Review of The Lexical Basis of Sentence Processing
Date: Fri, 20 Dec 2002 12:17:04 +0000
From: C A Ankerstein
Subject: Merlo & Stevenson (2002) Lexical Basis of Sentence Processing
Merlo, P. and Stevenson, S., ed. (2002). The Lexical Basis of Sentence
Processing: Formal, Computational and Experimental Issues. John Benjamins,
hardback ISBN 1588111563 (US) 9027249873 (EUR), viii + 363pp, USD 100.00 /
EUR 110.00, Natural Language Processing Volume 4.
Carrie Ankerstein, Department of Human Communication Sciences, University of
This book contains a selection of the papers given by participants of the 11th
Annual CUNY Conference on Human Sentence Processing, March 1998 entitled "The
Lexical Basis of Sentence Processing: Formal and Computational Issues".
The primary motivation of the conference was to gain a deeper understanding of
the lexicon and its impact on processing by elaborating on the structure and
probabilistic content of lexical representations, which together have an
influence on sentence processing. The multi-disciplinary nature of the
conference is reflected in the resulting collection of papers, including
perspectives from psycholinguistics, computational and theoretical linguistics.
The body of the book is divided into 15 chapters, which are organized into
three parts. Part I outlines the fundamental issues surrounding the lexicon and
sentence processing, the relationship between syntax and the lexicon are
discussed in Part II and in Part III the details of lexical entries are
The introductory chapter by editors Suzanne Stevenson and Paola Merlo entitled
"Words, numbers and all that: the lexicon in sentence understanding" is
intended to assimilate the various viewpoints of the disciplines contributing
to the volume. They give an introduction to the research that has been done in
various fields on sentence processing. They also outline and draw similarities
between the following chapters and discuss a lot of the issues that pervade
this area of research.
Part I: Fundamental Issues
Chapter 2: "The lexicon in Optimality Theory" by Joan Bresnan. In Optimality
Theory (OT) the lexicon is the result of syntactic variation. She discusses the
dependency of the content of the lexicon on syntax with the example of dialect
variation in the inversion of the copula in negation, e.g. aren't I vs. *I
aren't (Standard English) and Amn't I vs. *I amn't (Scots), arguing that this
arises not out of a simple lexical preference, but derives from the grammatical
properties of the dialect.
Chapter 3: "Optimality-theoretic Lexical Functional Grammar" by Mark Johnson.
This chapter discusses the impact of OT on the "classical" Lexical Functional
Grammar and focuses on the relevance of Bresnan's proposal for sentence
processing. The discussion of linguistic universals and markedness in terms of
optimalisation fits well with the probabilistic language models of
Chapter 4: "The lexicon and the laundromat" by Jerry Fodor. Fodor discusses the
content of the lexicon: "nothing belongs to a lexical entry for a lexical item
except what that item contributes to the grammatical representation of its
hosts". His discussion of the lexicon is guided by two principles:
compositionality and reverse compositionality. Fodor discusses the impact of
compositionality on frequency: frequency information, for example, cannot be a
property of a lexical entry because that would violate reverse compositionality
which states: the grammar of the constituents is exhausted by what they
contribute to the context.
Chapter 5: "Semantics in the spin cycle: competence and performance criteria
for the creation of lexical entries" by Amy Weinberg. Weinberg's chapter
reflects on Fodor's reverse compositionality criterion and the apparent
conflict with probabilistic models of language processing, which by nature are
based on frequency.
Chapter 6: "Connectionist and symbolist sentence processing" by Mark Steedman.
Connectionist models have been built to successfully simulate syntactic
parsing. Steedman argues that the relevance of connectionist models in language
processing lies in their predictive approach to processing, thus putting
emphasis again on frequency of associations in the lexicon.
Part II: Division of labour between syntax and the lexicon
Chapter 7: "A computational model of the grammatical aspects of word
recognition as supertagging" by Albert E. Kim, Bangalore Srinivas and John C.
Trueswell. In their chapter, Kim et al. discuss a model for disambiguation
based on a constraint-based theory of sentence processing and argue that much
of the syntactic ambiguity of language can be understood as lexical ambiguity,
which is resolved during word recognition.
Chapter 8: "Incrementality and the lexicon" by Vincenzo Lombardo and Patrick
Sturt. Lombardo and Sturt's chapter discusses some of the consequences of
strong incremental parsing. If parsing is incremental, the parser needs to be
able to integrate a new word into the structure via a connection path, thus
syntactic structure is created through the links between the individual lexical
items. Since the nodes of these links may not be fully analyzed, the
projections may be headless projections. Lombardo and Sturt discuss two
potential problems of headless projections in connection paths: what knowledge
is needed beyond traditional lexical projection and what is the extent of
headless projections in the connection paths.
Chapter 9: "Modular architectures and statistical mechanisms: the case from
lexical category disambiguation" by Matthew W. Crocker and Steffan Corley. In
their paper, Crocker and Corley review a modular, statistical model of human
lexical category disambiguation (Corley and Crocker 2000). Lexical frequency is
an important factor in their model, but unlike Kim et al. they argue that
lexical and syntactic processing are separate processes. They present empirical
evidence to support this.
