(Cambridge studies in linguistics 82) john m anderson a notional theory of syntactic categories cambridge university press (1997)

This book presents an innovative theory of syntactic categories and
the lexical classes they define. It revives the traditional idea that
these are to be distinguished notionally (semantically). It allows
for there to be peripheral members of a lexical class which may
not obviously conform to the general definition. The author proposes
a notation based on semantic features which accounts for the
syntactic behaviour of classes. The book also presents a case for
considering this classification - again in rather traditional vein - to
be basic to determining the syntactic structure of sentences.
Syntactic structure is thus erected in a very restricted fashion, without recourse to movement or empty elements.



CAMBRIDGE STUDIES IN LINGUISTICS
General Editors: s.

R. ANDERSON, J. BRESNAN, B. COMRIE,

W. DRESSLER, C. EWEN, R. HUDDLESTON, R. LASS,
D. LIGHTFOOT, J. LYONS, P. H. MATTHEWS, R. POSNER,
S. ROMAINE, N. V. SMITH, N. VINCENT

A notional theory of syntactic categories


In this series

52 MICHAEL s. ROCHEMONT and PETER w. CULLICOVER: English focus constructions

and the theory of grammar
53 PHILIP CARR: Linguistic realities: an autonomist metatheory for the generative

enterprise
54 EVE SWEETSER: From etymology to pragmatics: Metaphorical and cultural aspects
of semantic structure
55 REGINA BLASS: Relevance relations in discourse: a study with special reference to
Sissala
56 ANDREW CHESTERMAN: On definiteness: a study with special reference to English
and Finnish
57 ALESSANDRA GIORGI and GIUSEPPE LONGOBARDI: The syntax ofnoun phrases:

configuration, parameters and empty categories
58 MONIK CHARETTE: Conditions on phonological government
59 M. H. KLAIMAN: Grammatical voice
60 SARAH M. B. FAGAN: The syntax and semantics of middle constructions: a study
with special reference to German
61 ANJUM P. SALEEMI: Universal Grammar and language learnability
62 STEPHEN R. ANDERSON: A-Morphous Morphology

63 LESLEY STIRLING: Switch reference and discourse representation
64 HENK J. VERKUYL: A theory of aspectuality: the interaction between temporal and
atemporal structure
65 EVE v. CLARK: The lexicon in acquisition
66 ANTHONY R. WARNER: English auxiliaries: structure and history
67 P. H. MATTHEWS: Grammatical theory in the United States from Bloomfield to
Chomsky
68 LJILJANA PROGOVAC: Negative and positive polarity: a binding approach
69 R. M. w. DIXON: Ergativity

70 YAN HUANG: The syntax and pragmatics of anaphora
71 KNUD LAMBRECHT: Information structure and sentence form: Topic, focus, and the
mental representations of discourse referents

72 LUIGI BURZIO: Principles of English stress
73 JOHN A. HAWKINS: A performance theory of order and constituency
74 ALICE c. HARRIS and LYLE CAMPBELL: Historical syntax in cross-linguistic perspective
75 LILIANE HAEGEMAN: The syntax of negation
76 PAUL GORRELL: Syntax and parsing

77 GUGLIELMO CINQUE: Italian syntax and Universal Grammar
78 HENRY SMITH: Restrictiveness in case theory
79 D. ROBERT LADD: Intonational phonology

80 ANDREA MORO: The raising of predicates: predicative noun phrases and the theory
of clause structure
81 ROGER LASS: Historical linguistics and language change
82 JOHN M. ANDERSON: A notional theory of syntactic categories

Supplementary volumes
LILIANE HAEGEMAN: Theory and description in generative syntax: a case study in
West Flemish
A. E. BACKHOUSE: The lexical field of taste: a semantic study of Japanese taste terms
NIKOLAUS RITT: Quantity adjustment: vowel lengthening and shortening in early
Middle English

Earlier issues not listed are also available


A NOTIONAL THEORY OF
SYNTACTIC CATEGORIES
JOHN M. ANDERSON
Professor of English Language,
Department of English Language, University of Edinburgh

