Generative Morphology [Reprint 2011 ed.] 9783110877328, 9783110131185

Citation preview

Generative Morphology

Studies in Generative Grammar T h e g o a l of t h i s s e r i e s is t o p u b l i s h t h o s e t e x t s t h a t a r e r e p r e s e n t a t i v e of r e c e n t a d v a n c e s in t h e t h e o r y of f o r m a l grammar. T o o m a n y s t u d i e s d o n o t r e a c h t h e p u b l i c t h e y d e s e r v e b e c a u s e of t h e d e p t h a n d d e t a i l t h a t m a k e t h e m u n s u i t a b l e f o r p u b l i c a t i o n in article form. W e h o p e t h a t t h e p r e s e n t s e r i e s will m a k e t h e s e s t u d i e s a v a i l a b l e to a w i d e r a u d i e n c e t h a n has h i t h e r t o b e e n p o s s i b l e .

Editors: Jan Köster Henk van Riemsdijk Other books in this series: 1. WimZonneveld A Formal Theory of Exceptions in Generative

Phonology

2. Pieter Muysken Syntactic Developments in the Verb Phrase of Ecuadorian 3. GeertBooij Dutch Morphology 4. Henk van Riemsdijk A Case Study in Syntactic

Markedness

5. Jan Köster Locality Principles in Syntax 6. Pieter Muysken (ed.) Generative Studies on Creole Languages 7. AnnekeNeijt Gapping 8. Christer Platzack The Semantic Interpretation

of Aspect and

9. Noam Chomsky Lectures on Government and Binding 10. Robert May and Jan Koster (eds.) Levels of Syntactic Representation 11. Luigi Rizzi Issues in Italian Syntax 12. Osvaldo Jaeggli Topics in Romance Syntax 13. Hagit Borer Parametric Syntax 14. Denis Bouchard On the Content of Empty Categories 15. Hilda Koopman The Syntax of Verbs 16. RichardS. Kayne Connectedness and Binary 17. Jerzy Rubach Cyclic and Lexical

Branching

Phonology

Aktionsarten

Quechua

Sergio Scalise

Generative Morphology

¥

1986 FORIS PUBLICATIONS Dordrecht - Holland/Riverton - U.S.A.

Published by: Foris Publications H o l l a n d P.O. Box 509 3300 A M D o r d r e c h t , T h e Netherlands Sole distributor for the U.S.A. and Foris Publications U.S.A. P.O. Box C-50 Riverton N . J . 08077 U.S.A.

Canada:

C IP-DAT A Scalise,

Sergio

Generative Morphology / Sergio Scalise. - Dordrecht [etc.] : Foris. - (Studies in Generative G r a m m a r ; 18) ISBN 90-6765-089-7 b o u n d ISBN 90-6765-090-0 paper SISO 805.3 U D C 801.55 Subject heading: m o r p h o l o g y ; generative g r a m m a r .

First edition 1984 Second edition 1986

The a u t h o r would like t o thank M a r i n a D r i g o for compiling the indexes. ISBN 90 6765 089 7 (Bound) ISBN 90 6765 090 0 (Paper) © 1984 Foris Publications - Dordrecht. No part of this publication may be reproduced or transmitted in any f o r m or by any means, electronic o r mechanical, including photocopy, recording, or any information storage and retrieval system, without permission f r o m the copyright owner. Printed in the Netherlands by I C G Printing, Dordrecht.

To Irene

Table of Contents Preface

IX

Acknowledgments

χI

Chapter I: The transformationalist treatment of word formation 1. The lexicon: from marginal to central 1.1. Syntactic Structures 1.2. The Standard Theory 1.2.1. Lexical Insertion 2. Word formation as transformations 2.1. The sentence as the source of compounds 2.2. Deletion of lexical material 2.3. Variability in the meaning of compounds 2.4. Absolute exceptions 3. Summary

ι 1 1 3 6 g 9 11 12 13 14

Chapter II: Lexicalist morphology 1. The Lexicalist Hypothesis (Chomsky 1970) 1.1. Consequences for derivation 1.2. Word stress rules 2. Prolegomena to a theory of word formation (Halle 1973) 2.1. The model 2.1.1. Word Formation Rules 2.1.2. The Dictionary and Lexical Insertion 2.1.3. Summary 2.2. Relevance of Halle's theory 2.3. Some criticisms of Halle's model 3. Summary

17 17 20 22 23 24 27 30 30 31 32 34

Chapter III: Word formation in generative morphology 1. Morphemes and words 1.1. The Word Based Hypothesis 1.2. Goals of a morphological theory 2. Word Formation Rules 3. Restrictions on Word Formation Rules 3.1. The base 3.1.1. Syntax and semantics 3.1.2. Phonology 3.1.3. Morphology

37 37 40 40 42 44 44 45 46 48

VI

Generative

Morphology

3.2. The output 4. Summary

51 54

Chapter IV: Readjustment rules 1. Readjustment Rules 1.1. Truncation Rules 1.2. Allomorphy Rules 2. Justification of Readjustment Rules 2.1. Readjustment Rules and Word Formation Rules 2.2. Readjustment Rules and Phonological Rules 3. Summary

57 57 58 60 61 63 66 67

Chapter V: Lexical formatives and word formation rules 1. Words and stems 1.1. Learned stems 2. Representation 2.1. External Boundaries 2.2. Formatives of the lexical component 2.2.1. Simple words 2.2.2. Stems 2.2.3. Affixes 2.2.4. Inflectional morphemes 2.3. Class I and Class II Affixes 2.3.1. Level Ordered Morphology 2.3.2. Level Ordered Morphology in Italian and Dutch 3. Compounding 3.1. The Variable R Condition 3.2. The "IS A" Condition 3.3. Boundaries in compounds and the Extended Level Ordering Hypothesis 4. Well formedness conditions 5. Summary

71 71 75 76 76 78 78 79 79 81 81 85 87 90 90 92 93 95 97

Chapter VI: Interplay between morphological rules 1. Strong Lexicalist Hypothesis 2. Derivation and Inflection 3. Compounding and Derivation 3.1. The Extended Ordering Hypothesis in English 3.2. The Extended Ordering Hypothesis in Italian 4. Compounding and Inflection 5. Some bordeline cases 5.1. The Past Participle 5.2. Evaluative Suffixes 6. Summary

101 101 102 115 116 119 122 127 127 131 133

Chapter VII: Constraining word formation rules 1. The Unitary Base Hypothesis

137 137

Table of Contents

2.

3. 4. 5.

6.

1.1. The Modified Unitary Base Hypothesis 1.2. Ν, V, A + suffix 1.3. N, V + ata 1.4. N, V + ino 1.5. One suffix or two? The Binary Branching Hypothesis 2.1. Parasynthetics 2.2. The suffix -istico The Ordering Hypothesis The N o Phrase Constraint Blocking 5.1. Productivity 5.2. Blocking and the Blocking Rule 5.2.1. Blocking and the Blocking Rule in French Summary

VII 138 140 141 143 144 146 147 150 151 154 156 157 158 160 163

Chapter VIII: Morphology and syntax 1. Word Formation Rules and Transformations 1.1. Locality 1.1.1. The Adjacency Condition 1.1.2. The Atom Condition 1.1.3. A comparison between the Adjacency Condition and the Atom Condition 1.2. Subcategorization Frames 1.2.1. Subcategorization Frame and Syntactic Category 1.2.2. Morphology and Argument Structure 2. Clitics 3. Interaction between Morphology and Syntax 3.1. Word Bar Theory 3.2. Inflection 4. Summary and conclusions

167 167 169 169 171

Symbols and Abbreviations

201

Subject Index

205

Affix Index

211

Word Index

215

Index of Names

231

Bibliography

233

173 178 178 181 183 185 186 191 197

Preface

For every scientific discipline there are fruitful periods and periods of decline. Morphology was a very central field in the structuralist period, both in the European and in the American tradition. Unfortunately, the pendulum swung back with early generative grammar, mainly because of the priority assigned to syntax. A sign of this lack of interest in morphology during the sixties is the fact that morphology was not supposed to account for a specific set of problems: sometimes it was attached to the syntactic component (morphosyntax) and sometimes to the phonological component (morphophonology). In the last ten years, on the contrary, morphology has received a great deal of attention. It has, in fact, become an entire subcomponent of the grammar (morphological component), which is now thought to operate in an autonomous way with respect to the other components of the grammar. Today, an explicit formal status is attributed to morphological rules, and the study of the properties of these rules has lead us to discover that in the lexicon there are many more regularities than we originally imagined. As has been pointed out, morphology today is a micro-system, with a dictionary of primitives (words, stems, affixes, etc.), formal rules (Word Formation Rules) and abstract principles that govern the form and the functioning of the rules (adjacency condition, unitary base hypothesis, etc.). In this book, we will examine the historical context in which the "new" generative morphology has evolved (Chapter I), the work that "founded" the field, Halle's 1973 proposal (Chapter II) and the first theoretical, non episodic, proposal, that of Aronoff 1976 (Chapters III and IV). Following this, a model of the organization of the lexical component is given, along with some well formedness conditions (Chapter V), In addition, a model is proposed for the interplay among the various types of morphological rules, namely derivation, inflection and compounding rules (Chapter VI). Next, a closer look is taken at word formation rules, and a number of constraints on these rules are examined (Chapter VII). Finally, it is argued that while morphology and syntax must be considered separate subcomponents of the grammar, they interact in interesting ways (Chapter VIII). As is clear throughout the book, the works of Halle (1973), Siegel (1974), Aronoff (1976), Allen (1978) are seen as basic in the development of the relatively homogeneous theory of generative morphology as it is

X

Generative

Morphology

understood today, and our debt to these works is great. Nevertheless, the morphological model proposed in this book differs from each of the models offered in the works cited above. Most differences derive from the facts that cross-linguistic evidence has been taken into account wherever possible, and that the final proposal also draws on more recent developments in morphology (e.g. Lieber 1980, Williams 1981a, Selkirk 1982) which have considerably improved our understanding of the field. Of course, much work remains to be done in the relatively new field of morphology.

Acknowledgments

First, I would like to thank Pieter Muysken of the University of Amsterdam for having called my attention to the need for a book on morphology. In fact, this book was begun during my stay from 1979 to 1980 at the Institute of General Linguistics at the University of Amsterdam, a true "paradise" for a theoretical linguist, because of the richness of the debate, and the diversity of the points of view which it offered, as well as its excellent library. I am indebted to the C N R / Z W O for a scholarship which helped make this year in Amsterdam possible. While in Holland, I benefited greatly from lengthy discussions with members of the Linguistic Institute in Amsterdam, in particular Hans den Besten, Simon Dik, Pieter Muysken, Henk van Riemsdijk, Norval Smith and Irene Vogel. I also had very interesting discussions with other D u t c h colleagues, in particular, Minne G. de Boer of the University of Utrecht, Geert Booij of the Free University of Amsterdam, and Harry van der Hulst and Michael Moortgat of the University of Leiden. I would like to extend my warmest thanks to Mark Aronoff, Pieter Muysken and Irene Vogel for having read an earlier version of this book from beginning to end; their comments and insights have been most valuable. In particular, I thank Irene Vogel for her help throughout every phase of this book's preparation, especially for the irreplaceable combination of encouragment and constructive criticism she offered me. I also thank Enzo Lo Cascio and Marina Nespor of the University of Amsterdam, Maria Luisa Altieri Biagi, Andrea Fasso and Luigi Rosiello of the University of Bologna, Giorgio Graffi and Paolo Ramat of the University of Pavia, and Giulio Lepschy of the Reading University who read and commented on earlier versions of parts of the manuscript. T h a n k s also go to my students in the courses from 1981 to 1982 and from 1983 to 1984 at the University of Venice for their helpful criticism. I am also greatly indebted to the person who taught me respect both for linguistics and for research in general, Luigi Heilmann of the University of Bologna, who has been a true "maestro" for me. Special thanks go to my son, Sandro, who helped me through critical periods, showing great patience and maturity. Finally, I would like to express my gratitute to the editor of Foris, Henk La Porte, for his infinite courtesy and to the directors of the series, Henk van Riemsdijk and Jan Köster, for their constant encouragement.

Chapter I

The transformationalist treatment of word formation In this chapter, we will first show how the treatment of the lexicon has evolved from Syntactic Structures to the "Standard Theory". We will then discuss the only systematic proposal for word formation advanced in the prelexicalist framework, that of Lees (1960); first the fundamental aspects of the theory will be presented and then its inadequacies will be pointed out.

1. The lexicon: from marginal to central The development of Transformational Generative Grammar from its beginning up to the present, can be seen, among other ways, as a progressive refinement of the structure of the lexical component'. This does not mean that the evolution within the theory was motivated by considerations having to do with the lexicon itself; in fact, the opposite is true. That is, the changes in the organization of the lexicon followed from changes proposed for the organization of the transformational component, the categorial component and even the phonological component. The fact remains, however, that the lexicon, in the beginning, was conceived of simply as a list of lexical formatives, while today it is thought of as having a complex internal structure which is capable of handling a wide variety of phenomena. It is for this reason that the organization of the lexicon has become an important part of the theory of grammar. 1.1. Syntactic Structures In Syntactic Structures, the lexicon is not an autonomous component; the rules that introduce lexical items are the last rules of the categorial component. The categorial component thus includes two types of rewriting rules: phrase structure rules and rules that insert lexical items. The two sets of rules are not formally differentiated, as can be seen in (1): (1)

(i) (ii) (iii) (iv) (v) (vi) (vii)

S NP VP ->· Aux-> Det -> Ν -» V

N P + Aux + VP Det + N V + NP pres., past the ... aunt, book ... take, read, walk

2

Generative

morphology

A phrase marker that can be constructed on the basis of the rules in (1) is given in (2):

The problems with this proposal, however, are obvious. That is, the grammar in (1) also generates ungrammatical sentences such as those in (3): (3)

(i) *the aunt walks the book (ii) *the book reads the aunt

In (3i), the non grammaticality arises from the fact that an intransitive verb is followed by an object and in (3ii), it arises from the fact that read has an inanimate subject and an animate object. As far as the goals of the then new theory were concerned, sentences such as those in (3) did not constitute a problem, and in Syntactic Structures, Chomsky already suggested some possible solutions to this problem. In relation to (3ii), Chomsky observed, in fact, that in order to develop not simply a fragment of a grammar but rather a complete grammar, it is necessary to impose many restrictions on the choice of the V(erb) in relation to the subject and object, such that sentences such as John admires sincerity are allowed but the inverse type of non-sentences such as sincerity admires John are excluded (Chomsky, 1957:42). As far as (3i) is concerned, on the other hand, the mechanisms presented in Syntactic Structures allowed for a solution of the type seen in (4), where (4iii) represents a modification of the original rule in (liii) and (4vii) represents a modification of the rule in (lvii): (4)

(iii) VP

Vtr+N ^intr

(vii) V t r take, eat ... Vintr -* walk, slip ... The modifications in (4) provide (a) a rewriting rule for VP with two possibilities: transitive verbs and intransitive verbs, and (b) two "lexical rewriting rules": one for transitive verbs and one for intransitive verbs. The solution proposed subsequently within the framework of the Standard

The transformationalist

treatment of word formation

3

Theory, however, was quite different since Chomsky, in Aspects, rejected the "syntactic" solution in (4), in favor of a "lexical" solution, which will be outlined in the next section. 1.2. The Standard Theory As far as morphology is concerned, the most important modification in the development of the Standard Theory in Aspects is the separation of the lexicon from the rewriting rules. It is worth noting that Chomsky considered this separation a substantial revision of the theory, one that affects its generative power and offers considerable advantages. Let us now briefly consider these changes. First of all, as Chomsky points out, many of the properties of the formatives can, with the revisions, be specified directly in the lexicon. This permits us to simplify the grammar significantly since many of the properties of lexical formatives are, in fact, irrelevant to the functioning of the base rules and, furthermore, are often idiosyncratic. For example, the fact that there are two classes of transitive verbs, those that allow deletion of the object and those that do not, no longer has to be handled by rewriting rules. Instead, verbs such as read and eat, that allow the deletion of the object, and verbs such as frighten and put, that d o not allow deletion, are specified in different ways in the lexicon with respect to the syntactic feature for the deletion of the object. The transformational rule that deletes the object thus applies only to those words specified positively for this feature 2 . A second point is the fact that in the Standard Theory, the lexicon forms part of the subcomponent of the base, but it is a (sub-sub) component separate from the rewriting rules; it consists of an unordered list of lexical items and a set of redundancy rules. Each lexical entry contains information about the syntactic, semantic and phonological properties of the specific lexical item, as well as any possible idiosyncratic information 3 . This information, together with redundant information, is specified in ordered sets of syntactic, semantic and phonological features, respectively. Limiting ourselves here to their syntactic properties, 4 we can represent the lexical entries of a small sample of lexical items as in (5), where anim = animate, hum = human, com=common, abstr = abstract, s t r = s t r o n g and prog = progressive. The symbol " + " before a feature or a category indicates that the item in question has that feature or is of that category; " —" indicates that the item does not have that feature or is not of that category; and "—" indicates the position in which the lexical item in question can occur in a given context.

(5)

lexical category

inherent features

strict subcategorization

John

[ + N]

[ +anim] [ +hum] [-com] [ + count] [ — abstr]

[-Det

]NP

selectional restrictions

4

Generative morphology lexical category

inherent features

strict subcategorization

boy

[ + N]

[ + Det

]NP

rabbit

[ + N]

[ + Det

]NP

book

[ + N]

[ + Det

]NP

patience

[ + N]

[ + Det

JNP

water

[ + N]

[ + Det

]NP

frighten

[ + V]

[ + anim] [ + hum] [ + c o m ] [ + c o u n t ] [ — abstr] [ + anim] [ — hum] [ + c o m ] [ + c o u n t ] [ —abstr] [ — anim] [ — hum] [ + c o m ] [ + count] [ — abstr] [ — anim] [ — hum] [ + c o m ] [ —count] [ + a b s t r ] [ — anim] [ — hum] [ + c o m ] [ — count] [ — abstr] [ - s t r ] [ + prog]

[+

NP]VP

climb

[ + V]

[ - s t r ] [ + prog]

[+

NP]VP

chase

[ + V]

[ - s t r ] [ + prog]

[+

NP]VP

read

[+V]

[ + s t r ] [ + prog]

[+

know

[+V]

[ + str][-prog]

[ + _

selectional restrictions

ΓΝΡ) CL

Ο

J

)]VP

NP [ + hum]

]VP

NP [ + anim]

fNPl „

Ο

J

NP [ + anim] NP [ + anim] NP NP [ +anim] [ +anim]

(5) is an approximation of how lexical items would be represented in the Standard Theory Lexicon. Summarizing, we can say that each lexical entry contains the following information: (6)

(a) lexical category (b) inherent features (c) contextual features (i) strict subcategorization (ii) selectional restrictions

The redundant phonological semantic and syntactic properties are specified in terms of redundancy rules (cf. Chomsky, 1965). For example, the syntactic regularity by which the feature [ + h u m a n ] implies the feature [ +animate] is not specified for each lexical item, but is established by a general rule. Consider now two specific cases: boy and frighten. As we see in (5), boy is a noun that can appear in the context Det , and together with a Det, forms a N P (as distinct from the proper noun John, which does not appear after a determiner, cf. *the John). It is, furthermore, a noun that has the features animate (cf. the boy chases the rabbit vs. *the book chases the rabbit), human (cf. the boy reads a book vs. *the rabbit reads a book), count (cf. did you see those six boys? vs. *did you see those six waters!), non abstract (cf. *the boy is a virtue vs. patience is a virtue), etc. Frighten, on the other hand, is a verb, one that is obligatorily transitive (cf. John frightens the boys vs. *John frightens), and requires an

The transformationalist treatment of word formation

5

animate object (cf. John frightens the rabbit vs. *John frightens the book). The inherent features of this verb include [ — strong] (ci. frighten/frightened/frightened) and [ +progressive] (cf. John is frightening the boy vs. *John is knowing the answer). A third revision introduced in the Standard Theory is the fact that lexical items are inserted into deep structures (or more correctly, in the place of dummy symbols in the deep structure) by the rule of lexical insertion (LI). This rule must take into account both the dummy symbol (whose place is filled by the lexical item) and the nature of the context. For example, LI inserts the verb frighten in the place of a dummy symbol dominated by V and followed by a N P (object) that has the feature [ +animate]. It should be emphasized that the representations in (5) are not complete, and, as they stand, generate a large number of ungrammatical sentences. An exhaustive proposal (which goes beyond the scope of this section) would have to enrich the framework with many additional specifications in order to avoid the creation of such sentences. That is, while the representations in (5) correctly generate the grammatical sentences in (7i) but not the ungrammatical ones in (7ii), they also incorrectly generate the ungrammatical sentences in (7iii): (7)

(i) generated, grammatical John knows the boy the boy frightens the rabbit the boy reads the book etc. (ii) not generated, ungrammatical *John is knowing the boy *the rabbit frightens the patience *the book chases the rabbit etc. (iii) generated, ungrammatical *the boy reads the patience *the rabbit climbs the water *John reads the rabbit etc.

It is clear that the proposal put forth in the Standard Theory represents an interesting step in the direction of a richer hypothesis about the organization of the lexicon. In particular, lexical entries are attributed with more grammatical information than in the Syntactic Structures model. This grammatical information (categorial labels, features, subcategorization frames) results in a more organized view of the lexicon; it allows a crossclassification of all lexical items and it determines certain aspects of their syntactic behavior. The information associated with the lexical items is also crucial for the operation of the Lexical Insertion Rule, whose properties will be examined briefly in the next section.

6

Generative morphology

1.2.1. Lexical

Insertion

The rule of Lexical Insertion inserts a lexical item in the position indicated by the dummy symbol Δ in the abstract phrase marker. The status of this rule, however, is not totally clear. It is often called the "first transformational rule", although it is obviously different in nature from the transformational rules. That is, while transformations move categories, LI introduces lexical material into certain positions. LI must, furthermore, apply more than once for each sentence. Finally, it should be noted that LI is also different from the phrase structure rules. That is, while phrase structure rules rewrite a symbol (e.g. NP) as a string of other symbols (e.g. Det + N) independently of the symbol that dominates the one that is rewritten (i.e. S or VP), LI, on the other hand, must take into account both the node that dominates Δ and the nature of the context of Δ. In other words, LI is a contextual rule. The following is a proposal for the formulation of the rule of Lexical Insertion made by Bach (1974:108-9): (8)

For every lexical entry Ε containing a phonological matrix P, a set of inherent features [ + Aj, αϊ A2,... α « - ΐ A n ], and contextual features [ + Xi — Yi, +X2-Y2> · · · , + X m _ Ym]> w e define a substitution transformation: SA: 1: W,Xi, [ Α ι Δ ] Υ ι , Ζ 2: W , X 2 , [ A l A ] Y 2 , Z

rn:W,Xm,[AlA]Ym,Z 1 2 12

3 4 Ε 4

5 => 5

According to this formulation, the contextual features provide the set of structural conditions for the transformation that replaces the dummy symbol (Δ) when this symbol is dominated by the lexical category of the entry in question. A concrete example of Lexical Insertion is given below, where the items to be inserted are John and read. (9)

(i) lexical entries: John, ([ + N ] , [ —Det

]>jp, [ + human], [ — common]

...) read, ([ + V], [ + [ + strong] . . . )

fNP] L h y p , [NP

],

[ + hum]

The transformationalist

treatment of word formation

(ii) deep structure (or abstract phrase marker): S NP

Aux

VP

I Δ

pres.

I Δ

(iii) lexical insertion of John5: SA: X, [ [Δ]ν]νρ, SC: 1 2 1 John + N [-Det JNP + human — common

(iv) lexical insertion of read: SA: Χ, [[Δ]ν]νΡ, SC: 1 2 1

Y 3: 3

Y 3 =

read +V NP + [-

]VP

NP_ [ + hum] + strong

(v) resulting initial phrase marker: S NP

Aux

VP

John

pres.

read

7

8

Generative

morphology

The whole procedure can be summarized in the following way: in (9i) we begin with two lexical items as they are represented in the lexicon, in (9ii) the deep structure is generated by the rewriting rules, in (9iii) and (9iv) the rules of Lexical Insertion replace the dummy symbols with the appropriate lexical items, and in (9v) the result is what is usually called the "initial phrase marker", that is, the structure which enters the transformational component. The rule of Lexical Insertion just seen must conform to the following principle proposed by Keyser and Postal (1976:182): (10)

Lexical Insertion Principle (a) Every occurrence of Δ must be replaced by an item from the lexicon. (b) An occurrence of the symbol Δ may be replaced by a lexical item only if the lexical item is marked as a member of the lexical category which immediately dominates that particular occurrence of Δ. (c) Otherwise, replacement of Δ by lexical elements from the lexicon is syntactically constrained only by contextual restrictions built into particular entries [ . . . ] .

This principle ensures that each dummy symbol is replaced by a lexical item (10a), that a lexical item such as read is inserted under the node V(10b), and that the insertion takes place in accordance with the contextual restrictions, so that, for example, an intransitive verb is not inserted before a noun object. In the rest of this book, we will not be concerned further with the Lexical Insertion rule. It should be born in mind, however, that LI is an important rule because it inserts into the initial phrase markers only those units that are defined as "words of the language", that is, no "more" (i.e. phrases) and no "less" (i.e. bound forms) 6 . Finally, it should be noted, that there are several important analogies (which we will not, however, explore further here) between LI and the rules of the lexical component. It suffices here to mention the following two points: (a) the type of information necessary for the operation of LI is the same that which is needed for the operation of morphological rules (cf. Chapters III and IV), and (b) it is possible that a rule of lexical insertion operates within the lexical component as well (that is, an operation that inserts "words" into the "word structures") similar to the one we have just seen (cf. also Lieber 1980).

2. Word formation as transformations Within the framework briefly outlined above, the only items in the lexicon were simple words; neither compounds nor derived words had a

The transformationalist

9

treatment of word formation

place there. Complex words, therefore, had to be formed by "rules": The only place where they could be constructed was the transformational component, the only device capable, at that time, of expressing relations. This was true for both compounds and derived words, as will be illustrated below. 2.1. The Sentence as the Source of Compounds The most exhaustive treatment of nominal compounds within a transformational framework is that of Lees (1960). Lees takes his lead from the following assumption: "English nominal compounds incorporate the grammatical forms of different sentence types, and of many internal grammatical relationships within sentences, such as subject-predicate, subject-verb, subject-object, verb-object, etc." (p. 119). In other words, Lees proposes that compounds are generated by transformations from underlying sentence structures in which the grammatical relations that hold, implicitly, between the two formatives of the compound are expressed explicitly. Consider the sentence Archie needs a manservant. The compound manservant, according to the proposal made by Lees, would have the structure seen below in (11); the deep structure of the entire sentence is given in (12) (cf. Botha, 1968:44). (11)

#

Δ

Δ

/

[the] [servant]

pres.

be

Δ

Δ

[a] [man]

s

(12)

NP

ν Ν

Det

Δ [Archie]

pres.

Δ

ί

[need]

[a]

Ν

S'

Α Δ [servant]

[the]

[servant] pres.

be [a] [man]

10

Generative morphology

In order to pass from the deep structure in (12) to the surface structure, according to the model proposed by Lees, the following procedure must be carried out 7 : (i) Apply a relative transformation to the embedded S', replacing its leftmost N P (the servant) with who and erasing the boundaries " # # " between who and the rightmost N P of the highest S (the servant). The result is # Archie pres. need a servant who pres. be a man #. (ii) Apply the Wh-deletion transformation to the result of (i), to give # Archie pres. need a servant a man # . (iii) Apply to the result of (ii) the Noun Shift transformation that deletes the DET (a) to the left of man and then shifts the Ν (i.e. man) to the position immediately to the left of the rightmost Ν of the highest S, in this case servant. The result is # Archie pres. need a man servant # . (iv) Apply a transformation to shift Aux (i.e. pres.) to the position to the right of the verb (need), yielding # Archie need pres. a man servant # . (v) Apply two transformations that introduce boundaries in the appropriate places: first, insert inter-word boundaries to give # Archie # need pres. # a # man servant # ; second, insert intraword boundaries "—" to give # Archie # need- pres. # a # man-servant #. As can be seen from this procedure for arriving at the compound manservant, the entire operation is extremely complicated, and it is difficult to imagine the set of operations that would be necessary to derive more complex compounds such as a wh movement rule applicability condition, or the Dutch compound landbouwmachineonderdelententoonstellingsgebouw "lit. agricultural machine parts exposition building" (cf. Booij, 1977). Before examining other criticisms, however, let us first consider the arguments Lees offers in favor of his transformational approach. Lees's arguments are essentially of a semantic and syntactic nature, and can be summarized in the following three points: (i) "Nominal compounds are understood on the basis of certain fixed syntactic relations (subject, object, etc.) which are specifiable only in terms of relations among constituents of underlying sentences" (Lees, 1960: xxxix); thus, in the derivation seen above it is understood that man and servant are in the relation "subject-predicate". (ii) The transformational treatment can explain the "multiple ambiguity" of compounds. That is, if the meaning of a compound is ambiguous, it is possible to make this ambiguity result from different deep structures corresponding to the different meanings. Thus, for example, the ambiguity of the compound snake poison can be accounted for in "grammatical" terms, without resorting to extralinguistic knowledge, by deriving the different meanings from the deep structures that underlie the following three sentences: (13)

X extracts poison from the snake The snake has the poison The poison is for the snake

(iii) The transformational treatment of compounds can account for the

The transformationalist treatment of word formation

11

intuition that windmill and flour mill represent different "grammatical" structures despite the fact that in surface structure this difference is lost and both compounds are of the form Noun + Noun. That is, the two compounds are derived from different deep structures, corresponding to the two sentences below: (14)

Wind powers the mill The mill grinds the flour

Such arguments are not very convincing, however, as has been observed, for example, by Booij (1977) and Allen (1978). The sentential origin of compounds is intended to account for the meaning of a given compound on the basis of the grammatical relations between the two (or more) constituents involved, but the observed regularities can, in fact, also be expressed by other types of rules, such as lexical rules. A more serious problem that arises from Lees's proposal, however, is a formal problem that concerns the deletion of lexical material and the excessive power of the transformations required by such an approach. This problem will be examined in some detail in the following section. 2.2. Deletion of Lexical Material In the series of transformations seen above in the derivation of manservant, nothing was said about the verb be in the embedded sentence, and it must be assumed that it was deleted. Similarly, to derive wind mill it is necessary to delete the verb power (cf. the first sentence in 14), and to derive the compound car thief, it is necessary to delete the verb steal, assuming that the deep structure of the compound is a sentence like the thief steals the car. Such deletion transformations, made necessary by the sentential origin of compounds, were also proposed by Meys (1975) to derive the word sea breeze, for example, from an expression such as breeze from the sea. It was already clear, however, by the mid 1960's, with Aspects, that this type of unrestricted transformations could not possibly bring us closer to an adequate characterization of the notion of "natural language", and that, furthermore, a grammar that incorporated such rules would lose any possibility of being "explanatory". In Katz and Fodor (1964) and Chomsky (1965), in fact, a principle was proposed to exclude from the grammar the type of unrestricted deletion operations proposed by Lees. This principle, called "recoverability of transformations", restricted deletions in the following way: (15)

[ . . . ] a deletion operation can eliminate only a dummy element, or a formative explicitly mentioned in the structure index (for example, you in imperatives), or the designated representative of a category [ . . . ] , or an element that is otherwise represented in the sentence in a fixed position (Chomsky, 1965:144-5).

12

Generative morphology

It should be noted, however, that even if it were possible to formulate the transformations involved in the derivation of compounds in such a way as to satisfy the recoverability principle, it would nevertheless be necessary to postulate at least as many transformations as the number of verbs that could be deleted, assuming, of course, that we want to maintain the sentential origin of compounds. It would be necessary, for example, in the cases seen above to make reference explicitly to transformations of "power deletion", "grind deletion", "steal deletion", etc. The deletion of lexical material is the most serious theoretical problem with Lees's proposal, especially in light of subsequent developments in Generative Grammar, where an effort was made to exclude unconstrained types of rules from the grammar. Constraining the rules of the grammar in syntax is a way of bringing us closer to the definition of "possible sentence"; similarly, in morphology, it is a way of bringing us closer to the definition of "possible word". In this respect, it is clear that Lees's proposal is not adequate. 2.3. Variability in the Meaning of Compounds In addition to the problem just discussed, there is another problem with Lees's proposal that makes it quite implausible. The paraphrasis given above for the compound wind mill "The wind powers the mill". Nothing, however, excludes other possible paraphrases such as "The wind activates the mill", "The wind makes the mill function", etc., or even a "passive" paraphrasis such as "The mill is activated by the wind". The answers of speakers asked what a compound such as information office means are, in fact, quite diversified. Instead of there being a single type of response, the answers seem to cluster around a range of possible paraphrases, as opposed to a range of impossible paraphrases (cf. Allen, 1978). Thus, information office might mean an office "that gives information", "(that is) for information", etc.; it can never mean an office "without information", "that destroys information", etc. It should be noted that the notion "range of possible meanings" renders the formulation of a transformation that deletes a specific verb or a preposition impossible. It is often the case, furthermore, that the meaning of a compound is somewhat different from the meaning that would be expected solely on the basis of the deep structure relations. For example, "a green black-board" is perfectly acceptable, while its supposed sentential source, "*a green board which is black", is not acceptable. There are, in addition, other types of idiosyncrasies, for example, the word pale face, that are problematic for any theory, but that are particularly problematic for a theory that postulates a sentential source for compounds. Specifically, a transformational treatment of compounds cannot account for the idiosyncrasies found in this area since there is no way in which a compound derived transformationally can acquire idiosyncratic features. Transformations are regular processes and therefore must not be bound by "lexical exceptions". In the lexicon, however, lexical exceptions

The transformationalist treatment of word formation

13

are the norm, both for compounding and for derivation. The fact that a compound such as lustraborse "purse shiner" is possible but non existent in Italian would require a mechanism to block the transformation that produces this particular word since other compounds of the same type do exist: portaborse "lit. purse carrier = lackey", lustrascarpe "shoe shiner" and portascarpe "shoe carrier". It is clear, however, that the non existence of lustraborse is not due to linguistic restrictions of any sort, but rather to some other nonlinguistic factors, perhaps of a pragmatic nature. A similar problem arises with derivation, where the non existence in Italian of the word mangiamento "eat + men", which is nevertheless a possible word, would require a special blocking mechanism to prevent the application of the nominalization transformation that normally applies to produce such existing words as trattamento "treatment" and avanzamento "advancement". Once again, it is clear that the nonexistence of mangiamento is the result of a lexical, and not a syntactic, idiosyncrasy. The problems mentioned in this section could all be eliminated, or at least more easily resolved, if the place in which idiosyncrasies are handled is taken to be the lexicon. It could thus be assumed that compounds are generated in the lexicon itself, and as a result, the deep structure of the sentence Archie needs a manservant would simply be as follows: (16)

S NP

Aux

VP

V / Det Δ Archie

pres.

NP \ Ν

Δ

Δ

Δ

need

a

manservant

The internal structure of the compound manservant must thus be represented within the lexicon, a point we will return to in Chapter V. 2.4. Absolute Exceptions The last criticism of the transformational treatment of word formation we will consider is the one raised by Jackendoff (1975:646). This criticism is concerned with the cases referred to as absolute exceptions by LakofF (1970:52 ff). In order to account for a word such as perdition, Lakoif proposes that the base is the underlying form perdite, an item which, however, never appears as a word in surface structure. Perdite must, therefore, be marked as a "positive absolute exception", in the sense that it must obligatorily undergo the nominalization transformation. Furthermore, when the same type of "positive exception" can undergo a set of

14

Generative

morphology

different, mutually exclusive, transformations, Lakoff is forced to intro0 duce Boolean combinations of exception features. Such a situation arises, for example, with an item such as aggress which not only is a positive absolute exception, as seen above, but is a form that can be nominalized with -ion (aggression) "or" with -or (aggressor). In addition to being quite complicated 9 , Lakoff's analysis must resort to "exception features" in order to prevent a lexical item from appearing in its basic form, a fact that Jackendoff concludes is "counter-intuitive". In fact, Lakoff is forced to claim that there is a hypothetical verb *king that corresponds to the noun king, just as the verb rule corresponds to the n o u n ruler. The "verb" king, however, would have the singular property that it never appears on the surface. In this case it would be considered an absolute exception. The introduction of the verb king would make the lexicon (in Lakoffs theory) simpler since it would eliminate an absolute exception. According to the transformationalist hypothesis, a lexicon where each n o u n with functional semantic information is related to a verb is, in fact, more highly valued than one in which this is not the case. As Jackendoff comments, however, "Since there is little evidence for such a preference, and since it is strongly counter-intuitive in the case of king, the transformational account - beside requiring a very powerful mechanism, the absolute exception, is incorrect at the level of explanatory adequacy" (1975:640) 10 .

3. Summary In this chapter, we have outlined the theoretical framework within which Lees's proposal was made. It was shown, furthermore, how the lexical component has become more central in the grammar as a result of the development of Lees's theory. Lees's proposal was presented in detail since this proposal was the first one to provide a systematic framework for handling the process of derivation. Such a transformationalist theory of word formation had a number of inadequacies, however, and these were discussed as well. In general, we can conclude that such a theory is no longer tenable today since it has become quite clear that transformations are not the appropriate instrument for handling word formation.

Notes 1. In this book, as in many other writings dealing with the same topic, the term "lexicon" is used with systematic ambiguity. It may refer to the component of the grammar where lexical facts are handled, or it may refer to a simple list of words. The two meanings are often referred to as "lexical component" and "dictionary", respectively. 2. More recently, as an alternative to the transformational treatment (i.e. a transformation deleting the object), a lexical treatment has been proposed by Williams (1981b) for verbs with deletable objects.

The transformationalist

treatment

of word

formation

15

3. In Chomsky's words: "In general, all properties of a formative that are essentially idiosyncratic will be specified in the lexicon. In particular, the lexical entry must specify: (a) aspects of phonetic structure that are not predictable by general rule (for example; in the case of bee, the phonological matrix of the lexical entry will specify that the first segment is a voiced labial stop and the second an acute vowel, but it will not specify the degree of aspiration of the stop or the fact that the vowel is voiced, tense, and unrounded; (b) properties relevant to the functioning of transformational rules [...]; (c) properties of the formative that are relevant for semantic interpretation (that is, components of the dictionary definition); (d) lexical features indicating the positions in which a lexical formative can be inserted (by the lexical rule) in a preterminal string" (Chomsky 1965:87-88). 4. Whether the following list of features is properly "syntactic" or "semantic" is a controversial matter. I will not enter into details concerning this question here, but it is important to point out, nevertheless, that a feature such as [ + a b s t r a c t ] is relevant to morphology, given the fact that it can block derivational processes (cf. Chapter VI). 5. Chomsky (1965:113 ff.) argues against the insertion of the verb before the insertion of the noun. 6. There can be exceptions to this statement. LI can insert also lexicalized phrases, listed as such in the dictionary. On this topic, cf. Scalise 1983. 7. In what follows, we will use the sentence Archie needs a manservant to illustrate the detailed procedure that Botha (1968:45) applies to the analogous sentence John needs a girlfriend. 8. A Boolean system is "a logic or system making use of connectives with the formal properties of not, or and and" (cf. Bach 1974:97). In this example, the Boolean combinations of exception features are intended to account for the fact that a form such as aggress must be transformed (in Lakoff's terms) either into aggression "or" into aggressor. 9. It must be noted that "Boolean exceptions state conditions which must be met in the course of an entire transformational derivation, rather than in the application of a single rule" (Lakoff 1970:68). Lakoff's account, thus, make crucial use of a very powerful device, i.e. "global rules", which has been severely criticized by Chomsky (1972) for syntax and by Jackendoff (1975) for morphology. 10. A definition of explanatory adequacy in a metatheory of the lexicon can be found in Jackendoff (1975:640).

Chapter

II

Lexicalist morphology In this chapter, we will outline the Lexicalist Hypothesis, the proposal that made it possible to treat word formation in a single place in the grammar, the lexicon. In addition, Halle's proposal, the first "lexicalist" theory of morphology, will be discussed in detail. This proposal is contained in a relatively short article, but one which has revolutionized the field of generative morphology. The central points of this work are that word formation occurs entirely within the lexicon and that it is handled by a specific mechanism, Word Formation Rules (WFR's). These two points have been adopted in almost all of the work on generative morphology in the decade following the original proposal. The large body of research inspired by Halle's theory has, however, revealed its programmatic, rather than systematic, nature as will be seen below in the discussion of some of the major criticisms of the theory. 1. The lexicalist hypothesis (Chomsky 1970) In addition to Lees's treatment of compound formation, a transformationalist treatment was proposed within the Standard Theory framework, for another type of word formation process, that which creates derived nominate. In brief, a phrase such as the criticism of the book was derived from the deep structure X criticizes the book. That is, derived nominate such as criticism, refusal, etc. were considered to be the result of a transformation that applied to verbal lexical items, criticize, refuse, etc., which were taken to be the basic forms. The reasons for this treatment were essentially two-fold: a general reason and a specific reason. The general reason was that, within the Standard Theory, any time forms were encountered for which it was thought that the grammar should express some type of relationship, the strategy was to identify a basic form and to derive the other form from this one by applying one or more rule(s). As has been mentioned, furthermore, the burden of expressing relations within the Standard Theory fell on the transformational (and phonological) rules, the obvious, though forced, result was that forms such as criticize and criticism were related by the transformational component. The specific reason was that by using transformations, it was possible to express generalizations about selectional restrictions of verbs and the corresponding nominate. For example, the verb refuse requires an animate subject, a restriction that also holds for the derived nominal, refusal (cf. John refused the offer and John's refusal of the offer). Thus, the non grammaticality of *the table refused the offer and *the table's refusal of the offer is accounted for by the same restriction.

18

Generative morphology

The transformational treatment of derived nominals was criticized, however, by Chomsky (1970) in Remarks on Nominalization particularly on the grounds that in English there are numerous systematic differences between "gerundive nominals" and "derived nominals". The "most surprising" differences are of three types: (a) syntactic productivity, (b) semantic regularity, and (c) the internal structure of the nominals. Each of these points will now be considered separately 1 . (a) Syntactic productivity. While almost all sentences can have a corresponding gerundive nominal, this not the case for derived nominals, as illustrated below: (1)

(i) sentences

(ii) gerundive nominals (iii) derived nominals

It is easy to please John. John amused the children with his stories. It seems that John is gone. its being easy to please John John's amusing the children with his stories its seeming that John is gone *the easiness to please John ""John's amusement of the children with his stories *its seem that John is gone

It is also the case, furthermore, that many derived nominals do not correspond to well formed sentences, or to sentences with corresponding gerundive nominals, as can be seen in (2): (2)

(i) sentences

*John doubted about their proposal. *It is probable John's leaving. *John advised to Bill. (ii) gerundive *John's doubting about their proposal nominals *its being probable John's leaving *John's advising to Bill (iii) derived John's doubts about their proposal nominals the probability of John's leaving John's advice to Bill

(b) Semantic regularity. While derived nominals often have meanings that are idiosyncratic with respect to the corresponding verb, the meaning of a gerundive nominals is always predictable on the basis of the meaning of the verb. Consider the examples mentioned by Chomsky: marriage, construction, actions, revolution, permutation, etc. These words demonstrate that in order to derive the nominal form from the corresponding verb it is necessary to introduce a great deal of idiosyncratic semantic material. In fact, it is difficult to imagine exactly what relation holds between the verb revolve and the noun revolution in the phrase the French revolution, or

Lexicalist

morphology

19

between the verb construct and the noun construction in the phrase the Anglo-Saxon genitive construction, (c) Internal structure. The internal structure of a derived nominal is that of a N P , while the internal structure of a gerundive nominal is that of a S. As a result of this difference, the former may have plural forms, but the latter may not. Thus, John's many refusals of the offer is grammatical, while * John's many refusings of the offer is not. Another result of the difference in the internal structures of derived and gerundive nominale is that the former may be modified by an adjective while the latter may be modified by an adverb. Thus, we have the two phrases: John's sarcastic criticism of the book and John's sarcastically criticizing the book. Finally, gerundive nominals, but not derived nominale, may undergo transformations, such as Dative Shift, that apply to sentences but not to NP's. For example, Dative Shift may apply to John's giving a book to Bill to yield John's giving Bill a book, but it may not apply to John's gift of a book to Bill to yield * John's gift of Bill a book. On the basis of these arguments, Chomsky concludes that derived nominals cannot be created by transformations from an underlying verb and he proposes, instead, a "lexical" treatment of such words, i.e. by means of morphological rules which operate within the lexical component. In fact, Chomsky's proposal to handle the creation of derived nominals in the base, was aimed at limiting the excessive power of transformations by assigning a richer structure to the base component. This shift of power from transformations to the base required a richer theory of phrase structure rules, the so-called X theory, as well as a lexical component that included rules of word formation. It should be noted that while Remarks on Nominalization is only concerned with nominalization, it is clear that nominalizations "are presented as a test case for the validity of the distinction [between lexical and transformational treatment of word formation], and that the methodology introduced here was supposed to apply, at least, to derivational processes in general" (Hoekstra et al. 1979:4). As a consequence of this modification, transformations were allowed to handle only totally regular processes, and not processes governed by lexical exceptions. This led, in turn, to what can be called the "Extended Lexicalist Hypothesis", according to which transformations cannot introduce lexical material. Jackendoff (1972:13) formulates the E L H 2 explicitly as follows:

(3)

[ T r a n s f o r m a t i o n s cannot change node labels, and they cannot delete under identity or positive absolute exception [...]. This means that the only changes that transformations can make to lexical items is to add inflectional affixes such as number, gender, case, person and tense. Transformations will thus be restricted to movement rules and insertion and deletion of constants and closed sets of items.

20

Generative morphology

The Extended Lexicalist Hypothesis marks an important stage in the development of transformational grammar towards what, today, is commonly called the "modular" hypothesis of the organization of the grammar 3 . According to this hypothesis, the grammar consists of sets of "modules" that interact in the generation of sentences, but which carry out simple operations in accordance with specific principles. Of relevance here, is the fact that the ELH prevents transformations from operating on morphological material 4 , a task that is delegated to "morphological" rules. In this way, we avoid encumbering the transformational component with a series of non pertinent problems; instead each "module" of the grammar handles "specific" problems, and is subject to specific principles and constraints. As we have seen, Chomsky's analysis of nominalization had important consequences not only for linguistic theory in general, but also more specifically, for the development of "lexical morphology". As has been mentioned elsewhere (Hoekstra et al., 1979:1), Remarks on Nominalization was not "revolutionary" in itself, but rather, it led to subsequent revolutions. In fact, Remarks has had an effect on the grammar in all of its subparts. The base component has been especially affected, but the transformational component has also undergone changes. Furthermore, in Remarks on Nominalization it is suggested for the first time that the expressive power of a grammar does not necessarily have to be concentrated in the transformational component, but rather, the decision about which component should have which functions is, in the final analysis, an empirical question. Finally, Remarks created the theoretical space for an autonomous morphological component, a possibility that was explicitly excluded in the earliest works on transformational generative grammar. In the next two sections, we will consider two more problems that clearly illustrate the inadequacy of the transformational approach to word formation. 1.1. Consequences for Derivation It was mentioned above that Chomsky's criticism of the transformationalist treatment of nominalization is, in fact, applicable to all derivational processes. In this section, an analysis of the suffix -ificare in Italian 5 will be presented as an illustration of how Chomsky's objections can be extended to the derivation of verbs. Within a transformational framework, the suffix -ificare "-ify" would be introduced by a transformation that applies to an adjectival (or nominal) root as follows: umido —> umidificare "humid - humidify", pur ο -> purificare "pure - purify", solido->solidificare "solid - solidify", etc. The semantic relation in this case seems to be quite regular, that is, "A + ificare" = "make A" or "make more A". Thus, at first glance, there does not seem to be any reason to object to a transformational treatment of derivation in this case. A closer examination of the facts, however, reveals three types of problems that force us to rule out a transformational treatment. Each point is considered separately below.

