Schönefeld, Doris (2001) Where Lexicon and Syntax Meet. Mouton de Gruyter,
vi+332pp, hardback ISBN: 3-11-017048-5, Trends in Linguistics 135.
David Parkinson, Natural Language Group, Microsoft Corporation.
OVERVIEW
No serious theory of the structure of the lexicon can afford to do without
at least some theory of syntax, just as no serious theory of syntax can
afford to ignore the lexicon. This is for the simple reason that the lexicon
serves the function in most linguistic theories of providing the raw
materials from which all syntactic and semantic structure is formed. Once
one begins to investigate the goals, claims, and theoretical devices of
competing theories, it becomes obvious that there are many ways in which the
division of labor between the lexicon and the syntax can be articulated. And
once one's scope is broadened to take in psycholinguistic data, empirical
questions about how lexical information is deployed in syntactic structure
in real-time linguistic events can be used to inform an evaluation of the
claims of competing theories concerning the type of information stored
lexically, and how this information is embedded into syntactic structure.
In Where Lexicon and Syntax Meet (henceforth WLSM), Schönefeld aims at
nothing less than a comprehensive evaluation of a wide range of theories of
syntax and the lexicon. Her aim is to formulate a competitive analysis of
these different theories against data from psycholinguistic findings, of
which she herself provides some, in the form of a corpus analysis of speech
errors and overlaps, as well as an experiment to evaluate priming effects of
collocational and idiomatic material. Along the way, she provides a useful
overview of most of the preoccupations of contemporary theories of the
lexicon and of syntax. The likeliest audience for this book is a
formally-inclined syntactician or lexicographer eager to know more about
theories of language processing, and how these theories might help constrain
a formal theory; or a psycholinguist looking for a good overview of formal
theories of the lexicon and syntax. WLSM is structured as follows:
Chapter 1: Introduction
In the brief introduction, Schönefeld sets out the overall aim of WLSM: to
identify the best theory of the interface between the lexicon and syntactic
components, where "best" is arrived at primarily via evaluation against
psycholinguistic data. In particular, she sets out the two main questions
(WLSM, p.3):
1. In what way are the selected linguistic models compatible with
psycholinguistic assumptions about the lexicon-syntax interaction in
language use?
2. How can the performance data I concentrated on, namely self-repairs,
overlaps and lexical patterns, be explained by the linguistic models
under analysis?
Chapter 2: Grounding and definitions
The aim of Chapter 2 is to provide the reader with Schönefeld's working
definitions of the key terms "lexicon" (or "mental lexicon") and "syntax".
She provides the following as a definition of syntax:
"Syntax describes the rules by which words combine in a verbal utterance,
what their contribution to the utterance meaning is, and the means by
which the intended combinations are signaled or expressed." (WLSM, p. 13)
It is unfortunate that Schönefeld's assumptions about the lexicon are not
spelled out so explicitly in this chapter; after a brief survey of the sorts
of things that the lexicon is supposed to contain, and the kinds of
functions it is supposed to perform, she points forward to Chapter 3, rather
than take a strong stance on her perception of what is intended by the term
"(mental) lexicon".
Chapter 3: Theories of language processing
This chapter consists of a lengthy discussion of language processing from
the standpoints of comprehension and production, naturally with an eye to
further refining our understanding of the interface between the two domains
of linguistic behavior. Among the issues which drive the discussion are:
a. Temporal structure of linguistic processing; i.e., whether serial vs.
parallel;
b. The nature of the flow of information; i.e., whether and to what extent
the lexicon and syntactic processing subsystems are modular (insulated
from one another except at tight interfaces) vs. intercooperative
(relatively open interfaces);
c. Heterogeneity of processing; i.e., whether and to what extent the same
processing is used in both comprehension and production;
d. Empirical support; i.e., how evidence from psycholinguistics can be used
to distinguish among the various theoretical claims made about the
subsystems supporting linguistic comprehension and production.
The high-level structure of this chapter being a division between production
models and comprehension models, Schönefeld ends each half of the chapter
with a set of assumptions that she takes away from the survey of competing
accounts: there are 9 assumptions regarding production & 10 regarding
comprehension. Overall Schönefeld's bias is towards a lexicon-dominated
interface, where information from lexical items takes precedence over
syntactic principles in construction of linguistic structure in real time.
Because this discussion is entirely drawn from psycholinguistic
experimentation, effects of frequency, context, priming, and so on, are key
contributors to the lexical-syntactic interface that Schönefeld is arguing
for; any reader who is unable or unwilling to put data from language use
ahead of purely formal argumentation might find this chapter tough going.
Nonetheless, it is the true theoretical center of the monograph.
