Cienki, Alan, Barbara J. Luka, and Michael B. Smith, ed. (2001)
Conceptual and Discourse Factors in Linguistic Structure.
CSLI Publications, paperback ISBN 1-57586-258-1, xiv+276pp.
Reviewed by Henri-Jose Deulofeu, Universite de Provence,
France <[email protected]>
DESCRIPTION
In this descriptive part, I will basically rely on the
extended preface written by the editors, which gives a
comprehensive overview of the book. I should mention that
the book includes a very useful index of authors quoted and
linguistic topics dealt with. This index contains also a
list of descriptive tools, theoretical notions and even
principles (e.g. discourse is motion along a path) used in
cognitive grammar, so that the nonspecialist but careful
reader can follow easily and finish the book perfectly
trained in the cognitive grammar approach.
Some editing mistakes should be noted,
e.g. 'it' instead of 'if', p. 2, l. 7; the ill formed
sequence 'atrribute the meaning to a the new target' p.
128, l. 8. More puzzling are the comments on some
figures (figure 1, p. 124) in which distinctions are made
between different types of lines (bold, dotted)which do not
appear in the figure itself. The presentation by Achard (p.
12) of French sentential complementation is obscured by the
fact that he refers to a missing figure 2. Figure 2,
which appears two pages later, is not about verb
complementation, but about reference point.
As for its content, this volume stems from the fourth
conference on the theme of Conceptual Structure, Discourse,
and Language (CSDL), which was held at Emory University in
Atlanta in October, 1998. The chapters of this volume are
based on the papers presented at this conference.
The common background can be summed up in two assumptions:
a) there is an integral relationship between meaning and
the form in which it is expressed in language.
b) linguists must give serious consideration to both
conceptual and discourse factors in order to achieve a more
complete account of linguistic structure
The works nevertheless reflect multiple perspectives, and
represent further progress in the dialog between cognitive
and functional approaches in linguistics. The range of
topics represented in this volume can be broadly subsumed
by four general topics:
1. the relationship of lexical and grammatical meaning;
2. metaphor and conceptual integration;
3. functional considerations; and
4. experimental and developmental approaches.
>From the point of view of data used,
if most of the papers rely on speaker's linguistic
intuition or on previous descriptions, part of them, and
not only those from experimental approaches (cf. Johnson),
use authentic data from electronic corpora. This new
methodological toll brings very convincing empirical
evidence to S�nchez and Tamer G. Amin' s papers.
The first set of papers demonstrates how lexical form and
grammatical form can be viewed as two points on one
continuum, with each representing different ways of
packaging meaning in language. Langacker tackles the
difficult case of analysing the meaning of the English wh-
formative, the basis of both question words and relative
pronouns (what, when, etc.). At the core of the analysis is
the cognitive notion of establishing mental contact with
one entity from a range of alternatives in the current
shared -(mental) discourse space specifically on the basis
of participation in a process (profiled by the clause where
the wh- functions). His account ultimately sheds light on
the parallelism between wh- forms and demonstrative th-
forms (that, then, etc.), the latter not crucially implying
process participation, and has broader implications for the
analysis of interrogative and relative clauses cross-
linguistically. A noteworthy new and stimulating approach
of headless relatives is proposed (p. 147). Achard, using
data from French, challenges the belief "that the syntactic
behaviour of raising verbs can be directly ascribed to the
specific underlying structures in which they occur".
Rather, he argues that the syntactic differences inherent
in the raised versus unraised verb constructions are not
merely structural, but are also a matter of cognitive
construal (more objective versus more subjective
construal), and thus he provides a motivated explanation
for the syntactic differences without appealing to
underlying formal levels of representation. Pustet's
typological study investigates the factors which determine
whether lexical items may or may not combine with copulas
and finds that the relevant criteria do not necessarily
coincide with the dividing lines imposed by the traditional
classification of nouns, verbs, and adjectives. Rather they
correspond to a number of semantic classes which cut across
the traditional parts of speech, but nevertheless reflect
the hierarchy of time-stability of the concepts
represented. Smith's paper examines the separability of
verb prefixes in German, and asks the question: Is there a
semantic motivation for prefix separation, or is the
separability versus non-separability of the prefixes
simply an arbitrary feature of a particular construction?
