| Review: |
Date: Sat, 27 Nov 2004 19:27:58 +0200
From: Galit W Sassoon
Subject: Fuzzy Grammar, a reader
EDITORS: Aarts, Bas; Denison, David; Keizer, Evelien; Popova, Gergana
TITLE: Fuzzy Grammar, a reader
PUBLISHER: Oxford University Press
YEAR: 2004
Galit W. Sassoon, Department of Linguistics, Tel-Aviv University
PURPOSE AND CONTENTS OF THE BOOK
This book is intended as a stimulating reading material for beginners and
specialists, who are interested in the question of whether grammatical
categories are fuzzy or discrete. The book consists of 28 sections,
presenting mainly classical texts, by authors from a range of disciplines
and perspectives. The introduction (pp. 1-28) highlights main points in
the texts, and in several newer papers which are not included in the
volume. It is followed by 5 thematic parts.
PART 1, "PHILOSOPHICAL BACKGROUND", consists of 4 very short sections (pp.
29-44) and a main section (pp. 45-64):
1) "CATEGORIES": In 2 sections from "The categories" (ARISTOTLE
350BC/1963) number and language are said to be discrete quantities, as
opposed to surfaces, time and place which are continuous; In 2 more
sections from "Metaphysics" (ARISTOTLE 350BC/1971) a contrast is observed:
On the one hand, everything can be either true or false (not both
simultaneously). On the other hand, everything is true (or false) to a
degree ("more truthlike" or "less in error").
2) "CONCEPTS" continues with FREGE's view of concepts (1903 p. 159) as
having sharp boundaries (by the law of excluded middle). The lack of sharp
boundaries in natural language "quasi-concepts" is considered a defect: It
doesn't allow us to decide whether certain things fall under them or not.
3) "VAGUENESS", RUSSELL's 1923 paper, defines Vagueness as a property of
representations (photos, maps, symbols etc.) A system of terms (connected
by relations) is a vague representation of another system iff there is no
1-1 relation between them, such that when terms of one system are
connected by a relation, the corresponding terms in the other system are
connected by the corresponding relation. For ex. a map is vague when
several courses of roads or rivers are compatible with each sign in it.
Similarly, Vagueness in language is a one-many meaning relation.
4) "FAMILY-RESEMBLANCES" (WITTGENSTEIN 1953, sections 66-78) puts forth
the idea that concepts are characterized by best examples and by family
resemblance between instances. Definitions and sharp boundaries do not
exist.
5) "THE PHENOMENA OF VAGUENESS" (ch. 1 in KEEFE 2000) reviews the space of
possible vagueness accounts: The pessimistic approach maps vague language
outside the realm of Logic; The epistemic approach maintains classical
logics: Every proposition, including ones with vague concepts, has a truth-
value, but we do not always know it; The degree approach uses a multi-
valued logic: That "Dan is tall" is true to some degree in the real
interval [0,1]. Finally, the supervaluation approach, which Keefe defends
in her book, maintains classical logic in "total-valuations", but allows
for a gap in "partial-valuations", e.g. Dan is neither tall nor non-tall
in any partial-valuation from which are accessible both a total-valuation
where Dan is tall and a total-valuation where Dan is not-tall. Keefe
describes a variety of phenomena that any valid vagueness theory has to
account for: Existing borderline-cases; The lack of sharp boundaries (the
mere possibility of existence of borderline cases); Higher order vagueness
(lack of sharp boundaries between P [or not-P] and the gap); The Sorites
paradox (roughly, is there a height at which one is 'suddenly' considered
tall) etc.
PART 2, "CATEGORIES IN COGNITION" (pp. 65-177), consist of 2 classical
texts (sections 6-7) which played a central role in the foundation of a
framework for cognitive experimental research of natural-concepts, and 3
texts by prominent Cognitive linguists (sections 8-10) who consider
linguistic concepts to be part of our overall conceptual system:
6) "THE BOUNDARIES OF WORDS AND THEIR MEANINGS" (LABOV 1973) shows that
words like 'cup' have no clear boundaries. The probability of naming a
container 'cup', 'bowl' or 'vase', depends on the container's width and
height, and crucially, on the context of its use (coffee, flowers, food).
Labov concludes that linguistic categories are NOT: discrete, invariant,
qualitatively distinct, conjunctively defined and composed of atomic
primes (P. 68).