Chapter 4: Linguistic models under scrutiny
Chapter 4 is the longest chapter in the WLSM, but I find it to be less
critical to Schönefeld's argumentation than the preceding chapter. In it,
Schönefeld surveys a reasonably broad range of linguistic theories in order
to discuss how they conceptualize the lexicon, syntax, and the interface
between the two. The high-level structure of the survey is a three-way split
between functional, generative, and cognitive approaches. The two main
functionally-oriented theories under consideration are Dik's functional
linguistics and Halliday's systemic linguistics. Generative approaches are
mainly represented by Chomskyan generative linguistics (Government &
Binding; Principles & Parameters), Lexical Functional Grammar, Diehl's
Lexical-Generative Grammar, and Head-Driven Phrase Structure Grammar.
Finally, cognitive approaches are represented by Deane's cognitive syntax,
Goldberg's Construction Grammar, and Langacker's Cognitive Grammar.
Schönefeld reaches the tentative conclusion that Langacker's theory provides
the most natural theoretical framework within which to account for the sorts
of performance facts discussed in Chapter 3. Tellingly, this is the case to
no small extent because of the theoretical commitment of Langacker's
schema-based approach to a rejection of arbitrary boundaries between lexicon
and syntax. Within the CG framework there are fixed schemas acquired through
automatization; depending on the size and type of the units entering in some
schema, we may call it a morpheme, collocation, syntactic idiom,
construction, word order generalization, or any number of other theoretical
entity. From this it follows that the commonalities between "lexicon" and
"syntax" are susceptible of being much more numerous than in many other
theories under discussion, in which a more or less rigorous -- if artificial
- division of labor is maintained between the lexicon as warehouse of raw
material and syntax as construction methods.
Chapter 5: Performance data
In the first of two chapters in which Schönefeld interprets the theoretical
background in light of psycholinguistic information, she uses a naturalistic
corpus-based approach, examining three types of data for clues to the
processes underlying them: reformulations/self-repairs in speech, overlaps
in conversation, and lexical co-occurrences. The general methodology is to
use these data as indicators of the extent to which lexical choices are
determined or constrained by syntactic structure (and therefore must be
accommodated to syntax when a disruption of one kind or another takes
place), or conversely of the extent to which syntactic structure is
determined or constrained by lexical choices. Although the data admit a
range of interpretations, Schönefeld argues throughout that they suggest a
lexicon-syntax interface which is more strongly constrained by lexical
choices, and that syntax is more often accommodated to the demands of some
lexical items than vice versa. Furthermore, she adduces some evidence
suggesting that the boundary between lexicon and syntax is somewhat porous,
since some linguistic units with syntactic characteristics (collocations,
co-occurrences, idiomatic constructions) appear to pattern in some respects
with common-garden lexical units like simple words.
Chapter 6: In the psycholinguist's laboratory
In Chapter 6, Schönefeld presents the results of an experiment designed to
further investigate the special characteristics of collocations under
controlled circumstances. More specifically, her experiment uses a lexical
decision task to measure the effect of repetition of collocations and
non-collocational but syntactically comparable phrases. The result of this
experiment suggests that collocations such as "seal X's fate" are processed
differently from syntactically productive phrases like "drink X's milk".
Here, the discussion of results is quite interesting and provocative, and
Schönefeld rightly indicates the need for further experimentation to answer
many questions left unanswered. Again, the results are used to argue for a
lexically dominated interface, with frequency and co-occurrence information
having a much greater importance than in many standard theories, in which
these effects are relegated to "mere performance".
Chapter 7: The finale
In this final chapter, Schönefeld ties together all previous discussion from
the psycholinguistic literature, from linguistic theory, and from her own
investigation and experimentation, to reach the conclusion that Langacker's
theory of Cognitive Grammar is the most psychologically plausible linguistic
theory of the interface between lexicon and syntax.
DISCUSSION
Schönefeld has undertaken a notoriously difficult task: that of using
psycholinguistic data as a way of putting various linguistic theories to the
test. At least in North American linguistic circles, there is a
long-standing reluctance to allow one's theory to become too beholden to
facts from the laboratory, or (worse yet) from some corpus of linguistic
behavior. So in this sense, Schönefeld may find herself preaching to the
choir: her likeliest audience is to be the psycholinguist looking for some
linguistic theory to use as a framework to which one's experiments can make
a contribution, or the adherent of Cognitive Grammar or some closely-related
linguistic theory, looking for some empirical justification of her/his
theoretical predilections. And there is simply not enough strong and
compelling data or argumentation in WLSM to accomplish the greater task of
compelling the generativist or theoretically-disinclined psycholinguistic to
take notice. It is to be hoped that Schönefeld and others will pursue the
line of experimental investigation that she opens up. Nonetheless,
Schönefeld is to be applauded for making a brave effort where so many have
failed, and for being unafraid to build bridges between formal linguistic
theory and psycholinguistics. Anyone who is interested in the leaky areas
between lexicon and syntax will find something of interest in WLSM.
ABOUT THE REVIEWER
David Parkinson is a computational syntactician in the Natural Language
Group at the Microsoft Corporation.
|