He argues for the former view, proposing a semantically-
based account for the word-order facts in German from the
perspective of cognitive grammar. Takahashi's paper
examines Talmy's notion of access paths (elsewhere known as
'fictive motion' or 'virtual motion') and contrasts an
analysis of them in English with how they are encoded in
Thai. Takahashi considers both the cognitive and functional
motivations for access path expressions, describing how
they are construed differently by speakers of Thai and of
English, providing a cross-linguistic comparison of the
semantic constraints on the form and the use of these
expressions.
A second subset of the papers focuses on semantics in
relation to processes of metaphorical mapping and
conceptual integration. These phenomena can provide
diachronic explanations of semantic change, as well as
explications of synchronic polysemy. Amin's paper provides
an account of the English-speaking layperson's idealized
cognitive model of thermal phenomena with a careful
analysis of English data. Theoretically, the work
represents a blend of approaches, beginning with Dowty's
system of verb classification to explore the semantics" of
heat and temperature, then moving to an exploration of
metaphors for heath...and concluding with an analysis of
how we make sense of seemingly contradictory understandings
of 'heat' through processes of conceptual integration.
Moore's paper addresses a refinement in our understanding
of the linguistic expression of temporal metaphor. Using
data from English, he proposes that what has previously
been analysed as a simple moving Time metaphor is actually
two metaphors, with the difference critically depending
upon recognizing the existence of both deictic and non
deictic expressions of time. He also considers additional
data from Wolof and Japanese which pro vide evidence that
deixis is not always relevant to the sequential relations
implicit in the metaphor.
De Haan's paper surveys languages with visual evidentials
and shows how evidentials typically do not derive from
vision words, as might at first be thought, but tend to
develop diachronically from demonstratives and tense/aspect
markers. The author then provides plausible cognitive
semantic motivations for this kind of grammaticalization
process, whereby the markers likely evolved from deictic
expressions, which in some way linked the speaker to the
event being witnessed visually. Bergen and Plauch�'s paper
presents an analysis of French constructions containing the
deictic locative expression voil� 'there is' and shows how
such constructions are related to each other semantically
as members of a radial complex conceptual category. The
various senses of the constructions are motivated by
metaphor, metonymy, constructional grounding, and other
cognitive mechanisms.
Research by a third subset of authors focuses on functional
considerations in explaining grammatical phenomena, and
highlights how factors involved in a discourse event play a
role in determining linguistic structure. Sanchez presents
empirical data which support the role of intonation units
as a means of coding semantic integration between the two
clausal constituents of complementation constructions in
spoken Spanish. Prosody, in terms of intonation units,
serves the dual function of signalling both the integration
of complex constructions and the expression of canonical
combinations of given and new information.
The papers by
Corrigan and Tao pay particular attention to the social
context of language use. Corrigan reveals that in
attributing causality in a verbal argument, we do not
necessarily assume that the person instigating an event
(usually coded as the grammatical subject) actually caused
the event to occur. Rather, there are features independent
of grammatical or thematic roles, such as contextual
factors concerning the social identities of event
participants, which play an important role when assigning
causality. Tao, in a cross-linguistic study of the switch-
reference pattern, proposes that grammatical coding of the
pattern, versus the lack of it, may reflect how much
inference speakers and listeners use in tracking
referential meaning. Therefore, the more grammatical coding
a language uses in its presentation of referential meaning,
the less its speakers may rely on inference in processing
discourse information, and vice versa: Elliptical reference
or zero anaphora, entails greater reliance on inference by
the listener.
Experimental and developmental evidence for certain kinds
of linguistic phenomena provides the basis for a fourth
group of papers. Hillert and Swinny address the question of
how idiomatical or literal meanings of compound nouns in
German are accessed during real-time sentence processing.