7) "PRINCIPLES OF CATEGORIZATION" (ROSCH 1978) shows that in the
conceptual hierarchy, basic level concepts (like 'dog') are acquired
earlier than superordinate ('animal') or subordinate ('terrier') ones;
they are associated with more common properties, a motor plan, perceptual
image, and simpler word forms. Within a concept, the more prototypical
exemplars (e.g. 'robin' in 'bird') are acquired earlier than less typical
ones, they are categorized faster, remembered best, retrieved faster,
share more common properties and fewer properties of other categories, and
have simpler word-forms. Rosch concludes that a concept structure reflects
Cognitive Economy (maximal information in minimal processing effort) and
Perceived world structure (the properties that tend to co-occur).
8) "CATEGORIZATION, FUZZINESS AND FAMILY RESEMBLANCES": JACKENDOFF 1983
(ch. 5 and sections 7.3-7.4) views categorization as a comparison of the
inner structure of 2 relata. 'Types' (non-referring expressions like the
predicate in: "The leftmost man is a reporter") are represented by a
partial list of their 'tokens' (referring expressions, like the subject in
the above example) and tokens are represented by a partial list of their
types. The application of a concept in new cases is governed by rules.
Jackendoff criticizes the notion of 'extension' (a set of past, present,
future and 'possible' instances) as neither cognitively real nor
computationally working.
9) In "DISCRETENESS" (LANGACKER 1987, section 1.1.4) we find cases which
fall between dichotomies such as: grammatical/ non-grammatical (graded
judgments); grammar/ lexicon (we would like to generate words
like 'stapler' by productive patterns like V+er, but their meaning is much
richer than: "something that staples"); semantics/ pragmatics; literal/
non-literal meaning (there is a spectrum of possible connections between
meanings in a polysemy, for ex. between a 'ring' as a sound, a boxing
arena and a piece of jewelry) etc. Langacker proposes to represent full
knowledge about a category with many typicality properties and the
holistic connections between them (for ex. the timing of features like -
consonantal, +high etc. in uttering the vowel sound [i]). Typicality
degrees determine membership.
10) "THE IMPORTANCE OF CATEGORIZATION" (LAKOFF 1987 ch.1-2) is a
historical review: It goes back to Wittgenstein and Austin (who, it is
claimed, noticed for words the kind of things that Wittgenstein noticed
for concepts), continues with Cognitive Anthropologists such as Lounsbury,
Berlin, Kay, Brown and Eckman, who explored biological taxonomies, colors,
emotions and kinship concepts. They established the universality (and
neural basis) of the basic colors' prototypical examples and the non-
universality of the boundaries between colors. They used Zadeh's multi-
valued Fuzzy-Logic to represent the meaning of terms like 'red'. Finally,
Lakoff considers Rosch to have established the non-discrete, psychological
and embodied nature of concepts: Their relation to mental images, motor
interactions, gestalt perception, learning, memory etc. Concepts are
viewed as cognitive models (or 'theories').
The heart of the book is in parts 3 and 4 (pp. 179-446).PART
3, "CATEGORIES IN GRAMMAR", concerns with the inherent difficulties in any
attempt to classify parts of speech into discrete categories with clear-
cut definitions:
11) "PARTS OF SPEECH", (JESPERSEN 1924 ch. 4): Proper names are normally
considered rigid designators of objects, unlike common names which are
depicting a sense or connotation. Conversely, it is suggested that proper
names have even more elaborated connotations than common nouns. This
explains how proper names are created from common nouns ("Central-Park")
and vice versa ('Caesars').
12) "ENGLISH WORD CLASSES", a methodological paper by CRYSTAL 1967, shows
that frequently-used dichotomies (like full-/empty-word or lexical/
grammatical) do not capture any naturally distinct sets. In order to
improve the classification, the criteria are divided to Phonological,
Morphological, Lexical, Semantic-Notional and Syntactic types. Crystal
considers the syntactic criteria (mainly substitution frames like
DET___NOUN for adjectives) the most central. He proposes to further rank
the centrality of each criterion by means of statistical significance
within and between categories. He assumes that categories overlap, and
their vague boundaries form "bridge classes".
13) "A NOTIONAL APPROACH TO THE PARTS OF SPEECH" (LYONS 1968, section
7.6): The Categorial-Grammar view of categories is presented. Categories
are considered universal, sentential position (subject, predicate, object
etc.) being the central classification criterion (only language specific
sub-categories are fixed semantically or morphologically). The name (Noun,
Adj. and Verb) is fixed by the reference type of most category members
(things, states or actions). Nouns are degree 1 categories, verbs and
adjectives degree 2 and adverbs degree 3. Sentence construction proceeds
by categories of higher degrees successively modifying categories of lower
degrees up to 0.