Lexicalist

21

morphology

First of all, in addition to the cases citated above, there are many other cases in which, given A, it is not possible to form the verb "A + ificare", for example, *tristificare "sad + ify", *bellificare "beautiful + ify", *cattivificare "bad + ify", etc. This is true not only for "simple" adjectives such as triste "sad", bello "beautiful", cattivo "bad", etc., but also for "complex" adjectives (i.e. adjectives that are themselves derived). Thus, starting with mangiabile "edible" and contradditorio "contradictory", for example, we cannot derive *mangiabilificare or *contradditorificare. It is nevertheless possible to formulate the same type of periphrastic verbal expression for both of these types of adjectives as was given for the well-formed verbs in -ificare seen above (cfr. "make high", "make edible"). It should be noted, furthermore, that among those adjectives that do not take -ificare, the first type (triste, etc.), however, often undergoes parasynthetic derivation (a point we will return to in Chapter VII). This process is formally different from derivation with -ificare, but is substantially the same semantically. Thus, we can form intristire, abbellire and incattivire, which mean, respectively, "make sad", "make beautiful" and "make bad". Secondly, there are verbs such as giustificare "justify" and verificare "verify" that have the same morphological structure as verbs such as umidificare and purificare but which do not mean "make just" and "make true", respectively. This shows, once again, that derivational processes are systematically subject to idiosyncracies and accidental gaps, a fact that can be illustrated by the following list: (4)

(i) dolcificare umidificare (ii) *bellificare *vecchificare (iii) solidificare falsificare (iv) *onestificare *correttificare

"sweeten" "moisten" "make beautiful" 'make old" invecchiare "age" 'solidify" *insolidire "make solid" "make false" *infalsire 'falsify" 'make honest" *inonestire "make honest" "correct" *incorrettire "correct" "dulcify" 'humidify" 'beautify"

addolcire inumidire abbellire

In (4i), the base adjectives can undergo both processes of derivation, that is, the addition of -ificare and parasynthetic derivation. In (4ii), the adjective can undergo only parasynthetic derivation, while in (4iii), only derivation with -ificare is possible. Finally, in (4iv), the base adjectives in question cannot undergo either of the two processes. The third problem has to do with the claim that the transformational treatment of derivation allowed selectional restrictions to be formulated only for both the base and the derived word. Thus, in the case of the pair distruggere/distruzione "destroy-destruction", the selectional restrictions formulated for the verb do not have to be repeated for the noun (cf. distruggere una citta "destroy a city", la distruzione della citta "the destruction of the city" vs. *distruggere un assassinio "destroy a murder", *la distruzione deWassassinio "the destruction of the murder"). It is also

22

Generative morphology

true of the adjectives under consideration in this section, that there are cases in which the same selectional restrictions hold for both the base and the derived form (cf. la stanza e umida "the room is humid", umidificare la stanza "humidify the room"; Vacqua e pura "the water is pure", purificare Tacqua "purify the water"). There are, however, other cases in which such a "savings" is not possible. For example, for the pair chiaro/chiarificare "light-lighten", it is not possible to formulate the selectional restrictions only once, as the following examples show. (5)

(i) i suoi capelli sono chiari "her hair is light" (ii) *il parrucchiere le ha chiarificato i capelli "The hairdresser lightened her hair."

The only way in which the transformational approach can account for the idiosyncratic semantic and syntactic differences between a base and a derived form is by listing all the irregularities in the lexicon, thus giving up the possibility of establishing a syntactic relation between the two items. Such a solution has two disadvantages, however. First of all, it multiplies the number of lexical entries which, as Corbin notes, nullifies any economy gained by the transformational approach in the first place. Secondly, it cannot account for the "relation" between a base and an idiosyncratic derived form, which may well be part of the linguistic competence of native speakers. 1.2. Word Stress Rules While Chomsky arrived at the conclusion that derived nominals could not be treated transformationally on the basis of syntactic considerations, Bresnan (1971) arrived at the same conclusion on the basis of phonological considerations. An examination of a number of exceptions to the Nuclear Stress Rule (NSR) led Bresnan to propose a modification of the status of this rule. This modification has had important consequences for the transformationalist/lexicalist" debate, and in particular, for morphological theory. The Nuclear Stress Rule is a cyclic rule that applies after all the stress rules have applied to individual lexical items. It is formulated as follows 6 : (6)

1 V-l

1 stress/[XVY_Z] A

ι where Ζ contains no V, and where A ranges over major categories such as NP, VP, S. Given the convention that the reassignment of 1 stress within a cycle reduces all other stress values by 1, the application of the NSR has the following results in the sentence "Mary teaches engineering".

Lexicalist (7)

morphology

23

[Mary [teaches engineering] y p ] s 1

2

1

1

word stress

2

1

I cycle: NSR

3

II cycle: NSR

The NSR must apply after the transformations, at the end of each cycle. Since it is necessary that word stress already be assigned in order for the NSR to operate, it is clear that the rules of word stress assignment must be ordered before the NSR. That is, they must apply in the lexicon or on the cycle prior to the application of the NSR. It is also the case that transformations cannot change the stress patterns of words after the NSR has applied. These facts were in conflict, however, with the claim that the relation between words such as derive and derivation was a transformational one, since such a claim would require that transformations be able to change the stress pattern of a word (cfr. derive-*derivation). This observation led Bresnan to the following conclusion: (8)

.. .[I]t appears that the stress contours of English sentences are determined in a simple and regular way by their underlying syntactic structures. Further, because prosodic stress rules like NSR require prior assignment of word stress, the latter must occur either on deep structure or in the lexicon. But if word stress is assigned prior to the syntactic transformations, then it follows automatically that transformationally attached affixes are stressneutral. (p. 269)

Bresnan is referring here to a distinction that is often made in English between those affixes that "shift stress" and those that are "stress-neutral" and thus do not shift stress (cfr. Chomsky and Halle, 1968; Siegel, 1974; Aronoff, 1976). The first group includes suffixes such as -ion and -al (cf. prohibit prohibition, ornament-* ornamental) and the second group includes suffixes such as -ment and -ness (cf. advance-*advancement, thdrough-*thbroughness). Bresnan's point is that, from a purely phonological point of view, it is possible to introduce "stress neutral" affixes transformationally but it is not possible to d o the same for those affixes that "shift stress". This is another important step in the direction of the lexicalist hypothesis in morphology. The main argument underlying this development can be summarized as follows: there exists a class of affixes that cannot be introduced transformationally, and therefore, at least for such affixes, the transformational treatment of word formation cannot be maintained. 2. Prolegomena to a theory of word formation (Halle, 1973) The linguist who was actually able to draw the necessary conclusions from

24

Generative

morphology

the criticisms that were being amassed against the transformational treatment of word formation was Morris Halle. In his article, "Prolegomena to a theory of word formation" (1973), Halle advanced the first proposal of an autonomous morphological component within the framework of the lexicalist theory. Although the article was relatively short and "programmatic" in nature, outlining the construction of a model of morphology rather than examining its details, the proposal was, nevertheless, clearly spelled out. It is, in fact, this work that has served as the foundation and inspiration for all the current research on morphology within the generative framework. Given the importance of Halle's proposal, we will examine this work in some detail in the following sections 7 . 2.1. The Model Halle takes his lead from the observation that, if a grammar is a formal representation of a native speaker's knowledge of his language, then there must be a component some place in the grammar that accounts for the speaker's lexical knowledge. A speaker knows, for example, (a) what the words of his language are (e.g. read is a word of English, but lezen is not), (b) that certain words have internal structure (e.g. un-drink-able) and, furthermore, (c) that the internal structure respects a specific order of concatenation of the morphemes involved (e.g. un-drink-able is a possible word but *un-able-drink and *drink-un-able are not). Such knowledge can be represented in the form of a model such as the one below 8 :

List of Morphemes

*

I

Word Formation Rules

Dictionary of Words

Output

Filter

Phonology

-

->

Syntax

Figure 1: Halle's model (1973) In this Figure, the first subcomponent is the List of Morphemes. Thus, according to Halle, the basic units of the lexicon are morphemes, a point we will return to below. Each morpheme is represented as a sequence of phonological segments and is provided with a labelled bracketing, so in the list of morphemes, a verb is labelled V, a noun N, etc. It should be noted that Halle also proposes a labelled bracketing for affixes, but one without a syntactic category label. That is, a suffix is simply labelled "suffix", without any further information. The representations of a noun, a verb and a suffix are exemplified in (9) a, b, and c, respectively.

Lexicalist (9)

morphology a. [home]Ν

25 b. [discuss] v

c. [-ity] Su ffix

In addition to bearing a syntactic label, morphemes are also specified with respect to certain other grammatical properties. Halle proposes that the verb write, for example, must contain information to the effect that (a) it is a verbal root, (b) it is a member of the "non Latinate" part of the list of morphemes (a fact that allows the word formation rules to combine it with certain affixes but not with others), and (c) it belongs to that subclass of verbs referred to as "strong verbs" (i.e. wrote not *writed). The representation of this verb in the "list of morphemes" is as follows: (10)

[write] [ — Latinate] [ + strong]

The second component of Figure 1 is the one that contains the Word Formation Rules (WFR's). The WFR's indicate, among other things, how the morphemes of a language are arranged in sequences to form the actual words of that language. They must, therefore, allow a sequence such as undrink-able and rule out a sequence such as *drink-un-able. In other words, WFR's must be able to generate all the well-formed words of a language and exclude all the ill-formed ones. Not all the words of a language, however, can be derived by means of "regular" and general rules; instead, there exist numerous exceptions in a language. These can be grouped into three types of idiosyncracies: semantic, phonological and lexical. Consider first the following words, examples of semantic idiosyncrasies (cf. Halle 1973:4): (11)

(i) approval recital proposal transmittal

(ii) arrival refusal rehearsal acquittal

The words in (11) have many features in common in their individual meanings (most of them have the meaning "the act of V-ing" and (if transitive) also "the act of being V-ed"), but this is not always the case. For example, recital commonly refers to a public performance given by a soloist and transmittal is used to refer to the transfer of official documents or information, but not of other things that are quite normally said to be transmitted. Consider next the examples of phonological idiosyncracies in the words below. (12)

divine-»divinity obscene -> obscenity profane -»profanity

dev[äy]ne-»div[I]nity obs[iy]ne->obs[e]nity prof [ey ] ne prof [ae] nity

26

Generative

morphology

N o u n s like the ones in (12) are normally subject to the trisyllabic laxing rule as seen in their phonetic transcription, but other nouns with the same form such as entirity, nicety, obesity and scarcity exhibit a different phonological behavior in that they are not subject to the laxing rule. Consider, finally, the examples of lexical idiosyncrasies (or "accidental gaps") given in (13): at-ion approbation recitation proposition derivation description conversion *arrivation *refusation *rehearsion

al approval recital proposal *derival *describal *conversal arrival refusal rehearsal

As these examples show, there are a number of verbal stems from which nominals in both -at-ion (or -ion) and -al can be derived. It is not always the case, however, that both forms are possible: there is a set of verbal stems that only allows nominals to be formed in -at-ion (as in (13ii) and another set that only allows nominals to be formed in -al (as in (13iii))). This last type of idiosyncrasy, in particular, cannot be handled easily within the list of morphemes, nor can it be handled by WFR's, without giving rise to an extremely long and undesirable list of exceptions, precisely in the component that is supposed to operate "regularly". Since these idiosyncrasies are associated with individual formations and are not expressible in terms of regular processes, Halle proposes to treat them with a different mechanism, a "Filter", the third component in Figure 1. That is, WFR's are allowed to apply freely and, together with the List of Morphemes, they define the set of "potential" words of a language. The Filter, then, specifies which are exceptions and adds any necessary idiosyncratic features. In other words, W F R ' s can form words that are phonologically, morphologically, syntactically and semantically well-formed but which, nevertheless, never appear on the surface. These are "possible" but "non existent" words, for example, milkery "place that sells milk", tangerineade "drink made from tangerines", etc. It is the j o b of the Filter to mark such formations with the feature [-Lexical Insertion], The words that actually pass through the Filter constitute the Dictionary of the language, the final component of the model of morphology presented in Figure 1. The Dictionary includes, therefore, the regular formations that the Filter has not modified by adding or removing any features, as well as the idiosyncratic formations that have been modified in some way by the Filter. We will now examine in more detail how the morphological component proposed by Halle works.

Lexicalist

27

morphology

2.1.1. Word Formation

Rules

As far as morphology is concerned, the most relevant innovation of Halle's proposal is the introduction of Word Formation Rules, the mechanism that has been the orienting and unifying force in the research on morphology within the generative framework over the past ten years. Halle distinguishes two types of WFR's, (a) those that apply to stems and form linear sequences of "stem + one or more morphemes", with or without internal structure, and (b) those that apply to words. This difference is illustrated in the examples below, where the W F R in (14i) attaches an affix to a stem 9 and the WFR in (14ii) attaches an affix to a word (cf. Halle 1973:10): (14)

(i) [STEM + i + t y ] N [STEM + ant]A [STEM + al] A [STEM + ther] N [STEM + some] a [be + S T E M ] γ

(ii) [VERB + a l ] N [VERB + a t i o n ] N [ADJECTIVE + (i) + t y ] N [ADJECTIVE + e n ] v [ N O U N + ish] A

According to Halle, there are thus two types of derived words: the type that has the structure shown in (15i), and the type in (15ii) that is derived from other words and therefore has a "somewhat more complex structure": (15)

(i) serendip-I-i + ty vac + ant tot + al bro + ther hand + some be + lieve

(ii) arriv + al condens + ation obes + ity dark + en child + ish

Beyond mentioning that there are two types of WFR's and, therefore, also two types of derived words, Halle did not make use of this distinction. It was Aronoff who, several years later, made an important theoretical contribution on the basis of the distinction. That is, Aronoff proposed that the set of WFR's include only rules of the second type described by Halle, thus founding what he called the "word-based" theory of morphology, a point we will return to in Chapter III. First, however, let us consider in more detail the WFR's of Halle's system. In Halle's model, all WFR's perform a number of tasks, and have the following characteristics: (a) WFR's specify the sequence in which morphemes must be arranged in order to form the words of a language. Thus, as we have seen above, they must allow the grammatical sequence un-drink-able, but exclude the ungrammatical one *drink-un-able.

28

Generative morphology

(b) WFR's specify (i) the syntactic category of the base (e.g. Noun, Verb), (ii) the syntactic category of the output (e.g. Noun, Verb), (iii) the internal boundaries in the output (e.g. " + "), and (iv) regular semantic information. (c) WFR's specify any semantic and syntactic properties of a derived word that do not coincide with those of the base. In this regard, Halle examines the following three cases: (i) cases in which WFR's not only change the lexical category, but also the syntactic features associated with an item. For example, the suffix -hood is attached to nouns with the feature [ + h u m a n ] , however, the result is a noun with the feature [ + abstract] (e.g. boy-»boyhood, priest-»priesthood). In cases of this type, the WFR must assign the feature [ + abstract] to the derived noun, since it is not found in the base, (ii) cases in which pairs of words with different syntactic and semantic properties are not derived one from the other. While in the pre-lexicalist framework a pair of words such as refuse/refusal would have been related derivationally, Halle accepts the "lexicalist" treatment proposed by Chomsky (1970). That is, both items have their own syntactic and semantic properties and there is, therefore, no reason to derive one from the other, as had been assumed in the prelexicalist hypothesis, (iii) cases in which the same selectional restrictions apply at one point in the subcategorization frame of the base and at another point in the subcategorization frame of the derived word. For example, the verbs derived from adjectives by the addition of the prefix en(e.g. rich^enrich, larger enlarge) can have as direct object complements the same nouns for which the adjective can serve as a predicate (e.g. the picture is large, enlarge the picture). It is, therefore, necessary for WFR's to be able to handle selectional restrictions independently of the subcategorization frames in which these restrictions are embedded 10 . (d) WFR's operate in the same way for derivational morphology and for inflectional morphology. Halle claims, as justification of this claim, that certain idiosyncratic behaviors observed in derivation have parallels in inflection, a point we will return to in Chapter VI. Thus, the list of morphemes must contain both derivational affixes and inflectional affixes, or inflectional morphemes, as they are sometimes called. Halle does not propose a specific representation for the inflectional morphemes, but it can be assumed that their representation is similar to that of the affixes, that is, a labeled bracketing with the label "inflectional affix". (e) WFR's are more powerful than phonological rules. That is, while the latter make use only of information explicitly present in the string at the point in the derivation in which the phonological rules apply, the former have access, according to Halle, to diiferent stages in a derivation, in particular, to (i) the Dictionary and (ii) the output of the phonological component. Consider first the case in which WFR's have access to the Dictionary. WFR's define, in part, the content of the Dictionary; however,

Lexicalist

29

morphology

WFR's of the type in 14(ii) must have access to the content of the Dictionary as well, in order to operate. For example, the word transformational, which has internal structure and semantic compositionality, is derived by a rule of the type seen in 14(ii). Such rules specify the category of the base, in this case [ N O U N + al], but the noun transformation is not contained in the List of Morphemes since it, too, is generated by a regular W F R [VERB + ation]. Thus, some of the information necessary for the application of the rule that forms adjectives in -al is found in the Dictionary. This leads Halle to conclude that WFR's must be able to make use not only of the information contained in the string to which they apply, but also of information contained in another sub-component, the Dictionary. That is, "certain words presuppose the existence of other words" (Halle, 1973: 13). It should be noted, further, that if a complex (i.e. derived) word contains as a subpart a sequence of morphemes already listed in the Dictionary as an existing word, any idiosyncratic features associated with that word will also be contained in the newly derived word. For example, since the word transformational has a special meaning in linguistics, this (idiosyncratic) meaning will also be present in the derived word transformationalist. Consider next the case in which WFR's have access to the output of the phonological component. For example, the incohative verbs in English formed by adding the suffix -en to a monosyllabic adjective are subject to the additional phonological restriction that the base must end in an obstruent, optionally preceded by a sonorant. Thus, the words in (16i) are well formed, while those in (16ii) are not: (16)

(i) blacken harden dampen

(ii) *dryen *greenen * laxen

The phonological restriction, however, seems to be a condition not on the string to which the morphological rule adds the suffix, but rather on the string that results after the application of the phonological rules. Consider the examples in 17: (17)

soften fasten moisten

[sofan] [fasan] [moysan]

The strings soft + en, fast + en and moist + en violate the phonological restriction mentioned above since they end in an obstruent preceded by another obstruent, not by a sonorant. The suffix -en, therefore, should not be able to attach to these bases. If, however, the phonological rule that deletes the t is allowed to operate, the result is a single obstruent, s or f

30

Generative morphology

(preceded by a vowel), a context that is acceptable for the suffixation of -en. According to Halle, if this is the correct explanation for the well formedness of moisten, as opposed to the ill formedness of laxen11, it means that WFR's must have access not only to the Dictionary but also to the output of the phonological component. In other words, Halle is making the claim that a WFR produces acceptable words if the words created by a given rule conform to certain phonological conditions only after the phonological rules have applied. (f) While WFR's are undoubtedly rules that create new words, Halle makes the claim, implicitly, that they also serve to analyze already existing words. Halle suggests, furthermore, at the end of his article, that although the content of the Dictionary is entirely determined by the List of Morphemes, WFR's and the Filter, it is not necessary to assume that all of these sub-components operate in the production of every speech act (p. 16). Instead, it can be assumed that a large part of the Dictionary is stored in the long term memory of a speaker and that WFR's are activated only when the speaker hears a word that is unfamiliar to him, or when he invents a new word. 2.1.2. The Dictionary and Lexical

Insertion

The Dictionary contains only (and all) the inflected forms of the words of a language. This proposal, however, quickly runs into difficulty, as Halle, himself, points out. For example, certain inflectional forms (e.g. case markers) are determined by the position of a word in surface structure, whereas Lexical Insertion is assumed to take place at a much earlier stage 19 in derivation . To overcome this difficulty, Halle proposes that the Lexical Insertion rules, instead of inserting a single inflected form, insert "partial or entire" paradigms, that is, "certain or all" inflected forms of a given word, and then a general convention eliminates all the forms except the one that fits syntactically into the configuration in which the item is actually found in surface structure. Suppose, for example, that the grammar must generate the sentence The student's pranks scandalized the teacher. If we limit ourselves to the major lexical categories (i.e. noun, verb, adjective), Halle's proposal requires that LI insert each word together with its inflectional paradigm, in this case: student/(s), prank/s, scandalize/{s)/ed/(ing), teacher/(s). The general convention would eliminate the inflectional endings not required by the sentence above (i.e. the ones enclosed in brackets) 13 . 2.1.3. Summary Figure 2 provides a schematic summary of the way in which Halle's model works; it contains only those parts of Figure 1, above, that are directly of interest here, that is, the elements of the morphological component.

Lexicalist

morphology

List of Morphemes

31 WFR's

Filter

Dictionary

Figure 2: The operation of the morphological component In this figure, we can follow what happens to five types of words: (1) The word friend ends up in the Dictionary as it is. That is, in the sentence My friend will come tomorrow, friend passes unchanged through the WFR's and the Filter. It must, however, be included in the List of Morphemes, since it is needed for the creation of other words, such as friendly. (2) The word boyhood is not in the List of Morphemes, but rather, there we find only boy and -hood. These elements are combined by the WFR's, and the output, boyhood, goes on to the Dictionary without acquiring any idiosyncratic features; it is syntactically and semantically regular. The change of the feature [-abstract] of the base (boy) to [ + a b s t r a c t ] in the output is carried out by the WFR's, according to Halle. (3) The word recital is formed regularly by the WFR's, as is boyhood. Before recital arrives to the Dictionary, however, the Filter assigns it certain idiosyncratic features with regard to meaning (i.e. "performance of a soloist"). (4) The word ignoration is formed by the WFR's, but it is blocked by the Filter, which assigns it the feature [-Lexical Insertion], That is, this word is considered a "possible" but "non-existent" word, and thus is not listed in the Dictionary. (5) The word mountainal is not formed by the WFR's, since -al is attached by rule to verbs (cf. refuse/refusal, approve/approval, etc.) not to nouns. Mountainal is both a non-existent and an "impossible" word. 2.2. Relevance of Halle's Theory Before we examine some criticisms of Halle's model in the next section, it should be emphasized that Halle's proposal constituted a fundamental step in the development of a theory of morphology: for the first time a single place, i.e. the lexicon, was proposed to handle morphological phenomena, and specific rules, i.e. WFR's, were posited to account for the formation of words. The modifications served not only to lighten the

32

Generative

morphology

burden of the transformational component by removing those operations that involved numerous lexically governed exceptions; they also provide a way to account for a fundamental difference between syntax and morphology. That is, the notion "possible but non existent" became a crucial notion in morphology, one which is, however, totally absent in syntax, where, in fact, it makes no sense to say that a sentence is "possible but non existent". With respect to earlier models of morphology, such as the "Item and Arrangement" model 1 4 that joined morphemes by the simple operation of concatenation, Halle's proposal represented a significant innovation in that it contains a special mechanism for creating words (i.e. WFR's), a mechanism that makes use of more linguistic information and carries out more abstract operations then a concatenation model. T o see why concatenation is not an adeguate way of creating words, consider the French word restructuration "restructuring" 1 5 . This word is not simply built up by concatenation of the morphemes re + structur + ation\ it must have internal structure since it is not possible to attach just any prefix or suffix to any base. T h a t is, the prefix re- must be attached to verbs, not to nouns (cf. *re-verite) or adjectives (cf. *re-grand), and the suffix -ation must also be attached to verbs, not to nouns (cf. *veritation) or adjectives (cf. *grandation). Thus, the base of the derived word restructuration must be a verb (structurer "(to) structure"), and the derivation must be carried out in two steps: (1) structure-^restructurer, (2) restructurer^restructuration. The word in question thus has the following internal structure: (18)

[ [re + [structur] v ] v + a t i o n ] N

It should be noted that the structure in (18) is the only possible one for restructuration:; the other bracketing, represented in (19), is ruled out by the WFR's: (19)

*[re[ [ s t r u c t u r ] v + ation]K(] N

This structure is impossible since, as we have said above, the prefix reattaches to verbs, not to nouns. It should be remembered, finally, that Halle's main purpose in the paper under consideration was to stimulate discussion about a neglected field. We can see today that this goal has, in fact, been completely achieved. 2.3. Some Criticisms of Halle's Model The large body of research on morphology inspired by Halle's article has raised questions about every subpart of the model seen above: the List of Morphemes, the WFR's, the Filter and the Dictionary. We will now examine each of these points separately.

Lexicalist

morphology

33

I. The List of Morphemes. There are two types of problems related to this subcomponent, a general one and a more specific one. The general problem is that the basic units of the system are morphemes, a choice that is decidedly "English-oriented", since in English, words and morphemes coincide most of the time 1 6 . As will be shown in the next chapter, it is possible to assume that the basic units of all word formation processes are "words" and not morphemes, a solution that is not only more general, in that it accounts for a larger body of cross-linguistic data, but also simpler and more elegant, in that it allows us to do away with Halle's Dictionary. The more specific problem, on the other hand, has to d o with the fact that derivational and inflectional affixes, since they too are morphemes, are included in the List of Morphemes, thus obscuring the difference between words (lexical morphemes) and affixes (or derivational morphemes). If the affixes are incorporated into the WFR's themselves, however, it is possible to distinguish between these two types of elements, a point that will be discussed in Chapter III. II. Word Formation Rules. A general observation that can be made in relation to Halle's WFR's is that they are quite unrestricted. Specifically, Halle's WFR's have access to information contained in later stages in a derivation. In fact, the Dictionary actually functions as a type of global condition on the input and output of WFR's. For example, in order to form the word constitutionalism, the WFR that attaches -ism must have access to the Dictionary since that is the only place in which the base to which it attaches, constitutional, is found. An item such as constitutional is not found in the List of Morphemes since it consists of more than one morpheme 1 7 . Given the general aversion to global rules, in recent developments in generative grammar on the grounds that such rules increase inordinately the power of the grammar, most linguists working on morphology have attempted to resolve this problem, as will be seen in the following chapters. Another problem with the WFR's in Halle's model is that they can generate a large number of ungrammatical forms. For example, they can generate both the grammatical forms in (20i) and the ungrammatical forms in (20ii): (20)

(i) constitutionalism capitalism (ii) *dogmatismal *fatalismal

It should be noted that the items in (20ii) cannot be ruled out on either syntactic or phonological grounds; in fact, -al usually attaches to abstract nouns (cf. inspiration + al), and the same sequences of segments are found in other words (cf. baptismal). This problem has received attention in more

34

Generative morphology

recent work, most notably, in that of Siegel (1974) and Aronoff (1976), as will be seen in Chapter III. III. The Filter. The major objection to the Filter is that it is not a finite mechanism. The set of possible but non existent words is not finite in the sense that there are no principles that restrict the degree of complexity of derived and compound words 18 . In fact, as Booij (1977) points out, whatever limits may be placed on compounds are not of a grammatical nature, but rather depend on performance considerations such as memory. Consider the following examples of a type of compound commonly found in certain Germanic languages: (21)

(i) Dutch: brandweerladderwagenknipperlichtinstallatiemonteurs "Fire department ladder truck blinking light installation mechanics" (ii) German: Donaudampschilfahrgeselleschaft "Danube steam ship company" (iii) English: junior high school teachers association curriculum planning committee

IV. The Dictionary. As was mentioned above, the Dictionary and the List of Morphemes introduce redundancy into the system. As will be seen in the following chapters, a word based morphology makes it unnecessary to maintain two distinct levels of representation.

3. Summary In this chapter, we have seen how morphological theory evolved into "lexical" morphology, leaving behind the "transformationlist" morphology of Lees (seen in the preceding chapter). It was seen, furthermore, how the move towards a morphology that operates entirely within the lexical component was made possible by Chomsky's Remarks on Nominalization. In addition, the first model of lexical morphology, that proposed by Halle, was discussed in detail. Specifically, the individual subcomponents, that is the List of Morphemes, the Word Formation Rules, the Filter and the Dictionary were all examined separately. In general, it can be said that Halle's model was quite elegant, and it served, above all, to set in motion a wave of research in the field of morphology, a field that had been neglected for quite a while among generative linguists. Subsequent research has revealed a number of difficulties with Halle's original model, however, and several specific criticisms were discussed, in particular, the facts that (a) the model is morpheme-oriented, (b) the WFR's are too powerful, (c) the Filter is not finite, and (d) the system is redundant.

Lexicalist

morphology

35

Notes 1. The brief account that follows is based on Chomsky (1970) and Jackendoff (1977). 2. Hoekstra et al. (1979:4) point out that "in the literature one can find a variety of terms referring to this position (e.g. strong, extended, generalized lexicalism) or to crucial principles involved (e.g. lexical integrity, spelling prohibition)". In this book, we will use only the term "Extended Lexicalist Hypothesis". The term "Strong Lexicalist Hypothesis" will be used, instead, to refer to those theories which allow inflectional processes in the lexical component (cf. Chapter VI). 3. See, for example, Chomsky (1981:135 ff.). 4. See the discussion on the "Strong Lexicalist Hypothesis", Chapter VI. 5. This analysis is based on Corbin's (1976 b) analysis of the corresponding suffix in French, though somewhat modified and extended here. 6. Cf. Bresnan (1971:257). 7. This discussion in based in part on De Boer and Scalise (1978). 8. In the model in Figure 1, the four subcomponents enclosed in large boxes form part of the lexical component. 9. For a definition of stem, cf Chapter V. 10. A transformational treatment of selectional restrictions has been proposed by Vergnaud (1973). But, on this topic, see Chapter VI. 11. It should be noted that when we are not dealing with the specific case of incohative verbs, the phonological restriction no longer holds (cf. flaxen). 12. Recall that Chomsky (1965) defined Lexical Insertion as the first transformation. 13. The burden of this operation may change drastically in different languages. In Italian, for example, there are, as a rule, two inflected forms for nouns, four (or in some cases two) for adjectives and approximately a hundred for verbs. This means that in a sentence such as Se quel ragazzo avesse comperato il libro, I'avrebbe letto anche mia sorella "If that boy had bought the book, also my sister would have read it", the lexical insertion rule would insert over 200 inflected forms (only for the major lexical categories) and then the convention would eliminate all except seven of these. 14. On this and other structuralist models of morphology, cf. Hockett (1954) and, more recently, Matthews (1970). 15. The source of this example is Corbin (1976b:53). 16. For further discussion of this issue, see Chapter V. 17. For the moment, it suffices to mention that Siegel's solution is to order the rule that attaches -ism after the rule that attaches -al. Aronoff proposes, instead, to use a negative condition according to which the rule that attaches -al may not apply to words of the form [X + ism], Both proposals correctly rule out the form *X + ism + al but see Chapter III (section 3.1.4.) for further discussion of negative conditions, and Chapter VII for further discussion of ordering. 18. Note that Halle does not consider compounds, but in a coherent lexical framework, compounds must also be generated by WFR's (cf. Chapter V).

Chapter

III

Word formation in generative morphology This chapter is devoted to a discussion of AronofT's (1976) monograph, the first systematic work on generative morphology 1 . AronofT's work is particularly noteworthy in that it succeeds in the difficult task of unifying the proposal made in several earlier works on lexical morphology, in particular Halle (1973) Siegel (1974) and Jackendoff (1975). Aronoff proposes a morphological theory that is consistent with the lexicalist hypothesis (cf. the word-based hypothesis) suggests an interesting way of refining morphological rules, (i.e. Word Formation Rules), and formulates a number of restrictions on WFR's in order to constrain their power. He proposes finally a set of rules that readjust the output of WFR's (Readjustment Rules). In this chapter, we will discuss separately some of these important innovations of AronofT's proposals.

1. Morphemes and words A well known definition oT morpheme is the one given by Hockett (1958:123): "The smallest individually meaningful element in the utterance of a language". In Saussurean terms, a morpheme is thus a minimal sign. A minimal sign, however, cannot be defined once and for all; rather, it has to be defined in relation to the pertinent linguistic level. Works such as Chomsky (1970) and Postal (1969) established that for the purpose of syntax the word is the minimal sign. Aronoff provides evidence that the word must also be considered the minimal sign for morphology. That is, Aronoff argues that there are morphemes, which must be considered linguistic units below the word level, though these morphemes have no independent meaning outside the specific words in which they occur. AronofT's demonstration oT this point is quite eleaborate; we will consider here only the main arguments. Consider, first of all, words such as the following: (1)

cranberry boysenberry huckleberry

We can analyze these three words as being composed of a constant element, berry and something else, namely the morphemes in (2):

38

Generative morphology

(2)

cran boysen huckle

The items in (2) do not occur independently or in other combinations in English; they are hapax legomena. This means that they have no meaning by themselves; rather, their meanings are intimately connected with those of the specific words in which they occur. Consider, next, another series of berry words such as the following: (3)

strawberry blackberry blueberry gooseberry

Analyzing these words by the same segmentation procedure, we are left again with a constant element, berry, plus the following morphemes: (4)

straw black blue goose

The morphemes in (4), contrary to the ones in (2), do occur elsewhere in English as independent words. Their meanings as independent words, however, are unrelated to their meanings when they occur in the words in (3). A blackberry is not necessarily a berry which is "black" there are green or red blackberries), and the meaning relation between straw and strawberry or between goose and gooseberry is not very clear. Aronoff examines several possible ways of solving the problem (e.g., giving morphemes underdetermined meanings, with contextually determined allo-meanings) but finally discards these solutions in favor of the simple observation that there are morphemes which have no constant meaning. The third set of data analyzed by Aronoff consists of Latinate verbs with bound stems and prefixes, stressed on the stem: X=fer refer defer prefer infer confer transfer

X = mit remit demit

X = sume resume

X = ceive receive deceive

X = duce reduce deduce

conceive

induce conduce transduce

presume commit transmit submit admit permit

consume subsume assume

adduce perceive

Word formation

in generative

39

morphology

For the sake of the argument, Aronoff begins by assuming that the prefixes do have constant meaning (which would make this case similar to the berry words). The question is, again, whether is possible or not to assign a constant meaning to the morphemes fer, mit, etc. If we try to assign a constant meaning to each occurrence of fer, mit etc. in the lists above, it becomes clear quite quickly that this is not possible. Prefer does not have even a vague meaning in common with confer and transfer, and commit does not share any meaning with remit and submit, etc. AronofF's conclusion is that "each stem occurs in different verbs, but never with the same sense. Rather the sense is determined by the individual verb", (p. 12). At this point, one may argue that the occurrences of the stem mit are not occurrences of the same morpheme mit but rather different stems mit ι, mit2, • • • mitn. But this solution too appears to be incorrect. Aronoff shows very convincingly that the stems in question d o share a common feature, though a formal one: all instances of the mit verbs exhibit the same allomorphy before the suffixes -ion, -ory, -or, -ive, and -able. That is, mit becomes mis, as can be seen in (6): (6)

permit remit

permission remission

permissive remissive

This alternation is restricted to the Latinate stems mit. In fact vomit, which is not analyzable as Prefix + stem does not exhibit the same allomorphy (cf. vomit/vomitory). It is clear, furthermore, that the alternation is not due to phonological factors: prohibit does not undergo the same change (cf. prohibition, prohibitive). This means that the factors which govern such an alternation must be established at a different level, that is, at the level of the morpheme or stem. We must conclude, therefore, that all the Latinate stems miti, miti... mitn are, at some level, different occurences of the same mit. The relevant level is that of the morpheme. Aronoff gives two reasons for this conclusion: first, features such as [ + Latinate] can be shown to be a property of morphemes, and second, the rule that changes t to s is a rule that applies to morphemes. Aronoff considers, next, the problem of whether the prefixes have a constant meaning. He lists the following data: re = X repel remit refer resume receive reduce

con = X compel commit confer consume conceive conduce

in — X impel

de = X

infer

demit defer

induce

deceive deduce

The prefix re- often has a meaning connected with "back" (cf. Williams 1973). It is easy to show, however, that this meaning is not really present in the re = X verbs listed above.

40

Generative morphology

But if re does not have a constant meaning, and neither does duce, how can we analyze the meaning of a word such as reduce as consisting of two parts, one associated with the prefix and the other associated with the verb stem? The answer is uncontroversial: it is simply not possible since we are dealing with morphemes without any meaning. Aronoff concludes, therefore, that "[t]he hypothesis that morphemes are 'the minimal meaningful elements of language' cannot be maintained even in any of its most contorted variants. In many cases this role of the minimal sign must be moved one level up, to the level of the word" (p. 14). 1.1. The Word Based Hypothesis The morpheme had already been rejected in earlier work as the basis of inflectional morphology (cf. Chomsky 1965:170-4). Aronoff went a step further, in the same direction, and rejected the morpheme as basis of derivational morphology as well. Now, given the fact that morpheme and sign do not share the same properties, and given the fact that WFR's, which are "regular" rules, can only derive meaningful words from meaningful bases, it follows that only words can be the basic units of Aronoff's theory of morphology. The Word Based Hypothesis (WBH) is formulated in the following terms (Aronoff, 1976:21): (8)

All regular word-formation processes are word-based. A new word is formed by applying a regular rule to a single already existing word. Both the new word and the existing one are members of major lexical categories.

This hypothesis makes a number of claims: 1. The bases of WFR's are words. 2. These words must be existing words. Thus, a possible but non existent word, according the hypothesis, cannot be the base of a WFR. 3. WFR's can take as a base only a single word, no more (e.g. phrases) and no less (e.g. bound forms). 4. and 5. Both the input and the output of a WFR must be members of a major lexical category. Finally, it should be noted that the WBH is intended to be universally valid and not as a language specific property of English word formation. We will discuss below each of these claims, their consequences and the possibility of maintaining all of them in a morphological theory consistent with the data and problems we are aware of today. 1.2. Goals of a Morphological Theory Aronoff establishes the following (minimal) parallel between syntax and morphology: just as syntax must define the class of possible sentences in a language, morphology must define the class of possible words in a

Word formation

in generative

morphology

41

language. There is also a basic difference, however, between syntax and morphology: the latter, but not the former, must distinguish between the two notions "possible" and "existent". As we have seen in the discussion of Halle's proposal, when we are dealing with words, we must take into account three categories: (a) actual words, (b) possible but non existent words, and (c) impossible and non existent words. By contrast, in syntax we have only two categories: (a) possible sentences and (b) impossible sentences. We have also seen how Halle's model was intended to handle this problem. AronofTs objection to Halle's solution is that in a morpheme-based theory there are two lexicons: the List of Morphemes (which, together with WFR's, defines the class of the possible words) and the Dictionary (where only the actual words of the language are stored). Aronoff's objection is not only based on economy, however. As he points out, words are peculiar in several respects. First of all, not all the words that could exist do, in fact, exist, and secondly, the existent words d o not always mean what they are expected to mean. Thus, the actual words (i.e. the members of Halle's Dictionary) cannot simply be a subset of the items produced by a regular morphology that generates words and their meanings on the basis of meaningful morphemes. In fact, as AronofT (1976:18) states: (9)

This is the basic trouble with morphemes. Because words, though they may be formed by regular rules, persist and change once they are in the lexicon, the morphemes out of which words seem to have been formed, and into which they seem to be analyzable, d o not have constant meanings and in some cases they have no meaning at all. It is this persistence which forces us to adopt a lexicalist hypothesis.

Of course, Halle was well aware of the problem and proposed that all phonological, syntactic and semantic properties not derivable by "regular" rules be handled by the Filter. According to Aronoff, however, Halle's model cannot account for words whose meaning is totally different from the expected one. A word such as transmission, for example, which should be an abstract nominal meaning "the action of transmitting" according the rules, has quite a different meaning when it refers to a part of a car. Transmission must, therefore, be listed in the lexicon. It should be noted that listing all the words with unpredictable idiosyncrasies in the lexicon does not deprive morphology of its content and value. The goal of a morphological theory becomes that of defining the "new" words that speakers can form, or more specifically, the "regular" rules by which new words are formed. While it seems reasonable to assume that the idiosyncrasies we find in words that have been in existence for a long time are not typical of the new words speakers create, it is also true that the words listed in the lexicon often exhibit regularities in addition to the idiosyncrasies. For

42

Generative

morphology

example, they have structure: even if the m o r p h e m e s that constitute them exhibit idiosyncrasies, these morphemes are, nevertheless, recognizable. A speaker w h o hears a word for the first time recognizes it as a word of his language, and has intuitions about its structure and a b o u t its meaning. A morphological theory, concludes Aronoff, must also say something a b o u t these facts, in particular about the relation between formal mechanisms that create new words and the analysis of already existing words.

2. Word formation rules A W F R specifies the set of words on which it can operate; this set is called the "base" of that rule. Every W F R specifies, furthermore, a unique phonological operation performed on the base. 2 Finally, every W F R specifies the syntactic label and subcategorization frame of the resulting word, along with a semantic reading which is a function of the semantic reading of the base. A first sketch of how a W F R can be represented is given in (10) 3 : (10)

[ W ] x - [ [ W ] x + Af]Y [Fa] [Fa] [Ffl

"semantics of Y"

(10) reads as follows: a word with the lexical category X and Features α is rewritten as a complex word with internal structure, consisting of the base word, a b o u n d a r y " + " 4 and an affix. T h e resulting word has the lexical category Y and the features β 5 . Associated with this "formal" part of the rule, is also a semantic part which is compositional in nature and usually is given in the form of a paraphrasis. T h e W F R ' s that form the words readable and boyhood can, thus, be formulated as follows: (11)

(i) [ r e a d ] y - » [ [read] v + a b l e ] A "capable of being read" [ + tr] [ + tr] (ii) [ b o y ] Ν - • [ [boy]Ν + h o o d ] Ν "the quality of being a b o y " [ - abstr] [ - abstr] [ + a b s t r ]

In (Hi), the W F R attaches the suffix -able to the transitive verb read, forming the derived adjective readable. In (llii), the W F R attaches the suffix -hood to the non-abstract n o u n boy, forming the derived noun boyhood. In (1 li), the rule changes the lexical category V into A, while in (1 Iii) it changes the feature [-abstract] into the feature [ + abstract]. In both cases, the rules provide the semantic reading of the derived words. The mechanism of WFR's, which will be further illustrated below, rests on several assumptions. T h e first assumption is that W F R ' s are "lexical" rules, that is, rules of the lexical c o m p o n e n t . A W F R can refer, as we will see below, to the syntactic, semantic, morphological and phonological properties of words,

Word formation in generative morphology

43

but not to syntactic, semantic or phonological rules. In other words, WFR's have access to the information associated with words in the lexicon, but not to other components of the grammar. That is, they operate entirely within the lexical component. This assumption makes WFR's a more constrained type of rule than those proposed by Halle which also had access to the phonological component. The second assumption is that WFR's are different from other rules of the grammar also in the way in which they are used. While syntactic and phonological rules are necessary for the generation of every sentence, WFR's are optional. Thus, it is possible to generate a sentence without using any WFR's, though this is obviously not the case for syntactic or phonological rules 6 . Aronoff proposes, therefore, to consider WFR's as "once-only rules". That is, a speaker does not apply a WFR every time he uses a complex word; instead, once a new complex word is formed, it is stored as such in the lexicon. It should be noted that the "once-only" character of WFR's does not imply that they are a superfluous device in the grammar. In Aronoff's proposal, in fact, WFR's have two functions: they not only account for "new" words, they also account for the internal structure of already existing words. In the latter case, since they relate fully specified existing words, they can be thought of as a type of redundancy rule, similar to those proposed by Jackendoff (1975). A third assumption is implicitly present in the form of the WFR's themselves. A rule such as the one in (10), in fact, implies a basic distinction between the notions of "word" and "affix", at least as far as the "level" of representation is concerned. Words are listed in the lexicon (or, to be more precise, in the list of words frequently referred to as the "dictionary"); affixes are located one level "below", that is, in the WFR component. The organisation of the lexical component can be represented as follows7:

Lexical Component Dictionary

i

i

WFRs

Figure 1: The organisation of the lexical component (I). The different locations of words and affixes amounts to saying that the two are, in fact, different in nature. While the former carry "categorical" information (e.g., boy "is a" noun), the latter carry "relational" information

44

Generative morphology

(e.g., the suffix -able forms adjectives from verbs). The information V—>A, as one can see, is a WFR, even if abridged in form. In other words, in this proposal, the representation of an affix is the WFR which attaches the affix to its base. Thus, contrary to Halle's proposal, where lexical morphemes and other (i.e. derivational and inflectional) morphemes were all represented at the same level in the List of Morphemes, here there are two different levels of representation. That is, "words" are represented in the Dictionary and "affixes" are represented in the WFR component.

3. Restrictions on WFR's Thus far, the differences observed between Halle's and Aronoff's theories of morphology are basically the following: (a) Halle's system rests on a morpheme-based hypothesis, while Aronoff's system rests on a wordbased hypothesis, and (b) WFR's have a global power in Halle's proposal, while they have no such power in Aronoff's proposal. As has been said above, one of the main targets of a morphological theory is to define the notion "possible word in a language". In order to achieve this goal, one must formulate restrictions on WFR's, in order to determine (a) what sort of information WFR's can have access to, and (b) what sort of operations they can perform. Aronoff's proposal, in this respect, represents the first systematic, non episodic, picture of what kind of restrictions a WFR can be subject to. Since WFR's apply to a base and generate an output, we will first examine the restrictions on the base (section 3.1), and then the restrictions on the output (section 3.2.). 3.1. The Base The base of a WFR is the set of words to which the rule can apply. We can represent this by means of the following diagram (cf. Booij 1977:41): (12)

Δ

Sui*

The base of the rule to which the suffix attaches is the entire set of words that can replace the dummy symbol dominated by X. What is needed, further, are appropriate restrictions on the attachment of a suffix to its base, so that the WFR is able to exclude any ungrammatical combinations, and the string "X + Suf" (i.e. "Y"), is a well formed word in a given language. Aronoff, in fact, proposed a series of syntactic, semantic, phonological and morphological restrictions on the base of a WFR, which we will examine in the following sections; another type of restriction, the Unitary Base Hypothesis, will be discussed in Chapter VII.