CAMBRIDGE
UNIVERSITY PRESS


Published by the Press Syndicate of the University of Cambridge
The Pitt Building, Trumpington Street, Cambridge CB2 IRP
40 West 20th Street, New York, NY 10011-4211, USA
10 Stamford Road, Oakleigh, Melbourne 3166, Australia
© Cambridge University Press 1997
First published 1997
A catalogue record for this book is available from the British Library
Library of Congress cataloguing in publication data
Anderson, John M. (John Mathieson), 1941A notional theory of syntactic categories / John M. Anderson.
p. cm. - (Cambridge studies in linguistics: 82)
Includes bibliographical references and index.
ISBN o 521 58023 4 (hardback)
1. Grammar. Comparative and general - syntax. 2. Grammar,
Comparative and general - Grammatical categories. 3. Semantics.
I. Title. II. Series.
P291.A53 1997
4i5-dc2O 96-21789 CIP
ISBN 0 521 58023 4 hardback
Transferred to digital printing 2004

CP


Much, then, that is considered by the generality of grammarians as
syntax, can either be omitted altogether, or else be better studied

under another name.
(Latham 1862: 577)

'A9f]vas



Contents

Preface
List of abbreviations

page x
xii

1
i. i
1.2
1.3

Prelude
Notionalism
Analogism
Minimalism

2
2.1
2.2
2.3
2.4

2.5
2.6
2.7
2.8

Fundamentals of a notional theory
Syntactic categories and notional features
Relations between elements
Further categories: the role of feature dependencies
Markedness and category continuity
Cross-classification
Gradience and second-order categories
Secondary categories
Non-complements

13
13
29
43
61
64
73
104
132

3
3.1
3.2
3.3
3.4

3.5
3.6
3.7

The syntax of categories
Verbal valencies
The content of the functor category
The basic syntax of predications
The formation of ditransitives
Variation in argument structure
Verbals as arguments
The structure of primary arguments

146
149
168
174
236
244
252
292

References
Index

i
2
6
8

320
345



Preface

There were the bits he understood. They were bad
enough. But the bits he didn }t understand were worse.

The initial quotation here and the chapter epigraphs are drawn from Patrick
White's novel The solid mandala, which and particularly Arthur therein have
much to say about grammar and meaning. The present work was concluded by
its author when the bits he didn't understand were/are still overwhelming; and
drawing a line at this point is, as is normal, relatively arbitrary. This means that
a number of issues which many might regard as crucial to present-day concern
with syntax have scarcely been touched on in what follows, or have even been
ignored, and those areas treated have received vastly varying degrees of attention. Also, the recognition here of previous work remains partial, and adapted
to the needs of the arguments put forward in particular sections. But there is for
such as the present enterprise no great virtue in comprehensiveness of bibliographical reference (or even a possibility thereof). Nor have I striven to establish throughout a consistent temporal endpoint for reference: in a work whose
writing extends over any length of time, achieving this would be equivalent to
having all of the Forth Bridge freshly painted at the same time. It is my hope that
nevertheless - or rather, as a result, to some extent - the following spells out sufficiently, over a wide enough area, and in appropriate sub-areas in enough detail,
the general structure of a notional theory of syntactic categories and the major
consequences for the syntax of adopting the views that syntactic categories are
so based and that syntax itself involves the interaction of structures projected by
these categories with pragmatically based requirements involving crucially the
organisation of information - though it should be conceded that the existence of
the latter is asserted here rather than fully motivated and articulated. The discussion which follows thus involves an attempt to support (what has been for
most, in the context of the latter half of the twentieth century) an unfamiliar

strategy, both by detailed investigation of some subsystems and with reference
xi


xii Preface
to a wide variety of language types; as such, its reception will no doubt founder
between and among the Scylla of 'true believers' (responsory motto: That's not
what we do (anymore)'), the Charybdis of the 'bug collector' (That doesn't
happen in (say) Amharic') and the Deep Blue Sea of scholarly inertia (The literature is what my friends tell me about'). I am, however, and of course, solely
responsible for failures to launch things worth receiving, despite the help of
those whom it is my pleasure in the following paragraph to both acknowledge
and absolve.
Most of the writing of the book was carried out during the academic years
1991-2 and 1992-3 while the author held a British Academy Readership, and
would not otherwise have been carried out: this does not overstate the debt. The
first semester of 1991-2 was spent as an associate of the Department of
Theoretical and Applied Linguistics in the School of English at Aristotle
University, Thessaloniki: I am grateful to Professors Efstathiadis and
Kakouriotis and their colleagues for their hospitality. I have also had the benefit of presenting aspects of the research reported on here to audiences at the
Universities of Goteborg, Helsinki, Manchester and Umea. And a number of
other people have contributed to the preparation of this work in diverse ways,
from providing the bottle of NAOY22A to talk over to offering criticisms to
chew over, to helping me avoid some garden paths. I should especially like to
thank the following: Nick Kontos, Scott McGlashan, Peter Matthews, Nigel
Vincent, and anon (who also writes some pretty good stuff of his own, particularly the early music); and, most particularly, Roger Bohm, whose influence
will be seen to be pervasive. And who could by-pass Fran Colman, longestsuffering in the process of the book's getting written? Not
John Anderson
Edinburgh

