16.2277, Review: Ling Theories/S yntax: Fried & Östman, 2nd review

[LINGUIST List]

LINGUIST List: Vol-16-2277. Thu Jul 28 2005. ISSN: 1068 - 4875.

Subject: 16.2277, Review: Ling Theories/Syntax: Fried & Östman, 2nd review

Moderators: Anthony Aristar, Wayne State U <aristar at linguistlist.org>
            Helen Aristar-Dry, Eastern Michigan U <hdry at linguistlist.org>
 
Reviews (reviews at linguistlist.org) 
        Sheila Dooley, U of Arizona  
        Terry Langendoen, U of Arizona  

Homepage: http://linguistlist.org/

The LINGUIST List is funded by Eastern Michigan University, Wayne
State University, and donations from subscribers and publishers.

Editor for this issue: Naomi Ogasawara <naomi at linguistlist.org>
================================================================  

What follows is a review or discussion note contributed to our 
Book Discussion Forum. We expect discussions to be informal and 
interactive; and the author of the book discussed is cordially 
invited to join in. If you are interested in leading a book 
discussion, look for books announced on LINGUIST as "available 
for review." Then contact Sheila Dooley at collberg at linguistlist.org. 

===========================Directory==============================  

1)
Date: 27-Jul-2005
From: Valeria Quochi < valeria.quochi at ilc.cnr.it >
Subject: Construction Grammar in a Cross-Language Perspective 

	
-------------------------Message 1 ---------------------------------- 
Date: Thu, 28 Jul 2005 03:32:04
From: Valeria Quochi < valeria.quochi at ilc.cnr.it >
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
YEAR: 2004
PUBLISHER: John Benjamins
Announced at http://linguistlist.org/issues/16/16-124.html


Valeria Quochi, Linguistics Department, University of Pisa

[For another review of this book, 
see http://linguistlist.org/issues/16/16-1913.html  --Eds.]

BOOK DESCRIPTION

This book consists of 5 chapters by 4 different authors: M. Fried, J. 
Östman, S. Fujii and K. Lambrecht. The book's goal is to present 
Construction Grammar to the broader linguistic community both as a valid 
formal theory of language, and as a useful set of tools for the 
explanation and description of linguistic facts across languages. The book 
can be divided into two main sections: The first two chapters give a nice 
introduction to Construction Grammar as a model of grammar in general. The 
authors reconstruct the historical roots of the theory as well as its 
similarities with other theories. The second section, formed by the three 
remaining chapters, shows how the formal tools previously presented can be 
applied to explain and describe specific linguistic expressions in 
different languages, namely Czech, Japanese and French. 

Chapter 1: J. Östman and M. Fried, "Historical and intellectual background 
of Construction Grammar". In this chapter the authors reconstruct the 
origins of Construction Grammar, its theoretical roots and its present-day 
developments, as well as give a nice introduction to the whole volume. The 
authors first attempt to clarify the notion of Construction, which has 
been abused in recent linguistic analyses of various orientations, and 
then to establish it as the main building block of Construction Grammar. 

Construction Grammar originates from the works and theories elaborated by 
C. Fillmore, his colleagues and students at Berkeley in the 1980s, 
evolving  particularly from Case Grammar (Fillmore 1968) and 
Lakoff's "Gestalt" Grammar (1977). During the 90s it was further 
elaborated by various Berkeley linguists and formalized, especially by 
Paul Kay. The application of constructional principles originally started 
with the so-called peripheral, idiomatic expressions and has subsequently 
been applied to other parts of grammar. The book as a whole tries to bring 
together what the authors believe are the most significant "branches" of 
Construction Grammar still active and valuable today.

Among the current lexical semantic frameworks, the authors highlight how 
Frame Semantics and Construction Grammar naturally and logically 
complement each other. Frame Semantics, in fact, represents the meaning of 
words: specifically, it is about the relationship between form and 
function from the perspective of lexical semantics. Construction Grammar 
has the same concerns but from the broader perspective of grammar.

Relative to the question whether Construction Grammar is a universal model 
of grammatical knowledge, the authors state that what can be considered 
universal in Construction Grammar is the set of formal mechanisms and 
representational apparatus, given that Construction Grammar notions are 
grounded in general cognition. Several studies of various phenomena in 
different languages seem to suggest that Construction Grammar is adequate 
for relating to a universal aspect of language as well as language-
specific facts. No assumption is made, however, as to what linguistic 
structures or properties should be considered universal.
 