14) "SYNTACTIC CATEGORIES AND NOTIONAL FEATURES" (ANDERSON 1997, section
2.1): A minimal set of 4 universal categories is proposed: Predicates {P}
(relations which are structuring events; For ex. verbs); Nominals {N} (for
ex. proper names, which are time stable, are always referential and never
predicative); Mixed {P,N} (for ex. indefinite nouns which assert existence
but do not identify the entity are said to be both referential and
predicative); and Functors {} (labels associating arguments and
predicates; for ex. prepositions and complementizers). This accounts for
languages with no noun/ verb distinction, where affixes or word order
distinguish predicates from arguments.
15) "BOUNDED REGIONS": LANGACKER 1987 (section 5.2) argues that
grammatical categories are semantically definable. A noun instantiates the
schema THING: A region in a domain like time, space or a kinship system.
Typical nouns depict physical objects. Count nouns depict bounded regions.
Past experiences shape future ones such that unbounded regions (for ex. a
dotted circle) or regions in abstract domains ('a team', "the prime
numbers") are cognized as bounded.
16) "THE DISCOURSE BASIS FOR LEXICAL CATEGORIES IN UNIVERSAL GRAMMAR"
(HOPPER AND THOMPSON 1984): Morpho-syntactic features of a category (noun
or verb) are argued to be manifested only when the discourse function is
typical of that category: Introducing a discourse entity is typical of
nouns. Introducing an event is typical of verbs. For ex. since the
incorporated word 'fox' in the utterance "we went fox-hunting" is not
introducing an entity to the discourse, it is phonologically reduced, it
is not marked for plurality or case, it lacks a determiner etc. The idea
is illustrated in a range of constructions and languages.
17) "GRAMMATICAL CATEGORIES" (TAYLOR 1989, ch. 10): In addition to
reviewing some of the papers in part 3, Taylor argues that typicality is
inherent to grammar. The lack of clear-cut conditions of any sort is
illustrated with the concepts 'word', 'affix' and 'clitic'. Words, unlike
affixes, are the minimal independent units (the affix being dependent on
the stem). We can only pause between words. Words are relatively stable
phonologically, can undergo transformation etc. The definite clitic 'the'
falls in the middle: we can pause before or after it but it does undergo
some phonological integration (it is pronounced differently in "the man"
and in "the earth") and it cannot move alone.
PART 4, "GRADIENCE IN GRAMMAR", concerns with descriptions of grammar by
gradient categories and rules:18) In "GRADIENCE", BOLINGER (1961 ch.1-2)
draws a distinction between Ambiguity (for ex. the discontinuity between
the two referents of 'rent') and Generality (the continuity found between
two 'apple' types). In some contexts, blending of two normally separate
meanings occur: We cannot tell the difference between 'didn't'
and 'haven't' in "I ___ put the book on the table".
19) In "DEGREES OF GRAMMATICALNESS", CHOMSKY (1961 section 5) accounts for
graded grammaticality judgments with a hierarchy of category levels. For
ex. level 1 may contain the category 'word', level 2 - the set {S, N, V,
Adj, Adv}, level 3 may distinguish transitive from intransitive Vs,
animate from inanimate Ns etc. The structure S: [[[Honesty]N [loves]v]v
[sincerity]N]S is ruled out by a level 3 rule ('love' selects for an
animate noun), so S is semi-grammatical in level 2. S is less bad
than "goes loves sincerity" since S's level of semi-grammaticality is
higher (S violates only more fine-grained category-distinctions).
20) "DESCRIPTIVE STATEMENT AND SERIAL RELATIONSHIP" (QUIRK 1965):
Grammatical categories are described in tables: A different item is
described in each row and a linguistic feature in each column. A [+] in a
table-slot means that the feature in that column holds of the item in that
row. Normally in such tables, different items are not characterized by
different feature sets (as a classical definitional theory would predict).
The items can be sorted so as to form a different type of regularity
called 'Gradience', i.e. a sequence of [+] signs which begins from the
leftmost slot in each row and gradually shortens from the top row to the
bottom row (forming a triangle of [+] signs). Quirk is concerned with yet
a third case, called Serial relationships, in which the [+] sequence both
begins and ends in an earlier column in each row from top to bottom
(forming a parallelogram of [+] signs). An ex. is the passive. Passive is
derived by transformation, except in verbs lacking an active form, where
it is derived by analogy to verbs with an active form. The analogy occurs
since other features are shared by the "analogical" verbs.