They present the results of two lexical priming experiments
which support the conclusion that, for German nominal
compounds, all meanings, both idiomatic and literal, are
simultaneously and exhaustively accessed when the nominal
compound is understood. Phrases with fixed meanings seem to
be processed in much the same way as lexical ambiguities,
blurring the distinction between syntactic, morphological,
and lexical representation. Johnson's research addresses
the processes children use to represent relations between
constructions, using one form-meaning pair to motivate the
acquisition of another related form in an overlapping
discourse context. He illustrates the process of
'constructional grounding' with deictic and existential
THERE-constructions. He proposes that children acquire the
more concrete deictic constructions first and then extend
these constructions to the more abstract existential uses.
He finds support for this proposal in corpora of children's
language usage. Papafragou provides conceptual motivation
and empirical support for the hypothesis that central
semantic and pragmatic aspects of language acquisition, in
particular, the acquisition of epistemic modals and
evidentials, presupposes certain advancements in children's
ability to attribute mental representations to themselves
and to others. Budwig's paper presents a developmental
functional approach in its investigation of the grammatical
acquisition of perspective in German and in English. Her
research shows that errors children make in marking agency
encode semantic and pragmatic aspects of scenes in
systematic ways that vary across languages: children begin
by using case markers to demarcate agency, and later use
voice to signify more subtle distinctions. This research
emphasizes that from the earliest stages of word
combination, children represent and express viewpoints and
perspectives, but that these expressions are sensitive to
linguistic forms, cultural cues, and social contexts.
CRITICAL EVALUATION
As a general positive appraisal, I share the editor's
comments when they say: "We believe the papers in this
volume demonstrate the need to incorporate both conceptual
and discourse factors in descriptions of linguistic form,
and that, as such, they represent a valuable step forward
in understanding human language". Nevertheless and perhaps
against the authors expectations, I want to point out that
I have been more convinced by the descriptive import of the
approach than by its explanatory ambitions. I share the
implicit belief that formal approaches of linguistic
phenomena have a tendency to narrow the scope of properties
observed to those which are crucial to model internal
theoretical issues. So, by focussing on the up to now
poorly explored domain of "natural" semantics and
pragmatics and on the specific properties of individual
constructions, these studies make a step forward in our
understanding of how lexicon is structured and grammatical
constructions interrelated: they reveal new properties of
linguistics units and urge us to explore new empirical
domains (discourse factors, sociolinguistic features of
speech situation), which extends our possibility of testing
analytical hypothesis. Papers by Tamer G. Amin, S�nchez,
Michael B. Smith are particularly representative of this
trend.
As for the explanatory import of the framework, and
focussing on syntax, which is my specialty, I would say
that it stands now as a valuable alternative to formal
frameworks, so that it opens, at least, the possibility of
stimulating comparisons between formal and motivated or
functional solutions. Nevertheless, the solutions to
syntactic problems proposed here will not always convince
formal syntacticians. The main problems that this approach
is to face are, from the point of view of a descriptivist,
the following:
a) problems of data
The progress of corpus linguistics urge us to think that a
linguistic descriptions should no more be based on mere
intuitions about acceptability of utterances. The
hypothesis should be checked against a blend of intuitions
and speaker's use as revealed by authentic corpora. I
already said that some papers use this methodological tool.
When pure intuition is used the reader may sometimes
disagree with the author on crucial data, and discussion
becomes impossible. Let's take the case of Achard's paper.
I basically agree with his semantic analysis, very
convincingly developed p. 10, with relevant examples (20)
and (21), but I disagree with plenty of the data used in
the discussions on the syntax of French clitic 'en', which
is supposed to be crucial for his analysis. Achard himself
points out that "as a matter of fact, several of my
consultants often questioned the validity of such
paradigmatic examples as (12b)" -- in fact the correct
reference seems to be (13a) -- 'l'auteur de ce livre
semble en �tre g�nial'. Nevertheless these unusual French
utterances are discussed at length, as well as other
questionable examples like (36) and (38). One is surprised
to find that example (36b), judged "pragmatically strange"
is marked by ??, whereas the "infelicitous" (38a and b) is
starred. There seems to be lack of consistency in the use
of the marks. Furthermore I definitely think that the
contrast between (30a) and (30b) would be impossible to
support by any kind of authentic data.