Word formation

in generative

3.1.1. Syntax and

45

morphology

Semantics

WFR's have access to all and to only the information contained in their base. This information is specified syntactically, and therefore the base must always be a member of a major syntactic category. In particular, WFR's are sensitive to (a) the syntactic category of the base: the suffix -able, for example, attaches to verbs, not to nouns or adjectives (cf. [ [ w a s h ] y + able]A versus *[ [book] ν + able] or *[ [black] A + able]), (b) the syntactic features of the base and, in particular, bi) subcategorization features: -able attaches to verbs marked as [ +transitive], not to verbs marked as [-transitive] (cf. [ [ w a s h ] v + able]Aversus*[[seem]v + able]), [ + tr] [-tr] and b(ii) selectional restrictions: the suffix -ee attaches to verbs which allow animate objects or indirect objects (cf. [ [employ] y + ee] ν versus * [ [ t e a r ] v + ee]) 9 . In general, relatively little work has been done in the area of semantic restrictions on the base. One suggestion was made, however, by Williams (1973), quoted in Aronoff (1976:47), to the effect that the prefix reattaches only to verbs whose meaning entails a change of state, generally in the object of the verb (cf. John punched Bill/* John repunched Bill versus John punched the holes in the paper/John repunched the holes in the paper). The ungrammaticality of the second sentence is apparently due to the fact that punch in the sense of 'to hit' does not entail any change of state in its object. In this book, we will not take up the issue of semantic restrictions again, except briefly in relation to other points. It should be noted that semantics in WFR's is a weak point in the theory, in general, because it is a relatively underdeveloped area. As an illustration of the type of problems encountered in this area, consider how different suffixes "choose" different meanings of the same base. The French word juste "fair" can be derived by adding two different suffixes, giving in both cases a noun: justice "justice" or justesse "fairness". A similar case is found with the adjective large "wide", from which one can get largesse "generosity" or largeur "width". What is interesting in these cases (quoted by Dell 1979), is the fact that the two "rival" suffixes, in both cases, seem to take into account different meanings of the same word; this can be visualized in the following way: (13)

^^.^^-"equitable"

-> justice

juste ~~-"fair" "generous"

justesse -»• largesse

large "

"wide"

largeur

This case, however, does not represent the only possible situation. That is, it is not always the case that when a word has more than one meaning all ι

46

Generative

morphology

meanings undergo derivation. Consider the following example: (14)


impopulaire

"of the people"

Here, the negation of the adjective (populaire "popular"), can be accomplished only with one meaning, i.e. the meaning "which many people like", giving impopulaire "unpopular"; negation does not affect the other meaning "of the people". In other words, the negation impopulaire can only mean "which many people d o not like". What W F R ' s can express in relation to the negation of adjectives is, for example, the fact that a negative prefix cannot attach to adjectives with negative semantic content (cf. in English *un + bad, *un + evil and Allen's (1978) discussion of this problem). There is at the moment, however, no way of encoding the fact that affixes can behave differently with words having more than one meaning, as in (13) and (14) above 1 0 . 3.1.2.

Phonology

Another type of restriction on W F R ' s can be expressed in terms of the phonological shape of the base. Siegel (1971), for example, observed that the suffix -en, which forms verbs from adjectives, attaches to words ending in t or d (cf. brighten, harden, versus *calmen). In fact, Marchand (1969:272) (quoted by Aronoff 1976:79) noted that the exceptions to this condition (cf. toughen, freshen, weaken) date back to an earlier "more liberal" period, while in the last 200 years only adjectives ending in dental consonants have been the input to this rule. Siegel (1971) actually proposes a further refinement of the condition, observing that -en attaches to forms that on the surface end in a single obstruent preceded by a vowel, with the possibility of an intervening sonorant between the vowel and the obstruent (glisten, harden, dampen, whiten, frighten)11. Other phonological condition have been proposed by Schultink (1962), Zwarts (1975) and Booij (1977) for Dutch. According to Zwarts, for example, the base of the suffix -sei cannot be a dental strident [s] or [z], as the following d a t a show: (15)

raad "to guess" wind "to wind" stijf "to stiffen" plavei "to pave" eis "to demand" kras "to scratch" kneus "to bruise" lees "to read"

raadsel "riddle" windsel "bandage" stijfsel "paste" plaveisel "pavement" *eissel *krassel *kneussel *leessel

Word formation

in generative morphology

47

Booij (1977:123) observed, however, that this restriction can be formulated in more general terms. In the first place, there is another suffix in Dutch which shows the same restriction, namely the suffix -zaam, which never attaches to words with a final dental strident: (16)

werk "to work" groei "to grow" buig "to bend"

werkzaam "active" groeizaam "growing" buigzaam "flexible"

was "to wash" ruis "to rustle" verhuis "to move"

*waszaam *ruiszaam *verhuiszaam

The restriction cannot, however, be predicted by a general condition that rules out the derivation of words that would have the sequence [ss] or [zz], Booij shows, in fact, that such a condition would be too strong, since there are words ending with a dental strident that do take a suffix with an initial [s]: (17)

-schap -ster

baas "boss" baasschap "lit. bosshood" secretaris "secretary" secretarisschap "secretariate" kaats "to play at ball" kaatsster "female ball player" fiets "to cycle" fietsster "female cyclist"

Booij (1977:124) thus proposes the following principle for Dutch, where he takes advantage of the fact that the first two suffixes begin with a single consonant and the other two with a consonant cluster: (18)

A suffix with the phonological form attached to a basis

word with the phonological form X

C + cont + cor

VC 0 cannot be

C. + cont + cor

In this way, Booij succeeds in avoiding the formulation of a rule-specific condition. Another case in which the attachment of an affix is subjected to phonological restrictions is that of the Italian negative prefix s-. Consider the data below: (19)

(i) fortunato leale

sfortunato sleale

"lucky - unlucky" "loyal - disloyal"

48

Generative

morphology

corretto gradevole (ii) umano onesto educato abitato

scorretto sgradevole *sumano *sonesto *seducato *sabitato

"correct - incorrect" "pleasant -- unpleasant" disumano "human - inhuman" disonesto "honest - dishonest" diseducato "well mannered - ill-mannered disabitato "inhabited - uninhabited"

*seivile *sgiusto *ssano *ssensibile *ssicuro

incivile ingiusto insano insensibile insicuro

(iii) civile giusto sano sensibile sicuro

"polite - impolite" "just - unjust" "healthy - unhealthy" "sensitive - insensitive" "safe - unsafe"

W h a t t h e s e d a t a s h o w is t h a t the prefix s- d o e s n o t a t t a c h t o w o r d s b e g i n n i n g w i t h a vowel (cf. (19ii)), a n d it a t t a c h e s t o w o r d s b e g i n n i n g with a c o n s o n a n t o n l y if t h e resulting p h o n o l o g i c a l s e q u e n c e d o e s n o t violate t h e g e n e r a l p h o n o l o g i c a l r e s t r i c t i o n s o n p o s s i b l e initial c l u s t e r s in Italian (cf. M u l j a c i c , 1969; Vogel, 1982). T h e d e r i v e d w o r d s in (19i) d o n o t violate these r e s t r i c t i o n s (cf. the w o r d s sforzo "effort", slitta "sleigh", sconto " d i s c o u n t " , sgravio "lightening"), while t h o s e in (19iii) d o , since / stJ" / , / s d j / , / s s / a r e , in fact, all i m p o s s i b l e initial clusters in I t a l i a n . T h e w o r d s in (19ii), h o w e v e r , d o n o t violate a n y p h o n o l o g i c a l res t r i c t i o n s of I t a l i a n : t h e y a r e " p h o n o l o g i c a l l y " perfectly a c c e p t a b l e f o r m s (cf. succo "juice", sonante " r e s o u n d i n g " , seducente " t e m p t i n g " , sabino " S a b i n e " ) . T h e y m u s t t h e r e f o r e be e x c l u d e d o n o t h e r g r o u n d s . W e c a n t h u s c o n c l u d e t h a t t h e r e s t r i c t i o n t h a t rules o u t t h e d e r i v e d f o r m s in (19iii) s h o u l d be f o r m u l a t e d in m o r e g e n e r a l t e r m s 1 2 t h a n t h e restriction t h a t rules o u t t h e d e r i v e d w o r d s in (19ii), since t h e f o r m e r f o l l o w s f r o m a fact a b o u t I t a l i a n p h o n o l o g y in general, while t h e latter is a "rule-specific" limitation. 3.1.3.

Morphology

In a d d i t i o n t o t h e syntactic, s e m a n t i c a n d p h o n o l o g i c a l c o n d i t i o n s , t h e r e a r e a l s o m o r p h o l o g i c a l c o n d i t i o n s o n t h e b a s e of W F R ' s . T h e s e c a n be d i v i d e d i n t o t w o m a i n types: c o n d i t i o n s t h a t h a v e t o d o w i t h a b s t r a c t m o r p h o l o g i c a l f e a t u r e s a n d c o n d i t i o n s s t a t a b l e o n specific m o r p h e m e s . T h i s latter t y p e c a n be f u r t h e r divided i n t o t w o s u b t y p e s : positive c o n ditions and negative conditions. L e t u s c o n s i d e r first t h e a b s t r a c t m o r p h o l o g i c a l f e a t u r e s . T h e n e e d for a b s t r a c t f e a t u r e s h a s been l o n g recognized, a n d s u c h f e a t u r e s are n o w used in m o r p h o l o g y as well as in p h o n o l o g y . In p h o n o l o g y it is a well e s t a b l i s h e d f a c t t h a t " f o r e i g n w o r d s " (i.e. n o n n a t i v e f o r m s ) " o f t e n exhibit p h o n e m e s w h i c h d o n o t a p p e a r at all in the native layer of t h e lexicon" (Saciuk 1969:465). But, m o r e i m p o r t a n t , t h e r e a r e a l s o p h o n o l o g i c a l rules w h i c h a r e sensitive t o t h e n a t i v e / n o n - n a t i v e d i s t i n c t i o n . C h o m s k y a n d H a l l e (1968:219), f o r e x a m p l e , s h o w h o w the Velar S o f t e n i n g R u l e is limited "only t o c e r t a i n lexically m a r k e d elements".

Word formation

in generative

morphology

49

This rule, which changes [g] and [J] and [k] to [s] before non low front vowels, operates in Latinate forms such as criti[s]ism (cf. criii[k]a/), medi\s]ine (cf. medi\X]ate), ri[j]id (cf. n'[g]or), but it does not operate in the native layer of the lexicon (cf. black ^>bla\k~]ish/*bla[s\ish, pig pi id] ish/*pi[fi ish). In morphology, Bloomfield (1933:252) observed that " [ n ] o r m a l roots combine with normal affixes" and "learned roots with learned affixes", where "normal" and "learned" can be reworded as "native" and "nonnative", respectively. The same feature is usually referred to by Aronoff as[ +Latinate]. The suffix -ity, for example, attaches to [ + L a t i n a t e ] words (cf. felicity, vivacity) and not to "native" or [-Latinate] words (cf. *widity, *strongity). In this sense, -ity contrasts with the suffix -ness, which does not discriminate between [ -I- Latinate] and [-Latinate] words (cf. common + ness, strange + ness and happy + ness, white + ness). Contrary to Bloomfield's claim, however, Aronoff also observed that WFR's restricted only to native words are quite rare. One such case in English is the suffix -hood, found in words such as motherhood, brotherhood. The distinction native/non-native is not the only relevant distinction. In fact, Saciuk (1969) proposed an elaborate division of the lexicon according to what he called "stratal features", as can be seen (with minor modification) in (20): (20)

Homogeneous

[ +Latin] [ +Greek] [...] We will not discuss this classification here; it is sufficient to note that features such the ones given above have always been used, more or less explicitly, in the literature on phonology and morphology. Postal (1968:121 IT), for example, showed the necessity of the feature [ + F r e n c h ] in order to assign stress to the French borrowings in Mohawk. The suffixes -iteit and -or in Dutch are sensitive to the features [ + F r e n c h ] and [ + L a t i n ] , respectively (cf. Schultink 1962:185-6 and Booij 1977:127). Moreover, the stratal division of the lexicon is shown by Booij (1977) to be of importance in Dutch for the Segment Structure Condition, inflexion rules and phonological rules. There are also other issues connected with the notion of "stratal features". We will not discuss them here, however, but simply mention one major theoretical point, that is, the question of whether stratal features are

50

Generative morphology

properties of morphemes or of words, for which the reader can refer to Aronoff (1976:51-52) 13 . Turning now to the conditions statable on specific morphemes, we will consider first the positive conditions, which are connected with the productivity of WFR's and are meant to encode the preference of a given affix for certain types of words. The general condition, for example, that the suffix -ity attaches to [ + latinate] words is not sufficient to express the preference of this suffix for bases of the form X-ic, X-al, X-able, X-id (cf. specificity, globality, readability, lucidity). Positive conditions are expressed as part of a WFR. For example, the rule which attaches the negative prefix un- can be stated in the following way (Aronoff 1976:62): (21)

Rule of negative un(a) [X] A -»[un + [X] A ]A semantics (roughly): un + X = not X (b) Forms of the base (where en is 1. Xyen 2. Xy + ing 3. Xy + able (worthy) 4. x + y (seemly) 5. X + ly (mindful) 6. X + ful (conditional) 7. X + al (warlike) 8. X -1- like

The two parts of (21) are to be interpreted as follows: (21a) is the WFR for the attachment of the prefix un-: the "formal" part says that the rule forms adjectives from adjectives and the "semantic" part states the meaning of the derived word in compositional terms. (21b) is the list of the "positive" conditions on the WFR itself. The list states the classes of bases to which the suffix attaches and is given (roughly) in order of productivity. It is understood that the absence from the list of some other class does not prevent the rule from applying to it, unless it is marked with a negative condition. All the examples of positive conditions seen above can be stated on adjacent morphemes (e.g. the condition on the attachment of -ity 'involves' this suffix and the adjacent suffixes -ic, -al, etc.; the condition on the attachment of un- involves this prefix and the non adjacent suffixes -en, ing, etc.). According to Aronoff (1976:53) there are also positive conditions which cannot be stated on adjacent morphemes. A well known example of this is the nominal suffix -ment which attaches most productively to verbs of the form en + X, be + X (cf. encroachment, bewilderment). There are thus reasons to believe that the adjacency condition is too powerful. As will be seen below, a number of problems can be resolved if we replace the notion of "linear" adjacency with the more interesting notion of "cyclic" adjacency as proposed by Siegel (1977).14

Word formation

in generative

51

morphology

Negative conditions express the fact that a given W F R does not operate with words belonging to a certain morphological class. F o r example, the suffix -al does not attach to the class of nouns of the form Xyment, where X is an independently occurring verb, as can be seen from the following examples (c). Aronoff 1976:54): (22)

(i) *ornay (ii) *fragy (iii) employ (iv) discern

ornament fragment employment discernment

ornamental fragmental *employmental *discernmental

The restriction at work here is not syntactic because -al attaches quite freely to other deverbal abstract nomimals (cf. organizational, confessional). It is not phonological, either, because the phonological sequence Xmental is not forbidden (as can be seen in (22i) and (22ii). This means that the restriction depends on the internal structure of the base, and Aronoff proposes to express such a condition in the following way: (23)

[[X]N-al]A Condition: X ^ Y y m e n t

The condition in (23) states that the base of the -al rule cannot be a complex word consisting of an independently existing verb plus the suffix -ment. As one can easily see, this negative condition, like the positive condition for the attachement of -ment, is not formulated on adjacent morphemes, a fact which is better seen in the following way: (24)

(i) [ [en + [X] + ment] (ϋ) [ [ [ Y ] v + m e n t ] + al]

In (24i), the suffix is allowed to 'see' the non-adjacent prefix en-, and in (24ii) it is allowed to 'see' the non-adjacent labeled bracketing V. We will return to these issues below. Finally, Aronoff proposes a special type of negative condition that he calls "Blocking". Since it has been suggested, for example by Allen (1978) and Zwanenburg (1981), that a special status should be given to this type of condition, namely that of a "principle" of morphological theory, we will discuss it in more detail below (cf. Chapter VII). 3.2. The Output There are, basically, two types of constraints on the output of WFR's: a syntactic one and a semantic one. Syntactically, every new word created by a W F R must be a member of a major lexical category. The category of the output is determined by the W F R itself. The following is a list of possible inputs and outputs in English word formation:

52

Generative morphology

(25)

Ν-»Ν V->N A->N N->V A->V V^V Ν -* A V->A A—• A A->Adv

king + dom invert + ion fastidious+ ness glory+ ify short + en 0 education + al drink + able green+ ish beautiful+ ly

The outputs of WFR's may vary, of course, from language to language. The list in (25) tells us, for example, that there are no suffixes that derive verbs from verbs in English. The same is not true in French, however, where such suffixes do exist, (cf. mordre/mordiller "bite-nibble"). It should be noted that the effect of the syntactic restriction is that only "words" can be "outputs" of WFR's or, as Carrier (1979:21) puts it, "WFR's do not produce immediate forms that are not complete words". The informal notion complete word must, however, be qualified. If we assume that WFR's produce complete words, we must also accept two caveats: (a) the word may need some readjustement, as is the case, for example, with the word cited above, invert + ion, which must be changed into inversion, and (b) in some languages, the outputs of some WFR's require overt inflectional markers before they can appear in surface structure (as is the case in Tagalog, according to Carrier (1979)). From a formal point of view, these two caveats can easily be incorporated into what can be called a strong version of the Lexicalist Hypothesis. In this case, it must be assumed that the Inflectional Rules (IR's) are part of the lexical component (as are the Readjustment Rules (RR's)), and furthermore, that WFR's produce words at some abstract level of analysis (i.e. words that cannot occur in surface structure). The lexical component (which incorporates WFR's, IR's, RR's), however, always produces complete words (i.e. words that can occur in surface structure). This point, as Carrier observes, cannot be made clearly in English, where words without overt inflections often occur in sentences. The claim that the output of WFR's is an inflected word can only be maintained if a clear distinction is made between the notions uninflected word (or stem) and inflected word (word form). The output of a WFR is a structure with a labeled bracketing, where both the category of the input and the category of the output are specified, along with the boundary required by the rule. 15 An output such as the word educational can thus be represented in the following way: (26)

[ [education] Ν + al ] A

Word formation

in generative

morphology

53

When a base undergoes several WFR's, all of the structure built up in the derivation is preserved. An abstract example of this case is given in 27): (27)

[ [ [ [ [ ] x + Suf] y + S u f ] z + S u f ] w + S u f ] N

The internal structure of the output is required in order to establish general conditions on WFR's, such as the Adjacency Condition, and to provide the information necessary for the operation of Readjustment Rules. 16 According to the semantic constraint, the meaning of the output of a WFR must be a function of the meaning of the base. This function, says Aronoff (1976:50), is the meaning of the W F R itself. The meaning of the output is often represented by a paraphrasis containing a variable, as the examples below show: (28)

(') Ε [X]v + able]A lovable (ii) [un + [ X ] A ] A unhappy (iii) [ [ X ] v + e r ] N diver

"capable of being X-ed' ! "capable of being loved" "not X" "not happy" "a person .who professionally or habitually X's" "a person who professionally or habitually dives"

Paraphrases such as the ones given above are, admittedly, no more than a useful way to represent the meaning of the output of WFR's. They have no theoretical significance. As we have already observed, the semantics of WFR's is an area that has yet to be properly investigated. Another observation Aronoff makes is that paraphrases can be misleading at times, in particular, when they appear to be more language specific than they actually are. The -er WFR, seen in (28iii), for example, is not only a rule of English, completely unrelated to similar rules in other languages, as can be seen from the following list: (29)

Latin French Italian Spanish Dutch Rumenian German

observa"tor" observa"teur" osserva"tore" observa"dor" waarnem"er" observa"tor" Beobachter"

• "one who habitually observes''

The meaning of a complex word is componentional only when it is created by a synchronically productive rule. Over time, a word can acquire unpredictable meanings, that is meanings that can no longer be derived

54

Generative

morphology

from the meanings of its components. The Italian word cited in (29), for example, along with its "regular meaning" given in the paraphrase, can also mean "Delegate to an international conference at which he is not a participant, or to a country to participate in juridical affairs" (Zingarelli, 1970-translation S.S.). Words are thus formed by regular rules but, once formed, they may take on impredictable meanings or, in Aronoff's terms, they may undergo semantic drift. This also explains the case of transmission, seen above, which no longer has its compositional meaning (i.e. "the action of transmitting") when it refers to a part of a car. In this section, we have briefly discussed the syntax and the semantics of the output of W F R ' s but, as Aronoff points out (1976:49), there is another area of research connected with this topic that must also be considered. That is, "the relation between the syntax and semantics of the base and that of the output of a WFR's, the common properties which the two share, and the ways in which these relations and communalities can be accounted for."

4. Summary In this chapter, we have examined some of the more important innovations found in Aronoff's (1976) work. Particular attention was given to the Word Based Hypothesis, WFR's and several restrictions on WFR's (restrictions on the base and on the output). It was also shown how Aronoff's theoretical framework has the advantage, with respect to Halle's, of restricting quite considerably the power of WFR's. In the rest of this book, we will assume Aronoff's framework as outlined in this Chapter, even though, as we proceed, we will discuss and sometimes accept modifications that have been proposed since the publication of the original work. In the next chapter, we will discuss another important innovation found in Aronoff's proposal, Readjustment Rules. It will be shown that these, too, are "lexical" rules, though different in nature from WFR's.

Notes 1. Two other works that were also very important in the development of generative morphology are Siegel's thesis (1974) and an article by Jackendoff (1975). In fact, AronofTs monograph develops many of the ideas presented in these two works. 2. According to Aronoff, this operation usually consists of an "adjunction" of an affix. But, in certain cases, the operation "can - - - b e null, and it may be more subtle" (p. 22). A case of "null adjunction" may be what has been called "implicit transposition" or "conversion" (Lieber 1980) or also "zero morpheme", that is a change of the lexical category without the adjunction of a manifest affix (cf. [ p a i n t ] N / [ p a i n t ] v , German [ R u f ] N / [ r u f ] v "call", French [ p o u v o i r ] N / [ p o u v o i r ] v "power - to be able", etc.). "More subtle" cases may be represented

Word formation

in generative

morphology

55

by modification of the root vowels such as sing/song, Dutch drink/drank "(to) drink-(the) drink" or by "subtraction" instead of adjunction, as in French arreter-arret "(to) stop/(the) stop'. A classical description in terms of "subtraction" (later on criticized by Schane 1968) is the one given by Bloomfield for the adjective in French. Observing masculine/feminine alternations such as plat [pla]/platte [plat] "flat", laid [l£~\/laide [led] "ugly", Bloomfield (1933:217) proposed to take the feminine form as basic and to derive the masculine "by means of a minus-feature, namely, loss of the final consonant [ . . . ] . " Finally, an other example of "subtraction" (or "morphological deletion rules", as Aronoff (1976:78) calls them) is the case of the Danish imperative, studied by Basb0ll (1970), which in some cases is derived from the infinitive by deletion of a final [a] (cf. bade [bae:de] "to bathe" ->bad [bae?] "bathe!"). 3. It should be noted that although Aronoff mentions the relevance of the syntactic features for the operation of WFR's, he does not include them in the formal representation of WFR's. Extensive use is made of such features, however, in Scalise (1983). 4. " + " is a morpheme boundary, as opposed to " # " , a word boundary. Cf. the discussion of boundaries below, Chapter V. 5. This way of representing the rule is somewhat misleading, in that it implies that a WFR "always" changes the lexical category of its base and the syntactic features associated with it. In a sense this is true, but it must be demonstrated. On this topic, cf. Williams (1981a) and Scalise (1983). 6. In a sentence such as Mary watched the boys and the girls in the garden, it is clear that WFR's are optional (and, in fact, do not apply here), as opposed to Inflection Rules which, on the contrary, are obligatory. Differences between derivation and inflection are discussed below, Chapter VI. 7. The model represented in Figure 1 is not proposed by Aronoff, who never, in fact, offers an explicit model of his morphological theory. This model was proposed in Scalise (1980); it is based on the models proposed in Siegel (1974) and Allen (1978). 8. This representation exemplifies a word plus a suffix. We assume here that an analogous structure exists for prefixes, although there are some basic differences between prefixes and suffixes, as will be shown in Chapter V below. 9. Among the syntactic features, Aronoff lists only the ones given here. It should be clear, however, that, in principle, all syntactic features can play a role in derivation. In Italian, for example, the suffix -aio (with the paraphrasis "person who has an activity connected with X") attaches only to nouns marked as [-abstract], not to nouns marked as [ +abstract] (cf. giornale + aio -»giornalaio "newspaper seller", vino + aio -» vinaio "wine seller", vs. gloria + aio-^*gloriaio "glory seller",futuro + aio-**futuraio "future seller"). A rare exception to this restriction in usuraio 'usurer'. For a more detailed analysis of the role of features in derivation, cf. Scalise (1980). 10. This problem, and many others, have yet to receive adequate treatment, though lately there has also been some work done in this neglected area (cf. Miller (1978), Booij (1979), Zwanenburg (1980b)). 11. This phonological condition raises other problems which we will not examine here. See Aronoff (1976:83). 12. We will not discuss here whether this restriction should be formulated as a sort of filter (as Booij proposes), or in other ways. 13. Cf. also Williams (1981a). 14. On this notion, cf. Chapter VIII. 15. Cf. Chapter V, sections 2.1 and 2.3. 16. On Adjacency, cf. Chapter VIII, and on Readjustment Rules, cf. Chapter IV.

Chapter IV

Readjustment rules It can be argued that current theories of generative morphology have made an effort to unify the domain of morphology by "subtracting" all morphological rules from the other components of the grammar and stipulating that they should operate only in one place, the lexical component. In the preceding chapters, we have seen which processes have been subtracted from the syntactic component (namely, derivation and compounding). In this chapter, we will see which processes have been subtracted from the phonological component. In particular, this chapter is concerned with the so called "Readjustment Rules", those rules that take care of the final phonological details of a word before it undergoes Lexical Insertion. First several types of Readjustment Rules will be examined, and then some theoretical problems associated with such rules will be discussed.

1. Readjustment rules Traditionally, two types of phonological alternations have been recognized: (a) those governed uniquely by phonetic or phonological factors, and (b) those governed by other factors as well. The second group includes alternations restricted to certain syntactic classes or governed lexically or morphologically (i.e. alternations that appear only in certain morphemes or classes of morphemes, or in the environment of certain morphemes or classes of morphemes). The rules that handle the second group of alternations are now claimed to belong to the lexical component, and they are usually referred to as "Readjustment Rules" (RR's). RR's were commonly used in SPE and in other standard sudies of generative phonology. In the S P E framework, RR's were used to rewrite a word (dominated by a categorial node) plus the syntactic features added to it during a derivation. A RR rewrites, for example, the sequence hit + past as / h i t / or the sequence take + past as / t u k / . In the SPE framework, RR's also had another function: they were used to "clean" surface structures in order to render them suitable entries to the phonological component. For example, they eliminated any boundaries that were not necessary for the phonological rules. Chomsky and Halle (1968) also suggested that RR's should operate when the output of a W F R is not an acceptable input to the phonological rules for other reasons, for example, in the case of the abstract nominal derived from the verb receive, which is not *receive + ion, but reception. To sum up, in the SPE framework, RR's performed operations of three

58

Generative

morphology

types: phonological, morphological and "transit" from one component of the grammar to another one. In most recent accounts, however, RR's have a much more constrained role, as we will see below. In this regard, let us consider Aronoff's general definition of RR's: (1)

[Readjustment rules are] those rules which are restricted to specific morphemes and take place only in the environment of specific morphemes (AronofF, 1976:87)

This definition more or less characterizes those rules which were traditionally referred to as "morphophonemic" rules. Aronoff proposes a further division of this set of rules, however, into Truncation Rules (TR's) and Allomorphy Rules (AR's), defined as follows: (2)

A truncation rule deletes a designated stem-final morpheme before a designated suffix. A Rule of allomorphy adjusts the shape of a designated morpheme or class of morphemes in the immediate environment of another designated morpheme or morpheme class (Aronoff, 1976:88).

In the next two sections, we will examine TR's and AR's more closely, and discuss examples of both types of rules. 1.1. Truncation Rules Let us first consider TR's, which have the following general form according to Aronoff (1976:88): (3)

[[root + A ] x + B ] Y 1 2 3 => 1 0 3 where X and Y are major lexical categories

We can illustrate the operation of TR's with the process that attaches the suffix -ico to a base in Italian. This rule of suffixation exhibits three different patterns (cf. D a r d a n o 1978:76): (a) the suffix attaches to a base without further complications (atomo-»atomico1 "atom - atomic"), (b) the attachment of the suffix modifies the base (diploma->diplomatico "diploma - diplomatic"), and (c) the suffix in question deletes another suffix contained in the base (prosodia^prosodico "prosody - prosodic"). The case exemplified in (b) is a case of allomorphy and the case exemplified in (c), which we will consider further here, is a case of truncation. F o r such cases, a TR can be formulated as follows: (4)

[ [ X + Suf] + i c o ] 1 2 3 => 1 ψ 3

Readjustment rules

59

The derivation of a word such as prosodico, in the framework of the theory presented above, will therefore be as follows: (5)

Dictionary WFR TR Output

[prosod + ia] [ [ p r o s o d + ia] + i c o ] 0 [prosodico]

The proposed derivation has crucial advantages over other possible derivations of the same word. It could be argued, for example, that the same output could be obtained by applying a Vowel Deletion rule twice, rather than a T R . In this case the derivation would be as follows: (6)

Dictionary WFR V. Del V. Del. Output

[ p r o s o d + ia] [ [prosod + ia] + ico] 0

Q

[prosodico]

The derivation in (6), however, is not plausible for two reasons. In the first place, it would require and ad hoc rule of vowel deletion for this specific case since the Vowel Deletion rule that applies elsewhere in Italian does not delete stressed vowels 2 , as can be seen in (7). It could not, therefore, delete the i of the dictionary form of prosodia, since it is stressed: (7)

(i) fama + oso giallo + astro (ii) virtu + oso blu + astro

famoso -> giallastro -» *virtoso *blastro

virtuoso bluastro

"famous" "yellowish'' "virtuous" "bluish"

Note that in (7i) Vowel Deletion operates deleting the last vowel of the stem, but it is blocked in (7ii), where the stem vowel is stressed. In the second place, the attachment of the suffic -ico causes the deletion of whole suffixes in other cases as well, where Vowel Deletion clearly cannot operate, given that the suffix also contains consonants (cf.

difterite / difterico

"diphteritis/diphteric"

or

metamorfosi / metamorfico

"metamorphosis / metamorphic". As it is formulated, however, the rule proposed in (4) is too general: it says that whenever there is a word with one (or more) suffixes, the last suffix will be deleted when the suffix -ico is attached to this word. This is too strong a statement since there are words such as evoluzionistico "evolutionistic" or intellettualistico "intellectualistic" which have the structure [ [ X + S u f ] + i c o ] but, nevertheless, do not undergo the T R in question. We must therefore reformulate (4) in such a way that the attachment of the "designated" morpheme -ico will cause the deletion of the "designated" stem-final morpheme -ia (and several other specific

60

Generative

morphology

morphemes) according to the general definition of TR's seen in (2). Regardless of the final form of the rule required to account for the example in question, it is nevertheless clear that it must be a Truncation Rule that deletes an entire suffix, not individual segments. 1.2. Allomorphy Rules As an example of an allomorphy rule we will examine the case of -ion nominals. Consider the following list of items (Aronoff, 1976:100): (8)

realize educate repeat commune resume resolve

realization *educatation *repetation *communation *resumation *resolvation

*realizion education *repetion communion resumption *resolvtion *resolvion

*realizition *educatition repetition *communition *resumition resolution

From this list, one can conclude that the suffix in question has five forms: ation, -ion, -tion, -ition, -ution. Aronoff observes that -ation is the unrestricted variant since there are no conditions on its attachment. That is, ation can be attached to words ending with labials, coronals and velars as can be seen in the following examples (cf. Aronoff 1976:100): (9)

coronal

labial perturbation formation exhumation unsurpation

cessation degradation elicitation accusation revelation declaration examination representation

deportation manifestation consultation affectation commendation sensation indorsation

velar evocation purgation prolongation

The -ion and -tion variants on the other hand, are governed by what Aronoff calls latinate roots. In particular, -tion attaches to non coronal roots and -ion to coronal roots: (10)

consume/consuption absorb/absorption

rebel/rebellion decide/decision

To account for these facts, Aronoff proposes the following allomorphy rule: (11)

Allomorphy of -ion:

Readjustment

61

rules

+ ation->< + l o n I { + tionj

/xJ

+cor| I — cor J

where Χ α cor is one of a set of specified latinate roots According to (11), the suffix -ation (the unrestricted variant) becomes -»-ion when it is attached to latinate roots ending with a coronal segment, and it becomes +tion when it is attached to latinate roots ending with a non coronal segment. Aronoff's analysis of this phenomenon is more complicated than this short summary shows and it has been called into question on several points 3 . The fact remains, however, that the attachment of a suffix to certain bases often requires readjustment of either the suffix (allomorphy of the suffix) or the base itself (allomorphy of the base). The literature offers a number of examples of such readjustments. In Dutch, for example, a case of suffix allomorphy is to be found in a class of suffixes which changes its vowel from [ — back] to [ + back] when the suffix is followed by another suffix marked as [ — native], as can be seen in the following examples (cf. Booij, 1982:128): (12)

eel aal4: air -» aar: eur -t oor: eus oos:

fundamenteel/fundamentalisme militair/militarist directeur/directoraat religieus/religiositeit

"fundamental/fundamentalist" "military/militarist" "director/directory" "religious/religiousness"

A similar vowel alternation was observed in French by Dell and Selkirk (1978), though in this case it applies to the base and not to the suffix as in vain/vanite "vain - vanity", fleur/florale "flower - floral". What must be stressed here is that these alternations are morphologically governed in the sense that, in both Dutch and French, the rule works only for a specified set of works. That is, it applies before suffixes marked [ — native] in Dutch and before suffixes marked, according to Dell and Selkirk, with a stratal feature [ + learned] in French. This ammounts to saying that the rules that handle these phenomena can be ascribed to the morphological component of the grammar or, in accordance with the terminology used here, to the set of Readjustment Rules that apply, in the lexicon, to the output of WFR's.

2. Justification of RR's It can be said that RR's are a byproduct of Aronoff's theory; that is, (a) TR's are necessary in a word-based morphology, and (b) AR's are necessary under the assumption that a WFR attaches one and only one affix, whose form is constant. In relation to the first point, consider another TR proposed by Aronoff

62

Generative morphology

in order to account for the surface form of -ion nominals derived from verbs ending with -ate. In order to derive a word such as equivocation, Aronoff proposes the following TR: (13)

equivoc + ate + ation 1 2 3 => 1 0 3

Such a rule is necessary in a theory which requires that every base of a derivational process be an existing word; it would not be necessary in a theory where the base can be a morpheme and the suffix can attach directly to the morpheme (e.g. equivoc + ation). The word-based theory has as many supporters as opponents 5 , and despite the problem just mentioned, additional arguments can be given, beyond those discussed in the literature, in favor of this theory. In particular, consider the following four points: 1. Even if we dispense with the word-based hypothesis, we cannot dispense altogether with RR's. They remain necessary, for example, in order to derive an Italian word such as mietitrebbiatrice "lit. reaper-thrasher" from mietitrice-trebbiatrice. In this case we need a TR such as the following: (14)

mieti +trice trebbia + trice 1 2 3 4 1 0 3 4 Condition: 2 = 4 6

It should be noted that this example shows that TR's can apply to the output of Compounding Rules as well as to the output of Derivation Rules. 2. At least in certain languages such as Italian, in which a morphemebased theory would be simpler for derivation and inflection (i.e. where the suffix can be attached to the base without further readjustments), such a theory would give rise to serious complications for compounds because it would require the implementation of "insertion rules". In a compound such as capostazione "station master", for example, a morpheme-based theory would start, presumably, with the two morphemes cap- and stazion-; the missing morphemes would have to be provided subsequently by insertion rules. 3. While a morpheme-based theory seems to have a relatively high degree of psychological reality for languages such as English, where morphemes and words coincide most of the time (e.g. boy, house, nice, quick, walk, laugh which are morphemes and words), this is not the case for languages such as Italian where morphemes and words do not coincide most of the time (e.g. the Italian translations of the English examples: ragazz + o, cas -I- a, carin + o, rapid + o, cammin + a + re, rid + e + re).

Readjustment

63

rules

4. RR's (in particular AR's) allow us to stipulate that the form of a given affix is constant. In the case of different variants of the "same" affix, all we must d o is posit one basic form and account for the others by means of AR's. Finally, let us consider the frequently cited argument that morphemes are most suitable as the basic units of the morphological system of a language because they are "primitive" units. As Matthews (1974:78) puts it: (15)

[ . . . ] the morpheme is established as the single minimal or primitive unit of grammar, the ultimate basis for our entire description of the primary articulation of the language [ . . . ] Words, phrases, etc. are all seen as larger, complex or nonprimitive units which are built up from morphemes in successive stages.

Even if morphemes are, in some sense, "minimal" units of the grammar, this does not necessarily mean that they must be considered "primitive". Every theory can posit its primitives and a particular choice is justified on the basis of the results it yields. Recall that at the beginning of Generative G r a m m a r , Chomsky posited as a primitive of his system the sentence, a unit which is by no mean "minimal". The validity of this choice is, nevertheless, quite clear. To summarize briefly, we have seen that while RR's are necessary in a word-based morphology, no theory can dispense with them altogether. There is, furthermore, evidence that RR's operate on different outputs of the morphological component. In the next two sections, we will motivate the decision to consider RR's a separate block of rules, distinct from both WFR's and phonological rules. 2.1. Readjustment Rules and Word Formation Rules As far as the status of RR's is concerned, two alternatives have been proposed in the literature: (a) RR's are incorporated into WFR's, that is, the same rule which attaches an affix will provide the necessary readjustments, (b) RR's are a different set of rules and, more specifically, they operate on the output of WFR's. Aronoff argues convincingly for the second alternative in a discussion of Siegel's analysis of the suffix -ee which is based, instead, on the first assumption. As we have seen in Chapter III, the suffix -ee attaches to transitive verbs with animate objects. In certain case, the suffix attaches to a base without requiring any particular readjustments, as can be seen in (16i); in other cases, however, certain problems arise, as can be seen in (16ii): (16)

(i) employ->employee pay-> payee

(ii) nominate -»nominee evacuate -»evacuee

Specifically, in (16ii) the morpheme -ate must be deleted. Siegel's solution

64

Generative

morphology

to this problem is the following: (17)

-ee Attachment a. [ [ ] v - e e ] N [ + tr] [ + a.o.] b-

[[]x-ee]

N

where there exists [Y]v [ + tr] [ + a.o.] such that [ Υ ] ν = [ [ ] χ ] ν (17a) accounts for the data in (16i), and (17b) for the data in (16ii). As was indicated above, a solution such as the one in (17) makes use uniquely of WFR's, and not of RR's. In order to avoid RR's, however, the W F R in question must be divided into two parts. Aronoff's solution, on the other hand, is to postulate a W F R of the same form as the one in (17a), whose output will be either X + ee or X + ate + ee. A Truncation Rule (such as the one seen above in (13)) subsequently deletes the morpheme +ate, when present, yielding the desired form X + ee. The advantages of the solution making use of RR's are manyfold: 1. It allows us to formulate only one W F R , which expresses correctly the fact that nominee bears the same relationship to nominate that employee bears to employ. The solution seen in (17), on the contrary, does not express this fact, since the sub-parts of the rule, (a) and (b), are two distinct rules. 2. It rules out outputs such as *nominatee, *evacuatee, which the rule (17) cannot avoid. It would be necessary, in fact, to formulate a restriction on (17a) to the effect that this subpart of the rule could not operate where (17b) operates. This disjunction is, however, implicit in the proposal with both W F R ' s and RR's. 3. It does not introduce ad hoc bracketings, such as the bracketing [ ] χ in (17b), which has no other meaning than the one attributed to it by this rule. 4. It avoids the type of condition in (17b) which is somewhat peculiar in the sense that it states, essentially, that a given word can be created if a certain other word exists from which, however, it is not derived. In addition to these arguments, originally given by Aronoff, against incorporating Readjustment Rules as part of WFR's, there are several other arguments in favor of considering RR's as a separate set of rules. First of all, as Booij (1977:40) points out, a suffix-replacing W F R would have transformational power, thus preventing us from properly constraining both WFR's and transformations. Since W F R ' s are structure building rules (cf. Chapter VII), if they are attributed with transformational power, we can no longer constrain transformations by stipulating that they do not in general 7 build structure.

Readjustment

65

rules

Secondly, according to Carrier (1979), the separation between WFR's and RR's not only makes WFR's simpler, but it also simplifies the formulation of some allomorphy processes. Consider the following pairs of words, all of which appear to bear the same morphological relationship to each other: the (18ii) forms are abstract nominals formed by adding +ion to the corresponding verbs in (18i): (18)

(i) immerse subvert conceive

(ii) immersion subversion conception

Carrier's argument runs as follows: in these examples we are faced with a case of root allomorphy vert/vers, ceive/cept). If we were to incorporate this root allomorphy into the WFR that attaches the suffix -ion, we would have to posit three separate WFR's, each of which performs the same formal operation (attachment of the suffix -ion to a verb to form an abstract nominal): (19)

(i) [ X ] v - [[X]v -ion] N (ii) [X-ceive]y [[X-cept]y -ion] N (iii) [X-vert]y -*• [[X-verd]y -ion]>j

If, however, the processes that change vert to verd and ceive to cept are separate from the affixation of -ion, the same -ion WFR will handle the derivation of all three abstract nominals. Thirdly, as Carrier points out, there are also RR's which are triggered by Inflectional Rules, and these must clearly be separated from WFR's. Actually, a more general point can be made here. In Scalise (1983), it is shown that a single rule, Vowel Deletion, applies to the output of the whole set of morphological rules in Italian, that is, to Derivation Rules, Compounding Rules and Inflection Rules. Consider the following examples: (20)

Derivation

Compounding

Inflection

veloce + itä -> velocitä amaro + ezza -»amarezza gloria + oso->glorioso fuori + uscito ->fuoruscito sopra + abito ->soprabito porta + aerei -> portaerei casa + e->case buono + i->buoni vende + uto -> venduto

"speed" "bitterness" "glorious" "refugee" "overcoat" "aircraft carrier" "houses" "good" (m.pl.) "sold"

The fact that Vowel Deletion applies at all of the levels seen in (20) shows that it is not possible to assign (at least certain) RR's to specific WFR's for two reasons. First of all, since the same Vowel Deletion rule applies in relation to almost all Derivation Rules and to large number of

66

Generative

morphology

Compounding Rules, this would require that the same rule be repeated in each W F R . Secondly, since Vowel Deletion also applies to the output of Inflectional Rules, which are not WFR's at all, it means that in this case the RR in question is, by definition, not part of a WFR 8 . In conclusion, we have seen that there are several arguments that clearly favor separating RR's from W F R ' s These can be summed up as follows: (a) separating W F R ' s and RR's makes WFR's more general thus allowing them to express more accurately specific morphological relationships; (b) this separation also makes allomorphy processes simpler; (c) there are RR's which apply to outputs of different WFR's and by separating them from the individual WFR's we can express the generality of the process in question by formulating a single RR; (d) the separation enables us to constrain the rules of the grammar, in this case by preventing WFR's from having transformational power.

2.2. Readjustment Rules and Phonological Rules Readjustment Rules are commonly considered "morphological" rules. That is, they apply to the output of WFR's (within the lexicon), before all other phonological rules. Thus, RR's are considered to be different in nature from the rules that operate in the phonological component. While this claim can easily be justified for TR's, which are structurally quite different from phonological rules, it needs to be motivated for AR's, which clearly bear some resemblance to phonological rules. In order to distinguish in a principled way between phonological rules and AR's, Aronoff imposes three restrictions on AR's. In the first place, the context of AR's is quite restricted. Specifically, an AR is a rule "which effects a phonological change, but which only applies to certain morphemes in the immediate environment of certain other morphemes" (Aronoff 1976:98). Phonological rules are not subject to restrictions of this type. In the second place, AR's cannot introduce segments which are not found in the underlying representations of the language. In the third place, AR's are ordered (together with TR's) before the rules of the phonological component. In Aronoff's words: "the ordering of allomorphy rules before the rules of the phonology, strict limitations on the environment in which these rules may operate and the restriction to underlying phonological segments, greatly constrain this otherwise omnipotent device" (p. 99). The second point (differences in environment) is illustrated below with an example from Italian. Consider first the following basic rule form: (21)

A->B/

Ζ

Call "B" the "target" of the rule and "Z" the context of the rule. If A, Β and Ζ are phonological segments, then the rule is a phonological rule. If, on the other hand, Ζ and/or Β are morphological elements, then the rule, according the given definition, is a RR 9 .

Readjustment

rules

67

Consider now the following data from Italian: (22)

corretto + [ j ] o n e ammonito + [ j ] o n e ingiunto + [ j ] o n e descritto + [ j ] o n e corretto + [i]vo ingiunto + [i]vo descritto + [i]vo

-» -> -> -» ->

corre[ts]ione ammoni[ts]ione ingiun[ts]ione descri[ts]ione correttivo ingiuntivo -*• descrittivo

'correction" 'admonition 'injunction" 'description'' 'corrective" 'injunctive" 'descriptive"

On the basis of these examples, one may hypothesize a rule such as the following: (23)

(t)t-[ts]/

+j

According to this rule, a voiceless dental stop (simple or geminate) becomes an affricate before the semivowel j. In this form, the rule in question would be considered a "phonological" rule, according the definition given above, since A, B, and Ζ are "phonological" segments 10 . It can be demonstrated, however, that the rule that changes t to ts cannot be expressed in purely phonological terms. Consider, in fact, the following data: (24)

carretto + [j]ere carre[tt]iere argento + [j]ere argen[t]iere m o n e t a + [ j ] e r e -»· mone[t]iere

"cart driver" "silversmith" "minter"

These examples show that a formulation such as the one in (23) is not adequate. We see, instead, that the rule in question is triggered by the suffix (-ione) and not by another suffix (-iere), even though the phonological context is the same in both cases (i.e. -1-j). We must, therefore, formulate the rule that accounts for the data in (22) in a different way, for example as in (25): (25)

(t)t-[ts]/

+ione

In this formulation, Ζ is a "designated" morpheme, and the rule cannot, therefore, be a phonological rule, according the definition given above.

3. Summary In the preceding sections, it has been seen that the lexicon includes a set of rules, the Readjustment Rules, which apply to the output of WFR's. The RR's are divided into two subsets: Truncation Rules and Allomorphy Rules. Both types of rules are "morphological" rules, not phonological

68

Generative

morphology

rules, and, moreover, are separated from the WFR's themselves. At this point, the model of the lexical component given in Chapter III, sec. 2, can be modified as follows:

Lexical Component Dictionary I WFR's I RR's (TR's, AR's)

4 output

— > L.I.

Figure 1: The organization of the lexical component (II). Figure 1 has been extrapolated, as was Fig. 1 in Chapter III, from Aronoff's proposal. It represents a model of the lexical component of the grammar which contains a level of representation, the Dictionary, which consists only of words. It shows, furthermore, that words may be subjected to Word Formation Rules (only derivational rules in Aronoff's proposal). After the application of the WFR's, RR's may apply in order to adjust the phonological form of a word before it undergoes lexical insertion. In the following chapters, the model represented in Fig. 1 will be discussed further, and improved in several respects.