Abbreviations and conventions

ABL(lative)
ABS(olutive)
ACC(usative)
AG(en)T(ive)
ART(icle)
ASP(ect)
AUX(iliary)
BEN(enefactive)
CLASS(ifier)
C(o)MP(lementiser)
C(ou)NT
COMP(letive)
CONT(inous)
COP(ula)
DAT(ive)
DEF(inite)
DIR(ectional)
DIST(al)
E(rgative)/I(nstrumental)
ERG(ative)

FIN(ite)
FUT(ure)
GEN(itive)
GER(und)
I/II = first/second person
III = third person
INC(lusive)

INDIC(ative)
INF(initive)
INSTR(umental)
INT(erior)
INTRANS(itive)
I(mmediate)P(ast
L(ong)F(orm)
P(a)RT(itive)
LOC(ative)
MASC(uline)
NEG(a*ive)
NOM(inative)
N(on)F(inite)

NONFUT(ure)
N(eu)TR(al)
PART(iciple)
PASS(ive)
PERF(ect)
PL(ural)
POSS(essed)
POT(ential)
PRES(ent)
PREV(erb)
REFL(exive)
REL(ativiser)
S(u)B(or)D(inator)
S(u)BJ(ect)
S(hort)F(orm)
S(in)G(ular)

SUBJ(unctive)
T(e)NS(e)
V(erbal)N(oun)

These abbreviations are mainly used in glosses, where, as labels for morphosyntactic categories, they appear in SMALL CAPS. In the text lexemes/words
also appear in small caps, the names of inflexional categories are capitalised initially, and cited word and sentence forms are italicised. Important terms appear
in bold on their (re)introduction, as do text occurrences, not within braces, of the
semantic features (P, loc etc.) which identify syntactic categories.

xni



I

Prelude

'In the beginning was what word?' Arthur asked.

This book is concerned with word classes and their categorisation, where word
is taken to be the basic unit of the syntax and their classification is determined
by this syntax, i.e. how they combine to form sentences. It is specifically concerned with the 'substance' of class labels (traditionally, 'noun', 'verb', etc.)
and with what role this 'substance' plays in the syntax. The chapters which succeed this one are concerned to lay out and motivate a particular approach to the
'substance' of word categorisation, one which I have dubbed 'notional', in so far
as it is conceived of as an extension of those 'traditional' grammars that saw this
'substance' as ontologically based. A more immediate antecedent is some suggestive work of Lyons (e.g. 1966; 1977: ch. 11; 1989); and the approach seems
to me a natural extension of my earlier work on 'case grammar' (e.g. Anderson
1971a; 1977; 1980b; 1984b; 1986b), as I have indicated elsewhere (Anderson
1989b; 1992b). The present brief chapter is intended, as described by its title, as
a 'prelude' to the main discussion, in providing in brief some context for the

assumptions and intentions that inform the latter. It does not attempt to provide
a historiography of word-class studies, or even of notionalism: these warrant
substantial treatment on their own account.
What follows takes off from the combined proposition that there are no
specifically linguistic semantic categories, or level of representation, distinct
from the syntactic and that syntactic categories are themselves grammaticalisations of cognitive - or notional - constructs; what follows is intended to
explicate and to provide support for (particularly) this latter proposition. But it
should already be clear that such a viewpoint also imposes certain research
strategies. It means, for instance, that it is predicted to be ultimately unprofitable
to pursue a theory of syntactic categories autonomously, in isolation from their
semantics; a cross-linguistic syntactic, distributionally established category that
is not notionally defined is merely of interest as a potential counter-example to