(30a) quand m�me, cette maison, le toit en est vraiment
bien ab�m�
(30b) ??quand m�me cette maison, le toit est vraiment bien
ab�m�
I don't want to say that my intuitions are better
than Achard's, I am just saying that the discrepancy of
intuitions on these examples show that there is a problem
in establishing reliable empirical data about 'en'
referring to subjects. This problem deserves a specific
study based on corpora sampled by registers. If one do so,
along the lines of Blanche Benveniste (90), it will appear
that in spontaneous French speech there is only one
productive use of en "en" referring to a quantifier in
direct object position: J'en vois deux. Some examples with
a small group of lexical items in idiomlike constructions
are found: 'j'en vois le bout, j'en connais pas la fin'.
Other uses appear only in written or oral formal registers.
This means that the use of 'en' in the crucial examples is
not acquired by all speakers of French, but learned at
school, so that the whole question deserves careful
sociolinguistic analysis. So what I am saying is that
Achard could have best let aside these problematic data for
further examination and concentrate on the problem of
Control versus SSR constructions of the verbs. Strangely
enough these uncertain data are the only empirical evidence
supporting the claim that "differences ... [in syntactic
behaviour] ... rather than being the output of structural
mechanisms ... naturally emerge out of the verbs' different
conceptual configurations ...". Other uncontroversial data
could have been used, such as the possible vs. impossible
substitution by clitic pronouns:
Paul l'a promis, d'aller � Paris / * Paul le semble, �tre
malade.
As well as the following contrast:
Ce que Paul avait promis c'est d'aider Marie
*ce que Paul semble c'est �tre malade
Those contrasts seem to me to stand as the real challenge
for Achard's analysis, as far as they show that in the
control case the sequence Vinf + object is a VP
constituent, whereas it is no constituent at all in the SSR
case. This seem to be a good argument for positing two
distinct syntactic structures, which challenges Achard's
analysis on this particular point.
b) methodological problems:
b1) synonymy and polysemy in syntactic constructions.
Some papers convincingly demonstrate that two apparently
synonymous formal variants are in fact distinct
constructions with different meanings (e. g. Michael B.
Smith on Separable Verb constructions in German). On the
contrary, the limits of syntactic polysemy is not so
clearly dealt with. I will take the example of voil�
constructions in French in Bergen and Plauch�'s paper. In
Figure 1, p. 59, the authors distinguish 8 different
constructions. This constructions of voil� are convincingly
distinguished by their meaning as extensions from the
radial category "central deictic". But these differences in
meaning reflect in syntactic properties only for tree
subcases: Event deictic, Central Time Deictic and Span of
Time Deictic. For a descriptive linguist defining
linguistic units on the basis of form meaning concomitant
variation, Figure 1 appears as mixing up cases of polysemy
of a unique construction and real cases of distinct voil�
constructions. And even the distinction between Central
Deictic and Event Deictic (p. 5O) appears to be a simple
case of subcategorisation of a Predicate–Object
construction: voil� Marie (CD) voil� que Marie part (ED).
On descriptive grounds, I would say that the only
differences in construction, if any, that I would accept is
between the constructs with the interpretation "the entity
[in complement position]... is in the perceptual realm of the
speaker" and those which do not imply this meaning. In this
case indeed, these semantic differences can be associated
with different syntactic properties: the latter are only
possible in a restricted subset of "small clause"
complements with non verbal predicates: (22) voil� mon
prof au labo, (23a) voil� mon prof content. The former is
associated with single complements: 'voil� Jean', 'voil�
que Jean vient', and "small clauses" with tensed verbal
predicates: voil� Jean qui arrive. With verbal predicates,
the infinitival form is impossible contrary to what is said
in the text: (16b) voil� partir Marie is for me a crucial
non grammatical example, best revealing the consequences of
reanalysis of voil� into a deictic predicate (compare the
perfect: je vois jean r�ussir l'examen, where the process
expressed by the plain verb 'voir' can be out of the
perceptual realm of the speakers). As for the "parangon"
construction posited in order to capture the special
"exclamative" meaning of "en voil� des linguistes" (how
good linguists these guys are), I remind the alternative
proposal by J.C. Milner (81) that there are not in French
any exclamative constructions (constructions with only an
exclamative reading). The exclamative reading, is a default
reading of various types of constructions when no
referential reading is contextually relevant. And it is the
case that the clitic doubling version of the central
deictic: "voil� des linguistes" can have a descriptive
"non parangon" reading: "L1 Cite-moi des linguistes. L2 Eh
bien en voil� des linguistes: Cienki, Achard, Sanchez...