21) "ON THE ANALYSIS OF LINGUISTIC VAGUENESS": NEUSTUPNY 1966 proposes to
characterize grammatical categories by a variety of (possibly non-
necessary) weighted features. Peripheral instances share many features
with members of other categories, and few features with members in their
category. Extreme cases may be regarded Borderline. 22) "NOUNINESS": ROSS
1973 works within Generative grammar, but he demonstrates in great detail
a noun-squish: A gradient description of the noun category. For Ross each
speaker represents nouns differently, but always with a Gradient. Finally,
Ross formulates island-constraints which are sensitive to noun-degrees.
23) "THE COORDINATION-SUBORDINATION GRADIENT": QUIRK ET AL 1985 (13.6-
13.19) attempt to describe a gradient tripartite matrix with coordinators,
conjunctions and subordinators.
24) "THE NATURE OF GRADED JUDGMENTS": SCHUTZE 1996 (section 3.3) proposes
that gradation is introduced by performance and does not reflect the
structure of grammar. This is problematic because our basic data is
affected by factors of performance. However, given the poverty of the
stimuli, it is claimed that learning sequences is much harder than
learning discrete choices. Moreover, as typicality effects exist in
definitional concepts like "even/odd number" (Armstrong et al 1983), they
don't entail lack of defining conditions. Finally, Schutze describes
problems in the new experimental research framework of gradience in
grammar.
PART 5, "CRITICISMS AND RESPONSES" (pp. 447-509, 4 papers):
25) "DESCRIPTION OF LANGUAGE DESIGN": JOOS 1950 presents a discrete view
of grammar as a code. The semantic and phonetic domains are non-discrete.
For ex. there is a continuum of temperatures and of sounds. For each
state "cold to degree n" for n between say +10 and -10 degrees, language
could have mapped a word with a vowel in the very same proportion between
say the sound [o] and [u]. But this is not the case. The different [t]
sounds in utterances of the word 'hotel' are all mapped to the same
linguistic phoneme. These discrete atomic descriptions, claims Joos, are
elegant and fruitful.
26) "PROTOTYPES SAVE" (WIERZBICKA 1990): Wierzbicka argues that prototypes
are often used as an excuse for theoretical sloppiness. For ex. the pope's
being "an unmarried adult male person" but not a 'bachelor' is taken as
evidence against the definitional view. Wierzbicka, instead, adds to the
definition the feature: "a man thought of as someone who could marry".
Wierzbicka challenges Wittgenstein with a definition for 'games': "Things
that people do, when they do something for some time, for pleasure,
imagining that they are in a world where they want to cause some things to
happen, where they know what they can do and what they cannot do, and
where no one knows all that will happen".
27) "FUZZINESS AND CATEGORIZATION": BOUCHARD 1995 (section 1.5.1) argues
that fuzziness characterizes "the concepts expressed by language", but not
the grammar itself. For ex. a variety of features of 'nounier' nouns are
shown to actually depend on the noun's referentiality - a semantic or
psychological aspect, not a syntactic one.
28) "THE DISCRETE NATURE OF SYNTACTIC CATEGORIES: AGAINST A PROTOTYPE-
BASED ACCOUNT": NEWMEYER 2000 describes different hypotheses about
typicality in grammar: The strongest hypothesis is Ross's "direct mapping
view" (e.g. the output of grammatical rules varies with the nouns'
degree); The "cutoff point view" only assumes that operations which are
not applying of typical items will never or almost never apply of atypical
items (Croft 1990, Lakoff 1987); The weakest hypothesis predicts only a
non-random correlation between typicality and morpho-syntactic behavior.
Newmeyer discusses counterexamples to these views. For ex. the verb 'hang'
is less typical than 'notice', but "the picture is hanging on the wall" is
more grammatical than "I am noticing a truck passing by". Thus, verb-
typicality does not affect progressive licensing. Generally, richer
inflectional possibilities are proposed to depend on semantic complexity
(for ex. achievement verbs are more complex than activities), not
typicality. Newmeyer rejects fuzzy boundaries by assuming that categories
overlap ('near' is both an Adj. and a Preposition not a borderline case),
and by showing that Ross's data about nouns is not squishy as assumed and
can be predicted by universal rules.