2 Prelude
the more retrictive theory of notional identification of word classes. It also
means, for instance, that, with respect to interpretation, representations constructed out of notionally defined syntactic categories serve as the input to
general (not language-specific) deductive systems. What I am concerned with
here, however, is the development of a system of notionally based syntactic
categories sufficient, in principle, to subtend the expression of a range of syntactic generalisations, i.e. generalisations concerned with systematic aspects of
the distribution of words in sentences. I return later in this chapter to the assumptions I adopt concerning the syntactic status of such categories, and, indeed,
about the nature of syntactic structure and its relation to other aspects of the
grammar; let us at this point dwell a little on the notional basis for the categories.
I.I

Notionalism

The content of a syntactic category, such as 'noun', is notional: the crosslinguistic classes associated with the category are identified as such by the recurrence, as members of these classes, of linguistic items that share certain
conceptual properties, as discussed in §2.1 below. Crucially, in the case of
nouns, these items denote (what are perceived as) discrete physical objects. Such

categories are grammaticalised; as part of a cultural system, the class associated
with particular languages may contain 'eccentric' members - indeed, members
whose apparent conceptual equivalents belong to a different category in another
language. Thus, whereas in English it is arguably the case that the distribution
of MAY and MUST warrants recognition that they belong to a class - or at least
subclass - distinct from that associated with (other) verbs, their equivalents in
many other languages share their distribution with verbs. And, as we shall be
discussing in §2.3.1, there are, for example, languages where the equivalents of
most or all of those items that we might classify as adjectives in English are
instead (untypical) verbs and/or nouns. Thus, whereas, rather uncontroversially,
RED in English is distributionally not a noun, but occurs distinctively as an adjective (with nominal uses - That's a nice red - being derivative), in Ossetic, for
instance, the 'corresponding' term is, apparently, syntactically like any (other)
noun (Abaev 1964: §54). Any notional characterisation of syntactic categories
should not only identify the central membership of categories but also provide
a basis for understanding such variation.
The existence of notionally non-central members frustrates simple-minded
application of blanket notional definitions of categories, despite their currency
until quite recently, particularly in pedagogical grammars, in attempts to differ-


Notionalism 3
entiate those major syntactic distinctions traditionally embraced by the label
'the parts of speech'. Such definitions either fail to encompass all the exponents
of a particular category, or they are so vague and generalised as to be nondistinctive of that category; we are all familiar with Jespersen's (e.g. 1924: 59)
demolitions in this respect of the definitions of Sonnenschein and co. (definitions
such as 'Nouns name. Pronouns identify without naming.'). See too Lyons 1977:
§11.1. Jespersen further shows (1924: 60) that morphologically based ('formal')
definitions, though again of ancient ancestry, are also clearly inadequate 'as the
sole test' for distinguishing 'parts of speech' in a generally applicable manner.
And he concludes: 'In my opinion everything should be kept in view, form,

function, and meaning ...' (ibid). But, in my view, these aspects he distinguishes
contribute in different ways to the 'view': classes are to be distinguished on a
morphosyntactic ('form' and 'function') basis, but their cross-linguistic identification is based on 'meaning', the notional character of central members.
Moreover, despite varying degrees of grammaticalisation - or desemanticisation - much of the detailed syntax, as well as the nature of the morphological
distinctions associated with particular categories, reflects notional properties.
This is quite apart from those relationships in the syntax - notably topicalisation
and the like, but including many other variations in linearisation (e.g. Bolinger
1952) - which expound the organisation of information in an ongoing fashion.
What I have in mind rather are the kinds of phenomena we shall be exploring in
chapter 3 below, such that, for instance, the capacity for what I shall term
'ectopic placement' - roughly equivalent to what has usually been interpreted
as the result of 'movement' - is associated with the semantic role of the element
concerned, in particular whether it is inherently subcategorised-for or not, subcategorisation being, of course, on a semantic basis. Likewise, as we shall again
be looking at, the basic word order patterns in a language reflect the (semantic)
valency status of elements (as, semantically, head, complement, circumstantial/adjunct, specifier). Consider here, as a small-scale example of the notional
sensitivity of syntax, one aspect of the syntactic behaviour of 'endocentric
adjuncts' in English.
As has often been observed, there are interesting restrictions in English on
preverbal placement of endocentric adjuncts like that in (1.1):
(1.1)

a. Mabel very quietly deteriorated/improved
b. Bert very quietly ate the last pie
c. Felicity very quietly moved away

such that the sentences in (1.2) are not nearly as happy as those in (1.1):