b2) morphological and syntactic meaning in constructions.
As said before, Langacker study offers a good starting
point for capturing semantic generalisations in the
analysis of wh- words. But I think that the question of the
relation between the meaning of the morpheme and the
meaning of the syntactic constructions in which it is
involved should be more deeply investigated. As an example,
Langacker suggests that there is a semantic difference
between week demonstratives ('anaphoric' demonstratives)
and relative pronouns: the referent of week demonstrative
is "singled out independently of its role in the
proposition", the reference of the relative form is
constructed through this role. But there is a syntactic
context in which a relative functions and is interpreted as
a week demonstrative: the so called 'linking' relative
found in narratives (e.g. in Latin 'quibus actis, Caesar
hostem ingredit'). Quibus is equivalent to 'hiis rebus
actis...', with the only difference that quibus is strictly
logophoric. In this case, the main clause plays no role in
singling out the referent of the relative. My point is that
the 'week anaphoric meaning' for the relative is associated
with a specific syntactic context, namely when the relative
functions in a plain clause, without any 'movement' or
'left isolated' properties. Conversely, all the examples of
wh- uses given in the text with clause bound interpretation
are clear instances of 'movement type clauses'. So, one can
wonder what belongs to morphological meaning and what to
construction meaning in the proposed semantic analysis.
c) problems of argumentation and structure of the proof.
c1) use of outdated or incomplete alternative solutions as
basis of discussion.
As part of my personal experience, I know how difficult it
is to handle both formal and functional frameworks when
trying to evaluate a solution of a linguistic problem. Now,
if a linguist of one side doesn't want to unfairly simplify
the solutions of the other side, the solution I would
suggest is that he takes advice from a referee colleague
from the other side before publishing. Going back to
Achard's article as an example, I repeat that the coherence
of his semantic 'with only one syntactic structure'
solution is unquestionable, but what he doesn't really show
is that it is superior to an up to date 'two syntactic
structures' one. His only references are the current state
of GB grammar in the late seventies. I am not sure that it
would have been so easy to challenge a syntactic solution
using the notion of "flat" vs. "embedded" structures
(Emonds 2000) or, in other frameworks, infinitive
complement structure vs. infinitive modal structure
(Blanche-Benveniste 84).
c2) structure of the proof
In some papers, a possible alternative solution is not
taken into consideration even if it is strongly suggested
by the data. Sanchez argues for an hypothesis of "dynamic
nature of constituency in spoken discourse". One basic
argument is that the relationship of constituency between
two constituents can be modified by prosodic contours. The
relation is strengthened if the constituents are in the
same prosodic unit and loosened otherwise. However, the
detailed empirical study on complement structures leads to
the conclusion that the correlation is "always mediated by
the informational load of the construction". In fact the
discrepancy between actual and expected results should have
lead the author to reconsider the discarded hypothesis of
fixed constituent structure, modified as follows: what if
prosody instead of modifying constituency would add
structure to it? Within this alternative hypothesis, an
utterance will be formally analysed at two levels:
constituency and prosodic units. The resulting structures
will receive different pragmatic interpretation in terms of
topic and focus for instance. The principles found by
Sanchez could then be interpreted as flexible conditions
on topicalisation or focalisation of syntactic
constituents.
In such a framework, the two prosodic units
analysis of creo que fui or the one prosodic unit of 'creo
que le ha creado un trauma' both become structurally
possible, but require marked pragmatic contexts.