Chapter 2: by M. Fried and J-O. Östman, "Construction Grammar: A Thumbnail 
Sketch" is an overview of Construction Grammar that addresses both its 
theoretical foundations and its basic formal devices.

Construction Grammar is defined as "a non-modular, generative, non-
derivational, monostratal, unification-based grammatical approach (Kay 
1995:171). It is non-modular in that all types of information 
(morphosyntactic, phonetic/phonologic, semantic, pragmatic etc.) are to be 
represented at one and the same level, within a complex sign; and 
monostratal because there are no derivations, transformations or 
movements: all expressions are not generated, but 'licensed' by particular 
abstract constructions. It is generative in the sense that it attempts to 
account for all the exclusively grammatical utterances in a language. 
Construction Grammar is also said to be a Maximalist Theory and usage-
based approach because its goal is to cover all the grammatical facts of a 
language, and attempts to do so by analyzing real occurring data. Finally, 
Unification is the basic formal mechanism that "ensures that pieces of 
linguistic material that do not match along any number and types of 
properties (i.e. syntactic, semantic, pragmatic) will not be licensed as 
possible constructs of a given language" (Chapter 2: 25). 

The notion of Construction is the basic unit of analysis and 
representation. Constructions are considered to be cognitive entities out 
of which speakers, build, complex expressions; therefore, utterances are 
the product of the interaction between grammatical constructions and 
linguistic material (i.e. words, or lexical constructions). 

In Construction Grammar all linguistic expressions, from morphemes to 
sentences to prosodic elements, are represented in the same way as 
constructions, so that no type of expression can be considered as more 
central or peripheral than the others. The analytical tools described in 
this chapter are claimed not to need any a priori decisions about what is 
core and what is peripheral in language: idiomaticity, and semi-
productivity, and regularity are treated alike. 

Every construction, however, may specify  more or less detailed 
information about its characteristics depending on its nature. Among the 
things that a construction may specify are morphosyntactic properties like 
dependency relations and ordering constraints, prosodic and phonetic 
shape, meaning features like boundedness or animacy, information about 
context like register. No information is obligatory, however, and there is 
no minimal number of features that a construction has to specify.  Usually 
grammatical and lexical constructions interact and integrate in non-
trivial ways, and it is specifically to account for how they interact that 
the formalism of Construction Grammar has been and still is being 
developed. 

Constructions of varying degree of abstraction and complexity are 
organized into networks and families structured through inheritance and 
instantiation relations. Box notation is the graphical organizational tool 
that has been chosen to formalize grammatical knowledge. Notation relies 
on 3 main devices: nested boxes are used to represent constituent 
structure; feature structures are used for encoding grammatical 
information; and co-indexation is used to keep track of unification 
relations. Feature Structures, i.e. attribute-values matrices, are not 
posited a priori, rather the identification of the relevant categories and 
values for the description and representation of a given grammatical 
construction is data-driven, which means that no category in the framework 
can be considered as universal or primitive. 

The representation of complex and larger grammatical patterns is made 
through two main domains of representation: one is the external domain, 
which represents the characteristics of a construction as a whole; the 
other is the internal domain, which represents the constituents (i.e. 
smaller constructions) that make up the construction. Both domains may be 
more or less detailed, thus constraining at different degrees the 
inventory of lexical constructions or linguistic material that can be 
unified with a construction. 

Finally, there are two main principles for building larger grammatical 
patterns in Construction Grammar: Linking  and Instantiation. Linking 
constructions must specify how semantic arguments of lexical constructions 
unify in a given grammatical pattern. Depending on the particular language 
analyzed, linking constructions may involve case marking, grammatical 
functions, word order. The fundamental role of linking constructions in 
general is to integrate frame-semantic and valence information of lexical 
items with grammatical constructions.

Instantiation Principles, on the other hand, are described to serve the 
purpose of constraining how constituents are physically realized within a 
grammatical construction: depending again on the language, they determine 
for example the structural dependencies between constituents, the 
grammatical functions that have to be present, or the specific word order. 

All the representational devices sketched in this review receive a 
detailed description and motivation supported by exemplified analyses of 
interesting constructions, not only in English.