Notes 1. In order to get atomico from atomo + ico, the only rule we need is Vowel Deletion rule. On this rule, see footnote (2) below. 2. The Vowel Deletion rule in Italian has the following form (cf. Scalise 1983:287 IT.) V->0/ + V. That is, an unstressed vowel is deleted when it is followed by a [ — stress] morpheme boundary which is, in turn, followed by a vowel. Cf. also footnote (8) below. 3. See Lieber (1980), Zwanenburg (1980a) and Scalise (1983). 4. A double vowel is changed to a single one when it is in an open syllable according to a convention of Dutch othography. Thus, when the suffixes in question are followed by another suffix beginning with a vowel, we find a single vowel rather than the double vowel of the citation form. 5. Among the former are Jackendoff (1975), Booij (1977), Allen (1978), Dell (1979), Scalise (1980). Among the latter are Halle (1973), Lieber (1980), Bauer (1979), Selkirk (1982). An intermediate position would be to maintain that the hypothesis is not universally'valid

Readjustment

rules

69

but it should be verified for each language. Booij (1977:32), for example, defends the Word Based Hypothesis for Dutch, but he does not want to propose it as a universal principle. 6. The condition says that the suffixes must be identical. Even though this condition holds for a large number of cases (cf. carto-chirmante from cartomante-chiromante "lit. fortune-teller - chiromancer", carto-libreria from cartoleria-libreria "stationery-bookshop", etc.) there are other cases in which it seems that we must relax the condition (cf. Scalise 1983:314). 7. With the exception of the so-called "Chomsky Adjunction" transformation. 8. The rule of Vowel Deletion is not exactly a Readjustment Rule if we strictly observe Aronoff's definition. It is probable, however, that Aronoff's definition will, in many cases, need to be revised. It should be noted, though, that Vowel Deletion (in the form given in In. 2) must operate in the lexical component and cannot be shifted into the phonological component (cf. Scalise, in prep.). 9. There are actually somewhat diverging opinions about the requirements a rule must satisfy in order to be considered a RR. According to both Aronoff (1976) and Carrier (1979), the environment of a RR must be specified in morphological terms. As far as the "target" is concerned, Aronoff's definition (i.e. that the target must be morphological) is stronger than the one proposed by Carrier (i.e. that the target may also be phonological). It should be noted, however, that Aronoff actually writes Allomorphy Rules with a phonological target, for example, the rule d - » s / n + ive (p. 108), which accounts for alternations such as defend/defensive. 10. The presence of the boundary " + " does not change the "phonological" nature of rule (23), at least in some definitions of "phonological rule". See, for example, the following definitions given by Anderson (1975:42-43): "Morpholexical rules are those whose environments involve essential reference to the identity of specific morphemes, lexical items, or classes of morphemes [ . . . ] phonological rules are those whose environments involve reference only to the phonological composition of elements of the string, to boundary elements, and perhaps to major lexical classes [...] and which furthermore operate so as to alter the categorical ( + / —) value of some feature which is distinctively specified in underlying forms of the language".

Chapter V

Lexical formatives and word formation rules

In the preceding chapter, a schematic representation was given of a model of lexical morphology that consists of different levels, specifically, a Dictionary and two blocks of rules. In this chapter, a proposal will be made for how lexical items and lexical rules can be represented. As far as lexical items are concerned, it will be proposed that the units of the Dictionary are "words" and "stems". We will propose, furthermore, a representation of these dictionary units, such that they can function properly within the theoretical framework sketched above. As far as lexical rules are concerned, we will give a representation of prefixation rules, suffixation rules and compounding rules, showing how these rules use the information associated with the lexical items. Finally, some wellformedness conditions will also be provided for these three types of rules.

1. Words and stems Before discussing how the formatives of the Dictionary are to be represented, we must discuss briefly which are these formatives. As has already been seen, Aronoff argues in favor of a "Word Based Hypothesis". The definition of the WBH has already been given in Chapter III, section 1.1. Let us now, for the sake of the discussion, divide this definition in three different points: (1)

(i) All regular word formation processes are word based. (ii) A new word is formed by applying a regular rule to a single already existing word. (iii) Both the new word and the existing word are members of major lexical categories.

This definition has been questioned in all its aspects. Point (liii) has been called the Major Category Restriction (cf. Muysken, 1981), and it has been questioned by Bauer (1980:78) on the basis of "words" such as downer, inness, uppity. These words, however, are not created by "regular" processes of word formation, as required in point (li), but rather represent marginal forms in English, and thus are not a serious challenge to the WBH 1 . Point (Iii) has been questioned on both the requirements "single"

72

Generative morphology

and "already existing". As far as the requirement "single" is concerned, the original meaning of the WBH was, as we have already mentioned, that a WFR cannot apply to phrases (or sentences). Botha (1981:18) refers to this restriction as the No Phrase Constraint, and argues against it on the basis of Afrikaans data 2 . We will discuss the No Phrase Constraint below, in Chapter VII. As far as the requirement "already existing" is concerned, several scholars have proposed that a WFR can take as its base a possible but non existent word (cf. Allen 1978, Zwanenburg 1980, Scalise 1980). In particular, Allen (1978) has proposed an "overgenerating" morphology according to which WFR's can operate on non existent words. Point (li) is one of the most controversial topics in the recent literature on morphology. Among those works in which point (li) is accepted, are those listed in (2), though it must be noted that almost every work contains proposals for modifying the WBH. (2)

Jackendoff (1975) Booij (1977) Scalise (1980) Zwanenburg (1980) Muysken (1980)

(not as a universal principle) ("abstract" words must be allowed) ("possible" words must be allowed)

Other scholars have, more or less explicitly rejected point (li) of the WBH. Among these, we can list the following 3 : (3)

Siegel (1974) Allen (1978) Hoekstra et al. (1979) Lieber (1980) Williams (1981) Kiparsky (1982) Selkirk (1982)

We will not go into the arguments presented against the WBH here; we will limit ourselves, instead, to several general observations. If we apply Aronoff's definition of "word" to different languages, the results are often somewhat different. Consider the following "words with inflection" 4 : Engl. Fr. Sp. Dutch Germ. Rum. Sw.

boy-s jardin-s libro-s tafel-s Kind-er studen^-i ros-or

"gardens" "books" "tables" "children" "students" "roses"

Lexical formatives

and word formation

rules

73

If we remove the inflections from these "words with inflection", the forms we end up with are the following: (5)

boy jardin libro tafel Kind student ros

Each of these forms is a "word" in the sense of Aronoff's definition. Such words have the following properties: (a) they are "morphemes" in the sense that they cannot be further analyzed into smaller meaningful units, and (b) they are "free forms" in the sense that they can appear on the surface as they are without the application of additional morphological rules 5 . Consider next the following "words with inflections", some in different languages, some in the same languages as above: Lat. It. Rum. Sw. Gr.

lup-i can-i munt-i flik-or pater-es

"wolves" "dogs" "mountains' "girls" "fathers"

If we now apply the same procedure to the words in (6) that we applied to those in (4), the results are the forms in (7): (7)

lup can munt flick pater

These items are similar to those in (5) in that they are morphemes 6 , but they are different from them in that they are not free forms. Instead, the forms in (7) are bound forms that cannot appear on the surface as such (i.e. the singular forms are lupus, cane, munte, flicka, pateras, respectively). This means that in these languages (or certain word classes in certain languages) inflection is effected by "addition" as in English, in Dutch, etc., while in other languages (or certain word classes in certain languages) it is effected by "substitution", as in Latin, Italian, etc. In this sense, languages such as English and Dutch can be considered more "agglutinating" in nature and languages such as Latin and Italian more "inflecting". In relation to the point under discussion here, we can say in general that if we remove the inflectional element from inflected words in languages such as

74

Generative

morphology

English and Dutch, what remains is a free form which can be considered a "word", while the removal of the inflectional element from words in languages such as Latin and Italian results in a bound form which cannot be considered a "word". It has been seen, furthermore, that there are languages that behave in one way with respect to one word class and in another way with respect to another word class. Bloomfield (1933:224ff.) made a distinction between English and German on the grounds that English has word-inflection, word-derivation and word-composition. Thus in items such as plays, player, playground there is always the word play, a free form. German, on the contrary, has stem-inflection, stem-derivation and stem-composition. In the regular padigm of a German verb, we find a common element, for example lach [lax] which is not equal to any of the inflectional forms (cf. lachen [lax-en] "to laugh", (ich) lache [lax-e] "(I) laugh", (er) lacht [lax-t] "(he) laughs", (er) lachte [lax-te] "(he) laughed", gelacht [ge-lax-t] "laughed"). In derivation and compounding, the paradigm is represented by the same form (cf. Lacher [lax-er] "laugher" and Lachkrampf [lax-krampf] "laughingspasms". Thus, lach is not a free form, it is a bound form. A similar situation is found in languages which are said to have "thematic" derivatives such as Classical Greek, Latin, Sanskrit, etc. In Classical Greek, for example, in derivation and in compounding we find forms which d o not coincide with any inflected form. In the derivate [hippo-te:s] "horseman" and in the compound [hippok a n t h a r o s ] "horse-beetle", for example, we find the common element [ h i p p o ] "horse" which d o not coincide with any inflected form (cf. Bloomfield, 1933:224): nom. gen. dat. acc. voc.

[hipp-os] [hipp-ow] [hipp-o:j] [hipp-on] [hipp-e]

The form hippo cannot be considered a "word" in Aronoff's sense, but rather a "theme" (that is a root (hipp-) plus a thematic vowel (-o)), a bound form again. F r o m what has been said, one must not conclude, however, that all the languages in (6) must be considered non word based. There are, in fact, arguments, for considering Italian morphology as word based, but this requires a weakening of the original formation of the W B H (cf. Scalise 1983). Given our present state of knowledge, it seems clear that the W B H is "subject to typological differences" (Hoestra et al. 1979:20). In this chapter, we will take a position that represents a compromise between the two positions outlined above: on the one hand, we will consider stems to be units of the lexicon along with words, but on the other hand, we will propose that stems are to be considered a special subpart of the Dictionary 7 . The set of stems can be divided into two basic categories: (a) those which are no longer productive (e.g. sipid, maculate),

Lexical formatives

and word formation

rules

75

and (b) those which are productive in the formation of the learned lexicon (e.g. logy, graphy). While the former are generally recognized as stems, the latter often are not. In the next section, we will examine these items, and in particular, the reasons for considering them stems.

1.1. Learned stems Items such as anglo, bio, electro, franco, etc. are often referred to as prefixes, and items such as erat, phile, etc. as suffixes 8 . According to this analysis, a word such as Francophile would have the following structure: (9)

[ [franco] P r e [phile] S u f ]

This structure, however, is unsatisfactory for a number of reason: (a) An item such as phile can be the second element of a word, as in Francophile, but it can also be the first element as in philanthropist. A "true" affix is not this free; if it occurs to the left it is a prefix, if it occurs to the right it is a suffix. (b) The items in question can be "factored out" 9 . One can say, for example, it does not matter if they are philo- or anti-Soviet, but with "true" affixes this is not possible, as in *I do not know if he should be dis- or encouraged. (c) The items in question can combine quite freely with each other, as in an Ang lo-1 talo-Soviet production and an Italo-Anglo—Soviet production. Again, this is not the case with "true" affixes as in indeformable vs. *deinformable. (d) If we label the internal constituents of words such as Francophile as Pre + Suf, as in (9), most morphological theories would have problems in deriving the external label A(djective), which is necessary information for the proper functioning of the word. (e) A structure such as the one in (9) is highly marked and, furthermore, does not occur with "true" affixes (cf. *in + ic, *super + ous, *de + ous, etc.). (f) A structure such as the one in (9) behaves more like a compound than like a derived word. This can be seen, for example, in cases where an ο appears in compounds (e.g. halo-American) as well as in combinations in which the second element is one of the items in question, and in particular, one with the stratal feature [ + G r e e k ] (e.g. music+ logy->musicology, dialect + logy-*dialectology). It should be noted that the ο in the examples just given cannot be part of the Greek element (i.e. *-ology), since such an analysis would require that we posit forms with this ο also for items such as -ograph(y), -ophile, -ophobe, -ometer, etc. (cf. historiography, Germanophile, Germanophobe, galvanometer). By doing this, we obscure the relations between graph(y) in historiography and in graphology, graphomaniac, between phile in Germanophile and in philanthropist, etc. The problem raised in (f) can be easily resolved if we consider the items in question to be "stems" rather than affixes (cf. also sec. 2.2.2. below). Such stems have a basic form without a vowel, and when necessary, a

76

Generative

morphology

readjustment rule applies to provide one. By stating the phenomenon in this general way, we are able to express both the fact that it is ο that appears when the second element is [ + Greek], and the fact that it is i that appears when the second element is [ + Latinate], as seen in (lOi) and (lOii), respectively 10 : (10)

(i) music + logy -»musicology history + graphy -»historiography German + phile->Germanophile (ii) cone + fer -»conifer insect + cide->insecticide herb + vore-> herbivore

Considering the items in question stems rather than affixes has an additional advantage, and that is that we can now refer to an item such as phile in the samy way (i.e. stem) whether it is the first or second element of a word (cf. philanthropist, anglophile). Finally, we can also identify three different types of compounds in which at least one element is a stem (in addition to the usual type consisting of two wrds, e.g. windmill), as seen in (11): (11)

(i) [ [ S T E M ] [ S T E M ] ] (ii) [ [ S T E M ] [ W O R D ] ] (iii) [ [ W O R D ] [ S T E M ] ]

Anglophile, astronaut geophysics, graphomaniac musicology, insecticide

What all of these considerations show is that there must be a special subpart of the lexicon that accounts for the functioning of learned stems since they are, in fact, quite productive synchronically. We will accept here Booij's (1980) suggestion that the grammar incorporate a list of Greek and Latin stems for the formation of the learned lexicon.

2. Representation The list of lexical formatives includes: (a) underived words (or simple words), (b) stems, (c) affixes (i.e. prefixes and suffixes), and (d) inflectional morphemes. The pertinent question now is how these formatives are to be represented. Before answering this question, however, we must briefly examine the problem of boundaries associated with words.

2.1. External boundaries In the SPE theory, each lexical category was assumed to carry automatically a boundary symbol " # " to its right and to its left. An adjective such as sincere would, therefore, have been represented in the Dictionary as follows:

Lexical formatives and word formation (12)

rules

77

[ # sincere # ]

Assuming as a starting point the SPE representation, Siegel (1974) proposes the following structure for a derived word: (13)

[ # [ # W O R D # ] w [ + Suf]Suf#]w

According to this structure, a word is labeled W(ord), and it is preceded and followed by a word boundary " # " ; this is true for both underived and derived words. A suffix is labelled S(uffix), and it is preceded by a boundary. In (13), the suffix boundary is a morpheme boundary " + ", but we will see below that affixes can be introduced by both types of boundaries, that is morpheme boundaries and word boundaries. The word sincerity, for example, has in Siegel's account, the following representation: (14)

[#[#sincere#] [ + ity]#]

In order to delete the boundaries associated with the inner word, Siegel proposes two rules (following Selkirk 1972). The first one is the following: (15)

+

This rule reduces the representation of [ # [ # s i n c e r e / ] [ + i t y ] # ] to [ # [ # s i n c e r e ] [ + i t y ] # ] . To eliminate the second unnecessary boundary (the one immediately to the left of sincere), Siegel proposes another rule, which is given the name of SPEII: (16)

in the sequence Ζ # χ ] # γ ] or W [ Y # [ X # Z , where Y / S , delete the "inner" word boundary (p. 157).

The application of SPEII to [ # [ # s i n c e r e ] [ + i t y ] # ] results in [ # [ s i n cere] [ + i t y ] # ] . This machinery has been quite convincingly criticized by Booij (1977:92 ff), whose argument runs as follows: In the SPE framework, word internal boundaries are inserted after the transformations have operated, and these boundaries can subsequently be replaced by morpheme boundaries by means of affix-specific Readjustment Rules. Such a procedure, however, is a consequence of the transformationalist framework of SPE in that it is the only way to obtain word internal boundaries if complex words are derived by transformations. In a lexicalist framework which contains WFR's, however, the most natural solution is to have the WFR's specify the internal structure, including the internal boundaries, of complex words. Booij thus proposes a modification of the Boundary Insertion Convention (BIC):

78

Generative

morphology

(17)

T h e boundary # is automatically inserted at the beginning and the end of every string dominated by one of the lexical category "noun", "verb", "adjective" which is not dominated by a lexical category Ν , V, A or by a category such as "sentence", "noun phrase", "verb phrase" which dominates a lexical category (1977, p. 93).

Since in a word such as [ [sincere]^ + i t y ] N , the internal A is dominated by N , there will be no insertion of # and there will therefore be no need of deletion rules of the type postulated by Siegel. Booij's proposal can be further simplified as follows: (18)

T h e boundary # is automatically inserted at the beginning and at the end of the maximal string dominated by X. where X = any M a j o r Lexical Category.

According to (18), we would have the following derivation of sincerity: (19)

W

[sincere] A

DR

[[sincere]A

BIC

[#[sincere]A + i t y ] N # ]

In this derivation,

+ity]N

a WFR

(i.e. D R )

inserts the internal

morpheme

boundary " + ", which indicates that the word has internal structure and the B I C inserts the external w o r d boundaries which define the domain of the " w o r d " so that other rules of the grammar

(e.g. transformations)

cannot operate below this domain. W e can now return to the representation of lexical formatives. 2.2. Formatives

of the lexical

component

In what follows, a proposal will be made for representing the formatives of the lexical component, though the present discussion will be limited to the issues relating to syntactic category. 1 1 The other issues relating to syntactic features, subcategorization features, stratal features, subcategorization frame and argument structure will be handled separately. T h e formatives of the lexical component (simple words, stems, affixes and inflectional morphemes) must all be represented unequivocally, that is, in such a way that each type of formative has a type of representation different from all the others. Each type of formative will be discussed individually. 2.2.1. Simple

Words

Simple (or underived) words have the following characteristics: (i) they are labelled with a syntactic category (i.e. one of the categories mentioned by the phrase structure rules), and (ii) they contain no internal boundaries. Examples of the representation of simple words are: [ b o o k ] ^ , [honest] A, [play]v.

Lexical formatives 2.2.2.

and word formation

rules

79

Stems

The characteristics of stems are: (i) they are bracketed, (ii) they are not labeled with a syntactic category (cf. below), and (iii) they d o not contain internal boundaries. In relation to point (ii), it is necessary to make several remarks. Siegel has proposed that stems are not labelled with a syntactic category, but rather with the label S(tem). This proposal is not adequate, however, in the type of system under consideration since it provides no way to determining the syntactic category of words composed, for example, of a prefix and a stem. That is, if a word such as immaculate is bracketed as [ [in] p r e + [maculate]g], there is n o way to know that the whole word is an adjective. A similar point can be made about words composed of two stems, such as astronaut. Stems must therefore have a category label that (a) allows for the correct functioning of these items in the morphology (and thus allows us to determine the category of the entire word), and (b) prohibits them from undergoing lexical insertion. The solution proposed by Siegel satisfies (b) but not (a). An alternative solution would be to label stems as "sX", that is, as "stem of X", where X is a major lexical category. A speaker "knows" what syntactic category a stem is associated with. F o r example, phono is associated with the category noun since it is subject to a derivational rule that normally derives adjectives from nouns by adding -ic (cf. history—> historic and phono-aphonic). By the same token, maculate must be associated with the category adjective since it is subject ot the rule that adds the negative prefix in- to adjectives (cf. probable ^improbable, maculate->immaculate). F r o m a formal point of view, we can assume that the operation of WFR's takes into account the syntactic category X and subsequently deletes the label "s" once the new word has been created. 2.2.3.

Affixes

1 "J Affixes are introduced by the derivational rules. There is thus no difference between the representation of affixes and the rules that introduce them. Affixes, therefore, must (i) be bracketed not as a u t o n o m o u s formatives but rather as formatives that are added to other specific formatives, (ii) not be labelled with only one syntactic category since they, in fact, express a relation between two syntactic categories (e.g. -ic forms Adjectives from Nouns), and (iii) be provided with a boundary symbol (see below). These points can be illustrated in [ d e + [generate] γ ] ν a n < 3 [[music] ν +al]AThe set of affixes can be divided into two subsets, prefixes and suffixes, depending, among other things, on the position in which they are attached. The former precede the element to which they are added, while the latter follow it. It should be noted that prefixes are different from stems (which we have seen are sometimes also referred as prefixes); while prefixes are uniquely attached to the left of their base element, stems exhibit a variety of combinatorial possibilities. As the following examples from Italian illustrate, they may be (a) preceded by a prefix (e.g. a+fono

80

Generative morphology

"soundless"), (b) followed by a suffix (e.g. fono + ico "phonic"), and (c) combined with other stems or words in compounding formation (e.g. tele +fono "telephone", fono + gramma "phonogram", bio + chimica "biochemistry"). We will now consider more specific characteristics of prefixes and suffixes. In the case of prefixation, the syntactic category of a derived word is that of the word to which the prefix is added: (20)

[dis -I- [honest] A] A [dis +[charge] y ] v [dis + [ f a v o r ] N ] N

Similarly, when prefixes are attached to stems, the category of the derived word is that of the "X" of the stem, "sX": (21)

[in + [maculate] s A] A [de + [duce] s y]v

Prefixes, in contrast with suffixes, do not always rigidly select the syntactic category of the base to which they are attached. Thus, while the negative prefix in- always selects Adjectives (cf. in-capable, in-probable) and the prefix en-selects Verbs (cf. en-trust, en-close), the prefix super can attach to nouns, adjectives and verbs (cf. super-man, super-intelligent, super-saturate). The representation of a prefix will, therefore, be as follows: (22)

[Pre + [

]χ]χ

In order to express the fact that prefixes, contrary to suffixes, do not change the syntactic category of their base 13 , (22) can be represented more accurately as in (23): (23)

[Pre + [

]αΧ]αΧ

The BIC will later assign external word boundaries to (23) giving as a result the structure in (24): (24)

[#Pre + [

]αΧ#]αΧ

Suffixes differ from prefixes not only in that they follow the element to which they are attached, but also in that (a) they derive words that belong to a specific syntactic category (e.g. -ity always derives Nouns, -ous always derives Adjectives, etc.), and (b) they are attached to words which belong to a specific syntactic category (e.g. -ity is attached to Adjectives, as in serene-*serenity, obese->obesity, -ous is attached to Nouns, as in fame^famous, couragecourageous. We can thus accept a proposal made by Pesetsky (1979:21) for a well formedness condition of suffixation:

Lexical formatives (25)

and word formation

rules

81

apply Axy only t o the structure [ ] χ where it is assumed that each affix is specified as bearing the relation Δ to two categories, χ and y, where χ is the label on the bracketing to which the affix can attach, and y is the label on the new bracketing created by the attachment.

Suffixes are thus represented in the following way: (26)

[[

] x + Suf A xY]Y

The BIC will later assign external word boundaries to the structure in (26) in the usual way: (27)

[#[

]X + SufAXY#]Y

2.2.4. Inflectional

morphemes

We will assume that inflectional morphemes are represented in the same way as affixes. Inflectional morphemes, however, in contrast to prefixes, always follow the element to which they are attached and, in contrast to suffixes, do not change the syntactic category of their base. Inflectional morphemes, in contrast to both prefixes and suffixes, cannot be attached to stems. Examples of inflectional rules are seen in [ [t>oy] N + s]fj, [ [walk]v+ed]v· In conclusion, we can say that in the dictionary there are free forms (words) and bound forms (stems). There are also other b o u n d forms, that is, prefixes, suffixes and inflectional morphemes, but these are represented by means of "rules". Bound forms d o not undergo Lexical Insertion. Lexical Insertion can insert into deep structures only simple words or complex words, the former without any internal structure and the latter with internal structure. Both simple and complex words have external boundaries. F r o m what has been said above, we can conclude that a complex word is a derived word (or, as we will see below, a c o m p o u n d word) which (i) is bracketed, (ii) belongs to a syntactic category, and (iii) contains internal boundaries. 1 4 2.3. Class I and Class II

Affixes

In relation to the morphology of English, many linguists accept a distinction between two classes of affixes, referred to as Class I and Class II. The distinction is based on the following differences in the behavior of the two types of affixes 15 : (a) Class I affixes cause a stress shift, while class II affixes are stress neutral, as seen in (29): (29)

Class I ciirious -»curiosity lücid-»lucidity

Class II ciirious -»curiosness liicid-»lucidness

82

Generative

morphology

(b) Class I affixes allow nasal assimilation, while Class II affixes do not: (30)

Class I

legal -»illegal, *inlegal regular-»irregular, *inregular Class II

lawful->*ullawful, unlawful real *urreal, unreal (c) Class I affixes can attach to stems, while Class II affixes can attach only to words: Class I

Class

sub-mit de-duce frict-ion dict-ate leg-al

re-wash sub-human kind-ness peace-ful treat-ment

II

In order to account for these facts, it has been proposed that Class I affixes are introduced by a morpheme boundary " + " and Class II affixes by a word boundary " # " . The morpheme boundary allows the application of some phonological processes (e.g. stress shift, nasal assimilation), while the word boundary blocks them. The boundary is part of the representation of the affix itself, so in English, the two classes of affixes can be represented in the following way: (32)

Class I: +ion, +ity, +ate, -l-ive, +ic, sub + , de + , in + , etc. Class II: # y , #ly, # like, #ful, #some, #ish, # e d , #ing, #ness, r e # , u n # 1 6 , etc.

Accepting this distinction, along with the distinction between words and stems, we arrive at the following possible structures in English: (33)

[ P r e + [STEM]] immaculate deduce refract imply

[Pre# [STEM]] (none)

[ [ S T E M ] +Suf] friction legal probity vacate phonic

[ [STEM] # Suf ] gruesome hapless

Lexical formatives

and word formation

rules

[Pre+ [WORD]] inequality degenerate compassion

[Pre# [WORD]] rewash paramedical monosyllable superman

[[WORD] +Suf] profanity limitation musical supervisory acidity

[[WORD] #Suf] kindness peaceful refusal happyness heavily

83

The distinction between " + affixes" and " # affixes" is the basis of an important hypothesis concerning ordering in English word formation which can be stated in the following terms (Siegel 1974:152): (A) Class I affixation precedes Class II affixation; (B) Cyclic stress assignment rules follow Class I affixation and precede Class II affixation. The two theoretical claims can be represented by the following ordering: (34)

(1) Class I affixation (2) Cyclic phonological rules (3) Class II affixation

(34) represents a hypothesis (referred to as the Ordering Hypothesis) by which the morphological component is considered to consist of different blocks of rules, ordered among themselves (though nothing is said about the possible ordering inside each block) 17 . According to (34), + affixation applies first, then the cyclic phonological rules apply, and finally # affixation applies (after which no cyclic phonological rules may apply any more). Let us now examine concretely some of the consequences of (34). (34) predicts that the structures in (35i) should be possible structures while those in (35ii) should be ruled out: (35)

(i) [Pre # [ [ X ] + Suf]] [ [Pre + [X] ] # S u f ] [Pre # [Pre + [X] ] ] [[[X]+Suf] #Suf]

(ii) *[Pre + [ [ X ] # S u f ] ] *[ [Pre # [X] ] + Suf] * [Pre + [Pre # [ X ] ] ] * [ [ [ X ] # Suf] + Suf]

As an illustration of how the Ordering Hypothesis succeeds in accounting for differences in the behavior of certain affixes in a unified and elegant way, consider the following differences between the two negative prefixes in- and tin- (cf. Allen 1978): (a) In- causes a shift in the position of primary stress, while un- does not

84

Generative

morphology

affect stress: (36)

pious-»impious finite -»infinite

lawful -»unlawful grateful-»ungrateful

(b) In- permits nasal assimilation, while un- does not: ,„„ (37)

. , , i illegal in-legal-»"! ^ ^

un + lawful

f unlawful

(c) In- can be attached to stems, while un- usually cannot: (38)

implacable immaculate

*unplacable *unmaculate

(*placable) (*maculate)

(d) In complex words with the structure Pre-X-Suf, in- co-occurs with one set of suffixes and un- with another set of suffixes: in X ion inXitude inXity in X ate inXive inXic

indecision incertitude inability inanimate indecisive inorganic

(ii) u n X y unXly unXlike unXful u n X worthy unXsome unXish unXed unXing unXness

unlucky unfriendly unchildlike unfruitful unsea-worthy unwholesome unselfish unprecedented unsurprising unhappiness

All these differences can easily be accounted for if we assume that in- is a Class I prefix, represented as in + , and that un- is a Class II prefix, represented as un#. Thus, in relation to point (a), we see that the "weak" boundary " + " permits the stress shift, but the stronger " # " does not. Similarly, in (b) the assimilation rule applies across " + ", but not " # " . In relation to (c), the examples in (38) show that the structure Pre + S T E M is acceptable, while Pre # S T E M is not. With respect to point (d), it should be noted that all of the suffixes in (39i) are Class I suffixes, while those in (39ii) are Class II suffixes. Given the assumption that the morphological component is ordered such that Level I affixation processes precede Level II affixation processes, what this last point shows is that only those words derived with Level I 1 8 suffixes are available for further derivation with the Level I prefix in-. Since words derived by Level II suffixation are formed after Level I processes have applied, words with Level II suffixes cannot be further derived with in-. W o r d s derived with Level II suffixes can still, however, be subsequently derived with the Level II prefix un# (cf. Allen 1978).

Lexical formatives

and word formation

85

rules

To see more clearly how the ordering of the affixation rules works, consider the following derivations of three words (cf. Mohanan 1982:4): (40)

Diet. + suffix + pref. # suffix # pref.

[organ] [ [organ] ic] [in [ [organ] ic] ]

[organ] [ [organ] ic]

[un [ [organ] ic] ]

[comfort]

[ [comfort] able] [un [ [comfort] able] ]

From these derivations it is clear why a word such as *incomfortable is not possible: the word comfortable is no longer available to the in- prefixation rule. The explanatory adequacy of the ordering analysis offered by Allen derives from "the consistency with which combinatorial properties of affixes can be determined by their level-membership, and from the interesting correlations noted by Allen between level-membership and various phonological and semantic properties of affixes" (Pesetsky 1979:17). 2.3.1.

Level Ordered

Morphology

In order to facilitate the visualization of a model of morphology organized in blocks, a somewhat simplified version of Siegel's (1974) representation of such a model is given in (41): (41)

+ affixes underived words stems

Class I-derived Words (underived words)

I Cyclic Stress Assignment Rules

Stressed Underived Words Stressed Class I-derived Words I # affixes

Class II Affixation

i Output

86

Generative

morphology

This diagram shows that the input to Level I affixation consists of (a) + affixes, (b) underived words, and (c) stems. The output of Level I affixation consists of derived words, but no stems. The cyclic stress rules apply to this output. The input to Level II affixation consists of (a) Level # affixes, (b) stressed underived words, and (c) stressed Level I derived words. It should be noted that since WFR's are optional, it is possible for a word to pass through the entire component represented by the diagram in (41) without undergoing any affixation rules, but undergoing, instread, only those rules that are obligatory (i.e. the stress rules). The structures in (42i) may all be found after the application of the stress rules, while the one in (42ii) is excluded: (42)

(i) [ W ] [ [W] + S U F ] [PRE + [W]] [[S]+SUF] [PRE + [S]]

(ii) *[S]

Given that WFR's may apply more than once, in addition to the structures just given, the structures in (43) are also possible: (43)

[ [ [ W ] + SUF] + SUF] [PRE + [PRE + [ W ] ] ] [[[S] + SUF] + SUF] [PRE + [PRE + [S]]]

[PRE + [ [ W ] + S U F ] ] [[PRE + [W]] + SUF] [PRE + [ [S] + S U F ] ] [[PRE + [S]] + SUF]

In theory, the structures just given can also return for further application of WFR's, and thus create words with η number of affixes. In practice, however, it is probable that all languages impose restrictions both on the number of affixes and on the relative number of prefixes and suffixes, an issue that will not be discussed further here. The structures in (42) and (43) are "words" in every sense and, in fact, bear stress. They may either pass through the rest of the component without undergoing further affixation, or they may undergo affixation with # affixes, giving structures such as those in (44i); structures such as those in (44ii), however, are not possible: (44)

(i) [ [ W ] # S U F ] [PRE#[W]] [[[W] + SUF]#SUF] [PRE# [[W] + SUF]] [[PRE + [ W ] ] # S U F ] [ P R E # [PRE + [ W ] ] ] [[[S]+SUF]#SUF] [[PRE + [S]]#SUF] [ P R E # [PRE + [ S ] ] ]

(ii) * [ [ S ] # S U F ] * [ P R E # [S]] * [ [ [ W ] # S U F ] + SUF] * [PRE + [ [W] # S U F ] ] *[[PRE# [W]] + SUF] * [ P R E + [ P R E # [W] ] ] *[[[S]#SUF]+SUF] * [ [ P R E # [S]] + S U F ] *i *[PRE + [ P R E # [S]]]

Lexical formatives

and word formation

rules

87

To summarize, the diagram in (41), and the structures it generates in (42i), (43) and (44i) show, first of all, why Level II affixes cannot be attached to stems (i.e., there are no stems available at the point in which Level II affixation applies). They show, secondly, why Level II affixes may be attached only to stressed words (i.e. there are no unstressed words at the point Level II affixation applies). In addition, they show why Level II affixes are stress-neutral (i.e. the cyclic stress rules operate before Level II affixes are attached). Finally, they show why Level II affixes are always external to Level I affixes (i.e. Level II affixation can attach # affixes to words with derived + affixes, but not vice versa). In other words, the generalization here is that affixes of a level η m a y never be attached to items that have been formed by attaching affixes of a Level n + 1. 2.3.2. Level Ordered Morphology

in Italian and Dutch

The application of the model of level ordering morphology to other languages has, in general, had the following results: while it is possible to conceive of the morphological component in these languages, too, as a set of blocks of rules ordered with respect to each other, the means of differentiating + affixes and # affixes have not been found to be the same. We will now examine some examples from Italian and Dutch, two languages for which explicit proposals have been made. 1 9 Let us consider Italian, where it should be noted, first of all, that the same affixes can be attached to both words and stems, as seen in (45). (45)

in + [decente] A in + [sipido] s A [ a p o s t o l o ] N + ico [oftalmo] sN + ico

"indecent" "insipid" "apostolic" "ophthalmic"

Secondly, all prefixes are stress-neutral, while suffixes are stress-shifting, 20 as illustrated in (46): (46)

Prefixes in + colpevole in -I- morale co(n) + aut0re iper + säturo Suffixes

incolpevole immoräle coautore ipersäturo

"not guilty" "immoral" "co-author" "hyper-saturated''

colpevole + ezza m o r a l e + itä autore + izzare säturo + azione

colpevolezza moralitä autorizzäre saturazione

"guiltiness" "morality" "to authorize" "saturation"

Given that the basic differences between the two types of affixes in English is not found in Italian, it might seem promising to adopt the hypothesis advanced by Strauss (1979), according to which there is, in fact, no

88

Generative

morphology

distinction between + affixes and # affixes, but rather all affixes may be considered to have a " + " boundary. Strauss's hypothesis, however, has a rather undesirable consequence as will be shown. According to Strauss, the difference between boundary types is not necessary because it is the ordering itself that blocks the shifting of stress in words formed with # affixes: (47)

(1) Level I affixation (2) Stress rules (3) Level II affixation

If the stress rules apply after Level I affixation and before Level II affixation, the Level II affixes may also have a " + " boundary, given that the stress rules have already applied, and cannot apply again. Applying this reasoning to Italian, we arrive at the following order: (48)

(1) Suffixation (2) Stress Rules (3) Prefixation

Since the prefixes are attached after the stress rules have applied, it follows that the addition of a prefix cannot further modify the stress of the base, and prefixes may thus also be specified with a " + " boundary. This boundary, furthermore, correctly permits the application of the nasal assimilation rule, as in in + maturo^immaturo "immature", and in + legale^illegale "illegal". Even though this hypothesis is quite elegant and accounts for certain facts of Italian, it cannot be maintained since it makes incorrect predictions in a number of other cases. Consider, for example, words of the form Pre-Word-Suf. According to the order in (48), such words could only be formed in one way, and would obligatorily have the following structure: (49)

[Pre + [ [ W O R D ] + S u f ] ]

This structure is found in Italian in words such as immangiabile ( = [in [ [mangia] abile] ]) "inedible". It is not, however, the only structure that exists. In a word such as inutilitä "uselessness", it is necessary to posit the structure [ [in [utile] ] ita], a structure that is not permitted, however, by the order in (48), since it requires that prefixation apply before suffixation. As far as derivation is concerned in Italian, Scalise (1983) shows that it is necessary to postulate that all affixes must contain a + boundary. It will be shown later that an extended theory of extrinsic ordering allow us to make use of the difference between " + " and " # " also in Italian, but only to distinguish between derivation and compounding. Let us now consider some aspects of Dutch morphology. Booij (1977) has proposed that the distinction between + affixes and # affixes is a

Lexical formatives

and word formation

89

rules

relevant one for Dutch as far as the application of the stress rule is concerned. As the examples in (50) show, one group of affixes is stress shifting, while the other is stress neutral (the base is given in phonemic transcription): (50)

stress shifting

suffixes

/sinöd/

+aal

•synodaal

"synodal"

/streck/

+ age

• strijkage

"bow"

/miljard/

+air

• miljardair

"billionaire"

stress neutral

suffixes

/rod/

+achtig

• roodachtig

"reddish"

/hud/

+ loos

•hoedloos

"hat less"

/halv/

+ ling

•halfling

"hobbit"

From these lists (and the more complete ones given in Booij 1977:72-73), it is clear that the stress shifting suffixes always bear the primary word stress. That is, they modify the position of primary stress in a word. The stress neutral suffixes never bear primary stress, though they may bear secondary stress, indicated by "2" (cf. roodachting "reddish"). The difference between Level I and Level II affixes in Dutch can be motivated on other grounds as well, in particular in relation to syllabification rules and readjustment rules. For example, if a word is derived by attaching a Level I suffix to a base word with a final schwa (ortographically e), this schwa is always deleted, as seen in (51i). If, however, the suffix attached is a member of Level II, the schwa is not deleted, as seen in (51ii): (51)

(i) elite + air Syracuse+ aan ambassade + eur (ii) zijde actig jute actig groente actig

+elitair +Syracusaan >ambassadeur >zijdeactig +juteachtig • groenteachtig

"elitist" "Syracusian" "ambassador" "silklike" "jute like" "vegetable like"

As far as prefixes are concerned, however, Booij arrives at the conclusion that they must all be introduced by a " # " boundary. We will not go into his arguments here, but simply list the main points. Specifically, Booij

90

Generative

morphology

maintains that all prefixes have a # boundary because of (a) the stress patterns in complex words, (b) the patterns of syllabification of complex words, and (c) the restriction on schwa-deletion in complex words. What these different examples show is that the criteria for distinguishing categories among the affixes of a given language must be language specific, that is, specified separately for each case. On the whole, though, the idea of a morphology divided into various levels appears to be a valid one.

3. Compounding For the reasons mentioned in Chapter I, we will suppose here that Compounding Rules (CR's) are WFR's and, more precisely, that they are rules of linear concatenation 2 1 . The form of such rules is the following:

(52)

[

]χ, [

]γ-[[

]χ Q [

]γ]Ζ

Q in (52) stands for the internal boundary of the compound, an issue to which we will return below (cf. section 3.3.). The structure to the right of the arrow will receive external boundaries assigned by the BIC in the following way: (53)

[#[

]χ Q [

]γ#]

ζ

According to Allen (1980), compounds are characterized by two conditions, the Variable R Condition and the "IS A" Condition, each of which will be examined below.

3.1. The Variable R Condition The Variable R Condition refers to the fact that every "new" compound does not have one and only one meaning but rather a range of possible meanings, as distinct from a range of not possible meanings. The Variable R Condition is valid only for primary compounds (i.e. those nominal compounds in which the second element is not derived from a verb). Nonprimary compounds, called synthetic compounds or verbal nexus compounds, such as truckdriver, snow removal, in fact, contain a verbal form which usually determines unequivocally the meaning of the compound, thus ruling out other readings. The two primary compounds fire-man and water-mill, for example, have a range of possible meanings, as seen in (54i), although each has a conventional meaning. They also have a range of not possible meanings, such as those listed in (54ii) (cf. Allen 1978:92; 1980:10):

Lexical formatives (54)

and word formation

(i) fire-man

rules

91

water-mill

man who worships fire

mill for producing water

man who walks on fire

mill powered by water

man who sets fire

mill located near the water

man who puts out fires

mill for analyzing water

etc.

etc.

(ii) fire-man

water-mill

man who contains fire

mill which lives near the water

man who sells fire

mill which grinds water

man who dreams about fire

mill which drinks water mill made out of water

According to Allen, the Variable R Condition establishes a range of possible meanings in terms of the semantic feature sets of the constituent elements of the compound. T h e Variable R Condition "predicts that the complete semantic content of the first constituent element may fill any one of the available feature slots in the feature hierarchy

of the second

constituent element, as long as the feature slot to be filled corresponds to one of the features of the filler" (Allen 1978:93). This condition has been formalized as follows: (55)

Variable R

Condition

In the primary compound

_an_

where

«1

~ß\

~

A

"

Jm_

is the semantic content of A in terms of hierarchical semantic features

is the semantic content of Β in terms hierarchical semantic features

and Ε α η _ χ such that a n _ x s / ? m _ y then the meaning of X ranges from to

β ι ( α ϊ . . . an) ßm

(αι · ·

·an)

T h e condition presupposes that every lexical item has associated with it a list of feature slots specified approximately as in (56) (cf. Allen 1978:96):

92

Generative

morphology

(56)

X-man compounds:

dominant feature less dominant feature non existant feature

[ + occupation] [ + made out o f ] [ + container]

The more dominant feature slots, those higher in the hierarchy, are more likely to be filled. Thus, within the range of possible meanings, some are more likely than others. Impossible meanings can result from incompatibilities between two sets of hierarchical features. For an example, fireman cannot mean "man who trains fire", although lion-man can mean "man who trains lions". Incompatibilities can also result from filling nonexistent feature slots. Thus, fire-man cannot means "man who contains fire" because the semantic specification for man does not include the feature [ +container], as opposed to fire-box, where box may be said to include the feature [ +container]. 3.2. The '7S A" Condition The second condition, the "IS A" Condition, can be expressed in the following way for English 22 (cf. Allen 1978:105): (57)

In a compound

[[

]x [

]Y ] z

Ζ "IS A" Y This formulation is intentionally ambiguous with regard to syntax and semantics. F r o m the point of view of syntax, (57) means that the entire compound (Z) is of the same category as the second constituent (Y), as can be seen in the following examples: X Ν A Ρ Ρ A A Ν

y Ν Ν Ν V V A A

Ζ Ν Ν Ν V V Α Α

cable television blue jay underarm overshoot dryclean deaf-mute color blind

From the point of view of semantics, (57) indicates that "a semantic subset relationship holds between the compound Ζ and the compound constituent Y" (Allen 1980:11), as the following examples illustrate: (59)

black board mailman firetruck

IS A IS A IS A

"board" "man" "truck"

Lexical formatives

and word formation

rules

93

It should be noted that the "IS A" Condition captures an important generalization about compounds, and thus allows us to avoid redundancy. Specifically, without this condition it would be necessary to write a different rule for each type of coumpound, as the examples in (60) show: (60)

[[ [[ [[ [[

]N[ ]A[ ]p[ 1A [

]N]N 1N]N Ivlv ]A]A

sunshine highschool overdo wide-awake

Given the "IS A" Condition, instead, the information represented in the individual rules in (60) can be stated only once for all cases. Those compounds that satisfy both the "IS A" Condition and the Variable R Condition are semantically transparent or predictable compounds and must be treated differently from those compounds that d o not satisfy the "IS A" Condition (e.g. cotton-tail, paleface, etc.) or the Variable R Condition (e.g. cranberry, whinchat, etc.) While the first type can be constructed by regular word formation rules, we can suppose that the second and third types are stored directly in the Dictionary. They can thus be referred to as "lexicalized" compounds. It should be noted that compounds differ both from simple words and from phrases. They differ from simple words, first of all, in that they have internal structure, and secondly, because they behave differently as far as derivation is concerned (i.e., very few compounds can be derived, cf. Chapter VI). Compounds differ from phrases since the constituents of a compound cannot function as independent elements with respect to syntactic processes. That is, they cannot be "factored out", but behave instead as "islands" from which no material can be extracted, as can be seen by comparing the cases in 61 (i), where factoring out is possible and those in 61(ii), where it is not: (cf. Allen 1978:113): (61)

(i) John bought a red basket and I a blue one They picked black and blue pencils John bought a very green house (ii) *I don't want a breadbasket, I want an egg one *They picked black and blueberries *John bought a very greenhouse

3.3. Boundaries in Compounds and the Extended Level Ordering

Hypothesis

It was mentioned above that compounds contain internal structure and, in fact, this structure is indicated by the presence of internal boundaries. In the structures given above, the internal boundaries were represented with "Q". Let us now consider the hypothesis that, analogously to the case of derivation, the internal boundaries of compounds are accounted for directly by WFR's. That is, Compounding Rules generate structures of the following type:

94

Generative morphology

(62)

[[

]##[

]]

The boundary " # # " is a strong boundary that blocks semantic and phonological processes of amalgamation between W O R D 1 and WORD 2. Such a proposal has been formalized by Allen (1980:15) as follows: (63)

Strong Boundary Condition (Semantic Version) in the morphological structure X Bs Y no semantic amalgamation process can involve X and Y

where B s is a strong boundary # # where "semantic amalgamation process" refers to any process of meaning formation which is not completely generalizable

It follows from (63) that in compounds of the form # # , there is no semantic distortion, loss of information or semantic drift. Furthermore, on the grounds that the structure # # does not allow the application of certains phonological rules (which do, however, apply in the structure + ), Allen (1980:26) generalizes the Strong Boundary Condition as follows, such that it also accounts for the phonological facts: (64)

Strong Boundary

Condition

in the morphological structure X Bs Y no rule may involve X and Y

where B s , the strong boundary, is # # and where rule refers to both "semantic amalgamation process" and "phonological rule"

The structure # # is thus a structure that blocks phonological sandhi processes and semantic amalgamation processes between WORD 1 and W O R D 2. It is a well known fact, however, that the compositional nature of a compound may be obscured by factors such as frequency of use and "permanence in the lexicon". In order to account for this fact Hyman (1978) proposed the following boundary weakening rule: (65)

#

-»

+

If we apply Hyman's proposal to the structures seen above, we can postulate diachronic changes such as those in (66): (66)

# # #

+

#

Lexical formatives

and word formation

rules

95

According to (66), the strength of the internal boundaries of a compound can be gradually weakened over time. In the same direction, Allen (1975) proposed an interesting distinction, on the basis of Welsh data, between "strict compounds" and "loose compounds", which have the internal structures represented in (67i) and (67ii), respectively: (67)

(i) (ii)

+ # #

The structure in (67i) allows the operation of regular stress rules and certain vowel mutations, that is, rules of phonological amalgamation. The structure in (67ii), on the other hand, does not allow these rules to apply and thus accounts for the irregular stress patterns and absence of the vowel mutations in question in the second type of compounds. If we now incorporate compounding into the Level Ordering Hypothesis, following Allen (1978), we arrive at the Extended Ordering Hypothesis (EOH), which can be represented as follows: (68)

(1) Level I (2) Level II (3) Level III

+ affixation # affixation # # compounding

As was seen above, different languages can parametrize this Level order in different ways either by distributing affixes differently at the + and # Levels, or by omitting one of the levels. A version of the E O H that includes inflection can be represented as follows: (69)

(1) (2) (3) (4)

Level Level Level Level

I II III IV

+ derivation # derivation compounding inflection

More recent studies have confirmed the validity of the Ordering Hypothesis, and the E O H has, in fact, been adopted in relation to a variety of languages 23 . In the next chapter, we will discuss in detail the relationships between derivation and inflection, compounding and derivation and compounding and inflection.