A CRITICAL EVALUATION
I find the book interesting, stimulating, and highly relevant in the
current states of affairs in linguistics. It is rich in references, fully
indexed and comfortable to use. Most of the philosophical texts are in the
order of quotes (the e-description of the book might be somewhat
misleading in this respect), but they draw the readers' attention to
important historical papers, a full text of most of which can be found on
the web. The full texts by Russell and Keefe are, in my view, excellent
choices. They are clearly written, and they provide a basic idea of the
problem, and how inherent it is in thought and language. The papers in
part 2 are well chosen and the discussed matters are important and
intriguing. There is no doubt that linguists cannot ignore them. Parts 3-4
provide plenty of examples which refuse any simplistic classification and
indeed form stimulating puzzles.
Generally, the question whether grammatical categories are fuzzy or
discrete is empirical in nature, and in principle ought to deserve a
separate answer case by case. Yet, it is taken to pose a dichotomy which
splits linguists to 2 main approaches: those who are postulating discrete
grammatical categories, as opposed to those who are accepting the notion
of Fuzzy Grammar as a basic stance. The authors of this book, as its name
testifies, belong to the latter approach. This affects the book in several
ways. I will now make six critical comments, but they should not undermine
the general positive evaluation of the book.
1) The short philosophical part only hints at the substantial developments
that have been made in the research of language in philosophy, formal
semantics and linguistics since the invention of super-valuations (van
Fraassen 1969). Examples include the classical analysis of adjectives in
Kamp 1975 and Lewis's 1979 notion of standard of precision, which heavily
influenced our thinking about vagueness and gradability. More recent
developments include Barker 2002, Kennedy 1999, 2002 and references
therein, to mention but few. This reader cannot be complete without to at
least mention these prospects. Given the editors comment that: "a
formalism derived from predicate logic, or any other type of underlying
representation requiring clear cut syntactic categories ... will have
difficulties in representing prototype effects" (p. 9), the absence of the
above mentioned formal branch of linguistic thought is not surprising. Yet
it is disappointing. The editors themselves write in the preface that the
value of the reader lies in bringing together work from VARIOUS POINTS OF
VIEW.
2) Empirical evidence for the typicality effects in NATURAL CATEGORIES is
robust, and the cognitive psychological research has already developed
norms which indicate the reliability and validity of these effects.
However, we do not (yet) have equally solid grounds to believe that these
typicality effects exist systematically in the GRAMMATICAL CATEGORIES, as
the book might encourage the reader to infer. Evaluation of the robustness
of these effects in grammar still awaits a systematic experimental
examination. The experimental study of grammatical categories, however
promising a direction, is new, and hence still yields fewer robust valid
effects (SCHUTZE 1996).
3) The book does not discuss the problems and prospects in the development
of a proper typicality THEORY, despite the fact that the development of
such a theory is of utmost importance to the book's concerns. The lack of
a fully developed theoretical account for the structures and processing
underlying the typicality effects (which was pointed out by both Rosch
1978 and Lakoff 1987) makes it even harder to evaluate the hypothesis that
similar types of structures are underlying natural and grammatical
categories. Indeed Newmeyer 2000 found at least 4 types of hypotheses
about typicality effects in grammar.
The historical development of theoretical thought about the structures
underlying the typicality effects has known many obstacles. Fuzzy Logics
has been proven inadequate for language (Osherson et al 1981; Partee and
Kamp 1995). The analysis considerably improved by replacing it with a
three valued logic (ibid; Keefe 2000), but many problems are yet to be
solved (Sassoon in progress). It might have been relevant to mention that
gradability (scales) and vagueness (Keefe 2000) are distinct concepts,
though usually the former is accounted for in terms of the latter. Another
theoretical direction that the editors mention (pp. 20-23) are recent
texts on Optimality and probabilistic approaches to syntax. The inclusion
of these texts in the book (maybe instead of some overlapping material in
Part 2) could have been stimulating in important respects.
A third direction to go is sociolinguistics (p. 20). Note that the
correctness of a linguistic form in a given point in time depends on at
least 3 parameters (Myhill in press) -- the language forms in canonical
texts, the forms used by powerful social groups, and the forms advanced by
prescriptive authorities. Naturally, the different standards do not always
agree about the correctness of a form. I believe that such dissociations
may cause graded judgments and dependency on the context of use (home,
office etc.; recall Labov 1973). Thus, multiple standards may account for
the variance familiar to all linguists, within and between speakers.