Chapter 3: M. Fried " Predicate Semantics and Event Construal in Czech 
Case marking" describe an analysis of a semi-productive experiential 
construction in Czech, with respect to a similar but productive one. Both 
of them are impersonal constructions, i.e. they lack a nominative Noun 
Phrase, and both express some kind of experience located in a body part. 
However, whereas the fully productive pattern marks the Experiencer Noun 
Phrase with the Dative Case, there exists a range of similar expressions 
in which the Experiencer has the accusative case. 
 
The main claim of the paper is that case marking follows regular 
principles of Czech grammar even in the semi-productive Accusative-
Experiencer Construction, which is traditionally considered idiomatic 
precisely because case assignment seems not to follow the general rules of 
the grammar. Fried shows that, in this construction, case marking can be 
considered as regularly assigned if one considers that both head 
predicates and constructions synergistically contribute to its 
determination. In the Dative-Experiencer Construction, which requires zero-
place or unaccusative predicates, case assignment is fully predictable on 
the bases of the general principles of the grammar, even if its 
Experiencer and Locative arguments cannot be licensed by the predicate, 
but must be contributed by the construction. 

The Accusative Construction shares many of the constructional feature of 
the Dative one (i.e. the locative argument, the suppression of the agent 
etc.), but appears to be more restricted, which turns out to be the key 
for their different case marking. Among its restrictions, is the 
fundamental fact that this constructions appears to unify only with 
(semantically) transitive verbs, and the predicate tends to appear in the 
perfective aspect (another mark of transitivity and of complete 
affectedness of the patient, in Czech). These observations explain the 
tendency of the experiencer in this construction to receive accusative 
case, and, at the same time, it explains why, in some cases, dative 
marking is also possible. Moreover, the accusative marking of the 
experiencer is shown to be exploited elsewhere in the grammar, namely in 
an agent-demoting construction. The accusative construction is only 
partially productive because the possibility of marking the experiencer 
with the Dative Case is highly restricted on semantic and pragmatic 
grounds.

Case assignment is performed via the linking constructions that are 
inherited by the constructions, or by the head predicate. Constructions 
evoke the interpretive frames, which motivates the semantic properties in 
the external domain, in particular the inheritance of linking 
constructions. In Czech, transitive patients are linked to the accusative 
case and the dative case is assigned to not fully or directly affected 
endpoints. The "dative" assignment involved in the construction described 
is represented as the Dative-of-Interest linking construction; it is via 
the inheritance of this construction and a Locative construction, and via 
the integration of the semantic information of both the head predicate and 
the interpretive frame evoked by the construction, that the Dative-
Experiencer Construction correctly assign case roles to its arguments.

In the accusative constructions something different takes place: the 
Accusative Linking Construction is inherited in the internal domain, 
through the predicate, and therefore it licenses the patient to be linked 
to the accusative case. Still, the patient argument is construed as an 
experiencer because it is an external property of the construction. The 
unification is less straightforward, but it is nevertheless successful 
because there is no other candidate argument; however, in special 
situations or contexts there is the possibility for the speaker to give 
priority to the external properties of the construction, thus overwriting 
the predicate requirements and assigning dative case.

The difference observed between the "regular" dative-experiencer 
construction and the "idiomatic" accusative construction, in the end, is 
shown to depend on where the relevant linking constructions comes from: 
the external or internal domain.  Externally idiomatic expressions may, 
therefore, have a predictable internal organization, when one acknowledges 
that case marking in Czech is sensitive to the valence of the predicate. 

Chapter 4: Fujii "Lexically filled constructional schemes and types: 
Japanese modal conditional constructions" presents an interesting corpus-
based analysis and representation of a class of conditional constructions 
in Japanese. In order to give a unified account for the data the author 
posits two kinds of orthogonal constructions: Constructional Schemes and 
Construction Types, which capture all relevant generalizations over a set 
of similar expressions manifesting different degrees of idiosyncrasy.

Constructional Schemes are lexically unfilled templates that can be 
instantiated by particular lexical items (in this case by clause linkers). 
They represent generalizations over distinct construction types that share 
some semantic properties. Construction Types represent the specific formal 
properties of the constructions, along with the relevant restrictions, and 
therefore explicitly keep each class of construct distinct from the others.