4. Well Formedness Conditions In this chapter, three different Well Formedness Conditions were given for prefixation, suffixation and compounding. They are repeated in (70) for convenience:

96

Generative

morphology

(70)

Prefixation Suffixation Compounding

[Pre+ [ ]αΧ ]aX [ [ ]x + SufAXY]Y [[ ]X [ ] γ ] ζ = γ

(Scalise 1980) (Pesetsky 1979) (Allen 1978)

At least for some languages such as English, it is possible to collapse these three Well Formedness Conditions into a single condition. Before doing so, however, we must first discuss the notions of "head" of a complex word and "percolation". It has been recognized for a long time that compounds are morphological units that contain a head (i.e. endocentric compounds). Recently, Williams has proposed that not only compounds but all morphological complex words have a head. The following is Williams's definition of head (1981a:247): (71)

If both X and the head of X are eligible members of category C, then X C = h e a d of X C

According to this definition, the head in prefixed words such as re-write, in suffixed words such as lonely-ness and in compound words such as highschool, is always the right hand element, as can be seen in (72): (72)

(i)

/ \ re pre

writey (head)

(ii)

/ V lonelyA

(iii) nessN (head)

high a

school (head)

The head assigns to the entire word its category by means of a mechanism referred to as percolation. The following is Selkirk's (1982:21) definition of percolation 2 4 : (73)

If a constituent α is the head of a constituent β, α and β are associated with an identical set of features (syntactic and diacritic)

We can represent the percolation of the category label from the head to the entire complex word as follows: (74)

(i)

rep r e

writey'

lonelyA

(74) is an example of what Williams (1981a:248) calls the Righthand Rule (RHR): "In morphology, we define the head of a morphologically complex word to be the righthand member of that word". This rule is valid only for languages such as English, and not for languages with left-headed compounds (cf. Scalise 1983). Limiting our discussion to English for now, it is possible to express an interesting generalization about complex words,

Lexical formatives and word formation rules

97

which can be formalized as follows:

(75)

Well Formedness Condition for English complex words: in an English complex word X

Y

Together with the percolation convention, (75) can predict the category of every complex word in English: in prefixation (where X = Prefix and Y = Word), in suffixation (where W = Word and Y = Suffix) and in compounding (where X = Word 1 and Y = Word 2). F o r the moment, we will accept this Well Formedness Condition and incorporate it into our general framework.

5. Summary In the framework developed thus far, the lexical component of a grammar consists of the following parts: the Dictionary (i.e. a level of representation that contains words and stems), and blocks of Word Formation Rules (i.e. Compounding Rules and Derivation Rules). After W F R ' s , Inflection Rules, Readjustment Rules and a Convention for the insertion of the Boundaries apply to give the output of the morphological component. Such a model can be represented as in Figure l 2 5 : [WORD]x, [STEM]sX

Dictionary Morphological Component

DR's

CR's

[ [ W O R D ] x + Suf Δ Χ Υ ] Υ [Pre+ [WORD]aX]aX [ [STEM]sx + SufAXY]Y [Pre+ [STEM] asX]aX [ [ W O R D ] \ # S u f ΔΧΥΙΥ [Pre# [WORD] a X ] a X # #

[WORDJYJY

[ [WORD]X # # [ [STEM]sX # # [ [STEM]sX # #

[[WORD]

[STEM]sY]Y [WORD]Y]Y [STEM]sY]Y

X

IR's

[ [ [ W O R D ] x # # [ W O R D ] Y ] z + infl] z [ [ [ W O R D ] x + S u f A X Y ] y + inn]Y

RR's

Truncation Rules and Allomorphy Rules

BIC

[ # [ [ W O R D ] x # # [WORD]Y]z + infl#]z [ # [ [ W O R D ] x +SufAXY]Y +infl#]

Figure 1: The organization of the lexical component (III)

98

Generative

morphology

In relation to this model, there are a number of things that should now be clear in light of the material presented thus far. First of all, the Dictionary contains words and stems; both are bracketed and labeled, but contain no internal structure. Secondly, words are labeled with a syntactic category (i.e. they are free forms and as such can appear on the surface freely); stems are provided with a label that refers to a syntactic category (i.e. they are bound forms and as such cannot appear on the surface freely). In order for stems to appear on the surface, they must either be compounded or derived; in other words, the sequence "stem + inflection" is not grammatical. It was pointed out, further, that Derivation Rules are ordered in blocks, but the way in which these blocks are structured is language specific. Compounding Rules combine the units of the Dictionary into linear concatenations. In the Fig. 1 above, only productive CR's are represented, that is, the ones that generate # # structures. As far as boundaries are concerned, it was assumed, moreover, that the internal boundaries are taken care of by the WFR's themselves and that the external boundaries are taken care of by a special convention (BIC). Finally, there are well formedness conditions, incorporated in the Fig. 1 that predict (at least in English) the external syntactic category of both derived and compound words.

Notes 1. We d o not consider certain minor word formation processes such as the following to be serious counterexamples t o the W B H : acronyms (e.g. radar f r o m r a d i o detecting and ranging), blendings (motel from m o t o r hotel), clippings (lab from laboratory) and passages from proper to c o m m o n n o u n s (e.g. xerox), etc. 2. Cfr. also de Villiers (1979). 3. The scholars listed below d o not form a homogeneous group. F o r example, Selkirk argues in favor of two " b o u n d " elements, root and stem, while Siegel argues in favor of only stems. 4. In the list in (1), there are plural forms. In all the examples the plural m o r p h e m e is phonetically present except in the French example, where the s is only graphic [3ardi:]. Aside from "irregular" forms (e.g. cheval/chevaux "horse-horses"), the plural form in French is not marked phonetically in the n o u n s themselves but in the determiner (cf. [ ^ a r d e / k j a r d g ] "the garden - the gardens") and by "liaison" when the n o u n begins with a vowel (cf. [lami/lf.zami] "the friend - the friends"). 5. Cf. Bloomfield's definition (1933:160): "A linguistic f o r m which is never spoken alone is a bound form; all others (as, for instance, John ran, or run, or running) are free forms". 6. They are m o r p h e m e s because they are the result of a segmentation operation: Lat. lupi can be divided into a lexical m o r p h e m e lup- and a grammatical m o r p h e m e -i. 7. An interesting proposal has been put forth by Roeper and Siegel (1978). They propose that the lexical c o m p o n e n t of the g r a m m a r contains a Lexical C o r e and a set of W F R ' s . The Lexical C o r e is divided into two subparts: the Atomic C o r e and the Complex Core. The former is a list of words without internal structure (e.g. serene, boy); the latter is a list of words that have been formed by W F R ' s and, therefore, have morphological structure (e.g. serenity, boyhood). According to Roeper and Siegel, there are also words that seem t o fall

Lexical

formatives

and word formation

rules

99

between the two categories (e.g. possible, happy) but which must, however, be listed in the Atomic Core because they do not have compositional meaning. 8. Cf., recently, Bauer (1979). 9. This observation was made by Siegel (1974). Actually, Siegel observes that Class I Prefixes (see sec. 2.3.) cannot be factored out (cf. *ex and secretions) but Class II prefixes can be (cf. mono- and trisyllabic). 10. That the modification is induced by the stem can be demonstrated. In Italian, for example, with the [ + Latinate] stem coltura "culture", orto + coltura becomes orticultura "horticulture", and frutto + cultura becomes frutticultura "fruit growing". If we combine the two words, however, we get ortofrutticoltura, where only the vowel adjacent to coltura is subject to the said readjustment. 11. The following discussion is based on Siegel (1974) even though in many respects it departs from Siegel's proposal. 12. We accept here Aronoff's framework instead of Siegel's, a choice that has implications for how items are bracketed. In Siegel's view, where affixes are formatives of the lexicon, a word such as acidity is bracketed as [[acid] [ + ity]]. In Aronoff's view, where affixes "are" WFR's, and thus represented in a different component, the same word will be bracketed as [[acid] +ity]. Throughout the rest of this book, Aronoff's framework will be followed. 13. There are languages that do not exhibit any exceptions to this generalization and others that do. For example, in English, the prefix en- is said to derive Verbs from Adjectives (cf. [rich] A ->-[en + [rich] A ] v )· 14. In an overgenerating morphology (cf. Allen 1978), an item with internal structure is not necessarily a word. For instance, left-handed can be thought of as being generated from left and handed, where the second item has internal structure (hand + ed) but is not an existing word. 15. The following account is based on Siegel (1974) and Allen (1978). 16. Un is a # prefix according to Allen (1978). 17. On this problem, cf. Chapter VII. 18. Allen (1978) prefers the term Level to the term Class, a choice we will adopt here. 19. Cf. Scalise (1983) for Italian and Booij (1977) for Dutch. 20. There are no exceptions to this statement. Cases such as stbria^stbrico "historyhistorical" are only apparent exceptions. That is, it is not true that the suffix -ico does not affect the position of stress; in fact, it requires the placement of the stress on the immediately preceding syllable (cf. algebra-*algebrico "algebra-algebraic"), though in the case of storia this'cannot be seen. 21. This section on compounds is based on Allen (1978, 1980) and draws most of the examples from the same source. 22. This condition is strictly valid for languages where the head of the compound is WORD 2. For the notion "head", see section 4, below. 23. For Japanese, Kageyama (1982) has proposed a well argumented level hypothesis where the order is: (1) " + ", (2) " # " , (3) " # # " . He shows, however, that in Japanese, affixation and compounding are attested at all three levels. Other applications of the EOH, either in its original form or with minor modifications, have been suggested by Booij (1977) for Dutch, Pesetsky (1979) for Russian, Carrier (1979) for Tagalog, Scalise (1980) for Italian. An ordered model of morphology has also been adopted in more recent works (cf. Mohanan 1982, Kiparsky 1982) though within a different framework, that of Lexical Phonology. 24. Selkirk modifies this definition slightly later in the same work, but this modification is not relevant here. 25. For IR's and BIC only two examples are given.

Chapter VI

Interplay between morphological rules In the preceding chapter, it is suggested that the lexical component of a grammar is organized in blocks of rules, that is, Compounding Rules (CR's), Derivation Rules (DR's), and Inflection Rules (IR's). In this chapter, the problem of the interaction of these types of rules will be addressed. It will be claimed, in particular, that IR's are different from DR's and that they must be ordered after them. Next, we will examine the interplay between DR's and CR's and CR's and IR's and we will argue in favor of the Extend Ordering Hypothesis, the hypothesis according to which morphological rules have the following order: DR's, CR's and IR's. Finally, two borderline cases will be examined: the nature of the Past Participle and evaluative suffixes.

1. Strong Lexicalist Hypothesis Before discussing the relation between derivation and inflection, it is necessary to address a theoretical problem that is of direct relevance here. It should be noted that in Fig. 1 of the preceding chapter, the inflection rules are found in the lexical component, though such a position is not uncontroversial, as will be seen below. With regard to the position of inflection in a grammar, there are at least two possible hypotheses, referred to as the Strong Lexicalist Hypothesis (SLH) and the Weak Lexicalist Hypothesis (WLH). According to the first hypothesis, inflection operates entirely within the lexical component (as indicated in Fig. 1 in Chapter 5), and according to the second hypothesis, inflection operates in the syntactic component, or later. The SLH has its origin in a statement by Chomsky (1970), according to which: (1)

Syntactic rules cannot make reference to any aspect of the internal structure of a word.

In its more general form, this position, also referred to as the Generalized Lexicalist Hypothesis, can be formulated as follows: (2)

Syntactic transformations are never allowed morphological operations (Lapointe 1978:3).

to

perform

102

Generative

morphology

Another variant is Selkirk's Word Structure Autonomy Condition which can be stated as in (3): (3)

N o deletion or movement transformations may involve categories of both W-structure and S-structure (Selkirk 1982:70).

This issue has received much attention in recent years, to the extent that it is possible to identify at least two groups of linguists according to whether they accept the Strong or the Weak Lexicalist Hypothesis, as seen in (4): SLH Halle (1973) Jackendoff (1975) Booij (1977) Lapointe (1978) Brame (1978) Carrier (1979) McCarthy (1979) Lieber (1980) Scalise (1980) Muysken (1981) Williams (1982) Kiparsky (1982) Strauss (1982) Selkirk (1982)

WLH Chomsky and Halle (1968) Chomsky (1970) Siegel (1974) Aronoff (1976) Allen (1978) Thomas-Flinders (1981) Piatt (1981) Tuller (1981) Anderson (1982)

It is necessary, further, to distinguish two subgroups within the group of SLH supporters, those that maintain that derivation and inflection are essentially the same type of process (e.g. Halle, Jackendoff, Williams), and those that maintain they are different (e.g. Booij, Scalise, Kiparsky, Selkirk). The group of W L H supporters can also be divided into two subgroups, those that claim that inflection is a syntactic operation (e.g. Chomsky and Halle, Aronoff, Siegel), and those that claim that it is a phonological operation (Anderson, Flinders). In this chapter, we will adopt the SLH, for descriptive purposes. That is, it will be assumed that inflection operates entirely within the lexical component, and that this component has the goal of defining the notion "possible word" in the case of inflected words as well. A more detailed discussion and justification of this choice can be found in Chapter VIII, where the relation between morphology and syntax is examined. Let us now return to the issue of the relation between derivation and inflection.

2. Derivation and inflection Whether derivation and inflection must be separated or not is a problem

Interplay between morphological

rules

103

discussed by all those who have been concerned with morphology. Of course, those who defend the WLH, automatically assume that DR's and IR's are different in nature. Among those who defend the SLH, on the other hand, there are some who maintain that they are different and must therefore be separated, and others who do not. In this section, we will argue in favor of the division of these two blocks of rules on the basis of data drawn from Italian, a language with a rich inflectional system. Eight different arugments will be examined separately. (I) IR's never change the syntactic category of a word, while DR's change it. This point is easily demonstrated by the examples below: (5)

IR's:

DR's:

may

[ a t o m o ] N [ [atomo]N + i]N "atom - atoms" [ c o n t a ] v -> [ [conta]v + iamo]v"(he) counts (we) count" [atomojjsi [ [atomo]N + ico]A "atom - atomic" [contajy [ [conta]v + abile]A "to count - countable"

There are also cases where it is said that DR's do not change the category of a word, as in (6): (6) DR's

[avvocato]N - • [ [avvocato] >j + ura] ν "lawyer - lawyership" [cucchiaio]N [ [cucchiaio]N + ata]N "spoon - spoonful"

There are reasons (implicit in works such as Williams (1981a) and explicitly stated in Scalise (1983)), however, for believing that a DR always changes the syntactic category of its base, even when this change is not evident, as in the examples in (6). We will not go into this demonstration here; the reader is referred instead to the works just cited. (II) Inflection is always peripheral with respect to derivation. This point is actually a restatement of Greenberg's (1963) Universal n.28, and for right constructing languages such as Italian, it can be represented in the following way: (7)

Word - Derivation - Inflection *Word - Inflection - Derivation

According to (7), derivational affixes cannot be attached to inflected words, a position often referred to as the Uninflected Base Hypothesis (UBH). This claim has, however, been questioned. Moody (1978), for example, gives counterexamples from French, and wile most of his arguments are not crucial, the most serious case (i.e. adverb formation), in fact, appears to be common to a number of Romance languages, as seen in (8):

104

Generative

(8)

Fr. Sp. It.

morphology

maladroitement [maladrwatamä] conservadoramente [konservadoramente] certamente [tjertamente]

"awkwardly" "conservatively" "certainly"

In each of these cases, the base seems not to be the uninflected adjective in (9)i)), but rather the inflected (feminine) form in (9(ii)): (9)

(i) maladroit conservador certo

[maladrwa] [konservador] [tjerto]

(ii) maladroite conservadora certa

[maladrwata] [konservadora] [tjerta]

"awkward" "conservative" "certain"

In order to account for such cases in Spanish, Harris (1977) proposed that the Dictionary contains both the masculine and the feminine forms of adjectives. There are, reasons for not doing so, however, as can be seen in relation to the Italian example. First of all, listing the feminine forms of adjectives in the Dictionary causes us to miss a generalization since it requires that the inflectional endings of adjectives be listed in one part of the grammar and the inflectional endings of other categories in an another part of the grammar. Secondly, the adverbs in question contain no notion of "feminine", at least synchronically, although we know that the original Latin forms did incorporate such a notion. That is, certamente comes from a Latin phrase composed of an adjective (certa "certain") and a noun (mente "mind"), both of which are feminine and in the ablative case. The entire phrase has the literal meaning "with a certain (determined) mind". In modern Romance languages, however, this compositionality has been lost and mente is no longer an independent noun in such constructions, but rather an affix. In Italian, furthermore, the -mente formations require certain readjustment rules in the case of adjectives ending with e: (10)

(i) (ii) (iii) (iv)

maggiore maggiormente formale -> formalmente pedestre -> pedestremente veloce -> velocemente

"larger - more largely" "formal — formally" "dull - dully" "quick - quickly"

These examples fall into three groups: (a) cases where e is deleted, i.e. (lOi) and (lOii), because it is preceded by r or /, preceded in turn by a vowel; (b) cases where e is retained, i.e. (lOiii), because the preceding r (or /) is not preceded by a vowel; and (c) cases where e is retained, i.e. lOiv), because it is not preceded by r or /. A relatively simple way of handling all of these cases is to propose that all adverbials in -mente are subject to the following readjustment rules: (11)

(i) o - » a /

+mente

(ii) e - > ( / ) / V - r l

+ mente

Interplay between morphological

rules

105

According to this proposal, mente formations do not retain the properties of their ancestor formations but rather, synchronically, are derived words where the notion "feminine" is no longer present. The rules in (11) have the advantage of treating mente formations in a unified way, without introducing weakly motivated asymmetries into the Dictionary. A second type of counterexample that is sometimes cited (cf., among others, Booij 1977:47) is found when a derivational suffix is attached to a comparative, which is claimed to be an inflected form, as in It. maggioranza "majority" (i.e. maggiore "larger" + anza). It can be shown, however, that such items are not counterexamples to the UBH, but rather maggiore, the comparative form of grande "large", represents a case of suppletion. The two items can be seen to be two distinct units on the basis of the suffixes they can be combined with: (12)

grande "large" grandezza "largeness" *grandanza grandioso "grandiose" *grandenne ingrandire "enlarge" *grandato

maggiore "larger' *maggiorezza maggioranza majority" *maggioroso maggiorenne adult" *immaggiorire maggiorato ' buxom"

Since the two items in question exhibit different morphological behavior, they must be listed in the lexicon as separate (suppletive) forms; they cannot be derived one from the other. A form such as maggioranza is therefore not derived from an inflected form, but from a base, maggiore, and thus the comparative forms do not constitute a counter example to the UBH. (Ill) DR's and IR's are sensitive to different properties of their bases. The following table shows which types of information DR's and IR's are sensitive to: (13) (1) syntactic category (2) conjugation/declension (3) subcategorization features (4) selectional features (5) other features (5i) ± count (5ii) ± animate (5iii) + abstract (5iv) ± common

DR's + + + +

IR's + + + —

— + + +

+ — — —

Point 1 shows that both DR's and IR's are sensitive to the syntactic category of their base. It has already been seen that an affix selects a particular category to attach to, and it is a well known fact that the

106

Generative

morphology

inflectional endings of verbs are different, for example, from those of adjectives or nouns. Point 2 represents the fact that IR's are typically sensitive to conjugation class (for verbs) and declension class (for nouns and adjectives), and that DR's are sensitive to the same information, though in a more subtle way. Consider the following data, in which the form on the left of the arrow is a verb, and the one on the right a derived noun: (14)

(l) allargare affossare imboccare incollare sbarcare rincarare scarcerare assimilare atterrare allunare imbeccare infuriare alleggerire approfondire arricchire inasprire indurire impoverire

• allargamento • affossamento • imboccatura > incollatura •sbarco - > rincaro scarcerazione assimilazione atterraggio - > allunaggio imbeccata infuriata alleggerimento - > approfondimento arricchimento - inasprimento indurimento - > impoverimento

"widen - widening" "to ditch - ditching" "to feed - mouth, opening" "to glue - gluing" "disembark - disembarcation" "to raise prices - price raise" "to release (from prison) - release" "to assimilate - assimilation" "to land - landing" "to land (on the moon' - moon landing' "to feed (birds) - beakful" "to enrage - outburst" "to lighten - lightening" "to deepen - deepening" "to enrich - enrichment" "to embitter - embitterment" "to harden - hardening" "empoverish - empoverishment"

The fact that derivation rules are also sensitive to information such as conjugation class can be seen by comparing the examples in (14i) and (14ii). (14i) contains parasynthetic verbs 1 , all belonging to the first conjugation (the thematic vowel is a). From these verbs it is possible to derive nominals using almost the entire set of nominaliziag suffixes (-ura, -o, -zione, -aggio, -ata). If the verb belongs to the third conjugation, however, (as in (14ii)), where the thematic vowel is (i), it is possible to derive nominals only with the suffix -mento "-ment". Thus, the choice of the suffix is sensitive to the conjugation class of the verbs; in this particular case, first conjugation can take any nominalizing suffix, while the third class can take only -mento. In relation to Point 3, it should be noted that inflection depends heavily on subcategorization features as far as the choice of the auxiliary is concerned. In Italian, a verb is conjugated in compound tenses with the auxiliary avere "to have" if it is transitive (cf. (15i)). Reflexive verbs are conjugated, instead, with essere "to be" (cf. (15ii)), and some verbs (e.g. cominciare "to begin", durare "to last", finire "to finish") take avere when the object is expressed or implied (cf. (15iii)) and essere when used passively with no object (cf. (15iv)). (15)

(i) ho mangiato il pane ho visto Antonio

"I ate the bread" "I saw Antonio"

Interplay

between morphological

rules

(ii) mi sono lavato il viso mi sono alzato presto (iii) ho finito i compiti ho cominciato un nuovo libro (iv) e finito il primo tempo e cominciato alle due

107 "I washed my face" "I got up early" "I finished my homework" "I began a new book" "The first part (of a movie) is finished" "It began at two o'clock"

DR's, too, are sensitive to subcategorization features. F o r example, the suffix -abile/-ibile "-able" can be attached to transitive verbs, as shown in (16i), but not to intransitive verbs, as shown in (16ii): (i) contare prevedere mangiare portare (ii) comparire crescere morire restare

"to "to "to "to "to "to "to "to

count" predict" eat" carry" appear" grow" die" remain"

contabile prevedibile mangiabile portabile *comparibile *crescibile *moribile *restabile

"countable" "predictable" "edible" "portable"

Point 4 shows that DR's can be sensitive to selectional features, and that IR's, as far as we know, are not. The only example cited in the literature of a D R that is sensitive to selectional features is the case of the suffix -ee mentioned by Aronoff (1976). The example has been seen above in Chapter III, section 3.1.1. In point 5, it is shown that IR's and D R ' s are sensitive to different semantic features. In (5i), we see that IR's typically take into account the feature [ + c o u n t ] , Consider the following examples: (17)

singular (i) zucchero pattume polvere (ii) cappello balena uccello

"sugar" "garbage" "dust" "hat" "whale" "bird"

plural *zuccheri *pattumi *polveri cappelli balene uccelli

"hats" "whales" "birds"

The plural forms in (17i) are starred because their bases are mass nouns; zuccheri pattumi and polveri are not regular plural forms, although they may exist with idiosyncratic meanings. DR's, on the contrary, do not seem to distinguish between [ + c o u n t ] and [ - c o u n t ] nouns, as can be seer below: (18)

(i) zucchero pattume polvere

-»zuccheriera —»pattumiera ->polveriera

"sugar bowl" "garbage container" "powder house"

108

Generative

morphology

(ii) cappello -> cappelliera balena -> baleniera uccello -> uccelliera

"hat box" "whaler" "bird cage"

We see from the d a t a in (18) that the same suffix -iera can be attached to the [ - c o u n t ] nouns in (18i) and to the [ + c o u n t ] nouns in (18ii). In relation to point (5ii), it seems that IR's d o not distinguish between [ + a n i m a t e ] and [-animate]. Consider the items in (19), in which plurals are formed in the same way for both types of nouns: (19)

singular (i) vigna cagna (ii) gatto faggio

plural -> vigne -> cagne -> gatti faggi

"vineyard - vineyards" "bitch - bitches" "cat - cats" "beech - beeches"

DR's, on the contrary, are sensitive to this distinction. F o r example, there are suffixes, such as -eto, which can be attached only to [-animate] nouns: (20)

(i) vigna faggio (ii) cagna gatto

vigneto faggeto *cagneto *gatteto

"vineyard - grape grove" 'beech - beech grove"

Point (5iii) shows that DR's are sensitive to the feature [ ± abstract]. While there are suffixes that are not sensitive to this feature, such as -ista (cf. (21i)), there are others that are, such as -aio (cf. (21ii)): (21)

(i) macchina telefono simbolo allarme (ii) verdura orologio altezza speranza

—• macchinista telefonista -» simbolista allarmista verduraio -» orologiaio -> *altezzaio *speranzaio

"machine - machinist" "telephone — operator" "symbol - symbolist" "alarm — alarmist" "vegetable - green grocer" "watch - watchmaker" "height" "hope"

IR's, on the contrary, attach the same endings to [ + abstract] nouns as to [-abstract] nouns: (22)

(i) altezza altezze speranza -> speranze (ii) macchina macchine verdura -> verdure

"height - heights" "hope - hopes" "machine - machines" "vegetable - vegetables"

Finally, in relation to Point (5iv), it should be noted that, as for (5iii),

Interplay between morphological

109

rules

there are some suffixes that are not sensitive to the distinction [ ± common] (cf. (23i)), but there are others that are, such as -ese (cf. (23ii)): (23)

(i) Darwin-»darwinista Petrarca -> petrarchista umore -»umorista piano —>pianista (ii) Francia—»francese Siam -> siamese Milano -»milanese paese ->*paesese (but paesano)

"Darwin - Darwinist" "Petrarch - Petrarchist" "humor - humorist" "piano - pianist" "France - French" "Siam - Siamese" "Milan - Milanese" "village - villager"

maremma—>*maremmese (but maremmano)"Maremma - from the Maremma" diocesi —> *dioccsese (but diocesano) "diocese - diocesan"

IR's, on the other hand, as a rule, do not apply to [-common] nouns. Compare the items in (24i) and (24ii). In the marginal cases of first names, however, the inflection of [-common] nouns is the same as that of [ + c o m m o n ] nouns (cf. (24iii)): (i) Arno Dante Moravia (ii) libro stivale storia (iii) Piero Irene Maria

—>

>

—• —> —>

*Arni *Danti *Moravie libri stivali storie Pieri Ireni Marie

"Arno" "Dante" "Moravia" "book - books" "boot - boots" "story - stories" "Piero - pi." "Irene - pi." "Maria - Pi "

From what we have seen so far, we can conclude that DR's and IR's are sensitive to different properties of the bases, in particular, it seems that DR's are sensitive to a wider range of information than IR's. The distinction between DR's and IR's is further confirmed if we look at what these rules can "do". (IV) DR's and IR's "do" different things. If we take the same list of points seen in (13), and ask what kinds of information contained in the base can be changed by DR's and IR's, we end up with the results shown in the following table, where " + " indicates that change is possible and " —" that it is not possible. (25) (1) (2) (3) (4)

syntactic category conjugation/decl. class subcategorization features selectional features

Dr's + + + +

IR's — — — —

110

Generative

morphology

(5) other features 5i. + count 5ii. + animate 5iii. ± abstract 5iv. ± common

Dr's + + + +

IR's — — — —

As we have done above, we will look at each point separately. In relation to Point 1, we have already seen that DR's may change the syntactic category of their base while IR's cannot (cf. point I above). In relation to Point 2, it can be demonstrated that DR's can change the declensional class of nouns and the conjugation class of verbs (cf. (26i) and (26ii), respectively). The same is not true of IR's. ( 2 6 ) (i) bottega (pi. botteghe) Colombo (pi. colombi) (ii) dormire (III Con.) ridere (II Con)

-»bottegaio (pi. bottegai) -»colombaia (pi. colombaie) ->dormicchiare (I Con.) -»ridacchiare (I Con.)

"shop-shops" "columbary - columbaries" "to sleep - to doze" "to laugh - to chuckle"

In Point 3, we see that DR's can change subcategorization features, while IR's cannot, as the following example illustrates: (27)

ridere di Antonio ->*deridere di Antonio *ridere Antonio -»deridere Antonio

"laugh at A." "deride A."

In other words, the derivation of deridere from ridere changes the subcategorization of the verb from intransitive ( PP) to transitive ( NP).

In Point 4, we see that DR's can change selectional features, as illustrated in the sentence in (28). The same is not true of IR's. (28)

*Giorgio ruba Antonio Giorgio deruba Antonio Giorgio ruba i risparmi di Antonio *Giorgio deruba i risparmi di Antonio *Giorgio ruba Antonio dei risparmi Giorgio deruba Antonio dei risparmi

"*Giorgio steals Antonio" "Giorgio robs Antonio" "Giorgio steals the savings of A." "*Giorgio robs the savings of A." "*Giorgio steals A. of the savings" "Giorgio robs A. of the savings"

The verb derubare "to rob" is derived from rubare "to steal"; this derivation changes the selectional features in the following way: (29)

. rubare

NP, [-animate]

derubare

_NP, [ + animate]

PP [ +animate]

PP [-animate]

Interplay

between

morphological

111

rules

Actually, Derivation Rules in this respect seem to be a very powerful mechanism in that they can transfer selectional restrictions from one position to another. Consider the representation of a verb such as interpretare "interpret": NP

(30)

[

NP

[ + human]

[-animate]

]VP

(30) shows that the verb interpretare allows [ + h u m a n ] subjects and [-animate] objects. Thus, the sentences in (31) are grammatical while the ones in (32) are not: (31)

(32)

Giorgio interpreta le rune il ragazzo interpreta il volo degli uccelli

"Giorgio interprets the runes" "the boy interprets the flight of the birds"

il filologo interpreta il manoscritto *la volpe interpreta le rune *Giorgio interpreta Anna

"the philologist interprets manuscript" "the fox interprets the runes" "Giorgio interprets Anna"

the

In order to account for such facts, Vergnaud (1973) proposed to use a rule of the following type: (33)

[ + unit] [ + animate] 1

[ + unit]

interpretare

[ + unit] [-animate]

2

interpretabile

3

[-unit]

[-animate] 3

1

Specifically, rule (33) accounts for the fact that the selectional restrictions on the object of the verb are the same as those on the subject of the derived adjective. It shows, furthermore, that the verb has an object, indicated by [ + unit] to the right while the derived adjective does not, as indicated by [-unit]. There are m a n y more changes in the subcategorization frames that can be carried out by Derivational Rules, but for the moment we will limit ourselves to the examples above, and continue the examination of the list in (25). In relation to Point (5i), we see that DR's can change the [ + c o u n t ] feature of the base in both directions. Consider the items in (34): (34)

(i) [-count] pattume paglia

[ +count] • pattumiera • pagliaio

"garbage - garbage container" "hay - hay stack"

112

Generative

morphology

(ii) [ + count] [-count] polio pol lame foglia -»fogliame

"chicken poultry" "leaf - leafage"

In (5ii), we see that DR's can change the feature [ +animate]. The following examples show that the change works in both directions: (35)

(i) [ — animate] [ + a n i m a t e ] giornale —>giornalaio "newspaper - newspaper seller" dogana ->doganiere "customs - customs clerk" (ii) [ + animate] [ — animate] cane ->canile "dog — kennel" Colombo —•colombaia "dove - columbary"

(5iii) shows that DR's can change the feature [ +abstract], and this change, too, works in both directions: (36)

(i) [ - a b s t r a c t ] [ + abstract] artigiano -»artigianato magistrato ->magistratura (ii) [ + abstract] [ — abstract] simbolo -»· simbolisa comando comandante

"artisan - handicraft" "magistrate - magistracy" "symbol - symbolist" "command - commander"

Finally, (5iv) shows that DR's can change the feature [ + c o m m o n ] , though this works only in one way: (37)

[ — common] [ + c o m m o n ] Piemonte ->piemontese "Piedmont - of Piedmont" Perugia -+perugino "Perugia — of Perugia"

(V) DR's change the conceptual meaning of the base, IR's change only the grammatical meaning of the base. As we have seen in Chapter III, a WFR has two parts, a formal one and a semantic one. The semantic part of a WFR is expressed in compositional terms approximately in the following way: (38)

(i) X + aio

"person who performs an activity connected with X" (e.g. orologiaio "watch maker") (ii) X + abile "capable of being Xed" (e.g. desiderabile "desirable")

The meaning of a derived form is compositional only if the rule is productive. The attachment of an inflectional morpheme does not change the meaning in the same way, it changes only certain aspects of the grammatical meaning, as illustrated below:

Interplay between morphological (39)

(i) N + i (ii) V + o

rules

113

means "more than one N " (e.g. cani "dogs", where i indicates masch., plur.) means "Verb + past tense" (e.g. amo "he loved", where ö is the past tense marker)

Point V refers to the normal synchronic operation of the rules in question. It is also possible, however, for both derived words and inflected words to take on idiosyncratic meanings, as seen in (40): (40)

(i) ceppo "log"/ceppi "logs" and "blocks of wood used to immobilize prisoners" (ii) dirigire "to direcf'/dirigibile "capable of being directed" and "air vehicle"

Bearing this in mind, items which are often cited as counterexamples to point V can be accounted for. For example, English nut which in the plural, nuts, can assume the meaning of "crazy" (cf. Siegel 1974) and Dutch tafel "table", which in the plural tafelen takes on the meaning "the tables of the law" (alongside the regular plural tafels "tables") (cf. Booij 1977), are, in fact, not really counter-examples. They are either relics or idiomatic expressions, and do not represent syncronic processes. All words (simple as well as derived and compound words) may potentially acquire idiomatic or idiosyncratic meanings, but this is a diachronic phenomenon, and it is neither regular nor predictable 2 . (VI) DR's may reapply, IR's may not. The output of a DR may follow two paths: it may either undergo Lexical Insertion (after the application of IR's) or return to undergo additional applications of the DR's. The word utilitaristicamente "in an utilitarian way" illustrates the path of repeated applications of DR's; in this case different WFR's have applied: base: rule 1: rule 2: rule 3: rule 4: rule 5:

utile utilitä utilitario utilitarista utilitaristico utilitaristicamente

"useful" "utility" "utilitarian" "utilitarianist" "utilitarianistic" "in an utilitarian way'

Repeated application of IR's, however, does not seem possible. While it may be possible to imagine that an inflected word such as lavava "(he) was washing" is formed by a series of IR's, first lava + ν ( = imperfect tense), then lavav + a ( = third person, singular), there is nevertheless a crucial difference between this case and the case of repeated applications of DR's. That is, in derivation, but not in inflection, the result of each application of a DR must be an existing word (or, in some cases, in an overgenerating

114

Generative morphology

morphology, at least a possible word). Lavav is neither an existing nor a possible word in Italian. Note that in derivation, the same rule can apply twice (usually not in the adjacent cycle) to form a complex word: (42)

storia + ico + ista 4- ico (storicistico) "historicistic" uno + itä + ario + itä (unitarietä) (3) "unitariness" precipita + evole + issimo + evole + mente (precipitevolissimevolmente) "extremely precipituosly"

(VII) DR's are not fully productive, while IR's are. It is in general possible to attach to any word the entire set of inflectional endings associated with the word class in question. In a few marginal cases, this is not true, such as nouns that only have a plural form (pluralia tantum) or defective verbs. These cases are highly marked, however, and must, in any case, be specified in the Dictionary. We can thus say that inflection is paradigmatic. The same is not true, however, for derivation. While we can always attach all inflectional endings of the relevant class to a verb, in derivational morphology we cannot attach to a given verb all the suffixes that are potentially attachable to verbs. That is, derivational morphology tends to exhibit gaps, due to a variety of factors. In the following list, we see that specific verbs take specific nominalizing suffixes but not others, while there are other verbs which do not take any nominalizing suffix at all: pettinare -•pettinatura modanare ->modanatura profanare -> profanazione -•alienazione alienare allontanare -> allontanamento -•sbranamento sbranare

(*pettinazione) (*modanamento) (*profanata) (*lienatura) (*allontanazione) (*sbranata)

telefonare -•telefonata mangiare —»mangiata indovinare ancorare

(*telefonamento) (*mangiatura)

"to comb - hairdo" "mold - moulding" "to profane - profanation" "to alienate - alienation" "to remove - removal" "to tear to pieces - tearing to pieces" "to telephone - tel. call" "to eat - full mean" "to guess" "to anchor"

In other cases, a particular affix may not be used because of what Aronoff calls blocking 4 , as illustrated in (44). That is, a derivation is blocked when there already exists a word with the same base and the same meaning. Thus, the verbs in the following list do not take any nominalizing suffix because the base of the verb is already a noun: (44)

Noun Verb disciplina -»disciplinare rovina —»rovinare condanna ->condannare inganno —»ingannare

*disciplinamento *rovinamento *condannazione "ingannamento

"discipline - to discipline" "ruin - to riun" "condemnation - to condemn" "deception - to deceive"

Interplay between morphological

rules

115

(VIII) DR's are optional, while IR's are obligatory While there are sentences in which no WFR has applied, there are no sentences where no IR has applied. This is a trivial observation, but it leads to a more radical claim and that is that every word which occurs in a sentence has undergone inflectional rules. To see this point, consider briefly the term "word". When we speak of words, it is necessary to distinguish between "lexeme" (or dictionary entry or citation form) and word-form (a word which actually occurs in a sentence). In this book, we are using the term "word" ambiguously, as does Aronoff (1976:XI), that is to refer to both "lexeme" and "word form". There are cases where wordforms are clearly recognizable as in It. libri "books", Engl, boys, where both have an overt inflection marker, i and s respectively. In the case of libro and boy, however, we must ask whether such items are lexemes or word-forms. A possible answer is that they are ambiguously both lexemes and word forms; in the first case, they are units of the Dictionary and in the second case they are elements that we find in sentences. In Italian and English, certain words can occur in sentences without any overt inflectional marker, but in many languages this is not the case. For example, according to Carrier (1979:18), in Tagalog the words (lexemes) require inflection in order to occur in sentences 5 . Extending this view to Italian and English, we see that the difference between uninflected words (lexemes) such as libro and boy an inflected words (word forms) libri and boys is that the former have the abstract representations [libro]N, [boy] Ν and the latter, at some point of the derivation, have the representations [ [libro] + infl]fvj, [ [boy] + infl]^. In this case, inflection turns out to be zero. This proposal thus requires that every word which occurs in a sentence be inflected even if it is a matter of zero inflection. 3. Compounding and Derivation As we have seen above, a model such as the one in (45) has been proposed for a number of languages that have been studied in some detail within the framework of lexical morphology. (45)

116

Generative morphology

(45) predicts that words can be found that have first undergone derivation and then compounding, but not compounding then derivation. In other words, the ordering given in (45) allows structures such as those in (46i) to be generated, but not those in (46ii):6 (46)

(i) [ [ [ ] + S U F ] # [ ]] [[ ] # [ [ ] + SUF]] [ [PRE+ [ ] ] # [ ]] [[ ]#[PRE+[ ]]] (n) [ [ [ ] # [ ] ] + SUF] [PRE+ [ [ ] # [ ]]]

The ordering in (45) has been used, either in its original form or in a slight modified form, in accounting for a variety of languages: English (cf. Siegel 1974, Allen 1978), Welsh (Allen 1978), Dutch (Booij 1977), Malayalam (Mohanan 1982), etc. Since the ordering hypothesis is not uncontroversial, both with regard to the languages for which it was proposed and for other languages as well, we will examine some of the problems it raises. In particular, we will examine the order proposed in (45) in relation to English and Italian. 3.1. The Extended Ordering Hypothesis in English It will be remembered that the model of lexical morphology proposed by Allen (1978) includes the following three levels: (47)

Level I: + affixation Level II: # affixation Level III: compounding

Allen refers to this proposal as the Extended Ordering Hypothesis (EOH), a hypothesis according to which "Level III morphological processes are ordered after all rules of affixation" (1978:215). This claim is uncontroversial as far as the relation between Level I affixation and compounding is concerned. That is, + affixes do not occur on the outside of compounds, but they may occur, as (47) predicts, within compounds, as illustrated in the following examples (cf. Allen 1978: 215): (48)

+ affix outside compounds * [in + [fuel — injected] ] *[ [battle-hero] + ic] *[ [street-music] + al]

+ affix inside compounds [ [inhospitable] [looking] ] [ [heroic] [sounding] ] [ [musical] [sounding] ]

What is controversial, instead, is the relation between Level II affixation and compounding. Allen bases her claim that Level III processes must be ordered after Level II processes on both phonological and morphological arguments. We will consider here only the morphological

Interplay between morphological

rules

117

arguments; the reader is referred to Allen's work for the phonological arguments. There are, first of all, certain cases in which the proposed ordering is unproblematic, as can be seen in (49), where the Level II prefix cannot be attached to a compound. re#

de# V de#N mal#

re-clean re-mail re-wash de-magnetize de-sensitize de-forest de-mast mal-formation mal-nutrition

but

* re # *re# *re# *de# *de# *de# *de# *de# *de#

[vacuum-clean] [blackmail] [brainwash] [hand-magnetize] [self-sensitize] [pine-forest] [fore-mast] [home-formation] [child-nutrition]

etc. There are other cases that seem to be counterexamples such as those in (50), but which can be accounted for if the element external to the compound is not considered an affix but rather an element added by a process of compound formation. As the items in (50i) show, it is possible to find non before adjectival compounds, but only because non is attached, according to Allen, not by derivation rules but by rules that are essentially equivalent to compounding rules 7 . A level II prefix such as un- cannot, in fact, be added to compounds, as seen in (50ii): (50)

(i) non [class-conscious] non [color-blind] non [home-made]

(ii) *un [class-conscious] *un [color-blind] *un [home-made]

Finally, there are still other cases that raise more serious problems for the E O H as can be seen in (51): (51)

to to to to

re-air-condition re-proof-read pre-air-condition pre-proof-read

It is clear that the structure of these words cannot be that given in (52i), since the prefix re- and pre- attach to verbs, not to nouns. The only possibility is, therefore, the structure in (52ii): (52)

(i) [ [ P R E [ (ii) [ P R E [ [

]N][ ]v]v ]N[ ]V]V]

The structure in (52ii), however, violates the E O H which says that

118

Generative

morphology

prefixation may not take a compound word as a base. In cases those in (51), the nature of the prefix is not ambiguous, as it is in of non. According to Allen, in fact, words such as air condition generated by productive compound formation rules of the type seen, and repeated in (53): (53)

[

] N navi "ferry boat(s)"

traghetto

fondo assistenza->fondi assistenza "assistance fund(s)"

These two cases are quite straightforward as long as we have a general condition that requires Inflection Rules to "know" where the head of a complex word is. There are, however, other cases in both English and Italian which are more problematic. Consider the following cases 12 : (76)

(ϋ)

(i)

sale arm

s s

manager race

lava porta

piatt letter

i e

In both structures in (76) we find plural forms which are internal to the compound (i.e. sales, arms, lettere "letters", piatti "dishes"), since the entire compound in each case is a singular form (i.e. one sales manager, un portalettere "one mailman") 13 . The plural of the entire compound is formed in English and Italian according to the two structures in (77): (77)

W

sale arm

s s

manager race

s s

w

W

infl

Infi

lava porta

piatt letter

i e

0

126

Generative

morphology

In both these cases, the plural inflection of the compound is not simply the plural of the head. That is, in English the inflectional element is the sister node of the entire compound not of the head (i.e. manager, race). In Italian, the fact that the inflectional element is not simply attached to the head is even clearer, since the compounds in question are exocentric compounds 1 4 , that is compounds that are without an apparent head 1 5 . These cases of internal inflection are problematic not only for Allen's constraint but also for the alternative constraint proposed earlier in this section. It should be noted, moreover, that the examples under consideration also pose problems for Aronoff's Uninflected Base Hypothesis if we attempt to extend its application from derivation to compounding, since they show that certain compound formation rules do, in fact, take an inflected word as a base. Once again, it seems that the solution may be a loop of the type proposed by M o h a n a n (1982), which would allow the output of a given morphological level to return for further application at an adjacent level. Such a solution would clearly represent a weakening of the E O H , and it merits more thorough investigation before any definite conclusions can be drawn. Finally, there are two more points that are worth mentioning in relation to the interaction between compounding and inflection. First of all, there are actually two possible structures in Italian for an exocentric compound consisting of the sequence W O R D - I - W O R D + Inflection. One is that seen above for words such as portalettere, and repeated in (78i), and the other is that given in (78ii) for words such as portaceneri "ashtrays": (78)

(i)

While the plural is "inside" the compound in the first case, in the second case, it is "outside" the compound. This is demonstrated by the fact that cenere "ash" is a [-count] noun, and has the plural ceneri only in the idiosyncratic meaning of "mortal remains". The word portaceneri does not, however, mean "cremation urn", as we would expect if the i of ceneri were, in fact, the plural marker of W O R D 2 . Instead, it is the plural marker of the entire compound: "ash trays" 1 6 . In other cases such as portalettere, there is no overt plural for the entire compound since there is already a plural morpheme on the right (through it is internal to the compound). This type of situation is accounted for by a mechanism of "inflection blocking" proposed by Kiparsky (1982), on the model of Aronoff's (1976) blocking mechanism for derivation 1 7 . In the case of inflection blocking, sequences of consecutive inflectional elements are prohibited. The second, and final, point we will raise here is that there seems to be

Interplay between morphological

127

rules

a tendency to generalize certain patterns in a language. In certain old Italian compounds with the head on the left, a process of boundary weakening has taken place, reducing the strength of the internal boundary between the components of a compound and thus the compositional nature of the word. As a result, the position of inflection has shifted from that following the head (WORD1) to that following the entire compound. Thus, the contemporary plural of pomodoro "tomato" (lit. porno+ d(i) + oro "apple of gold") is pomodori, although until recently normative grammars recommended the plural pomidoro which reflects the etymology of the word according to which pomo is the head of the compound.