4) The evidence which the book brings in support of the "fuzzy grammar"
hypothesis differs in many cases from the evidence for fuzziness in
natural concepts. For ex. natural concepts usually exhibit only family
resemblance, not Gradience as was claimed for the grammatical concepts
(Ross 1973, Quirk et al 1985 etc.) In fact, Ross's detailed data suggest
that two disconnected poles, nominal and sentential (or clausal), exist
after all, and that variance between and within speakers characterizes
mostly items with mixed behavior (low family resemblance, if you like).
Even Quirk's 1965 weaker notion of serial relationships seems to be
stronger than family resemblance as we know it in natural concepts.
Neustupny's 1966 theory is the only one requiring nothing but family
resemblance to hold of grammatical forms, which indeed seems to describe
best the data about coordinators and subordinators (see p. 430), time
adverbials (p. 209) etc.
Chomsky's 1961 account for graded judgments is elegant, but is criticized
by Schutze 1996 in that it only accounts for judgments of UN-
grammaticality. Indeed, typicality effects in natural concepts are linked
to typicality effects in their negated concept (Smith et al 1988, Hampton
1997). In natural language, any vague adjective ('tall') is treated
as 'discrete' if context fixes for it a proper standard of precision.
Any 'discrete' noun ('student', 'prime number') is treated as scalar if
modified by adjectives like 'typical', 'normal' or 'relevant in the
context'. What would this fact predict of the grammatical categories?
(Maybe, they are in fact born discrete, but yet flexible enough to adapt
to non-standard contexts?) At any rate, it is important to keep in mind
that the possibility that natural and grammatical categories are quite
different in nature is still viable.
5) Many types of phenomena which the book brings in support for a fuzzy
grammar are actually coherent with a 'classical' or 'discrete' view of
grammar. For ex. CATEGORY OVERLAP is coherent with discrete categorization
(Aarts 2004) and similarly Bolinger's syntactic BLENDING is coherent with
this too; FEATURES THAT CROSS CATEGORIES (e.g. a connotation for both
common and proper names) are coherent with the idea that categories can be
distinguished by classical criteria; TYPICALITY ORDERING WITHIN A CATEGORY
is coherent with the existence of necessary and sufficient conditions
(Armstrong et al 1983). Actually, associating with discrete categories a
SET of criteria immediately enables us to order items by the number of
criteria that apply to them.
However, 2 types of phenomena are NOT coherent with a discrete view and
hence form evidence for a fuzzy grammar: The first type is the LACK OF
DEFINING CRITERIA. To take an example which was not discussed in the book,
many scholars from a variety of approaches agree that Theta roles are not
characterized by classical definitions, but indeed by prototypes (Dowty
1989) or by labels with very little content (Reinhart 2002). So this might
indeed be a case in point; That "not even one of the most robust
properties of NPs (...) shows up in all instances" (Taylor pp. 299-307)
might indeed show that NPs resist definitions, and they ought to be
associate only with prototypes. But another alternative is that, in the
future, the exceptions will be accounted for in systematic ways (Newmeyer
1998, 2000).
The second type of true evidence against discrete categories is formed by
BORDERLINE-CASES. For ex. a clitic which falls between a 'word' and
an 'affix' (Taylor 1989) might indeed show that these categories lack
distinguishing criteria. However, note that in many natural concepts
(those traditionally considered 'sharp'), most non-central instances fall
under the concept, but are in the periphery (for certain reasons). Real
borderline cases are rare. If future discrete grammatical classifications
will have but few rare border-line cases it would be possible to represent
them by a reasonably short list of exceptions. In sum, Wierzbicka (ch. 26)
is right in that prototypes should not be used as an excuse for
theoretical sloppiness. We should at least try to identify discrete
categories and clear-cut categorial criteria, the existence of which is
coherent with, and does not decrease the importance of, most of the
phenomena described in the book.
6) In any event, criticism against the prototype view should not be sloppy
too! I had expected the closing part "criticisms and responses" to expose
the reader to more serious proposals to account for the data presented in
parts 3-4 in discrete or classical terms (see for ex. Wasow 1997). The
criticisms included in the book were quite weak. For ex. Wierzbicka's
definition for 'game' considerably over-generates (in my free time, I
enjoy cleaning my room. This activity is clearly not a 'game' but it falls
under the proposed definition) and under-generates (unlike Wierzbicka
[note 4 p. 469] I consider "games played by mathematicians" to
be 'games').