The linguistic objects under investigation are three deontic modal 
conditional constructions, represented as three distinct Construction 
Types: a fully regular and productive bi-clausal construction, a more 
restricted constructions called Integrated Evaluative Conditional, and a 
fully idiosyncratic Reduced Construction. All of them may receive an 
interpretation of obligation, and all of them have the same structural 
first part (the conditional antecedent). The three constructions are also 
different with respect to restrictions on their form and degrees of 
productivity. One important difference lies in the way they get the 
relevant pragmatic interpretation: while both the Integrated and the 
Reduced Constructions are demonstrated to be conventionally associated 
with a particular modality meaning (depending on the clause linker), the 
productive bi-clausal construction may only receive that interpretation 
through conversational implicature. There is also a difference in how the 
semantic interpretation is obtained between the Integrated and the Reduced 
constructions: in the former the pragmatic function is claimed to be 
achieved compositionally, whereas the latter is completely idiosyncratic. 
Thus, the Reduced Construction is considered to be a special case of the 
Integrated Construction, and a conventionalization of the conversational 
implicature implicit in the bi-clausal construction. However, according to 
the author the constructional scheme, not the bi-clausal construction, is 
the source of the Reduced Construction, because it is the specific 
pragmatic meaning that uniquely links all construction types together.

Various clause linkers may be involved in the different constructions, and 
it is observed that different clause linkers are consistently associated 
with one and the same modal interpretation across the three constructions. 
This facts suggest that clause linkers are more appropriately associated 
with Construction Schemes, and thus provide the main motivation for their 
existence. 

The main theoretical point of the paper is that, positing these two types 
of constructions, the model becomes more economical: constructional 
schemes account for the differences in meaning and will have different 
associated clause linkers, whereas construction types account for the 
structural dissimilarities between constructs, which are independent from 
the particular modal interpretation. A Constructional Scheme together with 
its related Construction Types, constitute a family of constructions. 

Chapter 5: K. Lambrecht "Interaction of Information Structure and formal 
structure. French Right Detached 'comme'-N construction" analyses a very 
common, yet unaccounted for by traditional grammars and dictionaries, 
French spoken grammatical construction that belongs to the family of Right-
Detached constructions. The analysis demonstrates that the form of the 
construction directly reflects its information-structure requirements, 
while at the same time being formally, semantically and pragmatically 
motivated because its relevant features occur elsewhere in the grammar of 
French. Nevertheless, the combination of elements in the construction is 
shown to be not predictable from general principles of French grammar. 

The Right Detached 'comme'-N construction is a special kind of a general 
and basic French construction, the Preferred-Clause construction, from 
which it inherits its main characteristics: i.e. the predicate-focus 
information structure. In particular, The Right Detached 'comme'-N 
construction inherits features also from a general dislocation 
construction in French (the Right Topic construction). The Right-
Detached 'Comme'-N construction splits the complement of a standard copula 
construction (a kind of Preferred-Clause construction) leaving the 
adjective modifier of the Noun Phrase in it post-verbal position and 
dislocating the Noun to the right, i.e. after the clause boundary that is 
marked by the main sentence accent (ex. 'C'est marrant, comme histoire' 
vs. 'C'est une histoire marrante'). Sentence accent in French is used also 
to mark the focus domain, and topicalization cannot be marked through 
deaccentuation, which motivates the existence of dislocation constructions 
in French, that is as a structural means to topicalize an element the is 
within the focus domain. 

The Right-Topic construction and the Right Detached 'comme'-N 
construction, however, are two distinct constructions in that the detached 
constituent of the first one have a referential function, because it 
provides a referent for the bound pronoun that has to be present in the 
main clause. The detached element of the 'comme'-N construction, instead, 
is not syntactically or semantically related to the main sentence - no 
bounded pronoun is present in the main clause- nor it has a referential 
function. The detached element of the 'comme'-N construction has the 
pragmatic role of specifying the category of the referent of the subject 
of the main clause. Other restrictions apply, which make the construction 
both syntactically and semantically non compositional: The noun in 
the 'comme' phrase, for example, is necessarily a predicative, bare noun 
and cannot be freely modified; and the 'comme'-Noun phrase does not have 
the meaning it has elsewhere in the grammar, so that it must be considered 
a construction specific constituent, where 'comme' functions like a 
copula. The truth-conditional meaning of the construction, however, is 
equivalent to the standard copula construction, an instance of the 
Preferred- Clause construction, which is another proof of the fact that 
the physical appearance of the construction as a whole reflects its 
information structure.