5. Some borderline cases In the preceding sections, we have presented some arguments in favor of a clear distinction between derivation and inflection. There are, however, a number of cases in which this distinction is somewhat more subtle. In the next two sections, we will examine two such cases, that of past participle and that of evaluative suffixes in Italian. 5.1. The Past Participle According to some linguists (e.g. Matthews, 1974:53-54), the Past Participle (PP) is one of those cases which makes it difficult to draw a clear line between derivation and inflection: crowded may be classed as a derived "Participial Adjective" but heated cannot, since it is the P P of the verb heat. The reason for making such a distinction is that the first item may be modified by very (cf. a very crowded room) while the second may not (*a very heated room). In fact, a difference between the verbal and adjectival aspects of the P P has been recognized in a number of works among the more recent ones: Freidin (1975), Siegel (1974), Wasow (1977), Allen (1978), Lieber (1980), Williams (1981b). Let us consider the arguments given by Allen (1978) for establishing a distinction between -ed adjectives and -ed past participles since these bear directly on the relation between derivation and inflection under examination here. Allen's claim is that -ed adjectives are derivationally created, while -ed verbal forms are inflectionally created. In the first place, -ed adjectives can behave like simple adjectives as far as adjectival modification and complementation are concerned. Verbal (PP) -ed forms cannot, as is illustrated in (79): (79)

(i) Derived-ed Derived -ed

He was as

(ii) Passive -ed

new toy (*by his friends) He was delighted by his friends

(i)

Derived-ed

He was more

as ;a child with a

than appi apprehensive

(*by the prospects for advancement)

128

Generative

morphology

(ii) Passive-ed

He was excited advancement

(i) Derived-ed

He was too

(ii) Passive -ed

by

the

prospects

for

to act responsibly

(*by the strange turn of events) He was confused by the strange turn of events

Secondly, as Allen observes, in general, the meaning of -ed adjectives are different from those of the corresponding past participles. Consider the following sentences: (80)

(i) The door looks (ii) The door was closed by Bill

In (80i), closed only indicated the present state; it does not imply anything about previous states. That is, the door was closed (in the adjectival sense) is a true statement even if the door had never been open. In (80ii), however, closed indicates a change of state, and the statement the door was closed (in the verbal sense) can be true only if the door had previously been open. Furthermore, there are many participial forms such as determined, resolved, excited, etc. that describe a "mental state" in their adjectival sense, but a "change of state" in their verbal sense (cf. John is a very determined man vs. the elections were determined by the Catholic voters). Finally, there are -ed forms which derive from intransitive verbs and as such cannot have a passive (i.e. verbal) counterpart (cf. lapsed, traveled, etc.) To account for these facts, Allen proposes that the adjectival forms (such those in (79i) are derived by WFR's, while the passive forms (such as those in (79ii), are formed by IR's. Such a solution is not without its problems however, since it implies that the two types of -ed forms are totally unrelated to each other. For example, there would be no relation between determined, the adjective, and determined, the Past Participle. In relation to Allen's proposal, Lieber (1980:230) correctly points out that "if the adjectival participles and the verbal participles are derived from independent adjectival and verbal suffixes, there is no explanation in this system for the fact that the verbal and the adjectival forms are always identical". In response to this problem, Lieber proposes to derive the adjectival forms from the past participle. In addition to the points mentioned above, it can be seen in (81) that -ed adjectives behave in the same way as lexical adjectives in other respects as well (cf. Lieber 1980:229): (81)

(i) Antartica is uninhabited (ii) Joe seemed annoyed with Silvia

Interplay between morphological

rules

129

(iii) He is very excited about the idea (iv) Harry wanted me to finish the opened box before starting the new one In (81i), the P P is negated with the prefix -un, just as lexical adjectives can be negated (cf. un-safe). The example in (8Iii) shows that the P P may appear in a predicative position, as many adjectives (cf. Joe seemed happy). In (86iii), we see, furthermore, that PP's, like adjectives, may be modified by a degree particle (cf. He is very tall). Finally, in (81iv), the P P appears in prenominal position, as do lexical adjectives (cf. Harry bought the big box). On the grounds that adjectival PP's always behave in the same way lexical adjectives do, on the one hand, and that they are always identical to verbal PP's, on the other hand 1 8 , Lieber advances the proposal that adjectival PP's are derived from verbal PP's by means of a rule of zero affixation. Translated into the formalism adopted in this book, Lieber's proposal can be exemplified as follows, for the derivation of the adjectival form determined: (82)

(i) [determine] γ (ii) [ [determine] γ + ed]y (iii) [ [ [ d e t e r m i n e ] y + e d ] y + 0 ] a

The advantage of this proposal is that with zero affixation, as seen in (82iii), the identity between the verbal and adjectival forms of the P P is "both predicted and explained". Lieber demonstrates, furthermore, that the objections that are generally raised in relation to zero affixation are not valid in this case (cf. pp. 225-ff.). Nonetheless, this proposal, too, encounters a number of serious difficulties. The first problem is of a formal nature, and is a problem only within the type of morphological framework adopted here, not within the framework adopted by Lieber. The derivation in (82) presents an ordering which is contrary to that required by a level ordered morphology since an inflectional rule, (82ii)), precedes a derivation rule, (82iii). Since there are strong reasons for maintaining the opposite order, DR's before IR's, as we have seen, more convincing evidence than that provided by Lieber would be necessary before we can accept the alternative solution. Another problem arises in languages such as Italian where the type of process under consideration is even more widespread. In Italian, a P P can become an adjective, as in English, but it can also productively became a noun, as seen in (83): (83)

Verb morire "to die"

PP adjective morto G. e piü morto che vivo "dead" "G. is more dead than alive"

noun il morto fu lavato "the dead (person) washed"

was

130

Generative

morphology

Such a phenomenon would give rise, in Lieber's system, to a proliferation of zero affixation rules; in this case, one for deriving adjectives and one for deriving nouns. A final problem is that in languages with a richer inflectional system than that of English, the P P can also apper in inflected forms. Consider the following sentence in Italian: (84)

Queste conseguenze sono state determinate dalla sua ignoranza "These consequencies were caused by his ignorance"

In (84), the word determinate "determined" is a verbal P P (cf. the presence of the agent), and has the following structure: (85)

Verb determin

PP Infi, (fern., pi) at e

In Lieber's system, this word would have to be analyzed as consisting of "Verb + Inflection (PP) +Inflection (fem., pi.). The application of two consecutive inflection rules, however, violates the blocking principle proposed by Kiparsky (1982), already mentioned above. In the remainder of this section, we will advance an alternative proposal which makes use of the notion of zero affixation, but does not use the same category labels in the rules as Lieber does. Consider first the following classification of PP's as [ + V] and [ + N ] proposed by Aoun (1981) for standard Arabic: (86)

+ N

-Ν

+V

PP

V

-V

Ν

Ρ

This classification is intended to account for the fact that PP's have an ambiguous nature, being both verbal and adjectival. This ambiguity is easily seen in Italian if we examine some of the properties of a P P such as deciso "decided". (87)

adjectival properties (i) takes

clitics

(ii) is the base of adverb formation rules (iii) takes agent phrases (iv) takes degree

spec.

verbal properties

+

example decisosi "its having been decided" decisamente "decidedly" deciso da G. "decided by G." G. e molto deciso "G. is very decided"

Interplay between morphological

rules

131

What we propose in order to account for this double nature of PP's, is that Aoun's suggestion be incorporated directly in the WFr's themselves. That is, WFR's could be allowed to operate not on syntactic categories (Ν, V, etc.) but rather on syntactic category features ([ + N, + V ] , [ + N, —V], etc.). In this way, PP's would be derived as follows: (88)

[ [determine]

+ ed] +V -N

The feature [ + V] accounts for the verbal behavior of a P P such as determined, while the feature [ + N ] accounts for its adjectival behavior. It should be noted that the rule in (88) changes the syntactic category features involved and therefore must be a WFR, and not an IR. This proposal has several advantages. First of all, it explains the formal identity between verbal and adjectival PP's. Secondly, it does not require the use of zero affixation: when the output is an adjectival form, it will have the features [ + V, + N], and when it is a noun, as in the Italian case seen above, it will have the features [ — V, + N ] , Furthermore, the proposal advanced here does not require that an inflected P P such as determinate in Italian be derived by two consecutive applications of IR's. Instead, it allows us to analyze such forms as consisting of the sequence "Word + derivation + inflection", a pattern that is the usual one. Finally, it maintains the proposed ordering of derivational rules before inflectional rules. What the proposed analysis shows is that in the case of past participles, what seemed like an interaction between two morphological levels, derivation and inflection, is in fact, only a process of derivation, assuming that derivational rules may take into account syntactic category features, and not just syntactic categories. Such a modification may have consequencies for other processes as well, an area that must still be investigated. 5.2. Evaluative Suffixes Another case that appears to be on the borderline between derivation and inflection is that of the "evaluative suffixes", commonly found in Italian and other Romance languages. In Italian these suffixes are usually divided into the following four groups: diminutives (e.g. -ino), augmentatives (e.g. one), pejoratives (e.g. -accio, -ucolo) and others (e.g. -ello, -etto, -uzzo). The following discussion, will, however, be limited to an examination of the diminutive suffix -ino. Consider first the examples in (89): (89)

(i) [tavolo] N [tavolino] N "table - little table" (ii) [giallo]A —• [giallino]A "yellow - yellowish" (iii) [bene] Ad ν [benino]Adv "well - so so"

132

Generative

morphology

In (89), we see that -ino can be attached to nouns, adjectives and adverbs, and that its attachment does not change the syntactic category of the base 1 9 . Using the wellformedness condition proposed by Pesetsky (1979), and discussed above in Chapter V, section 2.2.3., we can write the following three rules to account for the three respective examples in (89): (90)

(i) [ [ (ϋ) [ [ (iii) [ [

ÜN+inoANNÜN ]A + i n o Λ Α Α ] Α

] Adv + i n o AAdv AdvÜAdv

It is clear that such an account contains a great deal of redundancy and does not capture the generalization underlying the three rules. Since the suffix -ino is transparent with respect to the syntactic category, we can, in fact, formulate the three rules in (90) as a single rule, given in (91): (91)

[[

Lx

+ino]aX

This rule is not only economical, it also captures the fact that the syntactic category of the output is always the same as the category of the input. Actually, an even stronger claim can be made here. That is, the attachment of an evaluative suffix not only does not change the syntactic category of an item, it does not change any of the features attached to the base either, as can be seen in (92i) for the syntactic feature [ +abstract], and in (92ii) for the subcategorization feature [ +transitive]: 2 0 (92)

(i) [letto] Ν -> [lettino] Ν [ - abstr] [ - abstr] [idea] Ν [ideuzza] Ν [ +abstr] [ +abstr] (ii) [giocare]v-»[giocherellare]v [ - trans] [ — trans] [mangiare] y -> [mangiucchiare] γ [ + trans] [ + trans]

'bed - little bed' "idea - little idea" "play - play around" "eat - nibble"

It will be recalled that WFR's can change the category of the base as well as its syntactic features and subcategorization frame, while IR's cannot effect such changes. The problem now is to determine whether the rules that attach the evaluative suffixes are WFR's or IR's. Let us now consider a more complete set of properties that characterize the behavior of evaluative suffixes in Italian (cf. Scalise 1983). (a) They change the semantics of the base (e.g. lume^lumino "lamp - little lamp") (b) They allow the consecutive application of more than one rule of the same type, and at every application the result is an existent word (cf. fuoco-+fuocherello^fuocherellino "fire - little fire - nice little fire").

Interplay between morphological

133

rules

(c) They are always external with respect to other derivational suffixes and internal with respect to inflectional morphemes (e.g. contrabbandierucolo "small-time smuggler" = Word (contrabbando "contraband") + Derivational suffix (-iere "agentive") + Evaluative suffix (-ucolo "pejorative") + Inflectional morpheme (-i "masch., pi."). (d) they allow, although to a limited extent, repeated application of the same rule on adjacent cycles (e.g. carinino "nice + dim. -I-dim.") 21 . (e) They do not change the syntactic category of the base they are attached to (cf. (89) above). (f) They d o not change the syntactic features or the subcategorization frame of the base (cf. (92) above). While the properties in (a) and (b) would lead us to group the evaluative suffixes with derivational affixes, the properties in (e) and (f) would lead us to group them with inflectional affixes. The properties in (c) and (d), however, are different from those of both derivational and inflectional affixes. On the basis of these last two properties, we can conclude that the evaluative suffixes cannot, in fact, be grouped with either of the other types of affixes. In a level ordered morphology such as the one assumed here, this situation can be handled easily by ordering a separate block of "Evaluative Rules" (ER's) after the derivational WFR's and before the Inflectional Rules, as represented in (93):

J.

I

ER's 1

I IR's

This ordering accounts for the linear distribution of the suffixes in question. Thus, we find sequences such as "Word-I-Derivational Suffix -I- Evaluative Suffix + Inflectional Affix" but we d o not find sequences such as "Word + Evaluative Suffix + Derivational Suffix + Inflectional Affix" (cf. point (c) above). The evaluative suffixes can be thus considered as a separate block of rules, different from both DR's and IR's.

6. Summary We have taken as a starting point in this chapter a Strong Lexicalist Hypothesis, and in particular, we have assumed that Inflection Rules are

134

Generative

morphology

part of the lexical component of a grammar. In this context, a number of morphological processes and interactions between different processes were examined. In this way, we were able to demonstrate that DR's and IR's are two distinct types of rules in that (a) they can change the syntactic category of their base while IR's cannot, (b) IR's are always peripheral with respect to DR's, (c) DR's and IR's are sensitive to different properties of their bases, (d) DR's and IR's "do" different things, (e) DR's change the conceptual meaning of their base while IR's only change the grammatical meaning, (f) DR's may apply more than once forming a "word" each time, while IR's cannot, (g) DR's are not fully productive, while IR's are, and (h) DR's are optional while Ir's are obligatory. In addition, it was shown on the basis of the interaction between CR's and DR's, that the EOH, according to which derivation precedes compounding is the most restricted hypothesis regarding the ordering of these two block of rule types. In a few cases, however, it was seen that there may also be a need for limited use of a loop device. An examination of the interaction between CR's and IR's showed, furthermore, that when IR's operate on compounds, they must take into consideration which element is the head of the compound. In general, the inflectional morphemes introduced by IR's are attached to the head, although occasionally we do find inflectional morphemes on non-head elements inside compounds. Finally, two bordeline cases were examined, that of PP's and that of Evaluative Suffixes in Italian. In the first case, it was proposed that PP's be handled by derivational WFR's, not by IR's, and in the second case, it was proposed that Evaluative Suffixes be considered distinct from both DR's and IR's and that they be ordered in the lexical component between these two types of rules. By way of summary of the main points of this chapter, consider Figure (1)·

Figure 1: The organization of the lexical component (IV).

Interplay

between morphological

rules

135

Figure (1) represents a somewhat simplified version of the lexical component we have developed thus far. 22 In the Dictionary we find words, and, in marginal area, also stems. The morphological subcomponent contains Word Formations Rules (i.e. DR's and CR's) and Inflection Rules. The WFR's are optional, while the IR's are obligatory. Furthermore, while DR's and CR's may return for further applications, IR's may not, as indicated by the arrow returning ( LT) under the first two blocks of rules. The broken lines represent loops which allow DR's to take a compound word as a base and CR's to take an inflected word as a base 2 3 . A number of possible structures generated by the model in Figure (1) are given in (94), along with examples: (94)

(i) (ii) (iii) (iv) (v) (vi) (vii) (viii) (ix) (x) (xi) (xii)

inflected word (boys) derived, inflected word (singers) compounded, inflected word (school boys) derived, compounded, inflected word (sewing machines) derived, derived, inflected word (modernization) derived, derived, derived, inflected word (structuralistic) derived", inflected word (structuralistically) compounded, compounded, inflected word (high school teachers) compounded, derived, inflected word (old maidish) inflected, compounded, inflected word (programs coordinators) compounded, compounded, compounded, inflected word (tuna finish salad sandwiches) compounded", inflected word (tuna fish salad sandwich platters).

Notes 1. On this type of verb, cf. Chapter VII. 2. Carstairs (1981:54) observes that in some languages a few stems exhibit alternative inflections (cf. Engl, fish, pi. fish or fishes, Lat. domus "house" dat. sing, domo or domui, Germ. Wort "word" pi. Worte or Wörter, etc.). He also observes that this variability sometimes can be assigned to different dialects or "registers". In order to account for these facts, Carstairs proposes the following Inflectional Parsimony Hypothesis: "Even when there is more than one inflexional realization available for a given combination of grammatical features, each stem must select only one of these realizations (or can select more than one only at high cost)". 3. The suffix -ita becomes -eta after [j]: Cf. Scalise 1983, Chapter V. 4. On the notion of blocking, cf. Chapter VII. 5. Carrier actually makes a stronger claim. She maintains that also the outputs of some WFR's require overt inflectional markers before they can occur in sentences. This point, too, cannot be demonstrated on the basis of English or Italian data. 6. In (46) we give also structures with prefixes, but in the following discussion we will only consider structures with suffixes. The structures given in (46) are only a few of the

136

Generative

morphology

possible ones. If the ordering in (45) is not constraint, in fact, it is possible to generate also structures such as the following [PRE + [ [ ] + S U F ] ] # [ ] ] ; [ [ [ P R E + [ [ ] + S U F ] ] # [ [ [ [ ] + S U F ] + S U F ] + S U F ] ] , etc. Intuitively such structures seem not highly probable. More data, however, would be necessary in order to decide whether they are sistematic or accidental gaps. 7. Other items "which may also attach to words by means of a compounding rather than a true affixation process are extra, super, over, all-" (Allen, 1978:288). 8. Allen (1978) also discusses other potential counterexamples and concludes in each case that they are not, in fact, violations of the EOH. 9. Cf. the preceding chapter, section 3.3. Arguments for the distinction between strict and loose compounds in Italian are given in Scalise (1983). 10. For example, Vogel and Scalise (1982) show how strict compounds in Italian undergo secondary stress RR's and how loose compounds do not undergo the same rules. 11. Other problems are discussed by Botha (1981), who actually argues against the EOH on the basis of Afrikaans data. 12. Other examples cited by Selkirk (1982) include: overseas investor, parks commissioner, sales receipt, parts distributor, arms conscious, sales oriented. Strauss (1982:76) cites still other examples: admissions office, jobs committee, materials center, rights organization, systems analysis. 13. Note that the plural is the same in Italian: due portalettere "two mailmen", tre lavapiatti "three dishwashers". 14. Portalettere "mailman" is a noun formed from a verb porta "carry" and a noun lettere "letters". We will see in Chapter VIII that not only the category but also the features associated with the head, such as [ ± animate], percolate from the head to the entire complex word. It suffices here to note that the noun lettere, however, cannot be the head of the compound since lettere is [ - a n i m a t e ] and portalettere is [ +animate]. 15. There have been attempts to interpret these compounds as endocentric (e.g. Marchand 1969, Zuffi 1981) though the results are not uncontroversial (cf. Scalise 1983). 16. The fact that we can distinguish in a compound between "inflection of Word 2" and "inflection of the whole compound" is an argument against Strauss" hypothesis (1982:4), according to which inflection is non-node building. 17. On blocking, cf. Chapter VII. 18. Lieber (1980:223) cites two cases where the adjectives and the verbal participles may differ in form, namely burnt vs. burned, and proven vs. proved. 19. Mayerthaler (1981:108) makes a distinction between "category changing" and "non category changing" morphological operations. He proposes that the German diminutives belong to the non category changing operations. In German, a diminutive can change the gender, not the category of an item (cf. [ F r a u ] N "Lady, fem." —> [Fraiilein] N "Miss, neuter"). 20. An evaluative suffix attached to a verb has usually a "frequentative" value. 21. Other examples of cyclic application of ER's are given by Schultink (1974) for Afrikaans (cf. stukkietjie "little little piece", kindjietjie "little, little child"). The fact that these rules apply on adjacent cycles distinguishes them from derivational rules which may also exhibit a certain degree of recursivity, but not on adjacent cycles (cf. Germ. Einheitlichkeit "unitariness", Freiheitlichkeit "liberalness" (Wurzel 1970), Dutch Kleurloosheidloos "without colorlessness" (Booij 1977). In all of these cases, in fact, between the two application of the same sufifixation rule, there is always an intervening suffix. Booij (1977:154) observes that prefixes may occasionally exhibit recursive attachment without another intervening prefix (cf. Dutch ultra-ultramodern "very, very modern", para-para normaal "paraparanormal", but these cases are limited to a small group of relatively "independent" prefixes. 22. To simplify the figure somewhat, we have not included the ER's, RR's and BIC. 23. It should be noted that the loop device must be subjected to strong restrictions to prevent the massive overgeneration of complex word structure.

Chapter

VII

Constraining word formation rules In this chapter, we will discuss several proposals that have been put forth in order to constrain Word Formation Rules, namely the Unitary Base Hypothesis (Aronoff, 1976), the Binary Branching Hypothesis (formulated originally by Aronoff 1976 as the "one affix, one rule hypothesis"), the Ordering Hypothesis (Chapin, 1967, 1970), the N o Phrase Constraint (Roeper and Siegel, 1978; Allen, 1978) and Blocking (Aronoff, 1976). Several other constraints proposed in recent years, in particular, the Adjacency Condition (Siegel, 1977) and the Atom Condition (Williams, 1981a) will be discussed in the next chapter.

1. The unitary base hypothesis According to the Unitary Base Hypothesis (UBH), the syntacticosemantic specification of the base of a word formation rule may be more or less complex, but it is always unique (Aronoff, 1976:48). For example, the base of a given affixation rule may be a noun or a verb, but the same affixation rule cannot take as its base both nouns and verbs. Aronoff maintains, at least for English, that all the apparent counterexamples he found could be analyzed as different rules whose operation happens to be homophonous. Consider, for example, the suffix -able which can be attached to nouns (e.g. fashionable) and to verbs (e.g. acceptable). According to Aronoff, we are dealing, in this case, with two different suffixes (and therefore with two different WFR's) for the following two reasons: (a) the denominal Xable adjectives can be further derived with the suffix -ness but not with the suffix -ity (cf. fashionableness vs. *fashionability), and (b) the two -able formations have different semantics: the denominal one means approximately "characterized by X", while the deverbal one means approximately "capable of being Xed". Before examining the empirical validity of the U B H , two observations should be made. First of all, besides the Unitary Base Hypothesis, it is necessary to postulate a Unitary Output Hypothesis. This second hypothesis does not allow a particular phonological form to be considered a single affix if it produces outputs with different category labels or different semantics. 1 F o r example, we cannot speak of single prefix s- in Italian in sbloccare "to unblock" and sgocciolare "to drip", even though in both cases the s- is attached to a verb, because the semantics are not the same in the two cases. In the first example, s- adds the meaning of "negation", while in the second example it adds the meaning of "repeated action".

138

Generative

morphology

The second observation is that in analyzing the bases of a given affix, we must be sure we are dealing with a synchronic phenomenon, since the productivity of an affix with specific bases often changes over time and may leave what appear to be counterexamples to a particular rule. 2 In the following sections, we will examine a number of cases of affixation, including some which violate the UBH. An alternative formulation of the hypothesis will be proposed along the lines of the X theory, and it will be shown that this formulation is, in fact, able to handle the cases that are problematic for the UHB. 1.1. The Modified Unitary Base Hypothesis Although Aronoff does not distinguish between prefixes and suffixes in his U B H , such a distinction appears to be necessary. While the U H B appears to hold fairly generally for suffixes, there are many prefixes which attach to three categories, Α, Ν, V, as can be seen in (1) and (2), which contain examples from English and Italian, respectively: (1)

English (i) inter-

(ii) co-

(iii) dis-

(iv) pre-

(2)

+ A + N + V +A + N +V + A + N + V +A +N +V

international interstate interweave coextensive coauthor coexist discontent displeasure discontinue premature prehistory presuppose

ltalian (i) contro-

(ii) inter-

im) sopra-

(iv) pre-

+ + + + + + + + + + + +

A N V A N V A N V A N V

controintuitivo controrivoluzione controargomentare interdisciplinare interfaccia interporre sopranazionale soprappeso sopra(v)valutare pr er in asci mentale preistoria preordinare

"counterintuitive" "counterrevolution' "counterargue" "interdisciplinary" "interface" "interpose" "supranational" "overweight" "over-estimate" "pre-renaissance" "prehistory" "prearrange"

The application of a given affix to more than one base is not limited to

Constraining

word formation

rules

139

prefixation, however. There are also examples of suffixation in which we find that the same affix attaches to difFerent bases, in violation of the U H B . Consider the following Italian examples: (3)

(i) -anza V A (ϋ) -esimo Ν A (iii) -ime V A (iv) -ia Ν A

->• Ν Ν Ν Ν Ν -> -> Ν Ν -> Ν

tolleranza tolleranza lontananza incantesimo umanesimo mangime grassime maestria cortesia

"tolerance" "distance" "enchantment' "humanism" "poultry feed" "manure" "mastery" "kindness"

( f r o m tollerare "tolerate") (from lontano "distant") (from incanto "enchantment") (from umano "human") (from mangiare "eat") (from grasso "fat") (from maestro "master") (from cortese "kind")

While these examples violate the U B H in that the same suffix is attached to two different bases in each case, closer examination reveals that there is, in fact, a pattern in the violation. T h a t is; in each case, the suffix attaches either to Adjectives and N o u n s or to Adjectives and Verbs. In terms of the syntactic features of the X theory (cf. C h o m s k y 1970), the first g r o u p m a y be characterized as [ + N ] , while the second g r o u p may be characterized as [ + V], Thus, while it is not true that suffixes are always attached to a unique base in terms of syntactic category, we see that they are always added to bases that can be characterized by a single syntactic category feature. C o m b i n i n g this observation with the observation m a d e a b o v e a b o u t the basic difference between prefix and suffixes, we m a y reformulate U B H as follows: (4)

Modified

Unitary Base Hypothesis

(MUBH)

A suffix may be attached only to bases that form a syntactic class specifiable in terms of a single syntactic category feature in X b a r theory terms. According to the M U B H , a given suffix may be attached t o b o t h Adjectives a n d N o u n s ([ + N]), or to b o t h Adjectives a n d Verbs ([ + V]). Only suffixes which attach to b o t h N o u n s and Verbs would violate t h e M U B H , since these two categories d o n o t have a syntactic category feature in c o m m o n . It should be noted that the point which distinguishes between prefixes and suffixes is the fact that the "single syntactic feature" restriction on suffixes does not hold for prefixes. As was seen in (1) a n d (2), a given prefix may be attached to b o t h N o u n s and Verbs (as well as Adjectives). While the M U B H allows us to account for the vast majority of suffixation rules, there are several cases which seem to violate the hypothesis. Consider the following suffixes in Italian (selected f r o m Tekavcic's (1972) list): (5)

(i) -ic-

A Ν V

" t o limp" "to s n o w " "to smoke"

140

Generative (ii) -ingo

A Ν V (iii) -ismo/ A -ista Ν V (iv) - a t a Ν V (ν) -ino Ν V Ν

morphology -» A A -> A -» Ν Ν ->• Ν -> Ν -» Ν -> Ν -»Ν -> A

solingo ramingo guardingo petrarchismo/ista socialismo/ista determinismo/ista cuchiaiata mangiata postino imbianchino pecorino

"solitary" "wandering" "cautious" "Petrarchism/ist' "Socialism/ist" "determinism/ist "spoonful" "meal" "mailman" "painter" "of sheep"

In (5i-iii), we see that the same suffix attaches to three different bases (A, Ν, V), and in (5iv-v) to two bases (Ν, V). In all these cases, it is impossible t o characterize the bases in terms of a single syntactic category feature since they include the c o m b i n a t i o n of N o u n and Verb. In the following sections, we will examine these problems in m o r e detail. 1.2. N,V, A + Suffix As we have seen in (5), there are several suffixes in Italian that a p p e a r to attach to bases of three different syntactic categories. A closer examination of these cases, however, reveals that they are not, in fact, counterexamples to the M U B H , but only a p p a r e n t counterexamples. Consider first the case of -ic-. It should be noted, first of all, that a l t h o u g h Tekavcic lists -ic- as a suffix, it is not really a suffix but an infix. T h a t is, it is only found within words (e.g. inciampare/inciampicare "to trip - to trip repeatedly"), but never word finally. F u r t h e r m o r e , the example of attachment of -ic- to a Verb ( a f f u m i c a r e "to smoke"), is actually a case of parasynthetic word formation (cf. sec. 2.1. below) and the base is not the verb fumare " t o smoke" but, more correctly, the n o u n fumo "smoke". The only cases in which -ic- attaches to verbs are those in which it is infixed to an already existing verb, adding only a frequentative value. In such cases, however, we are clearly not dealing with the same process that applies to n o u n s (e.g. neve/nevicare "snow - to snow") and adjectives (e.g. zoppo/zoppicare "crippled - to limp") in which -ic- a t t a c h m e n t functions as a true W F R , changing the semantics as well as the syntactic category of the base. We thus have two -ic- a t t a c h m e n t rules, one of frequentative infixation which applies to verbs and a W F R which applies to n o u n s and adjectives. This second rule, in fact, conforms to the M U H B since it applies to the set of words characterized by the single syntactic feature [ + N], The second suffix, -ingo, raises another problem, since the only words listed in a b a c k w a r d s dictionary of Italian (Alinei 1962) are the following: casalingo " h o m e y " and ramingo "wandering" f r o m the n o u n s casa "house" a n d ramo "branch", solingo "solitary" f r o m the adjective solo "alone" and guardingo "cautious" f r o m the verb guardare "to look". F r o m these few

Constraining

word formation

141

rules

examples, it is clear that -ingo attachment is not a productive rule of Italian, and as such does not constitute a real c o u n t e r e x a m p l e to the M U B H . In fact, as Aronoff (1976:48) points out, "less productive rules tend to be less coherent, and we should expect more variation a n d exceptional behavior with such a rule". A similar point can be m a d e in relation to the third type of suffix, -ista a n d -ismo. Of a very long list of n o u n s ending in -ista and -ismo, only five can be derived f r o m a verb: apprendista "apprentice" f r o m apprendere " t o learn", determinista/determinismo "determinist — determinism" f r o m determinare "to determine" and arrivista/arrivismo "social climber - social climbing" f r o m arrivare " t o arrive". Once again, the fact that so few cases of this rule can be found shows that it is not productive and therefore not in conflict with the general rule of -ista/-ismo a t t a c h m e n t . T h e general rule is a very productive process which attaches the suffixes in question to N o u n s and Adjectives, the combination characterized by [ + N ] , Thus, in this case, too, what appeared to be a counterexample to the M U B H is not actually a problem, but the result of a small n u m b e r of unproductive formations. 1.3. Ν, V +ata T h e next suffix we will examine is -ata, which attaches to b o t h n o u n s and verbs 3 in violation of b o t h the U B H and M U B H . Let us first consider the relatively complex semantics of this suffix: (6)

(i) piede

-»pedata

(ii) c u c c h i a i o (iii) cretino

-»cucchiaiata " s p o o n - spoonful" "idiot - act of an idiot" -»cretinata

(iv) cancello

-»cancellata

"gate - set of gates"

"set of X"

(v) a n n o

"foot - kick"

"blow d o n e with X" " a m o u n t c o n t a i n e d in X" "act of X"

-»annata

"year - period of a year"

"period X of time"

(vi) arancio

-»aranciata

"orange -

"product of X"

(vii) guardare

-»guardata

"to look - glance"

orangeade"

"single act of X"

W h e n -ata is attached to nouns, it has a wide variety of p a r a p h r a s e s (6ivi), but when it is attached to verbs, it has only one p a r a p h r a s e (6vii), which is different f r o m those given for the nouns. Consider, next, the following three n o u n s derived f r o m verbal bases: (7)

(i) c a m m i n a t a (ii) bevuta (iii) d o r m i t a

"a walk" "a d r i n k " "a sleep"

( f r o m c a m m i n a r e "to walk") ( f r o m bere "to drink") ( f r o m d o r m i r e " t o sleep")

All three w o r d s in (7) have the same paraphrase ("single act of X"), even t h o u g h the form of the suffix varies: -ata, -ita, -uta. In order to account for this variation, one could assume (as we have d o n e elsewhere) t h a t the base of verbal derivation is the " t h e m e " cammina, beve, dormi, that is, the verbal root plus the thematic vowel. This would account for the forms of the suffix in (7i) and (7iii). The example in (7ii) could be treated with an

142

Generative

morphology

allomorphy rule of the base, which carries out the change e->u. Such an allomorphy rule does not solve the problem, however, since it would apply before the Vowel Deletion Rule of Italian (cf. Scalise 1983), which would later delete the vowel in question, giving the wrong result, as seen in

(8): (8)

beve beve + ata bevu + ata

Dictionary WFR AR V.Del. Output

*bevata

It should be noted, however, that the distribution of vowels seen in (7) is the same distribution found in the past participle of the same verbs, as seen in (9):

(9)

(i) camminato (ϋ) dormito (iii) bevuto

"walked" "slept" "drank"

On the basis of this pattern, Herczeg (1972) proposes to derive the items in question by attaching the suffix -ata to a base consisting of the past participle of a given verb. Such a rule clearly leads to problems, however, as can be seen below: (10)

(i) camminato + ata (ii) dormito + ata (iii) bevuto + ata

The outputs in these cases are the incorrect forms *camminatata, *dormitata, and *bevutata, assuming that V. Del. deletes the final -o of the past participle. An alternative solution is to posit, instead of -ata, a suffix -a, which gives the following derivations:

(11)

(i) camminato + a (ii) dormito + a (iii) bevuto + a

V. Del. applies to these forms and the final output are camminata, dormita and bevuta, respectively, which are the correct results. Further confirmation of this analysis comes from the fact that such a suffix -a is actually found in other cases as well:

Constraining (12)

word formation

letto + a messo + a preso + a promesso + a

143

rules

->letta ->messa ->presa -»promessa

"read - (a) reading" "placed - (a) placement" "taken - (a) t a k i n g " "promised - (a) promise"

In each of these cases, the suffix -a is attached to a past participle (irregular form) and has the p a r a p h r a s e "single act of X". W e can t h u s conclude that the original problem of a single suffix -ata attaching to b o t h N o u n s and Verbs is not a problem at all. W e are, in fact, dealing with t w o different suffixes: -ata which attaches to n o u n s and -a which attaches to verbs (i.e. past participles). T h e M U B H is therefore not violated in this case. 1.4. Ν, V +ino The last suffix we will consider here is -ino, which attaches to b o t h n o u n s and verbs, as can be seen in (13). The suffix u n d e r examination here should not be confused with the diminutive suffix -ino, which has completely different properties, as we have seen in C h a p t e r VI, section 5.2. (i) cristallo sale asino alpe (i>) imbianca(re) scribacchia(re) attacca(re) spazza(re) ( i n scalda(re) frulla(re) macina(re) passa(re) (iii) posta bagno questura (ΗΓ)

vettura penna cero polso stoppa

-•cristallino ->salino -•asinino -•alpino -•imbianchino -»scribacchino -•attacchino ->spazzino ->scaldino ->frullino -»macinino -•passino -•postino -•bagnino -•questurino ->vetturino ->pennino -•cerino ->polsino -•stoppino

"crystal - crystalline" "salt - saline" "ass - assinine" "alp - alpine" "to paint - painter" "to scribble - scribbler" "to attach - bill-poster" "to sweep - sweeper" "to heat - heater" "to blend - blender" "to grind - grinder" "to strain - strainer" "mail - mailman" "bath - lifeguard" "police station - policeman' "vehicle - driver" "pen - pen nib" "wax - wax match" "wrist - cuff" "tow - wick"

The examples in (13) all have in c o m m o n the fact that they illustrate suffixation of -ino, although the different g r o u p s represent s o m e w h a t different patterns. First of all, the examples in (13i) can be distinguished f r o m all the others in that the o u t p u t in -ino is an adjective in the first case but a n o u n in the others. O n the basis of the U n i t a r y O u t p u t Hypothesis proposed above, we can therefore, establish that the rule in the first group, N - > A , must be separate f r o m the rule(s) characterizing the other groups. The U n i t a r y O u t p u t Hypothesis allows us t o distinguish, further-

144

Generative

morphology

more, between (13ii) and (13ii') as well as between (13iii) and (13iii'). In the cases of (13ii) and (13ii'), we can conclude that we are not dealing with a single rule because the outputs in the first group have the feature [ + a n i m a t e ] while those in the second group have the feature [-animate]. This same difference between [ + a n i m a t e ] in one group and [-animate] in the other is also found in (13iii) and (13iii'), respectively. In addition, all the examples in (13iii') are lexicalized forms in which the meaning of the derived word is not predictable. These words, thus, must be listed in the lexicon and not derived by rule. O n the basis of the Unitary Output Hypothesis, the only two sets of examples that could be handled by the same rule are those in (13ii) and (13iii), but these violate the M U B H in that -ino is attached to Verbs in the first case and to Nouns in the second. The results in both cases, however, are the same: [ + a n i m a t e ] N o u n s which can be paraphrased as "person who abitually or professionally has to d o with X". There are some ambiguous cases in both groups that at first glance make it appear possible to collapse the two sets of data. In (13ii), for example, attacchino may also be seen as deriving from the noun attacco "attachment", and in (13iii) a word such as bagnino may be seen as deriving from the verb bagnare "to bathe". It is not possible, however, to pick one category, either Nouns or Verbs and allow both groups to be derived from the same type of base, since there are other examples in each group which cannot be derived from a category other than the one originally posited. For example spazzino cannot be derived from anything other than the verb spazzare since there is no noun that could otherwise serve as its base. Similarly, postino must be derived from the noun posta, since there is no verb in Italian that could be taken as its base. To summarize briefly, what we have seen in the case of the suffix -ino is that the first set of examples in (13i) could be eliminated from consideration since the output is of a different syntactic category from the rest. The examples in (13ii') must also be treated separately since their output is [-animate], while the output in (13ii) is [ + a n i m a t e ] . The cases in (13iii') were eliminated not only because their output was [-animate], but more importantly, because they represent lexicalized forms that cannot be derived by a synchronically productive rule. The remaining cases, (13ii) and (13iii), in fact, violate the M U B H . At this point, we could simply say that we must be dealing with two different suffixes -ino since they attach to different types of bases, but this would not explain why the syntactic category and the semantics of the output are the same in both cases. In the next section, we will analyze this problem further from a more formal point of view. 1.5. One Suffix or Two? In relation to the problem just seen, H. van der Hulst and M. Moortgaat have observed (personal communication) that a similar situation exists in Dutch with the suffix -ist. Consider the words componist "composer" and

Constraining

word formation

rules

145

essayist "essayist", which have the structures given below:

(14)

(ϋ)

Ν

(i)

Ν

V

Suf

Ν

Suf

compon

ist

essay

ist

These two structures are identical, except for the fact that the left node in (14i) is a verb, and in (14ii) it is a noun. This difference determines, further, a difference in the paraphrases of the two items: "someone who composes" in (14i), and "someone who [ P R E D I C A T E ] essays" in (14ii). In the first case, the paraphrasis is unequivocally determined by the presence of the verb "to compose", while in the second case, we are dealing with an abstract predicate, such that an essayist could be someone who "writes, composes, publishes, etc., essays". The same point can be made about the difference between spazzino and postino in Italian. A spazzino is "someone who sweeps", while a postino is "someone who carries, distributes, delivers, etc. mail". The question now is whether such a difference is sufficient to establish two different affixes. While there are no clear answers to this question, let us consider the two criteria given by Aronoff (1976:48) for determining whether a given form should be analyzed as a single affix or as two affixes. The first criterion is whether in different cases the semantics of the affix are different, and the second criterion is whether the affix behaves differently from a formal point of view in different cases. As we have seen, it was on the basis of these criteria than Aronoff distinguishes between two -able suffixes: -able 1 can be followed by -ness but not -ity, while -able 2 can be followed by both -ness and -ity, and furthermore, -able 1 means "characterized by X", while -able 2 means "capable of being X-ed". In the case of -ino in Italian, the only difference observed was that of the two paraphrases mentioned above, and it is not clear whether only a semantic difference of this type is adequate for distinguishing between two suffixes. In this regard, let us consider an observation made by Lepschy (1981:71) about the suffix -are which attaches to nouns to form verbs in Italian. When -are is attached to the name of an animal, the resulting semantics differs according the animal named as the base, giving such meanings as (a) the sound made by the animal (e.g. chiocciare "to cluck" from chioccia "brooding-hen", cicalare "to chatter" from cicala "cicada", (b) the action of the animal (e.g. tarlare "to become worm-eaten" from tarlo "wood-worm", tarmare "to become moth-eaten" from tarma "moth"), and (c) other meanings (e.g. civettare "to flirt" from civetta "owl", scimmiottare "to ape" from scimmiotto "young monkey"). This observation seems to lead to the conclusion that there are η -are suffixes, with η different meanings. This is

146

Generative

morphology

clearly not a desirable solution, however, since it leads to an uncontrolled proliferation of suffixes. In the case of -ino, we are only talking about positing two separate suffixes, not η suffixes, but the point just made in relation to verb formation with -are is still relevant. That is, establishing separate affixes solely on the basis of different semantics seems to lead to the creation of an excessively large number of affixes, many of which are homophonous. Clearly a more adequate criterion will be needed for determining the number of separate affixes in a given case. Perhaps what is needed is a more precise specification of which aspects of the semantics must be the same or different in order to determine whether we are dealing with one or more affixes. In any case, for the time being, our specific problem of whether -ino is one suffix or two, must remain unresolved. It should be noted, in conclusion, that as far as the M U B H is concerned, if -ino is eventually determined to be a single suffix, this is the only counterexample to the hypothesis, since all the other apparent counterexamples were shown to be analyzable in terms of syntactic features once the marginal cases were separated from the productive rules.

2. The Binary Branching Hypothesis The Binary Branching Hypothesis (BBH) was originally formulated by Aronoff (1976) as the "one affix, one rule hypothesis", according to which a W F R attaches one and only one affix at a time. This amounts to saying that no matter how complex the internal structure of a derived word is, the structure will always be binary. According to this hypothesis, 4 morphological rules build structures such as the ones in (15i) but not such as the ones in (15ii):

A morphological three can, thus, branch to the left or to the right but it must always branch in a binary fashion, as illustrated by the Italian words utilitaristicamente "in an utilitarian way" and indeformabilita "indeformability":

Constraining (16)

word formation

rules

[ [ [ [ [ [utile]A

ita]N

147

(i)

ario]A

ista]N

ICO]A

mente]ADlv

(ii)

[[in

[[de

[forma]y]v

a

bile]AÜA

ita]N

Potential counterexamples to the B B H exist in Italian, where structures of the following types can be found: (a) X W Z is a word where W is a W o r d and X a n d Ζ two affixes, b u t where neither *XW nor * W Z are occurring words; (b) W X Z is a word where W is a W o r d and X and Ζ are two affixes which can be shown to be two independent affixes, but where *WX is not an occurring word. We will discuss these cases separately in the next two sections, showing that they are n o t real counterexamples to the B B H . 5 2.1. Parasynthetics Traditionally, 6 a parasynthetic word is defined as a complex w o r d derived by m e a n s of the simultaneous a t t a c h m e n t of a prefix and a suffix t o a single base, giving the form [ P r e + X + Suf]. "Simultaneous" a t t a c h m e n t refers to the fact that neither the sequence [ P r e + X ] nor the sequence [X + S u f ] exists alone; it is assumed that the two affixes must t h u s be added at the same time. Consider now the following examples of verbal parasynthetics 7 in Italian, where the prefixes in-, di- and s- are usually assumed to be a d d e d along with the suffixes -ire, and -are: (17)

brutto rozzo vecchio

-»imbruttire -»dirozzare ->invecchiare

"ugly - to m a k e ugly" "crude - to make less c r u d e " "old - to age"

The reason for considering these words parasynthetics is that, as mentioned above, neither [ P r e + X ] nor [X + S u f ] exists in these cases. T h a t is, we d o not find the adjectives *imbrutto, *dirozzo, *svecchio in Italian, just as we d o not find the Verbs *bruttire, *rozzare, *vecchiare. O n e can thus conclude that the derived words in (17) have the following structure, as illustrated with imbruttire:

148

Generative

(18)

morphology

V

Pre

A

Suf

in

brutto

ire

Such an analysis raises at least two problems, however, one of a descriptive n a t u r e and the other of a formal nature. The first problem is that in traditional analyses of parasynthetics, the suffix is usually described as an inflectional element. In the examples in (17), this would be the infinitive m o r p h e m e . According to such an analysis, the rules that form parasynthetics must attach two different types of affixes: a derivational affix (the prefix) and an inflectional affix (the suffix). This difficulty is not as serious as it seems, however. There are, in Italian, as in other languages, rules that normally derive verbs f r o m adjectives or n o u n s by adding an infinitive morpheme: (i) A->V zitto attivo calmo

—»zittire -»attivare ->calmare

"quiet - to quiet" "active - to activate' "calm - to calm"

->armare -»oliare ->firmare

" a r m - to a r m " "oil - to oil" "signature - to sign'

(ii) N->V arma olio firma

It should be recalled that a f u n d a m e n t a l difference between D R ' a and IR's is that the former, b u t not the latter, m a y change the category of the base in the process of word formation. T h e rules illustrated in (19i) and (19ii), in fact, change the syntactic category of the words in question. They cannot, therefore, be Inflectional Rules, but must, instead, be Derivational Rules. Returning to the formation of parasynthetics, we can conclude that we are dealing with a process of derivation with respect to the suffix as well as the prefix, since the suffixes in question are the same as those in (19). Even if we assume that the suffix of parasynthetic formations is a derivational suffix, we still have a n o t h e r problem to c o n t e n d with, the second type mentioned above. T h a t is, the structure given in (18) poses a problem with respect to the BBH. F u r t h e r m o r e , if we assume the analysis in (18), parasynthetics are the only case in Italian in which a ternary structure is posited in the morphology. There is thus ample reason to question the validity of the ternary structure posited above. Let us now examine the d a t a more closely. Consider first the word imbruttire "to make ugly". In analyzing the prefix, it should be recalled, first of all, that there are two different in-

Constraining

word formation

rules

149

prefixes in Italian. O n e of these is the negative prefix in- which, like the negative prefix un- in English, is never attached to an adjective with negative semantics. Thus, the prefix of imbruttire cannot be analyzed as the negative prefix that attaches to adjectives (brutto "ugly"). We are dealing, instead, with the in- that Tekavcic (1972) refers to as having "spatial" or "conceptual" value, or simply "intensive" value. Secondly, it should be noted that the "intensive" prefix attaches productively to verbs, while the negative prefix attaches to adjectives, as seen in (20): (20)

(i) in + Verb

(ii) in + Adjective

in + rompere in + mettere in + porre

"to break into" "to put in" "to impose"

in + elegante "inelegant" in + adatto "unsuitable" in + compiuto "unfinished"

The fact that the prefix in- with the "intensive" value attaches to verbs, not to adjectives, constitutes a problem for the ternary structure seen above in (18), where both the prefix and the suffix were attached simultaneously to an adjective. 8 This problem can be resolved, however, if we analyze the items in question, in terms of a binary structure in which the prefix is attached to a verb, rather than to an adjective: (21)

V

Pre

V

A

Suf

In this analysis, first a verb form is created and then a prefix is added to this form. The obvious difficulty with such an analysis is the fact that in the cases in question [X + Suf] refers to non existent words (i.e. *bruttire, *rozzare, *vecchiare. It should be remembered, though, that there are productive derivational processes in Italian that form precisely [A + Suf]v· The forms we are examining here, which have this structure, are, in fact, possible (though non existent) words in Italian. We can thus hypothesize that the parasynthetics of the traditional literature are actually formed in two steps: first, suffixation creates a possible, though not necessarily existent word, and second, prefixation generates the rest of the form, as illustrated in (22): (22)

1.

[

2.

[[

3.