Independent evidence for this hypothesis of autonomous
prosodic structure can be found in cases when prosodic
contours can even built syntactic structure in absence of
relations of constituency, for instance in this French
attested 'nominativus pendens' utterance:
Le piano les doigts c'est tr�s difficile
As for piano the fingers'(technique) it is very difficult
d) theoretical problems
d1)Sometime too ambitious theoretical goals are not
supported by the data and lead the author to miss
interesting descriptive conclusions. Tao wants to answer
two extremely general question: Why does language present
referents with different grammatical structures? and Do
people use an equal amount of inference to track reference
in discourse processing? The first question is supposed to
be answered by a modified version of the Gricean principle
of competing motivations to reach both clarity and economy.
The second by the Inference continuum (the more explicit
grammatical form the less speakers rely on inference). But
from the presented evidence, based on authentic spoken
data, the conclusion that emerges for me is that economy
and clarity are not principles structuring grammars, but
rather types of discourses. What is indeed shown is that
whether your grammatical structure has explicit reference
tracking devices (switch reference languages) or not (zero
anaphora languages), a speaker can built clear discourses
as well as elliptic ones needing inference from the
listener to be interpreted. The degree of inference used
seem to depend less on grammatical structure than on the
type of discourse or register: reference determination
seems more important in formal styles and narratives than
in familiar conversation. And indeed the Chinese utterance
the interpretation of which requires maximum of inference:
'have enough have find out' (I have enough watermelon have
they found out the guns (from extended context))
can be exactly paralleled by the following French pair of
utterances possible in informal conversation: j'en ai assez
merci alors ils ont fini par trouver �a? (the guns), where
the tracking of reference of neutral pronouns 'en' dans
'�a' requires strong inferences. It is even possible to
imagine such contexts for switch reference languages:
I have met Peter and John, X Diff good guy
The grammar tells us that the good guy X is different from
'I'. But it is only by inference that the listener will
find out whether X is Peter, John or even a character from
extended context.
d2) The status of the "conceptualisations" are uncertain for
me. In some papers they seem to be language dependant. For
instance when Takahashi shows that conceptualisation
underlining Path constructions are different in Thai and
English. This kind of studies appear in keeping with a
Sapir-Whorf conception of linguistic structure. In other
papers, conceptualisations are closer to universal semantic
representations. Strangely enough, it emerges that
conceptualisations in the domain of time and space often
belong to the Sapir-Whorf paradigm, whereas
conceptualisation in the field of human feelings or
reactions to the reality seem to be universal. As if for
linguist the Cartesian notion of a universal thinking
subject, culture independent, was an uncontroversial
notion. Many anthropologist would be surprised in front of
a paradigm which relativizes time and space representations
but maintain a unified vision of subjectivity. After formal
Cartesian linguistics are we getting semantic Cartesian
linguistics?
CONCLUSION
Following Lazard (2000), I want to point out that general
hypothesis about the structure of language should be based
on very detailed language descriptions including data from
large authentic corpora. But I acknowledge that cross
language generalisations are at least very effective
heuristic tools for descriptions. So that it is a good
thing for linguistic analysis progress that both types of
investigations be lead jointly, as it is definitely done
and well done in this valuable book.
REFERENCES
Blanche-Benveniste C. (1990) Grammaire premi�re et
grammaire seconde: l'exemple de EN. Recherches sur le
fran�ais parl�, 10, 51-73, Publications Universit� de
Provence
Blanche-Benveniste C., Deulofeu J. et alii (1984) L'approche
pronominale, Selaf, Paris
Emonds J., (2000) Lexicon and grammar: the english
syntacticon, Mouton de Gruyter, New York
Milner J-C. (1978) De la syntaxe � l'interpr�tation,
Seuil, Paris
Lazard G. (2000) Que cherchent les chercheurs,
Bulletin de la Soci�t� de Linguistique de Paris, Tome XCV
1, Peeters, Paris
ABOUT THE REVIEWER
Henri-Jos� Deulofeu is Professor in French Linguistics at
Universit� de Provence (France). His fields of interest are
descriptive syntax, discourse and utterance syntax of
spoken French, corpus linguistics.
|