4 Prelude
(1.2)

a. * Alfred very quietly arrived/danced
b. *Bert very quietly ate
c. *Felicity very quietly moved

even though variants with the adjunct in postverbal position are perfectly acceptable:
(1.3)

a. Alfred arrived/danced very quietly
b. Bert ate very quietly
c. Felicity moved very quietly

The verbs in (1.2a) differ from those in (1.1a) in being 'actions' rather than
'processes'. 'Actions' rather than 'processes' do not allow a preverbal adjunct
- unless they are presented, as in (i.ib-c), as semantically transitive (not necessarily 'syntactically transitive' - even if the latter concept is well-defined). Not
all 'processes' are conducive to the preposed adjunct, however:
(1.4)

*Basil quietly fell/tripped

The 'processes' in (1.1a) are specifically 'change-of-intrinsic-state' (rather than,
say, 'of-place' simply). And so on: other semantic variables supervene.
My point is that any understanding of the syntax of the adjunct is inextricable from the semantics of the adjunction. And this is characteristic of syntactic
generalisations: they refer to notional classes and to semantic relations between
them. More generally, syntactic properties are projected from notional. It is thus
inadequate to suggest that the regularities involved in (1.1-4), and elsewhere, are
simply non-syntactic precisely because they are semantically based; that they do
not demonstrate a semantic basis for syntax because the phenomena themselves
are not 'syntactic'. This is the 'selectional restrictions' strategy - on which see
e.g. Radford 1988: §7.9. On such a view, nothing of syntax (subcategorisation,

placement of complements, adjuncts and specifiers, 'ectopic' placements)
remains - as much of the content of the following chapters is intended to illustrate in some detail.
I shall argue, too, that the association with particular word classes of particular
'secondary' categories (such as Case, Tense), so that the latter are recurrently
realised in the 'formal' shape of the former, i.e. as part of their morphological
structure, is notionally non-arbitrary. The recurrence in association with nouns
of definiteness as a functional category - whether inflexionally or periphrastically expressed, i.e. as 'traditional' secondary category or 'article' (§2.7) - follows
from their notional characters: crudely (see further §§2.1, 3.7), a definite
article/affix is one of those elements that enables a noun to function as (part of)
a semantic argument (like a name or a personal pronoun) rather than simply as
a predicator.


Notionalism 5
Of course, there may be, as well as individual idiosyncrasies - lexicalisations
- and minor grammaticalisations of membership or relation, as well as farreaching relational grammaticalisations, such as that incorporated in the basis for
selecting subjects, as discussed in §3.3.1. But these remain desemanticisations;
they can be properly understood only against the backgound assumption of a
notional basis for syntactic concepts.
The basis of linguistic categorisation in meaning was the fundamental insight
of that long tradition which eventuated in the 'traditional' (and often pedagogical)
grammars which - despite e.g. Jespersen's attempts at toughening them up proved such soft targets for the 'structuralists'. In their different ways, and on different sides of the Atlantic, both Hjelmslev (1961 [1943]) and Harris (1946) reject
the notional content of syntax. Thus, for the former, 'the projection of the formhierarchy on the substance-hierarchy can differ essentially from language to
language' (1961: 97), where, on the plane of content, the 'form-hierarchy' is
roughly, in the terms being used here, the morphosyntactic categorisation - or
system of linguistic relations, in more Hjelmslevian terms - and the 'substancehierarchy' is a non-linguistic description of the 'substance' of meaning. With like
consequences, Harris (1946) identifies word - or rather 'morpheme' - classes on
the basis of language-particular diagnostic environments, so that the classes he
sets up for English and Hidatsa are strictly non-commensurate. So, both English
and Hidatsa have a class labelled W , but this is fortuitous, non-systematic:

whereas TV in English are 'morphemes which occur before plural -s or its alternatives' (Harris 1946: §4.1), in Hidatsa N are 'non-clause-final suffixes' (§5.1).
It seems to me that in this respect structuralist syntax was (and is) a barren
conceptual detour, and that the history of (transformational-) generative grammar represents a reluctant (and characteristically dissembled) but inexorable
retreat from that position. One facet of that retreat is part of the backgound to
the discussion of subcategorisation in §3.1.1. In this context, I should perhaps
make it clear that I do not regard the developments described as 'generative
semantics' (cf. e.g. the contributions to Seuren (ed.) 1974) as in themselves representing an embracing of notionalism: for most of the proponents of that
approach, a 'natural logic' was conceived of as forming the basis of syntax;
whereas I interpret a notional approach as excluding 'natural logic', as such,
from syntax (and indeed from grammar as a whole) and as having it operate
upon (amongst other things) interpretations of the categories and structures provided by the latter. To this extent, notionalism has something in common with
the Hjelmslevian position alluded to above, while, of course, still denying the
(essential) arbitrariness (as posited by Hjelmslev) of the relationship between
syntactic representation and logical semantics.