The information structural difference between a canonical copula 
construction and its equivalent 'comme' construction are demonstrated to 
depend on the different scope of the focus domain. In the canonical 
construction the entire denotation of the complement noun phrase is 
focused, whereas in the Right-Detached 'Comme'-N construction it is only 
the property predicated of the noun that is focused. Spoken French 
idiosyncratically mark this special information structure with a special 
type of dislocation, which basically tells the hearer that the category of 
the entity predicated is presupposed even if not previously introduced. 

Lambrecht analysis relies on two main assumptions: first of all it is 
assumed that pragmatic, information-structure features may be associated 
to constructions in the same way as semantic features, in order to 
restrict the possible discourse contexts in which they are used. Moreover, 
there must exist an information structure component in which these 
associations are formalized and that interacts directly with the formal 
and conceptual components to give rise to unique form-function pairings. 
In the case presented in the chapter, information structure is shown to 
directly determine the syntactic and prosodic shape of the whole 
construction.

CRITICAL EVALUATION

The book as a whole is an interesting introduction to Construction Grammar 
as a formal theory of language, and shows that the framework is 
sufficiently flexible and accurate to account for linguistic facts across 
languages. The volume, and in particular Chapter 2, will thus be an 
extremely useful reference for anyone interested in constructional 
approaches, especially since no textbook is publicly available yet 
(Fillmore and Kay's manual being still in manuscript form).

The three chapters dedicated to the study of special expressions in three 
different and unrelated languages demonstrate convincingly that 
construction grammar notions and formalism are useful in the analysis and 
representation of subtle facts about languages different from English, for 
many years the preferred language for constructional approaches. They show 
that a constructional approach is even more desirable than traditional 
approaches for other languages, because it allows for a uniform treatment 
of productive, less productive and highly idiosyncratic constructs, 
therefore capturing the generalizations that hold among them more 
accurately. The treatment of Case Marking, the construal of events in 
Chapter 3 and the investigation of the role of information structure on 
the shaping of constructions in Chapter 5 are of great interest.

Although the overall impression of the volume is highly positive, there 
are a few items that appear to be contradictory or unnecessary. 
Construction Grammar, by definition, is non-modular and monostratal; yet 
Fried's account requires different layers of semantic information: at 
least a lexical-level semantics, that would encode frame semantic 
information of single lexical items, and a clause-level semantics, that 
would encode constructional meaning. Moreover a semantic component is 
mentioned (chapter 3: 89). Given the rest of the discussion and analysis, 
I take this to be a terminological problem rather than a conceptual 
inconsistency, which might nevertheless be confusing. This possibly 
apparent contradiction is also present in Lambrecht's account, which 
posits an "information-structure component" that directly interacts with 
the other component of formal and conceptual structure. Again because this 
information is represented and formalized as constructions, I consider it 
to be a terminological ambiguity.

Fujii's analysis of the Japanese Conditional Constructions posits two 
types of constructions, Construction Schemes and Construction Types, as 
fundamental formal notions that make it possible to capture the relevant 
functional generalizations over different constructions. In my 
understanding of the general framework, constructions by definition can be 
more or less abstract, so that the distinction between construction kinds 
seems unnecessary, albeit at the terminological level for the sake of 
clarity. Constructional Schemes can be viewed as highly abstract 
constructions (by virtue of their associated pragmatic function) from 
which the more specific constructions will inherit the shared properties.

REFERENCES

Fillmore, Charles J. (1968) The case for case. In Bach, Emmon and Robert 
T. Harms (eds.): Universals in Linguistic Theory. New York: Holt, Rinehart 
& Winston. 1-88. 

Kay, Paul (1995) Construction Grammar. In J. Verschueren, J-O. Östman and 
J. Blommaert (eds.), Handbook of Pragmatics. Manual, 171-177. Amsterdam & 
Philadelphia: John Benjamins.

Lakoff, George (1977) Linguistic Gestalts. Chicago Linguistic Society 13, 
236-287. 

ABOUT THE REVIEWER

Valeria Quochi is a 3rd year PhD student in Linguistics at the University 
of Pisa, Italy. Her first degree is roughly equivalent to a major degree 
in English and German. She is interested in Computational Linguistics, and 
in particular in data-driven, cognitive approaches to language. Currently 
she is working on language acquisition of semi-productive constructions in 
Italian, with a constructional approach.





-----------------------------------------------------------
LINGUIST List: Vol-16-2277