Chapter 10: "Encoding and storage in working memory during sentence
comprehension" by Laurie A. Stowe, Rienk G. Withaar, Albertus A. Wijers, Cees
A.J. Broere and Anne M.J. Paans. Stowe et al. present brain imaging studies
that show the importance of three different areas in sentence processing. More
specifically, there are different areas of activation for the encoding of
lexical information, storage of lexical and phrasal information in memory, and
the processing of the syntactic structures, thus suggesting different areas for
storage and processing. They discuss the complex relation between the lexicon
and syntax in this regard and the impact of these findings on models of
Chapter 11: "The time course of information integration in sentence processing"
by Michael J. Spivey, Stanka A. Fitneva, Whitney Tabor and Sameer Ajmani.
Spivey et al. challenge previous papers in Part II and argue for an interactive
non-modular constraint-based model of sentence processing.
Part III: Details of lexical entries
Chapter 12: "The lexical source of unexpressed participants and their role in
sentence and discourse understanding" by Gail Mauner, Jean-Pierre Koenig,
Alissa Melinger and Breton Bienvenue. Mauner et al. discuss the role of syntax
in the lexicon regarding the content of lexical entries. Supported with
evidence from experiments involving the processing of agentless passives, they
explore the idea that unexpressed arguments are used in sentence processing
immediately at the verb as part of the argument structure of the verb.
Chapter 13: "Reduced relatives judged hard require constraint-based analyses"
by Hana Filip, Michael K. Tanenhaus, Gregory N. Carlson, Paul D. Allopenna and
Joshua Blatt. Filip et al. further explore the idea that the processing
difficulty that arises in reduced relative clauses is a result of the inherent
lexical semantic class of the verbs used as passive participles in reduced
relatives. They claim that the semantic variables play a much greater role in
the differences than syntactic complexity.
Chapter 14: "Predicting thematic role assignments in context" by Gerry T.M.
Altmann. Similar to the previous chapter, Altmann also develops a proposal that
involves finer-grained semantic information to define thematic roles with an
emphasis on probabilistic use of this information. Altmann presents
experimental data in support of this and discusses the implications within a
Chapter 15: "Lexical semantics as a basis for argument structure frequency
biases" by Vered Argaman and Neal J. Pearlmutter. Argaman and Pearlmutter
extend the lexical semantic theories of argument structure of Pinker (1989) and
Levin (1993) and propose that differences in argument structure biases are a
function of semantics. They present survey and corpus data in support of this.
They conclude with some further directions for the study of frequency effects
in language processing.
Chapter 16: "Verb sense and verb subcategorization probabilities" by Douglas
Roland and Daniel Jurafsky. Roland and Jurafsky argue for the "Lemma Argument
Probability hypothesis; a proposal that a separate set of subcategorization
probabilities found are associated with each sense of a word in the mental
lexicon". They provide corpus data to support this proposal for verbs, but
argue that this is also the case for adjectives and nouns. Their findings
converge with those of Argaman and Pearlmutter.
This volume provides a well-rounded and fair discussion of the issues at hand -
i.e. there is no bias for one side or the other, e.g. regarding the
relationship between syntax and the lexicon or the structure and content of the
lexicon. There are good links between chapters - the chapters in each section
generally acknowledge each other and integrate the issues surrounding the
lexicon discussed into the larger framework of processing, though this is more
prevalent in Part I.
This volume is probably not intended for beginners - some papers require a more
advanced knowledge of linguistics, computational linguistics and psychology.
Some of the papers may seem overwhelming in that they discuss similar issues,
e.g. frequency in the lexicon, from different perspectives and with very
different views, however, the introductory chapter by the editors should serve
as a good tool to integrate the issues and the views presented in this volume.
In general this is a fantastic discussion of issues in sentence processing such
as: what is the relationship between the lexicon and syntax: does one
constitute part of the other; are they separate processes; and what information
does the lexicon contain.
Corely, S. & Crocker, M. W. (2000). The Modular Statistical Hypothesis:
Exploring Lexical Category Ambiguity. In: Crocker, Pickering & Clifton (Eds.)
Architectures and Mechanisms for Language Processing, pp. 135-160. Cambridge
University Press: Cambridge.
Levin, B. (1993). English verb classes and alternations: a preliminary
investigation. Chicago, IL: University of Chicago Press.
Pinker, S. (1989). Learnability and Cognition: the acquisition of argument
structure. Cambridge, MA: MIT Press.
| ABOUT THE REVIEWER:
ABOUT THE REVIEWER Carrie Ankerstein is a PhD student in the department of Human Communication Sciences at the University of Sheffield, England. She has a Masters in Applied Linguistics from the University of Cambridge, England and a Bachelor's degree in German Linguistics from the University of Wisconsin-Madison, USA / University of Freiburg, Germany. Her research interests include the organization and representation of concepts in semantic memory and the relation of semantics and the lexicon in language processing.