[Pre + [ [

]A ] + Suf]v ] + Suf]v]v

[brutto] a [ [brutto] A +ire] ν [in + [ [ b r u t t o ] A + i r e ] v ] v

150

Generative

morphology

The proposal in (22), which is at this point still tentative, must be examined more closely. First of all, it should be clear that no matter how derivations of type A-»V or N - » V are treated, the verbal suffix is not an inflectional morpheme. It is a derivational suffix of the type seen in (19), since the rules in question create "new" words, not simply inflected forms. Secondly, the binary analysis implies that the base of a W F R can also be a possible but non existent word. This possibility, as we have already mentioned, was proposed by Allen (1978) under the name of "overgenerating morphology", as well as by Halle (1973), Booij (1977), Corbin (1980) and others. The need for an overgenerating morphology has already been demonstrated in relation to other problems, and the analysis proposed for the parasynthetics is only one more argument that can be added to the others. The weakening of Aronoff's original hypothesis according to which the base of a W F R must be an existing word such that it also allows the base to be a possible but not existing word, is therefore not an ad hoc or unmotivated modification needed only to handle parasynthetics. Instead, it is the result of a convergence of a variety of descriptive and formal necessities. 2.2. The Suffix -istico Another potential counterexample to the BBH arises in cases in which two independent affixes cluster in such a way that they seem to form a single affix. The Italian suffixes -ista and -ico illustrate this problem, as can be seen in (23): (23)

(i) umorista

"humorist"

(ii) metrico

"metric"

militarista

"militarist"

tisico

"phthisical"

corista

"chorister"

elettrico

"electric"

fiorista

"florist"

satirico

"satirical"

(iii) semplicistico

"simplistic"

(iv) calcistico

"rel. to football"

nudistico

"nudistic"

pugilistico

"pugilistic"

individualistico

"individualistic"

manualistico

"rel. to h a n d b o o k "

naturalistico

"naturalistic"

intellettualistico

"intellectualistic"

The lists in (23i) and (23ii) show that the suffixes -ista and -ico are two independently occurring suffixes (i.e., they attach to "simple" bases: umore + ista, metro + ico). Consider now the words in (23iii). They can be analyzed as consisting of a base (semplice "simple", nudo "nude", individuate "individual", naturale "natural"), plus the suffix -ista (semplicista "superficial person", nudista "nudist", individualista "individualist", naturalista "naturalist") plus the suffix -ico. This is not, however, the case for the words in (23iv), since the intermediate phase, " W + ista", is not an occurring word (i.e., *calcista, *pugilista, *manualista, *intellettualista are all non occurring words). N o t e that in this case we cannot solve the problem by saying that in Italian the two suffixes are reanalyzed as a single suffix, since we saw in (23i) and (23ii) that -ista and -ico are, in fact,

Constraining

word formation

rules

151

separately existing suffixes. If we a d o p t a strong version of W o r d - B a s e d M o r p h o l o g y (i.e., W F R take as a base only existing words), we would be obliged to conclude that the words in (23iv) are formed by the a t t a c h m e n t of two suffixes at the same time. An overgenerating m o r p h o l o g y , however, would solve this p r o b l e m in a way a n a l o g o u s to that seen above in the case of parasynthetic constructions. A word such as calcistico would, thus, be derived in two steps, according to the "one affix, one rule hypothesis": calcio-+calcista-^>calcistico. In order for this derivation to be possible, however, one must allow b o t h the o u t p u t of the first rule (the one which attaches -ista) and the input of the second rule (the one which attaches -ico) to be a possible but not existing word (i.e. calcista). In conclusion, the BBH has been shown t o be valid, provided t h a t the m o r p h o l o g y is an overgenerating m o r p h o l o g y . In any case it should be noted that even if there are some problematic cases in derivation, the area in which the hypothesis was originally proposed, there are, o n the contrary, n o conterexamples in c o m p o u n d i n g . C o m p o u n d s always exhibit a binary structure, as illustrated in the examples below: (24)

library

committee

meeting

shoe

roast

sandwich

platter

beef

department

sales

manager

Finally, it has been d e m o n s t r a t e d that by modifying the BBH we can account more adequately for the observed data. T h a t is, we must (a) modify the definition of the BBH in such a way t h a t it can also apply to c o m p o u n d s , and (b) modify the content of the W o r d Based Hypothesis in such a way that W F R ' s can produce and can apply to possible but n o n existent words, in accordance with an "overgenerating m o r p h o l o g y " .

3. The ordering hypothesis In the 1960's, a large b o d y of evidence was collected in favor of the possibility of establishing an order of application a m o n g b o t h phonological rules a n d transformations. T h e first a t t e m p t t o order morphological rules is that of C h a p i n (1967). W e will not discuss this proposal, however, since it is based on an assumption t h a t is no longer accepted today, a n d that is, that W F R ' s are a type of transformational rules. C h a p i n (1970) proposed a modification of the original proposal which we will consider

152

Generative

morphology

instead, b u t first it is necessary to clarify the distinction between intrinsic and extrinsic ordering. Intrinsic ordering obtains between two rules A and Β when the application of Β is logically dependent on the prior application of A. Suppose, for example, that there are the following two rules in a language: (25)

(i) [ [

]X + Suf]Y

(Rule A)

(ϋ) [ [

]Y + Suf]z

(Rule B)

In this case, it is obvious that Rule A must precede Rule B. Since Rule A generates an output, Y, which serves as the input for Rule B. O n the contrary, if Rule Β were to apply first, the structural analysis of Rule A would, subsequently, not be met and Rule A would fail t o apply. T h a t is, Rule Β generates an o u t p u t , Z, which cannot serve as the input for Rule A. Thus, the rules in (25) must apply in the order given, as can be illustrated with the following example: (26)

(i) [ [ o r g a n i z e ] y + a t i o n ] N (ii) [ [ o r g a n i z a t i o n ^ + a l ] A

Intrinsic ordering contributes to restricting the possible linear order of affixes in multiply complex words, though, by itself, it does not impose a d e q u a t e restrictions on possible outputs. Extrinsic ordering, on the other hand, is the ordering imposed on two rules, A and B, which logically could apply either in the order Α - B or in the order B - Α . Consider the following two rules: (27)

(i) [ [ (ii) [ [

]X + Suf]Y ]Y + S u f ] x

(Rule A) (Rule B)

In this case, Rule A may precede Rule B, but Rule Β m a y also precede Rule A. T h a t is, the o u t p u t of Rule A may be the input of Rule B, but the opposite is also true. This case can be illustrated with the following example: (28)

(i) [ [ (ϋ) [ [

] A + ism]N ] N + al]A

T h e two rules in (28) can theoretically apply in the order (i-ii) as well in the opposite order, (ii-i). In fact, however, the rules apply in only one order, (ii-i), since there are words of the form X-al-ism but not of the form *X-ism-al. O n e simple way of obtaining the desided o u t p u t is to stipulate an arbitrary extrinsic ordering between the two rules, which specifies the a p p r o p r i a t e order. Chapin's (1970) ordering hypothesis was proposed to account for relations a m o n g adjacent affixes, such as those just seen, as well as for the

Constraining

word formation

153

rules

distribution of "rival" afixes. In both cases, however, the hypothesis presents some problems. First of all, Chapin observes that a straight linear order of affixation rules is not possible. Consider two rules, A and B, such that the output of Β cannot be the input of A, and a third rule, C, whose output cannot be the input of B. According to the ordering hypothesis, C must be ordered after B. Furthermore, by the transitivity property, C should never be able to precede A. This prediction turns out to be false, however, as can be seen in the following example: (29)

(i) -ation (ii) -al (iii) -ize

precedes precedes precedes

-al -ize -ation

organizational industrialize industrialization

In (29iii), we see that -ize precedes -ation, while a consistent transitive ordering would predict that -ize should follow -ation. Given that a simple linear order is not possible, Chapin proposes that at least certain subgroups of WFR's should be ordered cyclically, according to what he calls the Epicycle Hypothesis (EH). This hypothesis requires that all W F R ' s work in linear order and that those WFR's which apply cyclically, such as those in (29), be placed adjacent to each other, as can be illustrated with the following diagram (Chapin 1970:61): (30)

A

Β

C

D

Ε

F

G

According to the EH, cyclic ordering is possible only among groups of alphabetically adjacent rules, such that & C, F- G, C - D - E , Α-B—C-D, etc. are possible cyclic groups but A - E , B - D - G , A - B - D - E are not. A theory of derivation which incorporates the EH would be more powerful than one in which r a n d o m sets of derivational rules apply cyclically. Even though it would be appealing to incorporate the E H into morphological theory, this move does not seem possible since there are too many counterexamples to the EH, noted mainly by AronofT. In the first place, the EH reveals itself inconsistent. Aronoff (1976), following Martin (1972), assumes that words of the form X-ory are derived from words of the form X-ion. Therefore, -ory must follow -ation in the cyclic group seen in (29). On the other hand, words of the form X-atorial (cf. dedicatorial) show that -al must follow -ory. This means that -ory should at the same time precede and follow a cyclic group. Note that it is not possible to place -ory in the sequence -ize, -ation, -ory, -al because in this way we would violate the convention imposed by the EH according to which we cannot place an affix between affixes which form a cycle. A second criticism of the EH has to d o with rival affixes. As far as the distribution of the "rival" affixes, such as -ation and -ment, is concerned, Aronoff observes that the ordering hypothesis runs into difficulty because the distribution of these affixes is not completely complementary. There

154

Generative

morphology

are in fact coexisting forms, that is, cases where different affixes can attach to the same stem, as can be seen in (31): (31)

consolement/consolation committal/commitment approval/approbation juncture/junction

In the third place, there is a type of condition which cannot be expressed by ordering at all, namely positive conditions. The suffix -ment, for example, attaches productively to verbs which begin with -en or be- (cf. enrichment, bewilderment) and Aronoff observes that there is no way for ordering to express the fact that a given WFR is more productive with certain bases than with others. It is possible, of course, to order the -ment rule after the be- and en- rules. In this way, the -ment rule will operate correctly, but there is no way to encode the fact that the -ment rule actually prefers words in be- and en-, unless recourse is made to some type of abstract feature. According to Aronoff, therefore, even if one accepts the Ordering Hypothesis, is not possible to dispense with abstract features altogether, not only for the reasons just seen (i.e. productivity), but also because there are other rules which must refer to abstract features mentioned in their base. The suffix -ity, for example, attaches only to words with the abstract feature [ + Latinate], This means that the use of abstract features is independent of ordering, or, in other words, that such conditions would be needed even in a theory involving ordering. The answers offered in relation to the problems raised in this section are of different types. In relation to the X-ism-al problem, for example, Siegel claimed that the problem can be easily solved if we assume that -al is a " +affix" and -ism is a " # affix", automatically blocking an output such as X-ism-al. Aronoff, on the other hand, claimed that it is necessary to establish specific conditions on word formation. Specifically, in relation to the problem in (29), if a condition is posited to the effect that -al cannot be the input to -ation and that -ize cannot be the input to -al, we can nevertheless form a word such as industrialization, since nothing prevents -ize from being the input to -ation. What remains today of such proposals is the claim that WFR's cannot be extrinsically ordered. Only blocks of rules may be ordered, in accordance with Siegel's proposal, a position we have explored in the preceding chapters.

4. The No Phrase Constraint As we have seen in Chapter V, the N o Phrase Constraint (NPC) 1 0 is a consequence of the original formulation of the Word Based Hypothesis. This notion has been further elaborated in Roeper and Siegel (1978) and

Constraining

word formation

rules

155

Allen (1978) a n d its m e a n i n g is very simple: a W F R can take as its base only m a j o r lexical categories (Ν, A, V) but not p h r a s e s ( N P , A P , VP) or sentences. Allen (1978:236) observes, however, t h a t there a r e suffixes t h a t a t t a c h t o " n o n m o r p h o l o g i c a l " units such as the following: (32)

(i)

-ness

(ii)

black and blueness every dayness fedupness

-ish at homeish stand-offish b l o o d and thunderish

(iii)

-y open-airy d o w n towny bread and buttery

In these cases, t h o u g h , t h e affixes in question " a t t a c h t o t h e p h r a s e - t y p e of the lexical c o n s t i t u e n t to which they n o r m a l l y a t t a c h " (Allen, 1978:238). T h u s , -ness a t t a c h e s t o adjectives a n d , in (32i), to adjective phrases; -ish a n d -y a t t a c h to n o u n s a n d , in (32ii-iii), t o n o u n phrases. Since in such cases the suffixes are in fact a t t a c h e d to the entire p h r a s e a n d not just to t h e second w o r d , t h e correct structures for t h e type of items in (32) are those seen in (33i), a n d n o t those in (33ii): (33)

(i) [ [ b l a c k a n d b l u e ] A n e s s ] ν [ [ b l o o d and t h u n g e r j ^ ish]^ (ii) *[black]A and [blueness]N * [ b l o o d ] N and [thunderish]A

T h e structures in (33i), however, seem to violate the N P C . In r e s p o n s e to this p r o b l e m , Allen argues convincingly t h a t such items are n o t , in fact, violations because t h e phrases in question, while syntactically regular, a r e not semantically regular, in the sense t h a t they are n o t c o m p o s i t i o n a l in m e a n i n g . In the case of fully c o m p o s i t i o n a l phrases, on the c o n t r a r y , affixation is not possible: (34)

(i) *[sad a n d ugly] ness * [black a n d p u r p l e ] ness (ii) * [ a t c h u r c h ] ish * [ s t a n d o u t ] ish (iii) * [ o p e n sea] y *[milk a n d b u t t e r ] y

Allen's conclusion is therefore t h a t those p h r a s e s t o which suffixes c a n be attached are "lexicalized or frozen phrases, mini idioms which m u s t be listed in the p e r m a n e n t lexicon with their c o m p o s i t i o n a l meanings, just like sentential a n d v e r b s - p h r a s e i d i o m s " (p. 238). Allen calls t h e suffixes t h a t can a t t a c h to lexicalized phrases "stretchable" suffixes a n d notes t h a t they are all Level II suffixes a n d that, in fact, very few suffixes are stretchable. In o t h e r words, t h e t w o m a i n p o i n t s are that: (a) only lexicalized p h r a s e s

156

Generative

morphology

may constitute the base of a W F R , and (b) only a few suffixes are stretchable. With these two points in mind, let us now briefly examine several counterexamples to the N P C cited in the literature. As we have already mentioned, according Bauer (1980:75) words such as up-to-date-ness, cut-downess, with-itness constitute a challenge to the N P C . But up-to-date, cut down, with it must all be considered lexicalized phrases. It can be seen, in fact that the same affixation process cannot apply to other non-lexicalized phrases with the same structure: (35)

*up to here ness *cut off ness *without it ness

In Italian, too, one can find cases where an affix can be attached to a non morphological unit: (36)

(i) guerrafondaio "war monger" (from guerra a fondo "total war") dirimpettaio "person living just opposite" ( f r o m dirimpetto "opposite") (ii) pressapochismo "lit. approximately + ism" (from press'a poco "approximately") radiosomaggismo "radiant May + ism" (from radioso maggio "radiant May") (iii) doppio lavorista "lit. double work + ist" (from doppio lavoro "double work") centometrista "hundred meter runner" (from cento metri "hundred meters")

As in the English cases seen above, it can be demonstrated that in the Italian examples the suffix attaches to the entire phrase and not just to the last word. F o r example, -aio attaches to nouns and not to adjectives or to adverbs, and thus *fondaio, from fondo "deep" is not a possible word. All the phrases in (36) must be seen as lexicalized phrases which are stored in the lexicon with their idiosyncratic meanings (e.g. cento metri, which, in this case, refers to a specific athletic feat). Furthermore, as in English, the possibility of an affix taking a phrase as its base is very restricted. In particular, in (36) we see that this process is limited to three suffixes, -ista, -ismo and -aio, and, in fact, we could reduce this list to the first two suffixes, since -aio is no longer productive in Italian. 1 1 What the examples in both English and Italian show is that there are strong restrictions on the application of WFR's to bases larger than a simple word. Specifically, this process seems to be limited to lexicalized phrases, a point which is relevant for the organization of the lexicon, since it requires that such phrases be stored in the lexicon, in addition to the usual words.

5. Blocking Another constraint on word formation is referred to as the "blocking principle", originally proposed by Aronoff (1976) and subsequently main-

Constraining

word formation

rules

157

tained by others such as Allen (1978) and Z w a n e n b u r g (1981). 12 In the following sections, we will show, however, that blocking is not really a general principle of lexical organization but rather the expression of a tendency t o w a r d s economy in the lexicon. Before discussing blocking itself, we will first show how Aronoff associates this n o t i o n with that of productivity, and second, how there are implicitly in Aronoff's p r o p o s a l two different notions of blocking, t h o u g h the distinction is not m a d e explicit and has lead to a certain a m o u n t of misunderstanding. 1 3 5.1. Productivity The productivity of a W F R cannot be identified simply on the basis of the frequency with which it applies. It is also necessary to t a k e into consideration the morphological restrictions on the w o r d s that can be t a k e n as the base of a given W F R (e.g., in English φ ness is more productive that + ity for forming n o u n s f r o m bases in -Xive, while -ity is more productive than φ ness with bases in Xile). Thus, it is not possible to classify W F R ' s as productive or not productive, but rather it is necessary to classify rules in terms of relative productivity with respect to other rules. While it seems possible to speak of productivity in a general sense when refering to rules that apply to a wide variety of bases, Aronoff points out that there are, in fact, very few W F R ' s that d o not have morphological restrictions on their base (e.g. -able). The possibility of applying to a wide variety of bases does not, however, necessarily m a k e a rule productive. F o r example, the suffix -ous seems not to have restrictions on its base, b u t it is clearly not what one would call a productive rule in English (cf. Aronoff, 1976:62). T o show how productivity is the result of an interaction of factors, Aronoff examines the two suffixes +ity a n d φ ness, when attached to the subclass of adjectives of the form Xous. Both -ity and # ness attach to the same category of base (i.e. adjectives) and yield the same category in o u t p u t (i.e. nouns) which is, furthermore, subcategorized in the same way (i.e. [ + a b s t r a c t ] ) . These facts m a k e it possible to directly c o m p a r e the productivity of the two rules in question. W h a t we find is that in adjectives of the structure Xous, φ ness is more productive since a t t a c h m e n t of +ity in such cases generally produces impossible or only marginally acceptable forms (cf. famousness vs. *famosity, dubiousness vs. ?dubiosity). An i m p o r t a n t factor in favor of the productivity of a given rule, in this case that of Φ ness attachment, is its semantic coherence and the intuitions of speakers in this regard. As Aronoff points out, "the surer one is of what a word will mean, the more likely one is to use it" (p. 39). Analogously, the phonological effects of a given affix on its base contribute to the intuitions of speakers a b o u t the rule and its productivity. In the case of Φ ness and + ity, in fact, the former does not produce any modification of the base, while the latter does, in particular, a stress shift (e.g. curious vs. curiosity), a difference encoded by the different types of boundaries ( " # " vs. " + ") associated with the two suffixes.

158

Generative

morphology

Another factor related to the (reduced) productivity of + ity is the fact that it sometimes requires the truncation of the preceding morpheme -ous (e.g. simultaneous^simultaneity vs. *simultaneosity). The truncation rule, given in (37), is itself not completely regular in that it does not always apply when its structural description is met (cf. Aronoff, 1976:40): (37)

Truncation

of -ous

os->0/

+ ity

The only regularity found with this rule is the fact that in the class of words of the form XVcious, the application of the rule depends on the vowel (V) that precedes -cious. T h a t is, it applies when the vowel is a or o, but not when it is e, as seen in (38) (cf. Aronoff, 1976:41): (38)

Xacious mordacious Xocious precocious

Xacity mordacity Xocity precocity

*Xaciosity *mordaciosity *X'ociosity *precociosity

Xecious specious

*Xecity *specity

Xeciosity speciosity

In all other cases, the rule seems to be governed lexically. This fact, too, is relevant to the discussion of productivity since speakers are often unsure of whether or not truncation applies, and in order to avoid mistakes, prefer the rule that does not require phonological changes, # ness attachment. Thus, in addition to factors of frequency and generality of application of a given W F R to different bases, the discussion of productivity must take into consideration the factors of semantic and phonological coherence, and the effects these may have on a speaker's intuitions. When a given rule is productive, the word formed by this rule does not have to be listed in the lexicon, but when there is any irregularity involved, the forms with information about their particular properties must be stored in the lexicon. 5.2. Blocking and the Blocking Rule In the case that words must be listed in the lexicon, the phenomenon of blocking can be observed: (39)

Blocking is the nonoccurrence of one form due to the simple existence of another (Aronoff, 1976:41).

As an illustration of this phenomenon, consider the following data given by Aronoff:

Constraining (40)

word formation Nominal

rules

159

Xous

-ity

-ness

various curious glorious furious specious precious gracious spacious

variety curiosity *gloriosity *furiosity speciosity preciosity *graciosity *spaciosity

vanousness curiousness gloriousness furiousness speciousness preciousness graciousness spaciousness

These examples show that if there exists an abstract nominal which is the base of an adjective of the form Xous, it is not possible to derive from this adjective another n o u n in -ity (cf. (40iii-iv-vii-viii)). The existence of the base noun "blocks" the derivation with -ity. If there is n o underlying abstract nominal, however, derivation with -ity is possible. It should be noted that blocking affects derivation with -ity, which is a non productive suffix, but not with -ness, a productive suffix. That is, only the lexicon is subject to blocking, and, according to Aronoff, the output of productive, regular rules is not stored in the lexicon. 14 Furthermore, still according to Aronoff, the lexicon is organized by stems, and for each stem there is one slot for each canonic meaning. In the case of glorious, for example, the W F R that attaches -ity would form gloriosity, but because of the non productivity of this rule, gloriosity would have to be stored in the lexicon. In the lexicon, however, it would find the slot for its canonic meaning (abstract nominal) already occupied by glory, and as a result, it would not be listed there. Extrapolating from Aronoff's proposal, we can describe the type of blocking just seen in the following terms: (41)

In a linear sequence

] χ ι + S u f l ] Y + Suf2]x2

X2 is blocked if XI and X2 are of the same syntactic category, have the same meaning and the same stem, and if Suf2 is not a productive suffix. This formulation represents a possible interpretation of the notion of blocking. 1 5 It should be noted that (41) involves three sets of labeled brackets; those of the blocking item, the first derived word and the second derived word, that is, the blocked item. On another occasion, however, Aronoff provides a somewhat different definition of blocking (pp. 55-6): (42)

An affix which is productive with a given morphological class will thus block the attachment of rival affixes to that class. [ . . . ] For example, we noted that #ness does not attach to bases of the form Xate, Xant, and Xent. This however, is merely a result of the

160

Generative

morphology

fact that the rival affix +cy does attach productively to these classes [ . . . ] . The productivity of +cy thus blocks the application of #ness in these cases. In this context, AronofT occasionally uses the expression "blocking rule", a term we will adopt here in order to distinguish systematically between the two uses, which, in fact, represent different phenomena. The notion "blocking rule" can be defined in an analogous way to that given for "blocking" in (41), as follows: (43)

In a non linear (i.e. paradigmatic) sequence

] x + Sufl]Y + Suf2] Y + Sufn] Y

] x + *Suf2] and ] x + *Sufn]y are blocked if Sufi is a productive suffix for the class of words ]χ. What the formulations in (41) and (43) show is that there are several basic differences between blocking and the blocking rule 16 : (a) blocking involves three sets of labeled brackets in a linear sequence, while the blocking rule involves two sets in linear order and others arranged paradigmatically; (b) in the case of blocking, the blocking item is an underived word, while in the case of the blocking rule it is a suffix; (c) in the case of blocking the relation between the blocking item and the blocked item is one between two forms listed in the lexicon (one because it is underived and the other because it is derived by a non productive WFR), while in the case of the blocking rule, the blocking item is a productive suffix. With these differences in mind, we will now examine some data in French.

5.2.J. Blocking and the Blocking Rule in French Let us first consider a case of blocking similar to that seen above in English: (44)

]Ni gloire espace harmonie * * *

-eux] A glorieux spacieux harmonieux curieux precieux anxieux

-ite]N2 * *

curiosite preciosite anxiete

This pattern is fairly common in French. It should be noted that according to the definition of blocking given above, it is necessary that both N1 and N2 be associated with the same stem. If this is not the case, the formation of N2 is not blocked, as can be seen in (45):

Constraining (45)

word formation

161

rules

(poison) veneneux venenosite "poison - poisonous - poisonousness"" (bois) ligneux lignosite "wood - woody - woodiness" (eau) acqueux acquosite "water - watery - wateriness"

It is also important to consider the function of the first suffix. In the case in question, -eux selects as its base a noun, but it does not distinguish between [ — abstract] and [ + a b s t r a c t ] nouns. This fact has the following consequence for blocking: if N1 is [ — abstract], then blocking does not apply, as illustrated below: (46)

monstre

monstrue^

monstruosite

"monster - m o n s t r o u s -

cal

calleux

callosite

"callus - callous - callousness"

monstrosity"

muscle

musculeux

muscolosite

"muscle - muscular -

eponge

spongieux

spongiosite

"sponge - s p o n g y - sponginess"

muscularness"

nceud

noueux

nodosite

"knot - knotty - knottiness"

While the pattern just presented is quite general, it is not always the case, however, that (a) if N1 is [ - a b s t r a c t ] there is automatically derivation giving N2, and (b) if N1 is [ + a b s t r a c t ] blocking always excludes N2, as seen in (47i) and (47ii) respectively: (47)

(i) laine

laineux

*

"wool - w o o l y "

gomme

gommeux *

"rubber - rubbery"

vent

venteux

*

"wind - windy"

(ii) religion

r^ligieux

r^ligiosite

"religion - religious -

vertueux

virtuosite

"virtue - virtuous -

vertu

religiousness"

virtuousness"

It is clear that religiosite and virtuosite have different meanings from religion and vertu, but one of the problems with blocking is precisely that of defining formally how much "semantic distance" there must be between XI and X2 in order for X2 to be able to be formed. Let us now examine the phenomenon we have referred to as the Blocking Rule, in light of an analysis proposed by Zwanenburg (1981) for the French suffixes -isme and -iste.17 Zwanenburg intentionally chose two productive suffixes since he maintains that blocking 1 8 also occurs with productive processes. What must be explained in this case is the fact that we d o not find certain words in -isme while their counterparts in -iste d o exist, and viceversa. First of all, it should be noted that words in -isme have two principal paraphrases: (a) "doctrine, attitude, tendency or current" in relation to the base, and (b) "quality" of that which is designated by the base, as illustrated in (48i) and (48ii), respectively (cf. Zwanenburg (1981: 66ff.)); (48)

(i) calvinisme ego'isme impressionisme existentialisme conformisme

"Calvinism" "egotism" "impressionism" "existentialism" "conformism"

162

Generative

morphology

(ii) amateurisme analphabetisme bilinguisme cretinisme athletisme

"amateurism" "illiteracy" "bilingualism" "idiocy" "athletism"

Z w a n e n b u r g points out that only those w o r d s with the first p a r a p h r a s e can have a corresponding f o r m in -iste; words with the second paraphrase cannot: (49)

(i) calviniste ego'iste impressioniste existentialiste conformiste (ii) *amateuriste *analphabetiste *bilinguiste *cretiniste *athletiste

"Calvinist" "egotist" "impressionist" "existentialist" "conformist"

The non existence of the words in (49ii) is explained by the fact that there is already a n o u n or an adjective referring to a person with the relevant characteristics (e.g. amateur "amateur", analphabete "illiterate", bilingue "bilingual" cretin "idiot", athlete "athlete"). These words are precisely the base forms for the words in -isme. It is not this fact, however, that Z w a n e n b u r g bases his analysis of blocking on, since it is handled, instead, by the W F R ' s themselves (cf. p. 68). Z w a n e n b u r g bases his proposal regarding blocking on a n o t h e r set of data, in particular, those words in -isme with the first p a r a p h r a s e given above, and the absence of corresponding words in -iste: (50)

(i) "doctrine": (ii) "illness": (iii) "sport":

academisme epicurism alcoolisme daltonisme nautisme hippisme

'academism" 'Epicurianism" 'alcholism" "Daltonism" 'seamanship" 'horsemanship"

Aside f r o m the examples in (50iii), which are fairly u n c o m m o n exceptions to the general pattern, the absence of words in -iste corresponding to the words in -isme in (50i) and (50ii) is accounted for by Z w a n e n b u r g by what he calls "le principe d u blocage", the Blocking Rule, in the present terminology. W h a t we find instead of these words, are other words that refer to a person with the relevant characteristics, but which are formed by the addition of a different suffix (e.g. academique "academic", epicurien

Constraining

word formation

rules

163

"Epicurian", alcoolise "alcholic", daltonien "daltonian"). Z w a n e n b u r g ' s conclusion is that the non existent words are blocked by existing w o r d s with different suffixes. Both Blocking and the Blocking Rule encounter a n u m b e r of difficulties, the most i m p o r t a n t of which are the following: (1) T h e notion of "different m e a n i n g " is crucial for b o t h Blocking and Blocking Rule but an a d e q u a t e definition has not yet been proposed. (2) It is not always clear whether a given suffix is productive or n o t . 1 9 (3) T h e r e is disagreement between those w h o propose that Blocking applies only to n o n productive processes and those who propose that is valid also for productive processes (thus requiring a modification of the original definition of Blocking). (4) T h e distinction m a d e above between Blocking and the Blocking Rule has not been clarified in sufficiently explicit terms, and it should be noted that, in addition to the differences already mentioned, there is a n o t h e r f u n d a m e n t a l difference between these two processes. T h a t is, the Blocking Rule (or the notion of "rival suffixes") introduces a contradiction into the system in the sense that it requires suffixes to " k n o w " their rival suffixes, or, in other words, that one W F R " k n o w s " a n o t h e r W F R , a possibility that is not considered in a model such as Aronoff's. (5) According to Siegel (1977), t h e claim on which the entire n o t i o n is based, that w o r d s formed by productive W F R ' s are not entered in the lexicon, is not valid. If it were valid, this would m e a n that we should never find w o r d s with m o r e t h a n one productive affix, t h o u g h such w o r d s are, in fact, quite c o m m o n (e.g. thanklessness, formlessness, sleeplessness). W h a t the various analyses of Blocking and the Blocking Rule seem to indicate is that they cannot, in fact, be considered formal principles restricting the W F R ' s , but rather the expression of fairly general tendencies, a pattern in the lexicon, the ultimate scope of which a p p e a r s to be economy.

6. Summary A m o n g the several constraints discussed in this chapter, we have seen that only the O r d e r i n g Hypothesis must be rejected altogether. At the present state of our knowledge, there seems to be n o way to impose an extrinsic order on W F R ' s . As far as the other constraints are concerned, we have seen: (a) T h e Unitary Base Hypothesis in its original formulation is t o o strong; a modified version of the U B H has been proposed that, using syntactic category features instead of syntactic labels, can handle a wider set of data. There are, however, some problems that remain to be solved, a m o n g which the most i m p o r t a n t is the one having to d o with the deciding whether a suffix is actually " t w o " suffixes, (b) The Binary Branching Hypothesis is, basically, a correct hypothesis. The large majority of the morphological structures built u p by W F R ' s are of a binary type. A potential counter-

164

Generative

morphology

example, the parasynthetic verbs in Italian, was examined and it was shown that these constructions, too, can be analyzed in bynary terms. In this case, however, one is forced to assume that WFR's can have "possible" but not existent words both as input and output, an assumption that implies a slight modification of the Word Based Hypothesis. We arrived at the same conclusions also in the analysis of some words in Italian where the sequence X-Sufl-Suf2 exists but where the sequence X-Sufl does not exist, (c) The N o Phrase Constraint seems to have some exceptions, but it is clear that there are strong restrictions on the application of WFR's to bases larger than words. In the large majority of cases, W F R ' s can apply to units larger than words, i.e. phrases, only when the phrases in question are lexicalized. This is consistent with the organization of the lexical component proposed in this book, in fact, since lexicalized phrases are assumed to be stored in the lexicon and not generated by Phrase Structure Rules, (d) Finally, Blocking was shown not to be a true principle of the lexicon, but only a general tendency towards the limitation of the formation of synonyms. We have seen that the notion of blocking can be fruitfully divided into two different notions: blocking, where the blocking effect is of a sintagmatic nature, and the blocking rule, where the blocking effect is of paradigmatic nature (i.e. "rival" affixes). In both cases, however, the two proposed principles can express a tendency of the lexicon towards "economy" but not a formal constraint on WFR's.

Notes 1.

This point will be discussed further in section 1.5.

2.

F o r example, Bauer ( 1 9 8 3 : 2 1 6 fT.) lists a set of nine English prefixes which appear to

violate the U B H (fore-,

re-, in-, mid-, ex-, circum-,

least six of these prefixes (fore-,

counter-,

re-, in-, mid-, circum-,

dis-, co-, inter-,

sub-) but for at

dis-) the violation is more apparent

than real because, as Bauer points out, a diachronic change has taken place (e.g. fore-

is

synchronically productive, if at all, only w h e n it is attached to nouns). 3.

In certain cases, it might seem that -ata

stupido^stupidata

can attach to adjectives as well (cf.

"stupid - act of a stupid person"). In such cases, however, the adjective in

question has the function of a n o u n (i.e. stupido = "stupid person"). Cf. Scalise (1983). 4.

T h e o n e affix, o n e rule hypothesis has been formulated as a "maximal nesting

condition" by Kiparsky (1973). See Strauss (1982) for a discussion of this condition. 5.

F o r a discussion of apparent c o u n t e r e x a m p l e s in Dutch, cf. B o o i j (1977:32).

6.

Cf. C o r b i n (1980), w h o discusses several definitions of parasynthetics.

7.

There are also n o m i n a l parasynthetics, formed with an adjectival or nominal base

(cf. Corbin, 1980). In this section we will be concerned only with verbal parasynthetics. 8.

N o t e that this problem is not restricted to in-, but arises also with the other prefixes

involved in parasynthetic formation: cf. ri- in ri-salire

"to g o up again", di-mettere

"remove",

s-legare

"untie", a(c)-conciare

"adjust".

9.

Booij (1977) points out s o m e cases in which t w o affixes, A and B, can occur in both

orders. F o r example, in D u t c h there are w o r d s such as roodachtigheid wijsheidachtig

"wisdom-like" or goddeloosheid

"reddishness" vs.

"godlessness" vs. goddelijkheidsloos

"without

divinity". 10.

We

constraint.

borrow

this term

from

Botha

(1981) w h o ,

actually,

argues

against

this

Constraining

word formation

rules

165

11. Botha (1981) and de Villiers (1979) provide additional examples in Africaans that seem to violate the N P C . They cite such structures as N P + p o s t p o s i t i o n (e.g. kerk-toe-stapery "church t o walk -ing"), Preposition + N P (e.g. agter-die-muur-rook-er "behind the wall smoke -er"), Adv + P P (e.g. laat-in-die-bed-kom-ery "late in the bed get -ing"), N P + P P (e.g. boek-in-die-bed-lees-ery "book in the bed reading"), and P P + P P (e.g. met-die-hand-in-diesak-staan-ery "with the h a n d in the pocket stand -ing"). While this process seems t o be fairly productive in Afrikaans, it is once again limited to a small g r o u p of suffixes. It should be noted, however, that the status of the words formed by such affixation processes is not clear. T h a t is, it is not clear whether they are completely n o r m a l formations or somewhat m a r k e d ones. 12. Actually, only Z w a n e n b u r g claims that blocking is a "principle" of the lexicon. Allen reformulates it as a "Filtering F u n c t i o n " and Aronoff does not offer any formal definition of it. 13. Actually, in Scalise et al. (1983) a third possibility is mentioned, t h o u g h we will not go into this here. In addition, we will not discuss another proposal m a d e by Allen (1978) for reformulating Blocking (cf. Scalise et al. (1983) for a critique of this proposal). 14. It should be noted that Aronoff uses the word "lexicon" with systematic ambiguity to refer to both the lexical c o m p o n e n t and the list of idiosyncratic forms of a language. In the discussion of Blocking, however, it is only the second meaning that is intended. 15. This is the same interpretation offered by Siegel (1977:11), t h o u g h Siegel uses the term root rather than stem. 16. F o r other differences, cf. Scalise et al. (1983). 17. W h a t is presented here is only the basic outline of Z w a n e n b u r g ' s analysis. 18. "Blocking Rule" in o u r terminology. 19. Cf. Siegel's (1977:17) discussion of the productivity of the suffixes -ic and -al

Chapter

VIII

Morphology and syntax

In this chapter, we will consider several aspects of the relationship between morphology and syntax. It will be shown, on the one hand, that the rules of morphology and syntax are distinct types of rules, characterized by different formal properties, and on the other hand, that there is a definite interaction between the morphological and syntactic components of the grammar. In the first part of the chapter, we will consider some formal properties of Worf Formation Rules such as locality and the way in which W F R ' s may change the subcategorization frame and/or argument structure of a word. In the second part of the chapter, we will consider two cases of interaction between morphology and syntax, in particular the order of the constituents in compounds and phrases and the point in the grammar in which inflection rules apply.

1. Word formation rules and transformations In a modular approach to the grammar, it makes sense to postulate the existence of a separate "module" when the facts that must be accounted for escape treatment by other already existent modules, and when it can be demonstrated that the proposed module is governed by specific principles. As far as morphology is concerned, it makes sense to postulate the existence of a new module (the morphological component) when it becomes clear that morphological data cannot be handled by other components of the grammar and that Word Formation Rules are governed by specific principles. That morphological data cannot be accounted for by syntactic transformations has been demonstrated in the preceding chapters. The alternative, Word Formation Rules, also leave room, however, for different interpretations about their nature. One can, in fact, accept the existence of WFR's but, at the same time, claim that their nature is transformational and, thus, that they are not governed by principles different from those that govern the transformational rules. 1 An opposing point of view is that WFR's and Transformations (T's) are completely different sets of rules. McCarthy (1979:201), for example, holds this second point of view, and proposed the following principle of Morphological Transformation Prohibition (MTP):

168

Generative

morphology

(1)

All morphological rules are of the form A->B/X, where A, Β and X are (possibly null) strings of elements

According to the MTP, morphological rules are context-sensitive rewriting rules, and thus differ from transformation rules. McCarthy observes that a theory that encompasses the M T P "strongly generates a much smaller class of grammars than a theory without this constraint." Lieber (1980) too, warns that morphological transformations open way to possibilities that presumably never occur in natural languages. 2 The most systematic attempt to differentiate empirically between WFR's and T's was made by Wasow (1977). 3 According to Wasow, the properties that distinguish these two sets of rules are the following:

(a) (b) (c) (d) (e) (0

can change base-generated structures can change syntactic categories are local involve phrasal categories operate before all transformations involve idiosyncratic information

WFR's no yes yes no yes yes

T's yes no no yes no no

Wasow's picture, however, must be corrected and enlarged. It must be corrected because points (a) and (e) can be considered corollaries since they follow by necessity from the position that is assigned to the WFR's in the grammar. In the lexicalist framework, W F R ' s are "lexical" rules, that is, they operate completely within the lexicon and, therefore, they cannot, by definition, change the structures that are generated by the base. Furthermore, they must, by definition, operate before all (syntactic) transformations: the WFR's, in fact, d o not have access to the output of Phrase Structure Rules or to any stage of a syntactic derivation. Wasow's picture must also be enlarged since the considerable amount of work done on morphology in the last few years has provided insights into the functioning and the nature of W F R ' s that were lacking in the late seventies. By excluding points (a) and (e), and adding several other points, we arrive at the result seen in (3):4

(a) (b) (c) (d) (e) (f) (g) (h)

can change syntactic categories can change subcategorization frames are local have binary branching properties involve idiosyncratic information involve phrasal categories are ordered include movement rules

WFR's yes yes yes yes yes no no no

T's no no no no no yes yes yes

Morphology

and

syntax

169

In this chapter, we will only discuss points under (3b) and (3c). Point (3a) was discussed in Chapter VI and points (3d), (3f) and (3g) were discussed in Chapter VII. As for point (3h), the reader is referred to Allen (1978) and Selkirk (1982) who, discussing Roeper and Siegel's (1978) proposal, demonstrate quite convincingly, that WFR's d o not include movement rules. As far as point (3e) is concerned, finally, we will limit ourselves here to the following observation: while the lexicon has traditionally been considered a list of irregularities, and while it is true that W F R ' s exhibit idiosyncrasies, it is also true that all recent studies in lexical morphology have shown a wide domain of regularities also in the lexicon. This way of considering the lexicon has been convincingly summarized by Jackendoff (1975: 668): (4)

We have thus abandoned the standard view that the lexicon is memorized and only syntax is creative. In its place, we have a somewhat flexible theory of linguistic creativity. Both creativity and memorization take place both in the syntactic and the lexical component.

1.1. Locality In the last years, the issue of locality in morphology has received much attention, especially following a proposal put forth by Siegel (1977). In the following three sections, we will consider Siegel's proposal, an alternative proposal made by Williams (1981a) and, finally, an empirical comparison between these two proposals. 1.1.1. Adjacency

Condition

F r o m a formal point of view, a W F R essentially attaches an affix to its base. A question that arises in relation to this operation is the following: what information of the base is accessible to a given W F R ? In other words, what does a W F R "see"? Aronoff (1976:52) discussed the possibility that W F R ' s might be sensitive only to those features associated with the morpheme that is adjacent to the morpheme added by the W F R . This "linear contiguity" hypothesis, which was finally discarded by Aronoff, predicts, for example, that a suffix will be sensitive to the last morpheme of the base, while a prefix will be sensitive to the first morpheme. It thus rules out the possibility, for example, that a suffix can be sensitive to the first morpheme of the base. Siegel (1977) shows convincingly that a "linear" condition is not adequate. Consider the following examples (p. 20): (5)

(i) * [ u n [ # d i s [ # h o n e s t ] ] ] * [ u n [ # d i s [ # loyal] ] ] (ii) * [ u n [ # d i s j o i n t ] ] * [ u n [ # discrete] ]

170

Generative

morphology

(iii) [ u n [ # [distinguish # ] e d ] ] [ u n [ # [discover # ]able] ] At first sight, it appears that the ungrammaticality of (5i) and (5ii) could be accounted for by the linear contiguity of un- and dis-. This possibility is ruled out, however, by the examples in (5iii), where un- is still adjacent to dis-. The difference between (5i) and (5ii), on the one hand, and (5iii), on the other hand, lies in the structure of the constituents. In (5i) and (5ii), dis- is contained in the cycle immediately adjacent to un-, while in (5iii), dis- is contained in a deeper cycle. 5 The specific condition proposed by Siegel (1977) for this case is the following: (6)

Condition on un- Affixation (revised) On- may not attach if dis- is uniquely contained in the cycle adjacent to un- (p. 22).

Siegel also proposes a more general condition on the form and functioning of WFR's, the Adjacency Condition: 6 (7)

Adjacency Condition N o W F R can involve X and Y, where X is an affix, unless Y is uniquely contained in the cycle adjacent to X (p. 23).

In general, we can say that given the following morphological structure, where X and Y are cyclic nodes within the word, an affix (X) has access to Y but not to Z: (8)

]

] Ζ

] Y

X

I 1 ι The structure in (8) is deliberately vague because Adjacency seems to be a general principle, valid for all the rules operating in the morphological component. 7 F o r example, Allen (1978) has proposed that the Adjacency Condition (AdC) also applies to compounds, Dell and Selkirk (1979) have used a type of AdC to account for the Learned Backing Rule in French, and Scalise (1980) has proposed that the AdC also applies to the derivation and inflection of compounds. In these last two cases, (8) must be interpreted as follows: Ζ is the bracket of W O R D 2 , Y is the bracket of the whole compound, and X is a derivational affix (for derivation of compounds) or an inflectional affix (for inflection of compounds). We will give here only an example of the latter case. In Italian, there are many compounds of the structure [V + N ] ^ . A compound such as portacenere "ashtray" is formed by a verb (porta "carry") and a noun (cenere "ash"). The noun is a [ — c o u n t ] noun and,

Morphology

and

syntax

171

therefore, it has no plural. 8 In spoken Italian, however, we find a plural form portaceneri.9 In this case, in fact, the plural inflectional affix does not "see" Z, but only Y, which is [ + c o u n t ] noun: (9)

]

porta cenere]

Z=N

-i]

Y= N

[ — count]

X = Af

[ +count]

In fact, portaceneri means "more than one ashtray"; in other words it is a regular plural form. If -i were attached to the node Z, than portaceneri would mean "cremation urn" instead. The Adjacency Condition has been generally considered an important step in the development of an adequate morphological theory because it limits the number of conceivable rules as well as their functioning. As any other theoretical claim, the AdC must utimately be verified empirically. We will d o this by comparing the AdC to an alternative proposal, the Atom Condition, but first we will outline the claims made by the Atom Condition. 1.1.2. The Atom

Condition

Williams (1981a) has observed that, since the AdC allows morphological rules to make reference only to "material" immediately adjacent to the affix, there are cases in which the AdC does not make correct predictions. As an example, Williams cites those verbs in English that form the past tense not by adding the morpheme -ed, but rather by a process of ablaut (e.g. stand->stood, take->took). In order to prevent -ed from being attached to these verbs, one can mark the base of regular -ed verbs with the feature [-ablaut]: (10)

[-ablaut]

ed

A verb such as understand, on the other hand, will be marked with the feature [ + a b l a u t ] , as illustrated by the following representation: (11)

[[under] [ s t a n d ] v ] v [ + abl]

T o form the past of understand, it would seem that the structure is the one

172

Generative

morphology

given in (12): (12)

[[under] [stand] v ]v + ed [ + abl]

In this structure, -ed is attached to the base verb because the feature [ + ablaut] is not contained in the cycle adjacent to the one that adds the past morpheme, and thus, according to the AdC, is not visible to the rule of past formation. In this case, (10) will not block -ed attachment, but rather will see to it that -ed is added, yielding the incorrect result: * und er st and ed. One way of blocking the incorrect output would be, according to Williams, to assume that the feature [ + a b l a u t ] percolates from the internal V to the external V, as represented in the following diagram: (13)

V

It is precisely to account for such cases that Williams proposed, as an alternative to the Adjacency Condition, what he called the Atom Condition (AtC): (14)

Atom Condition A restriction on the attachment of Af to Y can only refer to features realized on Y

The AtC accounts for the understood verbs by means of the mechanism of percolation: (15)

V

The affix "sees" the feature [ + ablaut] realized (by means of percolation) on the higher V node and thus does not apply, in this case, to understand. Comparing the AdC to the AtC, Williams remarks that the AtC is weaker than the AdC because it requires the mechanism of percolation.

