, and that, on the
whole, I find it very convincing and promising. My comments are thus
well-wishing and I am very much aware that many of the problems may
well be teething troubles of a young discipline or even
misunderstandings on my part.

The CxG approach advocated in this volume is the Fillmorean
approach and as such naturally incorporates many features of earlier
Fillmorean approaches, most prominently semantic roles and frame
elements. Both are very useful in the description and explanation of
many phenomena. However, I sometimes miss a certain awareness of
the fact that, by incorporating these notions into a new theory, one
does not only inherit their benefits, but also their problems. For
instance, it has turned out to be notoriously difficult to determine the
number and type of semantic roles and to draw clear distinctions
between them - Fillmore himself acknowledges this (Fillmore and Kay,
1995, p. 4-22). That this difficulty has not been overcome can be seen
by the fact that Fried and Östman, in the lexical construction of the
verb "to persuade" (p. 65), assign the semantic role of patient to
the "persuasion target", wheras Fillmore and Kay chose to assign to it
the semantic role of experiencer (Fillmore and Kay, 1995, p. 7-21).

Similarly, there is frequently some subjectivity involved when deciding
what frame elements are necessarily part of the meaning of a given
predicate. On p. 52, for example, one of the frame elements of WALK
is 'Companion'. Since just about any action can be performed together
with a companion, this can hardly be said to be necessary specifically
for the act of walking. I first suspected that Fried and Östman's reason
for assuming the 'Companion' as a frame element was that they need
it for explaining the construct "She'll walk you across the street" (p. 50)
(they argue - admittedly tentatively - that this comes about through
unification of an intransitive verb and the Affected Object construction,
which adds an object (patient) to the valence of WALK and links it to
the 'Companion'). However, since they go on to say that an added
argument need not be in the predicate's inventory of frame elements,
this may not have been the reason after all. Whatever it is: if the
notions of frame elements and semantic roles are to play a central part
in the formalization mechanims, we should not forget to address the
problematic issues this entails.

The Affected Object construction leads me to another point that I feel
is not always given the appropriate amount of attention in some CxG
publications and also in this volume, namely over-generation. On p.
24, Fried and Östman state that CxG "aims to account for all of the
grammatical sentences of the language and only those". Sometimes,
particularly on the more general levels, constructions are introduced
that solve the problem at hand but cause other problems elsewhere.
Take, for example, the Affected Object construction: how is it
accounted for that this construction does not unify with verbs like "kill"
or "eat" or "explode"? Also, as far as I can see, nothing prevents this
construction from licencing the ungrammatical construct "*I walked
him". It is repeatedly argued that this approach is "markedly different"
(p. 112, also p. 56) from Goldberg's argument structure constructions
(Goldberg, 1995), mainly because the latter is not flexible enough to
accommodate finer meaning distinctions, but also for some formal
reasons. Goldberg would see the construct "She'll walk you across the
street" as an instantiation of the "caused-motion construction"
(Goldberg, 1995, Chapter 7), which supplies the appropriate
arguments and imposes certain semantic constraints on its slots.
Whether or not it is true that Goldberg's approach could not be
adjusted to cope with the Czech phenomena presented by Fried, I
cannot judge. However, as long as the over-generation issues
mentioned above are not solved (or as long as I do not know how they
are solved), I find the solution offered by Goldberg intuitively more

The following remarks concern the "cross-language perspective"
announced in the title of the book. My own expectation when reading
the title was that the book would address the question whether and
how cross-linguistic studies could sensibly be conducted within the
CxG framework. Since CxG is sign-based and since signs are by
definition arbitrary and conventional and thus language-specific, this is
a legitimate question. Östman and Fried also point out that "the
question of universality requires serious attention" (p. 6) due to these
basic tenets of CxG. Croft (2001), who argues for the ultimate
language-specificity of constructions, claims that "their function in
structuring and communicating information is not [language-specific]"
(p. 60) and suggests that "valid cross-linguistic generalization are
generalizations about how function is encoded in linguistic form" (p.
363). Accordingly, one way of taking a cross-linguistic perspective in
the CxG framework could have been to concentrate on one particular
function and investigate its encoding in different languages.

However, the cross-language perspective in this volume is largely
restricted to applying the CxG framework to phenomena from different
languages. The long-term goal to find "cross-language
generalizations" (p.8) is mentioned but the phenomena under
investigation are from very different areas, so it seems that it was not
really attempted to achieve parts of this goal just yet. Maybe it is too
early to expect this kind of result from CxG: it has been characteristic
of this framework to investigate those phenomena that are traditionally
thought of as being on the periphery of grammar and to show that they
can be explained with the same type of mechanisms as the rest. This
was also done, and very expertly so, in the contributions in this book.
But it is in the nature of these "peripheral" phenomena to be highly
language-specific, and consequently they are not an obvious starting
point for cross-linguistic investigations. What this book has indeed
shown is that CxG is very suitable for describing, explaining and
representing the linguistic phenomena of different languages that have
little in common.

Finally, I noticed a couple of minor mistakes that must have escaped
the editors' attention: "18b" and "18c" on p. 55 should be "17b"
and "17c" respectively. In the gloss of example 20a on p. 67, two
words are marked as being nominative. I do not speak Czech, but from
the context I presume that "child" should really be marked dative. On
p. 122, one of the instances of "encoding idioms" should be "decoding
idioms". There is no outer box round construction type (iv) on p. 128. I
am not sure whether this is on purpose, but since there is one round
the corresponding constructional scheme on p. 130, I think there
should be an outer box on p. 128, too. In the constructional
scheme 'obligation' on the same page, there is a Kleene star next to
the right inner box, which does not make much sense. "Detained" on
p. 130 should be "detailed", and "idiomatitcity" on p. 151 "idiomaticity".
On p. 187, the PHON attribute of the adjective is assigned the
value "unacc", but, if I am not mistaken, the correct value would
be "acc".


Croft, William (2001). Radical Construction Grammar: Syntactic Theory
in Typological Perspective. Oxford: Oxford University Press.

Fillmore, Charles J. & Paul Kay (1995). Construction Grammar.
Lecture Notes, Lingustics X-20, University of California, Berkeley.

Fillmore, Charles J., Paul Kay, & Laura Michaelis (to appear).
Construction Grammar. Stanford: CSLI Publications.

Goldberg, Adele E. (1995). A Construction Grammar Approach to
Argument Structure. Chicago: The University of Chicago Press, 1995.


Lea Cyrus is a research assistant and PhD student at the English
Department at the University of Münster, Germany, where she teaches
1st and 2nd year undergraduate students. Her research interests
include descriptive grammar and bi- or multilingual treebank design.