6 Prelude
The positing of such arbitrariness frustrates any possibility of a general theory
of syntactic categories: overt syntactic diagnostics, just like the morphological,
are ineluctably language-particular. Of course, given a sufficiently abstract
conception of syntax, it is possible to make cross-language generalisations concerning the syntax of 'nouns', say. Thus, in any language with subjects, 'noun'
phrases will constitute the central constituent type with respect to capacity to
function as such. But, as we shall see, such generalisations flow from the notional
character of 'nouns', and are best formulated in these terms, with brute distribution itself as a consequence of such; syntax does not involve the arrangment of
autonomous, or notionally uninterpreted elements.
1.2

Analogism

I also share with Hjelmslev (and other - mostly dead - linguists) a conception

of linguistic structure as fundamentally bi-planar, as well as the assumption that
the two planes are structured in accordance with the same principles. My main
concern in the present section is to attempt to explicate the consequences of this
assumption, in particular; but let me turn first to bi-planarity and other aspects
of linguistic organisation.
We can differentiate different levels of representation of linguistic structure
(along the lines of Anderson 1992&: ch. 1) to the extent that the representations
assigned to different levels are governed by distinct regularities. Most basically,
or strongly (Anderson 1982a), we can distinguish representations based on a
distinct substantive alphabet. This is what distinguishes the phonological and
the syntactic planes: phonological representations are constructed out of an
alphabet of phonetically identifiable features; syntactic structures are erected on
the basis of notionally identifiable features. Morphology does not introduce a
distinct alphabet. In terms of a modified 'word-and-paradigm' approach to morphology, morphological structure proper interprets the syntactic ('secondary'
and 'primary') categories, such as 'Past' and 'verb', associated with a word, in
terms of the organisation of the phonological material associated with the item
into morphological units. Morphological structure itself involves unlabelled
relations: root, base, affix etc. are identified structurally. (For discussion see e.g.
Anderson 1984a; 1985a; 1992b: particularly §2.3; Colman 1990; 1991: ch. 2;
to appear.)
As is familiar, phonological structure also interprets plane-external structure,
indeed, both morphological and syntactic structure. There are thus asymmetries
in the relationships between the planes (including the sub- or inter-plane of morphology), as well as there being distinct (including, possibly, no) alphabets. This


Analogism 7
constrains the applicability of an otherwise very generally appropriate assumption concerning linguistic structure.
I espouse here the following assumption, given here in the form of a directive
for linguistic representations:
Structural analogy

Minimise (more strongly, eliminate) differences between levels that do not follow from a difference in alphabet or from the nature of the relationship between
the levels concerned. (Anderson 1992&: ch. 1)

I interpret this as, within the limits indicated, favouring analyses at different
levels - most significantly, on different planes - that deploy the same structural
relations and principles of combination. It is thus in accordance with the assumptions that, as in Anderson (1992&), the dependency, or head-modifier, relation
should be seen as basic to both syntactic and suprasegmental phonological
structure, as well as to the internal structure of segments and words, and that
segments and words should both be characterised categorially as complexes of
simplex features. These recurrences on the two planes are mutually supportive;
and, conversely, plane-specific proposals concerning structure must be shown to
follow from some independent difference between the planes. Anderson (1992Z?)
argues that this offers a preferable research strategy to the 'anomalisf position
advocated by Bromberger and Halle (1989), and, pursuing some earlier work
(especially Anderson 1985a; 1986b), attempts to illustrate the pervasiveness of
analogy of structure. Anderson (1987c) discusses some recent manifestations
of the 'analogist' view. However, such a view is of substantive interest only
if, as well as being appropriately constrained, it is pursued systematically and
comprehensively; a theory which permits sporadic borrowings of conceptual
apparatus from one level to another accords to the recurrences no more status
than coincidence.
I assume in the following chapters various aspects of the structure of syntax
and of the grammar that, whatever their internal motivations, receive further
confirmation from analogy. Thus, I shall propose specifically (§2.2) that syntactic structure proper is unlabelled: it is constituted by formal objects we can
represent as graphs, directed by the dependency relation and linear precedence;
and such graphs are projected from well-formed collections of categorial representations, each member of the collection being constituted by the set of notional
features associated with an individual lexical item. Syntax is an unlabelled projection of collections of the categorial properties of lexical items. In this it is
analogous to suprasegmental structure in the phonology, which is projected from
the categories of segmental structure. Units like the syllable, foot etc. are defined