As for Bouchard 1995, sweeping semantic (or interface) problems away from
Linguistics, like he proposes to do, is more a political statement than
serious criticism. The problems do not disappear by calling the people
dealing with them psychologists rather than linguists. Similarly, the
typicality effects need an account even if they are said to belong outside
of the structure of concepts. As for Joos's 1950 view of discrete
categories as elegant, note that a scientific rule of the form "all Ps are
Qs" is not stronger than a parametric correlation such as: "for any entity
d, d's degree in P predicts d's degree in Q" (Ross 1973).
Moreover, 'modules' may, in principle, like electric tools (using Joos's
metaphor), have multiple-state switches just as well as binary ones.
The "type of switch" can only be determined empirically. I find Newmeyer's
text the most serious attempt to propose an alternative to specific case
studies. But even this text remains somewhat on a meta-theoretical level.
It does not intend to describe a fully detailed study of one subject.
Nonetheless, the 500 pages (!) of the book raise lots of interesting
questions and intriguing possibilities, and as such the book is certainly
worth reading.
REFERENCES
Aarts, Bas, 2004, "Modeling Linguistic Gradience", Studies in Language 28
(1): 1-50.
Anderson, John M., 1997, A Notional Theory of Syntactic categories,
Cambridge Uni. Press.
Aristotle, 1963, "Categories and De Interpretatione", Translated by J.L.
Ackrill, Clarendon Aristotle series, Oxford Clarendon Press.
Aristotle, 1971, "Metaphysics", Translated by Cristopher Kirwan, Clarendon
Aristotle series, Oxford Clarendon Press.
Armstrong, S.L., Gleitman L.R. and Gleitman H., 1983, "What Some Concepts
Might Not Be", Cognition 13, 263-308.
Barker, Chris, 2002, "The Dynamics of Vagueness", Linguistic and
Philosophy 25, 1-36.
Bolinger, Dwight, 1961, Generality, Gradience, and the all or none, Mouton
de Gruyter.
Bouchard, Denis, 1995, The Semantics of Syntax: A Minimalist Approach to
Grammar, Chicago Uni. Press.
Carson T. Schutze, 1996, The Empirical Base of Linguistics: Grammaticality
judgments and Linguistic methodology, 61-81, Uni. of Chicago
Chomsky, Noam, 1961, "Some Methodological Remarks on Generative Grammar",
Word 17, 219-39.
Crystal, David, 1967, "English Word Classes", Lingua 7, 24-56.
Dowty, David, 1989, "On the Semantic Content of the Notion of Semantic
Roles", in G. Chierchia, B. Partee, and R. Turner (eds.), Properties,
types and meaning, 69-129.
Frege, Gottlob, 1903, "Grundgesetze der Arithmetic vol. ", in Peter Geach
and Max Black (eds.), 1970, Translations from the Philosophical Writings
of Gottlob Frege, Oxford, Blackwell, P. 159.
Hampton, J. A., 1997, "Conceptual Combination: Conjunction and Negation of
Natural Concepts", Memory & Cognition, 25 (6), 888-909.
Hopper, Paul and Sandra Thompson, 1984, "The Discourse Basis for Lexical
Categories in Universal Grammar", Language 60, 703-52.
Jackendoff, Ray, 1983, Semantic and Cognition, Mit Press, Cambridge, Mass.
Jespersen, Otto, 1924, The Philosophy of Grammar, George, Allen & Unwin,
London.
Joos, Martin, 1950, "Description of Language Design", Journal of the
acoustical Society of America 22, 701-8.
Kamp 1975, "Two theories of Adjectives", in: Formal Semantics of Natural
Language, ed. Edward Keenan, 123-155, Cambridge University Press.
Keefe, Rosanna, 2000, Theories of Vagueness, Cambridge Uni. Press, 6-36.
Kennedy, Christopher, 1999, Polar Opposition and the Semantics of Degree,
Kluwer academic press, Netherlands.
Kennedy, Christopher, 2002, The Landscape of Vagueness, Ms., Northwestern
University.
Labov, William, 1973, "The Boundaries of Words and their Meanings", in
C.J. Bailey & R.W.Shuy (eds.), New ways of Analyzing Variation in English,
Georgetown Uni. Press, Washington, 340-73
Lakoff, George, 1987, Women, Fire and Dangerous things: What Categories
reveal about the Mind, Chicago Uni. Press.
Langacker, Ronald W., 1987, Foundations of Cognitive Grammar v1:
Theoretical-Prerequisites, Stanford Uni. Press.
Lewis, David, 1979, "Score keeping in Language game", Journal of
Philosophical logic 8, 339-359.
Lyons, John, 1968, Introduction to Theoretical Linguistics, Cambridge Uni.
Press.
Myhill, John, in press, Multilingua.
Newmeyer, Frederick, 1998, Language form and Language function, The MIT
press, Cambridge & London.
Newmeyer, Frederick, 2000, "The Discrete nature of Syntactic Categories:
Against a Prototype-based Account", in Robert D. Borsley (ed.), The Nature
and Foundation of Syntactic Categories, Syntax and Semantics Vol. 32,
Academic Pres, New York, 221-50.
Neustupny, J. V., 1966, Travaux Linguistiques de Prague , 39-51.
Osherson, D.N. and E.E. Smith, 1981, "On the Adequacy Of Prototype Theory
As A Theory Of Concepts", Cognition 11, 7-262.
Partee, Barbara & H. Kamp, 1995, "Prototype Theory And Compositionality",
Cognition 57: 129-91.
Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech and Jan Svartvik, 1985,
A Comprehensive Grammar of the English Language, 921-8, Longman.
Quirk, Randolph, 1965, "Descriptive Statement and Serial Relationship",
Language 41(2), 205-17.
Reinhart, Tania, 2002, "The theta system -- An Overview", Theoretical
Linguistics 28(3).
Rosch, Eleanor, 1978, "Principles of Categorization" in Eleanor Rosch and
Barbara Lloyd (eds.) Cognition and Categorization, Hillsdale, 27-48 .
Ross, John Robert, 1973, "Nouniness", in: Osamu Fujimura (ed.), Three
Dimensions of Linguistic Research, Tec company Ltd, Tokyo, 137-257.
Russell, Bertrand, 1923, "Vagueness" Australian Journal of Philosophy and
Psychology 1, 84-92.
Sassoon, Galit, 2001, "Predicate Interpretations as Intensions Restricted
along Dimensions", MA Thesis, Tel Aviv University.
Sassoon, Galit, 2002, "Semantics with Clusters of Properties", In: Yehuda
Falk (Ed.) Proceedings of Israel Association for Theoretical Linguistics
18, Bar Ilan Uni.,
http://atar.mscc.huji.ac.il/~english/IATL/18/Sassoon.pdf.
Sassoon, in progress, "Semantics with Clusters of Properties", Doctoral
dissertation, Tel Aviv University.
Smith, Edward E., D. N.Osherson, J. Rips, M. Keane, 1988, "Combining
Prototypes, a Selective Modification Model", Cognitive Science 12, 485-527.
Taylor, John, 1989, 1995, Linguistic Categorization: Prototypes in
Linguistic Theory, Oxford Uni. Press.
Wasow, Thomas, 1997, "Transformations and the Lexicon", in: P. Culicover,
T. Wasow and A. Akmajian (eds.), Formal Syntax, 327-360, New York:
Academic Press.
Wierzbicka, Anna, 1990, "Prototypes Save: On the uses and Abuses of the
notion of 'Prototype' in Linguistics and Related Fields", in: Savas L.
Tsohatzidis (ed.) Meanings and Prototypes: Studies in Linguistic
Categorization, Routledge, London, 347-67.
Wittgenstein, Ludwig, 1953/58, Philosophical Investigations, translated by
G.E.M. Anscombe, Third Edition, 1968, Oxford, Blackwell.
Van Fraassen, Bas C., (1969), "Presuppositions, Supervaluations and Free
Logic", in K. Lambert (Ed.), The Logical Way of Doing Things, PP. 67-91,
Yale Uni. Press.
Zadeh, Lotfi, 1965, "Fuzzy Sets", Information and control 8: 338-53.
|
| |
ABOUT THE REVIEWER:
ABOUT THE REVIEWER
I am a PhD student in Tel Aviv University and I also teach "Guided reading
in linguistics" there. My main interest is in semantics. I am developing
an account for typicality, incorporating data and theoretical wisdom from
Cognitive psychology, Philosophy and Linguistics. Within a generative
formalist approach, I illustrate the effects of typicality in the analysis
of predicate meaning and of quantifying expressions (Sassoon 2001, 2002,
in progress).
|
|
|