8 Prelude
structurally; such 'categorial' labels (as 'syllable', 'foot' etc.) are unnecessary
and inappropriate (Anderson and Ewen 1987: chs. 2 and 3), just as phrasal
labels are irrelevant to the syntax.
Likewise, I shall propose that the exception with respect to this absence of
labelling in syntactic structure are those feature-attachments I shall refer to as
prosodic (§2.7.5): these are projected on to the syntactic tree from an element
associated with a particular configuration and thus acquire a syntactic domain;
they are manifested primarily as concord. Again, such attachments are appropriate to the characterisation of suprasegmental structure, notably of harmony
'processes' (Anderson, Ewen and Staun 1985; Anderson and Ewen 1987:
§7.6; Anderson 1987a; Anderson and Durand 1988). This is one indication that
syntactic prosodies - or autosegments, if the reader prefers - are not an arbitrary
device.
Perhaps most fundamentally, both syntactic and phonological categories are
represented, as indicated, as complexes of simplex features, where individual
features may simply combine (or be absent) or (where two are present) one may
be subsidiary to the other - there is asymmetry. I interpret this subsidiariness as
reflecting a (category-internal) dependency relation. We have, then, at least the
first-order categorial possibilities shown in (1.5), in the notation presented more
fully in §2.3:
(1.5)

a. {A} ('feature A appears alone in the categorial representation')
b. {A,B} ('features A and B combine')
c. {A;B} ('feature A governs feature B')

Much of chapter 2 in particular is devoted to showing that this provides an
appropriate notation for syntactic categories with respect to the range of 'behaviour' we can associate with such. So, cross-classification and various hierarchical
relations which categories contract are optimally characterised - i.e. in such a

way that simplicity of expression correlates with generality of the particular
phenomenon. However, a powerful (external) support derives from the replication
of like motivations for the same combinatory possibilities in the representation
of phonological segments (Anderson and Ewen 1987: ch. 1).
1.3

Minimalism

The basic unit of the syntax, the word, interfaces with the items listed in the
lexicon; indeed, word-lexical item is the unmarked mapping, to the extent that
others are typically interpreted as special, 'idiomatic'. An important facet of a
lexical item is the valencies and values it assumes as a word; in terms of valency


Minimalism 9
it imposes requirements on accompanying items, and it satisfies the value
required by other items. These valencies and values - subcategorisation and categorisation - are framed in terms of notional features, including semantic relations (a.k.a 'case relations', or 'theta-roles'). A well-formed predication
(realised ultimately as a sentence) has all its valencies satisfied, and only one
value - a predicator - that does not satisfy a valency. Syntactic structure dependency trees with nodes/realising-items ordered in linearity - is projected
from such well-formed predications, with results as shown schematically in
(1.6):
(1.6)

<

{catj}

where V means 'takes as a complement', and thus {category)j} is the value
that satisfies the valency of {catj}, and so can project a syntactic node dependent
on that projected by the latter. In a head-left (centrifugal) (sub-)language, {cat;}

precedes {catj}, as shown in (1.6). I take up again the character of syntactic projection more substantially in §2.2.
The character of these relations - lexical and projective - that the word enters
into introduces an asymmetry between the planes: individual phonological segments as such are not mapped onto lexical items, and suprasegmental structure
in phonology in part interprets morphological, syntactic and informational structures, as well as reflecting the categorial content of the segments comprising the
lexical, morphosyntactic or informational unit concerned (see Anderson 1987a;
Anderson and Ewen 1987: ch. 2). However, as noted, the projected object can
in both instances be represented as a directed graph, specifically a tree, without
labels except for prosodies.
I take it that projection of syntactic structure obeys some minimalist assumptions, apart from its restriction to the introduction of unlabelled nodes (with the
exception of the prosodic), which in itself has the consequence that syntax does
not introduce categorial information. In particular, the erection of syntactic
structure is monotonic: structure may not be erased or elements be changed in
placement ('moved'). This means that, among other things, a syntactic level like
'D-structure' ('associated with S-structure by the rule Move-a' (Chomsky 1981:
18)) cannot be motivated; it is not merely epiphenomenal but non-phenomenal.
We shall discuss such restrictions below in terms of inalterability:


io Prelude
Inalterability condition
The relations of dependency and sequence assigned to an element are inalterable

which forbids mutations, including reassignment of a position to an 'empty category'. I assume further that syntactic nodes cannot anyway be associated with
positions empty of lexical material. I formulate this informally as:
Lexical projection condition
Every syntactic node is projected by a category associated with a lexical item.
There is no 'pro', big or little.
On the other hand, a single item may be associated with more than one node
and these nodes may be dependent on different governors, leading to possible
non-projectivity, of the character of (1.7):

John

may

read

Rasselas

where the solid lines are again dependency arcs, linking head/governor (higher
node) to modifier (lower), and where I have also suppressed the categorial information (for a fuller representation see (3.137) in §3.3.4). The arc linking the
lower John node with that associated with read violates projectivity in so far as
it intersects an association line where there is no node to license this. A part of
the discussion of chapter 3 is devoted to establishing the basis for an understanding of the circumstances in which such non-projectivities are permitted,
and (eventually) to suggesting a rather restrictive condition on their occurrence
in apparently diverse circumstances.
The categorisations associated with particular lexical items may be of varying complexity, such that the categorisation for one item may include that appropriate to another, and sometimes this will be reflected in the morphology: where
one item corresponds to the base of another. Such categorial relationships
are regulated by redundancies. However, categorial redundancies may be of
several different types, of varying particularity. Thus, appropriate to many languages, at least, will be a redundancy, of the type just adumbrated, relating the
categorisation for an action verb to a nominal categorisation which includes the


Minimalism

11

categorisation for the verb and which is to be interpreted as fulfilling the actor
role in the action. Traditionally, this latter might be described as an 'agentnominalisation' (one function of -er in English - Anderson 1984Z?: §3.5), and its
formation be accounted the concern of derivational morphology. As permitting

apparently arbitrary exceptions and possibly showing non-compositionality
('semantic obscuration'), I count such redundancies as lexical: in particular
instances of the relationship, there may be no lexical items fully satisfying either
term of the redundancy. More regular are those redundancies which relate items
of a particular class to different detailed categorial possibilities, such as associating a verb with alternative participial (rather than deverbal adjectival - which
would be lexical) possibilities. Such redundancies, largely the concern, traditionally, of inflexional morphology, I shall term morphosyntactic. Some redundancies are satisfied only in the syntax: notable here is the requirement I shall
propose in chapter 3 that every predication must contain a particular semantic
relation, that I shall label 'absolutive' (a.k.a. 'neutral', roughly 'theme' among
the theta-roles of another tradition); not all predicators are subcategorised for
absolutive, but its presence is necessary to satisfy this particular syntactic redundancy. The satisfaction of the redundancy has interesting consequences for the
resolution of certain apparent problems in providing, within this minimalist
framework, for complex structures involving apparent 'movement'.
However, here, as with the other redundancy types, I am anticipating subsequent discussion which requires more intensive preparation than has been provided so far. In this section (and, indeed, chapter) I have baldly stated, in (for
the most part) a necessarily very provisional form, various assumptions concerning the nature of syntax and its relation to word categorisation which it must
be the role of the following discussion to attempt to justify. In the chapter which
follows we begin that discussion with an elaboration of the general outlines of
a particular version of notional theory and a provisional account of how the
categorial representations which most directly embody this theory induce the
relations of the syntax. Finally in this one, let me draw some further distinctions
which will be important for this elaboration.
Syntactic categories are distinguished in terms of their component combinations of notional features. To begin with, at least, it is useful to distinguish, in a
rather traditional way, and as anticipated in the preceding, between primary
and secondary features and categories. Primary categories are associated with
distinct distributional potentials. Secondary categories involve notional distinctions - such as Tense, or Number - which are reflected in the morphosyntax of
a primary category, such that the distribution of members of a secondary category is included in the distribution of the primary category. We should also note,