Morphology

and

syntax

173

On the other hand, the AtC is stronger (or at least as strong as the AdC) because it allows reference only to the features realized on the morpheme in head position, while the AdC allows reference only to the last attached morpheme. In the next section, we will compare the empirical validity of the AdC and AtC on the basis of data drawn from Italian. 1 0 1.1.3. A comparison Condition

between

the Adjacency

Condition

and the

Atom

The two conditions on WFR's make a number of predictions, some of which are the same and some of which are different. Let us begin with the cases where both conditions make the same predictions. In a structure such as the one in (16), both the AtC and the AdC predict that the attachment of Suf2 is sensitive to the presence of Sufi, and not to the presence of Z: 1 1 (16)

(i)

Ζ (ii)

[[[Z]

Sufi

Suf2

+Sufl]

+Suf2]

The AdC predicts that Suf2 is sensitive to Sufi because Sufi is the last morpheme attached to the base (Z). Suf2 does not "see" Ζ because it separated from Ζ by an intervening node (the one created by the attachment of Sufi). The AtC predicts the same thing but for a different reason. That is, Suf2 "sees" only the features realized on Y (cf. 16i)) and in the present case, these features percolate to Y from Sufi. The following data confirm the prediction: (17)

(i) Ζ organ *organ *favol favol (ii) odi *odi *lod lod

Sufi izza izza eggia eggia OS OS

evol evol

Suf2 zione mento zione mento itä ezza itä ezza

"organization"

"making u p tales" "odiousness"

"praiseworthiness"

Both (17i) and (17ii) show that Suf2 is sensitive to Sufi. In the first case, -zione attaches to X-izza while -mento attaches to X-eggia. In the second

174

Generative

morphology

case, -itä attaches to X-oso bases and not to X-evole bases, while -ezza attaches to X-evole bases but not to X-oso bases. Another structure in which both the AdC and AtC make the same predictions is the following:

Pre (ii)

[Pre+

Ζ [[Z]

Suf +Suf]Y]

Both the AdC and AtC make the prediction that in a structure such as the one in (18), Pre will be sensitive to Suf. This is the case for the same reasons seen above. In terms of the AdC, Pre "sees" Suf because Suf is "ciclycally" adjacent to Pre. In terms of the AtC, given the two elements Ζ and Suf, Suf is the head and thus the features associated with Suf percolate to Y, the only node "seen" by Pre. This point is clearly illustrated by Italian deverbal adjectives ending in -abile and -ato such as the following: (19)

(i) alienabile domabile controllabile mangiabile osservabile praticabile

"alienable" (ii) abitato adattato "tamable" "controllable" ancorato "edible" preparato "observable" disciplinato "practicable" vendicato

"inhabited" "adapted" "anchored" "prepared" "disciplined" "revenged"

Among the negative prefixes in Italian are the forms in- and dis-. If, in structures such as those in (18), the negative prefixes were not sensitive to the suffix of a word, we would expect to find all of the following combinations:

Morphology

and

syntax

175

N o t all of these possibilities are found, however. In fact, (20iii) must be excluded, as seen in (21):

(21)

(i) in-alienabile in-domabile in-controllabile (iii) *dis-mangiabile *dis-osservabile *dis-praticabile

"inalienable" "indomitable" "uncontrollable"

(ii) in-preparato in-disciplinato in-vendicato (iv) dis-abitato dis-adattato dis-ancorato

"unprepared" "undisciplined" "unrevenged" "uninhabited" "unadapted" "unanchored"

These examples show that, as predicted by the AdC and AtC, prefixation is sensitive to the suffix in certain cases. In this particular case, if the base of the prefix ends with -ato, then it can be negated with both the prefixes in- and dis-; however, if the base ends with -abile, it can be negated only with in-}2 Let us now turn to two structures for which the AdC and AtC make different predictions. Consider first the structure in (22):

Pre (ii)

[[Pre

Ζ

Suf

+[Z]]

+Suf]

In this case, the AdC predicts the Suf is sensitive to Pre because Pre is contained in the cycle adjacent to Suf. The AtC, however, predicts that Suf is sensitive only to Ζ because in the structure P r e + Z , Ζ is the head and the features percolate from Ζ to Y. The Italian data seem to confirm the AtC. In (23), in fact, we observe that if we allow Pre to vary in the type of structure under examination, this variation does not effect the nature of

176

Generative

morphology

Suf: (23)

a(c)-centrare de-centrare a(c)-clamare de-clamare a(p)-prezzare de-prezzare a(d)-durre de-durre

-> accentramento decentramento -» acclamazione -«· declamazione apprezzamento -> deprezzamento —• adduzione -> deduzione

"centralize - centralization" "decentralize - decentralization'' "acclaim - acclamation" "declaim - declamation" "appreciate - appreciation" "depreciate - depreciation" "adduce - adduction" "deduce - deduction"

What we see in (23) is not only that if we vary the prefix there is no corresponding variation in the suffix, but also that the suffix changes only when we change Z, that is, the head of the construction Pre + Z. Williams points out that there exist some exceptions to the prediction made by the AtC. For example, in English the prefix en- "potentiates" or favors the attachment of the suffix -ment in the formation of nouns (cf. the productive pattern of words such as en-rich-ment). It should be noted, however, that we are dealing, in this case, with a category changing prefix, one of the few prefixes of this type. Since the prefix changes the category of its base, in Williams' terms, it will be considered the head. In Italian, another situation is found, as illustrated by the examples in (24), where the words on the left of the arrow are verbs and those on the right are the corresponding nouns formed by adding -mento. The prefix in the second word of each pair of verbs is of the form aCx, where Cx takes its value from the consonant at the beginning of the base verb: (24)

cogliere accogliere crescere accrescere battere abbattere

-» -> -» -> ->

*coglimento accoglimento *crescimento accrescimento *battimento abbattimento

"gather" "welcome - (a) welcome" "grow" "increase - growth" "beat" "knock down - knocking down"

What we see in (24) is another case of "potentiation", in that the prefix, aCx, attached to the verb "potentiates" the attachment of the suffix -mento. There is however, a substantial difference between this case and the one mentioned by Williams in that in the Italian examples the potentiation is not brought about by a "head", since the prefix does not change the syntactic category of the words in question. That is, cogliere is a verb and so is accogliere. Furthermore, if we consider other data, we see that it is not a particular prefix which potentiates the attachment of -mento, but rather the presence of a prefix in the structure. In fact, -mento is also attached to verbs that contain one of several other prefixes 13 :

Morphology (25)

and

177

syntax

mostrare dimostrare mescolare rimescolare forare perforare

*mostramento dimostramento *mescolamento rimescolamento *foramento perforamento

"show" 'demonstrate - demonstration "mix" 'remix - mixing" ' m a k e a hole" 'perforate - perforation"

The last structure we will consider is the one in (26), for which the A d C and the AtC m a k e different predictions:

(ii)

Pre2

Prel

[Pre2

[Prel

Ζ [Z]]]

In this case, the AtC predicts the Pre2 is not sensitive to P r e l , the A d C predicts that the opposite is true. As can easily be seen, the structure in (26) represents the un-dis case, discussed above. P r e 2 must therefore be allowed to "see" P r e l , which m e a n s that in this case the A d C is superior to the AtC. T o summarize briefly, we have seen that in two cases b o t h the A d C and AtC m a k e the same predictions, and in two cases they m a k e different predictions. Of the latter two cases, one structure is best accounted for by the AtC and the other by the A d C . At present, therefore, it is difficult to choose one condition over the other. A final observation is w o r t h mentioning, however, in this regard. It should be noted that there is a basic difference between the A d C and AtC which m a k e s the two difficult to compare. Specifically, the A d C establishes which morphological material can be referred to by an affix, while the AtC establishes which features can be referred to by an affix. As a result, the AtC c a n n o t m a k e reference to specific morphemes, while the A d C allows this t o be d o n e . 1 4 Finally, it should be noted that the difference between a " h e a d " and the other elements of a complex word lies in the fact that a head "counts" in relation to the features it carries, while the other elements " c o u n t " in relation to the fact that they represent the presence of morphological material. We can conclude, therefore, that b o t h the A d C and the A t C define interesting sets of facts, but that neither succeeds in accounting for the entire d o m a i n of observable fenomena. W h a t is i m p o r t a n t , in any case, is

178

Generative morphology

that WFR's are subject to some principle of locality and thus esclude essential variables or long distance rules from morphology. 1.2. Subcategorization Frames According to Roeper and Siegel (1978:202ff.), WFR's can affect the subcategorization frames of their base in three ways. Subcategorization frames can be (a) inherited, (b) deleted, (c) added (by redundancy rules). Examples of these three possibilities are given below: (27)

(i) Inheritance John criticized the book John's criticism of the book (ii) Deletion They commenced the search They commenced to recite the poem They recommenced the search *They recommenced to recite the poem (iii) Addition the picture is large (*by the photographer) the picture is enlarged (by the photographer)

In (27i), we see that both the Subject and the Object of the Verb (criticize) can be inherited by the derived noun (criticism). In a case like this, what happens is that the subcategorization frame of the base remains the same in the derived word. In (27ii), we see that the verb commence can have two complements, a Noun Phrase (the serch) and a Sentence (to recite the poem), but the derived verb, recommence, can only take a Noun Phrase complement 15 . Finally, in (27iii), we see that the base cannot have an agent phrase, but the derived verb can; this agent phrase is created by redundancy rules, according to Roeper and Siegel. In more recent years, the problem of how derived word acquire their subcategorization frames has come to the attention of many scholars and there are many approaches to this problem. 16 We will outline here Williams' approach, but before doing so, let us consider how WFR's can be classified according to whether or not they change both the syntactic category and the subcategorization frame of their base. 1.2.1 Subcategorization

Frame and Syntactic

Category

Moortgat (1981) offers an interesting classification of WFR's in relation to the changes they can make (a) on both the syntactic category and the subcategorization frame, (b) on one of the two, or (c) on neither of the two. The table in (28) summarizes the changes that can take place when affixes are attached by WFR's. In this table, the column Suf provides information about the type of affix involved: " + Suf" = Suffix, " - S u f " = Prefix. The second columns, SC, tells whether the affix changes syntactic category (i.e.

Morphology and syntax

179

" + SC") or does not change syntactic category (i.e. " — SC"). The last column, SF, shows whether the affix changes subcategorization frame or not, " + S F " and " — SF", respectively:

1. 2. 3. 4. 5. 6. 7. 8

Suf +

SC

SF

+ +

+ +

+

-

-

+

-

+

-

+

•

•

-

+

-

-

Let us now consider each of this cases, and the Dutch examples given by Moortgat. Case 1. is a suffix that changes both the syntactic category and the subcategorization frame (cf. -izeer: [legal]^ "legal"—•[ [legal]a + izeer]y "legalize"). Case 2. is a prefix that changes both the syntactic category and the subcategorization frame (cf. be-: [zwanger] a "pregnant" [be + [zwanger] α] γ "make pregnant"). Case 3. is a suffix that changes the syntactic category but leaves the subcategorization frame, unchanged (cf. -heid: [gehoorzaam]^ "obedient"-•[ [gehoorzaam]^ + heid] ν "obedience"). Case 4. is a prefix that changes the syntactic category but leaves the subcategorization unchanged (cf. ge-: [hunker]y naar N P "hanker after NP"->[ge + [hunker] v ] ν naar N P "hankering after NP"). Case 5. is a suffix that does not change the syntactic category but does change the subcategorization frame. This possibility does not exist in Dutch, according to Moortgat. Case 6. is a prefix that does not change the syntactic category but does change the subcategorization frame (cf. ver-: [slapen]vi n tr "sleep"->[ver+ [slapen]y]vtr as in de dag verslapen "to sleep away the day"). Case 7. is a suffix that does not change either the syntactic category or the subcategorization frame, (cf. -el: N P [prikken]y "to prick N P " - > N P [prikkelen]y "to prickle NP"). Case 8. is a prefix that does not change either the syntactic category or the subcategorization frame (cf. ex-: [minnaar]]sj van N P "lover of NP"-»[ex + [ m i n n a a r ] N ] N van N P "ex-lover of NP"). This list lends itself to some interesting observations. We will limit ourselves to the following few remarks. In the first place, even though it is possible to find the whole range of possibilities in different languages, there must be a hierarchy of markedness. For example, case 7. seems more restricted than, say, case 1. By the same token, even though case 2. is

180

Generative

morphology

found in Dutch (and in English, as we have seen in relation to the prefix en-), it must be considered a marked option, compared to case 8., which seems to be the least marked possibility for a prefix. Secondly, we can note that the list in (28) has interesting consequencies for the notion of percolation. In Chapter V, we saw that the syntactic category percolates from the head, and it was assumed (following Williams, 1981a) that in suffixed words, the suffix is the head. Lieber (1980) has gone a step further in assuming that affixes form a unique lexical class (syntactic category in our terms) on their own, and impose a unique argument structure. In other words, Lieber assumes that not only the information regarding the syntactic category but also the information regarding the subcategorization frame percolates from the head. Cases such as 3., however, show that the subcategorization frame can be "inherited" from the base not percolated from the head. This case is illustrated in (29): (29)

Lieber's proposal, at first, seemed quite interesting in that it imposed strong constraints on the percolation convention, saying, in effect, that there could be no "double" percolation in a single process. This proposal turned out to be unacceptable, however, since double percolation, in fact, seems to exist. Furthermore, it is not limited only to suffixation (case 3. above), but it is also found in prefixation. In Italian, for example, the intransitive verb ridere "to laugh" is turned into a transitive verb when the prefix de- is added, forming deridere "to deride". As we see in (30), the feature [ + t r a n s ] does not percolate from the head: (30)

The prefix de- does not change the category, which remains Verb, but it does change the subcategorization of the verb, as seen in (31):17 (31)

ridere di Carlo *ridere Carlo deridere Carlo *deridere di Carlo

"laugh at Carlo" "deride Carlo"

Morphology

and

181

syntax

The problem with cases such as the one just seen is that they require a more refined definition of the notion "percolation". The question of how a new word acquires its subcategorization frame has not yet been answered systematically. In the following section, we will consider Williams' (1981a) proposal, which handles the problem in terms of argument structure. 1.2.2. Morphology

and Argument

Structure

In the model of morphology proposed by Williams (1981b), the notion of Argument Structure (AS) is used. The AS of a lexical item is a labeled listing of the arguments it can have. In his model, Williams adopts Gruber's (1976) categories: Actor, Theme, Goal, Source. A verb such as hit is represented in his terms as follows: (32)

hit (Actor, Theme)

According to (32), hit has two thematic relations, Actor and Theme, which correspond, for the present porpouses, to the notions of subject and object, respectively (e.g. Jonathan hits the ball). Williams complicates this representation slightly in order to distinguish between an external argument and an internal argument. The external argument corresponds, roughly, to the notion of subject, and it is underlined. The representation in (32) is thus rewritten according to Williams, as in (33): (33)

hit (Actor, Theme)

The problem that arises with this proposal is the following: suppose a lexical item has a given AS, and we apply to this lexical item a W F R which gives a new word with a dilTerent AS. H o w can we characterize this change? Williams' approach is to characterize the function that relates the old AS to the new AS. He claims, in fact, that there are two very simple rules by which the old AS and the new AS can be related: (a) externalize an internal argument, and (b) internalize an external argument. These two rule have the following forms, respectively (cf. Williams, 1981b:92): (34)

(i) Ε (X) (ii) I (X) where X is the name of one of the thematic relations

Williams defines E(X) as follows (p. 92): (35)

E(X):

erase the underline on the external argument, if there is one and underline X. If X = 0, then underline nothing.

In other words, the E(X) operation makes external an argument which was internal in the input word. As an example, consider the suffix -able that attaches to verbs. In the attachment of -able, the rule externalizes the

182

Generative

morphology

Theme: (36)

E(Th) ready (A, Th)

readableA (A, Th)

(36) accounts for the fact that the object of the verb becomes the subject of the derived adjective. Williams observes also that in this rule only the T h e m e can be externalized, as seen below: 1 8 (37)

(i) those things are promisable (ii) *those people are runnable *those people are promisable

(Theme externalized) (Actor externalized) (Goal externalized)

The function I(X) is a rule for causativization a n d nominalization. It is defined as follows (Williams, p. 99): (38)

I(X) (i) set the external argument of the input word "equal t o " X in the o u t p u t word; (ii) A d d a new external argument, A for verbs, R for nouns.

A case of I(X) is the a t t a c h m e n t of -ize. This rule is a I(Th) rule, as can be seen, for example, in the pair legal/legalize: (39)

I(Th) legal A (Th)

legalizey (A, Th = Th)

According t o (39), legalize has a new argument, A, and its T h e m e is the same as the T h e m e of the input word. Williams' proposal is m o r e complete t h a n the description we have just given. While we c a n n o t go into m o r e detail here, there is at least one consequence for the subcategorization frame t h a t is w o r t h mentioning. Given his theory of argument structure, Williams maintains that there is n o need of a set of rules for specifying the changes that take place in the subcategorization frames. According to Williams, in fact, there are realization rules for each argument; Goal, for example, can be realized in English as N P 2 or as N P in a P P t 0 (cf. give Bill the book and give to Bill). Williams proposes that the realizations for any type of argument can be r a n k e d according t o their markedness; in English, for example, P P t o is less m a r k e d t h a n N P 2 for Goal. If a new verb with G o a l were created, t h a n it would have the P P t 0 sub-categorization and not the N P 2 subcategorization. A n o t h e r example that Williams gives of this possibility is the following: a suffix such as -ize seems to create obligatory transitive verbs (cf. to modernize the agricultural system vs. *to modernize). In this case, too, it can be assumed that the u n m a r k e d choice for internal T h e m e is N P , while (NP) is the m a r k e d choice.

Morphology and syntax

183

In conclusion, according to Williams, the changes of the subcategorization frames need not to be specified for every case; they can follow, instead, from changes in argument structure. Williams' proposal, as we have seen, has the interesting characteristic of being very simple, though it has not yet been evaluated on the basis of a broad set of data, and therefore remains somewhat tentative. 19 2. Clitics Another area where there have been difficulties in drawing a clear cut division between syntax and morphology has to d o with the status of clitics. In what follows, we will discuss several criteria 2 0 that are relevant in distinguishing between derivational affixes and clitics. 1. Syntactic Category. As we have seen above, derivation rules can change the syntactic category of the words to which they apply. Clitics, on the contrary, do not change the category of the element to which they are cliticized. In Romance languages, for example, clitics normally attach to verbs, and the output is always a verb, as can be seen in the following Italian examples: (40)

V'

clitic

ascolta legger vedendo

mi lo ti

'listen to me' '(to) read it' 'seeing you'

2. Ordering. It has been suggested by Zwicky (1976:30) that alternative orders of morphemes within a word are associated with differences in cognitive meanings. While this may be true for derivational affixes, when alternative orderings are possible, as seen in (41i-ii) the same cannot be said of clitics. 21 In fact, alternative orderings of clitics are not usually possible, as seen in (41iii-iv). In some cases, where it seems that the ordering of clitics may be changed, what we have, in fact, is not two clitics, but one clitic and an independent pronoun (preceded by a preposition), as in (41v): (41)

(i) (ii) (iii) (iv) (v)

X - ic - ista X - ist - ico di - glie - lo 2 2 *di - lo - gli di(l)lo a lui

storicista artistico

"historicist" "artistic" "lit. said to him it' "lit. said it to him"

184

Generative

morphology

3. Internality. It has been observed by Zwicky, furthermore, that a morpheme "in construction with [ = internal] to an affix must be either a base or an affix" (p. 30). Consider, now, some possibilities of Italian cliticizations; (42)

(i) W ] v + Infl]v + C l i t i c ] γ + I n f l ] v ama r 1 i (ii)

W ] A + D e r ] v Infl] v + C l i t i c ] v concret izza r 1

amarli "(to) love them (masc., pi.)" +Infl]v e concretizzarle "(to) make them (fem. pi.) concrete"

(iii) * W ] + Clitic] + D e r ] + I n f l ] According to Zwicky's "Internality Principle", clitics should be considered affixes, since they appear "in construction" with other affixes (cf. (42i) and (42ii)). We have seen, though, that derivational affixes d o not occur to the right of inflectional morphemes. Moreover, it should be noted that the node Infi to the left of the clitic (both in (42i) and (42ii)) must belong to a definite set of Infi nodes. T h a t is, clitics can be attached to the right of a verb only when the verb is in the infinitive mode (cf. amarli in (42i)), in the past participle mode (cf. arresi-si "having surrended"), in the gerundive mode (cf. amando-li "loving them") or in the imperative mode (cf. amali "love them"). If the verb is in the indicative mode, on the other hand, clitics may not, in contemporary Italian be attached to the right, but appear on the left, instead: (43)

(i) *amoli *concretizzole (ii) Ii a m o le concretizzo

"(I) love them" "(I) make them concrete"

We must conclude, therefore, that clitics cannot be considered affixes since they normaly occur to the right of inflectional affixes, and since they are sensitive, in a way that derivational affixes are not, 2 3 to the type of inflection on the element to which they attach. 4. Phonology. It has been reported that vowel harmony rules apply to affixes, not to clitics 24 or, in other words, that affixes obey at least certain phonotactic restrictions and clitics do not. 2 5 There are also other differences between the phonology of clitics and the phonology of affixes. We will briefly discuss three points here in relation to Italian. The first point has to d o with primary stress. There is a general restriction in Italian according to which the primary stress of a word cannot fall further left than on the third to the last syllable. 26 If we build

Morphology

and syntax

185

up a word by means of derivation which would violate this restriction, certain readjustment rules operate to correct the situation. 2 7 With clitics, on the contrary, we find numerous examples in which the stress is farther to the left than the third to the last syllable since no readjustment rules apply to change these forms: (44)

contamelo mescolamela

"count it (for) me" "stir it (for) me"

Another phonological rule which is systematically violated in cliticization (cf. (45ii)), but not in affixation (cf. 45i)), is the s-voicing in Northern Italian: 2 8 (45)

(i)

Affixation di[s] + onesto ci[s] + alpino (ii) Cliticization arreso + [s]i pentito + [s]i

di[z]onesto ci[z]alpino

"dishonest" "from this side of the Alps"

arreso[s]i pentito [s]i

"having surrended" "having repented"

The third point is that clitics, contrary to affixes, behave like independent words with respect to the so called Raddoppiamento Sintattico of central and southern Italian. T h a t is, clitics undergo the rule whenever the phonological conditions are met, while affixes d o not: (46)

(i) Independent a [k]asa tre [g]atti (ii) Clitics d a + [m]i di + [l]o (iii) Affixes a + [t]ipico ri + [ f ] a r e

words -» a[kk]asa -> tre[gg]atti

"at home" "three cats'

da[mm]i di[ll]o

"give mi" "say it"

a[t]ipico -* ri[f]are

"atypical" "redo"

-

On the basis of the phenomena discussed in this section, 2 9 we can conclude that affixes and clitics constitute two separate sets of bound morphemes. We have seen, furthermore, that clitics are related to indepented words in several ways, a fact which support Lefebvre and Muysken's (in press) claim that clitics, or to be more precise, the positions in which clitics are inserted are base-generated. 3. Interaction

between morphology

and

syntax

In the preceding sections, we have seen that morphology and syntax are two different "subsystems" of the grammar, each of which can be defined

186

Generative

morphology

in relation to different formal properties. This does not mean, however, that these two subsystems d o not interact at all. In the following sections, in fact, we will consider some points of interaction between morphology and syntax. 3.1. Word bar Theory O n e of the main goals of the X-bar theory in syntax is to show that the internal structures of the major syntactic constituents (NP, AP, VP, PP) are essentially the same. In other words, the hierarchical structure of these constituents is the same; the difference lies in the different lexical categories of the words that can be inserted in place of the dummy symbols in deep structure. The basic structure for the different types of constituents is given in (47): (47)

X

Spec

Δ

X

X

Compl

Δ

Δ

This configuration expresses the generalization that any phrase consists of three elements: a specifier, the head of the phrase and a complement, as illustrated for the different types of syntactic constituents in (48):

(i) (ϋ) (iii) (iv)

Spec this having very much

X professor read useful behind

Compl of chemistry that book to smokers the target

(NP) (VP) (AP) (PP)

In (47) and (48), X is a variable that stands for a set of categories: Ν, V, A, P, the cases in (48i-iv), respectively. The level of the category is defined by the number of bars above X. In this regard, rewriting rules within the Xbar framework must obey the following restriction: 3 0 (49)

Xn

...X"-1...

According to this restriction, every syntactic category dominates a category of the same type, but of the immediately lower level in the X-bar hierarchy. Given that syntax and morphology have in common the level "X", that

Morphology

and

187

syntax

is the "word" (both syntactic and morphological), several linguists have suggested that the X-bar theory might be fruitfully extended to morphology (cf. Hoekstra et al. (1979), Toman (1980), Selkirk (1982)). We will consider here only Selkirk's proposal since it is the most detailed proposal and the one that promises the most. The morphological component in Selkirk's proposal is organized in the following subcomponents: (I) A set of context-free rewriting rules (the Word Structure which have the following form (Selkirk, 1982:4): (50)

Ρ Ρ Ρ

->· φ Af Q φ -> φ Q Af φ φ Q R ψ

Rules),

(Prefixation) (Suffixation) (Compounding)

where P, Q, R stand for individual category symbols, φ and φ are variables over category symbols These rules may also appear in the simplified form given in (51): (51)

Ρ Q R

-> ->

Λ Δ Δ

where P, Q, R stand for individual category symbols and Δ is the "dummy" symbol (cf. Chomsky 1965). (II) A list of lexical items of a language, the Extended Dictionary, which contains both free forms and bound forms such as affixes, stems and roots. (III) A (morpho) lexical insertion transformation which inserts the lexical items from the Extended Dictionary in the correct positions. This rule, like the Lexical Insertion rule (seen in Chapter I), is subject to any conditions imposed by the lexical item to be inserted. Such conditions are lexically specified, and are very similar to those required by the (syntactic) Lexical Insertion rule. The position Selkirk advances in her proposal is that "W-syntactic (i.e. morphological) categories are entities that are formally identical in character to syntactic categories [ . . . ] " (p. 7). Her hypothesis is, thus, that "all Wsyntactic categories, be they of the type of Word or "lower" than Word, are in the X hierarchy". In this system, the syntactic word is a category of level zero (i.e. without bars). Words of the categories Ν , A, V, P, therefore, have the category symbols Ν , A 0 , V°, P°, respectively. Given these basic characteristics of Selkirk's proposal, we will now proceed to examine two problematic aspects of the proposal, specifically, (i) the use Selkirk makes of the X-bar hierarchy, and (ii) the use of contextfree rewriting rules in morphology. 3 1

188

Generative

morphology

As far as Selkirk's use of the X-bar hierarchy is concerned, it should be noted that both simple and complex words are assigned to the same level and that affixes are not assigned to a negative (i.e. " — 1") level. As a result, the condition in (49) is violated in all Selkirk's rules. It would be possible, instead, to use the condition (49) to establish different morphological levels as follows: 32 (52)

Level 1 Level 0 Level — 1

morphologically complex words simple words affixes

A hierarchy such as the one in (52) corresponds better to the hierarchy of syntactic categories. In fact, the level 1 morphological categories, analoguously to the level 1 categories in syntax, are categories with complex internal structure. The level 0 categories are simple words, and constitute the point of contact between morphology and syntax. Finally, level — 1 represents preterminal categories that cannot occur on the surface. Given such a hierarchy, it is possible to write rules that respect the requirement express by the constraint in (49), as seen in (53): (53)

X1 X1

- Y° -» Χ " 1

X"1 Y°

(suffixation) (prefixation)

According to (53), the structures of morphologically complex words such as happiness and enrich are those in (54i) and (54ii), respectively:

Theses representations have the advantage that they clearly express the morphological complexity of the words in question. That is, they show a complex word which contains a free form and a bound form. As far as context free rewriting rules are concerned, according to Selkirk, these rules have several advantages with respect to other types of rules. T h a t is, (a) they are capable of generating all the words of a language and only those, (b) they express the intuitions of native speakers of a certain language that words have an internal constituent structure, and (c) they account for the recursiveness or self-embedding of morphological structure since "there is not principled upper bound to the length of words" (p. 3). In regard to point (a) and (b), it should be noted that Selkirk does not

189

Morphology and syntax

give any evidence that a Categorial Grammar (of morphology), such as the one adopted in this book is less adequate than the one she proposes. In fact, a Categorial G r a m m a r can express the points in (b) and (c) just as well as Selkirk's proposal can, as we have seen in the preceding chapters. As far as point (a) is concerned, according to Selkirk a context-free grammar seems to be the appropriate mechanism not only for generating all the words of English (weak generative capacity), but it also assigns the proper structural description to the strings it generates (strong generative capacity). For example, with respect to compounds, Selkirk claims that the possible combinations in English can be characterized as follows: schoolteacher highschool rattlesnake overdose nationwide icy cold overwide underfeed

N+ N A+ N V+ N P+ N N+A A+ A P+A P+V

Still according to Selkirk, given that the combinations V + A , N + V, A + V, V + V, N + P, A + Ρ and V + Ρ are not attested, we can write the following rewriting rules for English compounds:

fNl

(ii) A

-> < A > A

(iii) V

->

Ρ

V

What (56) represents, however, as Drigo (1983) has observed, is little more than a list of existing combinations in English. It does not explain why certain combinations are possible and others not. It does not even identify which combinations are productive and which are marginal, nor does it explain why this should be the case. Selkirk argues that the type of compounding rules proposed by Allen, which we have adopted in this book, is overgenerating, while the rules in (56) are not. It should be noted, however, that Selkirk's rules are not overgenerating in a rather weak sense since, in fact, they are explicitly derived from only the existing compounds in English. It does not seem, therefore, that the rules in (56) satisfy in any way the requirement of descriptive adequacy.

190

Generative morphology

Selkirk maintains, furthermore, that the gaps in English cannot be accounted for by universal principles because these gaps are, in fact, filled in other languages. By itself, this observation is not particularly relevant, since the use of the X-bar theory implies a certain parallelism between syntax and morphology, and thus offers the possibility of explaining morphological gaps in relation to the syntax. The X-bar theory can, in fact, be used as a structural restriction on the formation of compounds when we consider the relation between the constituents of a compound and the basic word order of a language. According to the X-bar theory, the constituents of a phrase must be analyzed, as we have seen above, in terms of a specifier, a head and a complement. As far as the position of the word within a phrase is concerned, the X-bar theory requires that if the basic word order of a language is SVO, then the complement is to the right of the head, and if the basic word order is SOV, then the complement is to the left. In other words, for each language, a parameter of the grammar fixes the recursive side with respect to the head of a phrase; 33 in the case of SVO languages this is the right side, while in the case of SOV languages, this is the left side. Applied to compounds, the X-bar theory predicts that in a SVO language V + N compounds, the complement, N, will be on the right, while in a SOV language it will be on the left. There is, in fact, interesting confirmation of this prediction in the history of Romance languages. 34 In Latin, SOV language, we find compounds of the type N + V, as illustrated in (57): (57)

agricola signifer piscicapi

"farmer" "standard banner" "fisher"

In modern Romance languages, which are of the type SVO, we find instead, V + N compounds, as seen in the Italian examples below: (58)

portascarpe portaband iera segnalibri

"shoe holder" "standard banner" "book mark"

"lit. carry+ shoes" "lit. carry + standard" "lit. mark + books"

It should be noted that in general, V + N compounds represent a problematic case for a X-bar morphology since they are exocentric constructions. The X-bar theory, nevertheless, makes predictions about these compounds that appear to be correct. The same seems to be true for other types of compounds. In Italian endocentric compounds of the form "Head + Complement" (cf. Lepschy and Lepschy 1977), in fact, we can observe the following regularities (cf. Drigo, 1983): (a) The category of the complement is always a category allowed by the X-bar theory for the type of head in question. In Italian, the endocentric combinations for "Head +

Morphology

and

syntax

191

Complement" are N + N, N + A and A + N . The head is always Ν in these cases, and the complement is Ν or A, precisely those categories that may function as a complement of Ν in the X-bar theory, (b) The complement follows the head, except in the combination A + N . It is interesting to note that Italian allows the marked sequence A Ν in a n o u n phrase (cf. grande libro "big book"). The same order in compounds is also the marked case. 3 5 (c) The non-existing compounds in Italian of the forms Ν + Ρ, A + Ρ and V + P are automatically accounted for by the fact that Ρ cannot be the complement of N, A or P. As these observations show, it appears that the X-bar theory can be used in interesting ways to express certain analogies between the principles of phrase construction in syntax and those of word construction in morphology, at least as far as compounds are concerned. 3 6 3.2. Inflection In this book, we have adopted a Strong Lexicalist Hypothesis, as was mentioned at the beginning of Chapter VI. Let us define the SLH more accurately now. According to the SLH, morphology is characterized by the following points (cf. Lefevbre and Muysken, in press): 1. Word formation takes place in the lexicon, and is carried out by rules which are different in nature from syntactic rules. The items that are inserted into phrase structure positions are completely formed words. 2. The relationship between the features of a given word and the structure in which the word appears is not of a transformational nature. 3. The "word", as defined by the lexicon, under normal circumstances, corresponds to the "phonological word". Some exceptions exist, such as clitics, generated by the phrase structure rules, and possibly, what can be called discontinuous words. 4. Inflection operates within the lexical component. Each of these points has been discussed at length in the literature. We have already discussed points 1, 2 and 3. In this section, we will, therefore, discuss only point 4. Recently, the claim under 4. has been challenged by Anderson (1982) in a paper entitled, significantly, "Where is morphology?" The main purpose of this paper is to demonstrate that claim under 4. is too strong. Anderson reformulates 4. in the following terms: (59)

Syntactic rules cannot make reference to any aspects of wordinternal structure.

The starting point for Anderson is the observation that, contrary to what is stated in (59), "some words are assigned morphological properties which depend directly on their position in larger constructions" (p. 573). Anderson lists three "morphological properties" which are dependent on syntax: (a) configurational properties, such as the assignment of case, which depends on "an interplay of grammatical structure [ . . . ] and

192

Generative

morphology

idiosyncratic properties of a particular case-assigning element (verbs, prepositions [...])" (p. 573), (b) agreement properties, such as agreement in number, case, etc. of an adjective with the head noun of the N P in which the adjective appears, (c) inherent properties. If an adjective must agree with a noun, the rule of agreement must have access to all features on which the agreement is based. Some of these features are assigned on a configurational basis (such as case), but other properties such as gender are inherent properties of the lexical items with which agreement takes place. Anderson illustrates his point with a detailed analysis of Breton data which brings him to the conclusion that Breton offers a convincing counterexample to the claim that syntactic rules may not make reference to morphological material internal to a (surface) word. Specifically, Anderson argues that derivational morphology is lexical, and since his Breton d a t a are drawn from the domain of inflection, it is necessary to find a principled way of distinguishing between inflection and derivation. In the absence of such a principle in the existent literature, Anderson proposes the following criterion (p. 587): (60)

Inflectional morphology is what is relevant to the syntax.

Establishing thus that (a) inflection is different from derivation and (b) inflection is dependent on syntax, Anderson proposes the model of morphology which is given below in a somewhat simplified form: Base Rules Lexicon D-structures Dictionary plus rules of derivation

Syntax (agreement)

S-structures

Lexical Items

(incl. morphosyntactic repr.)

Lexically interpreted S-structures

Phonology

Interpretation

(incl. inflection)

I P h o n o l o g i c a l form

Logical form

Figure 1: Anderson's "Extended Word and Paradigm" Model.

Morphology

and

syntax

193

This model is referred to by Anderson (and others) 3 7 as the "Extended W o r d and P a r a d i g m M o d e l " (EWP). Let us now examine several aspects of the model in m o r e detail. According to Anderson, the a d v a n t a g e s of his model are the following: 1. The separation of inflection f r o m derivation explains why inflectional m o r p h o l o g y a p p e a r s outside of derivational m o r p h o l o g y . 2. The location of inflection in the phonological c o m p o n e n t has the result that inflectional operations are not accessible to the rules creating Logical F o r m , and thus, they c a n n o t affect semantic interpretation. 3. There are interactions between inflectional process and p h o n o logy which are correctly predicted by the model. Anderson says he does not k n o w if there are similar interactions between p h o n o l o g y and derivation (in the affirmative, he is willing to accept a type of Lexical Phonology). 4. The fact that there are derivational rules that o p e r a t e on lexically restricted irregular stems, but not on fully inflected independent words is well accounted for by the model since derivational rules have access to any stem listed in the lexicon but not to the o u t p u t of the inflectional rules. This is so because (uninfected) lexical items are inserted at S-structure together with the morphosyntactic representation provided by the transformational rules. T h e input to the logical form is, thus, a lexical item with associated features. Anderson's claims a b o u t the advantages of his model are not, however, without their questionable aspects. First of all, as far as point 1. is concerned, it should be noted that separating inflection f r o m derivation does not automatically mean that the two must be put in different components. The model we presented above also m a k e s a distinction between inflection and derivation, b u t it assumes that they b o t h operate in the same s u b c o m p o n e n t , i.e. the lexicon. O u r model has two advantages, in this regard: (a) it allows us to claim that inflection and derivation, t h o u g h different, in m a n y respects, are subject t o the same general conditions (a point we will return below), and (b) it allows us to account more adequately for the relationship between inflection and composition. It is difficult to imagine how Anderson's model would account for a word such as programs coordinators, that is, a word with two "kinds" of inflection, one "inside" the c o m p o u n d and the other "outside" the comp o u n d . 3 8 In Anderson's model, this word would have to go back and forth t h r o u g h o u t the entire g r a m m a r . As for the second point, Lefebre and Muysken (in press) observed that an identical situation is found in the lexicalist model. T h a t is, it is only the feature list that may partecipate in interactions with other c o m p o n e n t s of the g r a m m a r . Thus, Anderson's model does not offer any a d v a n t a g e s in this respect. Anderson's third point is decidedly a weak point in the model since it is

194

Generative

morphology

clear today that there is just as much phonology involved with derivation as with inflection. In fact, "Lexicalist Phonology" 3 9 was developed to account for precisely these facts. Finally, point 4 does not really represent an advantage of the Weak Lexicalist Hypothesis advocated by Anderson over the Strong Lexicalist Hypothesis. It essentially has to do with the kind of lexical representation one is willing to hypothesize in the lexicon. One can also posit stems and roots in the lexicon and still defend the SLH 4 0 Let us now turn to some more general theoretical questions that can be raised in relation to Anderson's proposal, ones which are pertinent for the relationship between morphology and syntax. The first issue has to do with the notion of "autonomy of systems", defined in the following terms by Hale (1976:120): 41 (61)

A language consists of a number of distinct systems each possessing inherent principles of organization which are utterly independent of factors relating to any other linguistic system or to extralinguistic factors

Each of these systems (such as the categorial rules of the base, the lexicon, the transformational component, etc.) has its own primitives, its own properties, rule types, etc. These systems have highly specific properties but they are seen as interacting in the production of the final product (sentences in our case). Thus, the interaction between morphology and syntax is beyond doubt. What is crucial instead -as Lefevre and Muysken (in press) point out- is the way in which the two interact. The lexicalist hypothesis claims that word formation is an a u t o n o m o u s system located in the lexicon, and the SLH claims, furthermore, that derivational rules and inflectional rules are part of the same system. A consequence of this claim is that there must exist some general principles valid for both inflection and derivation. Anderson's model, on the other hand, according to which derivation and inflection belong to different "modules", implies that these two sets of rules are governed by separate and different principles. This is not true, however, since we have seen above that the Adjacency Condition, for example, holds for both derivation and inflection. 42 Furthermore there are a number of readjustment rules in Italian that operate in the same way in derivation and inflection. A similar point is also made by McCarthy who has demonstrated that in Semitic languages the same principles holds for inflection and derivation. 4 3 As far se the issue of how morphology and syntax interact is concerned, it can be said that in the lexicalist view, the point of interaction is at the insertion site. This is the point at which there is a list of feature specifications common to both morphology and syntax, as is illustrated in the following Italian example, amareggiavate "you (pi) were making (it) bitter":

Morphology

and

(62)

syntax

195

+ V + 1 conj + past + 11 pers + plural

amarA

egg'v

ai co „

v past

ate

Ilpers plural

Lefebvre and Muysken point out that entirely different theories govern the way in which the common list is build up from the syntax side and from the lexicon side. "In fact, a reasonable interpretation of the lexicalist hypothesis would be that the way the list is constituted from the syntax (e.g. whether a dative is "inherent", assigned by V or assigned by P) is independent of the feature in the word (e.g. through suppletion, affixation, etc.). Conversely, it is claimed that the way in which a feature is realized morphologically (go/went, walk/walked) does not affect the syntactic consequences of the feature being present in the morphosyntactic representation" (p. 23). Recall, furthermore, that one of the major points Anderson makes is that the claim made by the lexicalist hypothesis (Lapointe 1980) does not hold: (63)

Syntactic rules cannot make reference to any aspect of word internal structure

As Lefebvre and Muysken correctly observe, however, Anderson interprets this claim in such a way that it excludes even the list of c o m m o n features such as the one in (62) while, in fact, the lexicalist hypothesis excludes any reference to the internal structure of a word not to the dominating feature list. Syntactic rules may, in fact, refer to the dominating feature list, but not to the internal structure of a word. This difference can be seen in relation to the operation of percolation within the lexicalist framework. Consider a word such as nationalize, and its structure as given in (64):

196

Generative

morphology

This word has both internal structure and a dominating feature list (i.e. Verb, [ + t r a n s ] ) . The dominating feature list is created by the percolation of features from the head. A real counterexample to (63) would be a syntactic rule which refers to the node A or to the node N, that is, to the internal structure of the word. The dominant feature list is not part of the "internal structure" and is thus accessible to syntactic rules; this does not constitute a violation of (63). In conclusion, we can say that over the years, the relationship between morphology and syntax has gone back and forth between two poles: either the two fields were grouped together or they were considered as completely independent of each other. In the beginning, generative grammar reacted against the structuralist philosophy of "not mixing levels" and morphology and syntax tended to be combined and confused, with no advantage to either field; this.was the "morphosyntax" period. In a second phase, morphology was isolated into an a u t o n o m o u s component. In the period of "autonomous morphology", it was thus possible to study morphological facts without interference from other components of the grammar. In fact, this phase produced much research, most of which centered around the nature of morphological rules. In the present phase, which can be called "modular" period, a fruitful approach has been developed aiming at the mode of interaction between syntax and morphology. The situation today is more interesting for both field, in that each field has its own autonomy. This last phase has benefited from criticisms published in several important papers, such Anderson (1982) which deserves credit for having pointed out a number of major problems. As we have seen, however, Anderson's alternative fails to provide a substantial improvement over the SLH because it splits morphology into three different subcomponents, the lexicon, the transformational component and the phonological component. The weakness of Anderson's model, thus, lies exactly in the answer it gives to the question in the title of his paper, "Where is morphology?". His answer, as we have seen, would have to be "everywhere". Instead, what we have tried to show in this book is that morphological theory took a great step forward, starting with Halle, precisely when a single autonomous component was proposed to account for a wide variety of morphological phenomena, the lexicon.

Morphology

and

syntax

197

4. Summary and conclusions In this chapter, we have shown t h a t W o r d F o r m a t i o n Rules and transform a t i o n s differe with respect to a n u m b e r of formal properties. It was shown, first of all, that W F R ' s are constrained by a type of locality principle. In particular, two proposals advanced in order to express this constraint (i.e. the Adjacency C o n d i t i o n and the A t o m C o n d i t i o n ) were examined; a c o m p a r i s o n of these conditions revealed that each one is able to accound for a subset of the facts. Secondly, it was shown t h a t W o r d F o r m a t i o n Rules can change the subcategorization frame, a n d / o r the argument structure of an input word, a n d in this respect differ f r o m transformations. Several recent proposals advanced to handle such facts were considered, t h o u g h a n u m b e r of problems remain to be resolved in this area. Next, the position of clitics in the g r a m m a r was also briefly considered, and it was argued that clitics c a n n o t be considered a type of affix, but rather that they are base-generated elements. In accordance with a m o d u l a r a p p r o a c h to the g r a m m a r , we have assumed that the morphological c o m p o n e n t is an a u t o n o m o u s c o m p o n e n t as far as its internal structure is concerned, but also that it interacts with other c o m p o n e n t s of the g r a m m a r in a n u m b e r of ways. In the last two sections of this chapter, we have discussed two recent proposals bearing mainly on the relationship between m o r p h o l o g y a n d syntax. First, Selkirk's (1982) proposal to apply the principles of X-bar syntax to words was considered. While in general, the idea seems promising (if only for certain aspects of WFR's), Selkirk's proposal itself does not push the X-bar theory as far as might be possible in the direction of explanatory adequacy, in particular where the relationship between the basic (syntactic) word order of a language and the order of constituents in c o m p o u n d s is concerned. Finally, we have discussed Anderson's (1982) hypothesis that morphological rules d o not operate entirely within the lexical c o m p o n e n t , but rather they are found in several c o m p o n e n t s of the g r a m m a r . In particular, one of Anderson's claims, namely that inflection is sensitive to syntax, requires more careful consideration, and, in fact, p r o p o n e n t s of the Strong Lexicalist Hypothesis will have t o provide a more a d e q u a t e explanation of a n u m b e r of facts pointed out by Anderson. 4 4 It was argued, nevertheless, that Anderson's proposal is not any better t h a n current Strong Lexicalist Models and that it still appears more a p p r o p r i a t e that all morphological rules operate in the same c o m p o n e n t , that is, the lexical c o m p o n e n t . While several linguists have suggested that c o n t e m p o r a r y m o r p h o l o g y is a rather confused area to study, 4 5 in reality this a p p a r e n t confusion can be attributed to the fact that today m o r p h o l o g y is a discipline characterized by m u c h on-going research. We have t h u s seen, in recent years, the development of a variety of different proposals and theoretical approaches, as well as a large n u m b e r of empirical studies. In this book, we

198

Generative morphology

have attempted to offer a unified view of the major problems and areas of research in contemporary lexical morphology. If we have succeeded in this goal, it is at least in part due to a certain degree of simplification of the material covered. In particular, we have generally limited out attention to proposals and theories that follow a specific direction, that of lexical morphology, within the range of possible morphological theories. This, of course, has resulted in the exclusion from this book of several other theories advanced in recent years such as Dik's (1979) "functional" theory, Mayerthaler's (1981) "natural" theory, Dressler's (1977) "polycentristic" theory and McCarthy's (1979) "non-linear" theory, among others. 46 We have attempted, in particular, to identify a homogeneous thread of research as far as the fundamental "paradigms" are concerned. What has emerged, as we have already said, is a picture of a field characterized by considerable activity. Of course, not all the problems raised here have been resolved, and many of the problems facing morphology today have not even be raised. As Kuhn (1962) observed, after a period of "normal science", scholars begin to be confronted with "anomalies". In lexical morphology, too, we have already begun to see the discovery of anomalies, in the sense that research has already begun to reveal data that do not fit into the "normal" framework, but require, instead, solutions of a more or less ad hoc nature. This can be taken as an indication of a "healthy" field of study since it means that there are, in fact, proposals that make different empirical predictions that can be evaluated on the basis of a significant corpus of data. Thus, the broad range of research activity in the field of morphology is the sign that we are dealing with a mature scientific discipline.

Notes 1. Aronoff (1976:68) admits the possibility that copying rules could be stated as transformations, a position which was criticized by Booij (1977). 2. Lieber (1980:240) observes that Word Formation Rules d o not form words "by permuting the first and the last consonant of a string, or reversing the order of segments, or deleting every other segment." 3. In Wasow's account, the only type of WFR's are Derivation Rules. 4. The answers in (3) are given in yes/no terms, which gives a somewhat simplified picture. For example, Kayne (1981) argues that syntactic trees are binary, which may be relevant to point (3d). 5. The implicit assumption made here is that every WFR creates a cyclic domain, which is present at the point in which every WFR operates. Cf. also Booij (1979). This claim is contrary to the Bracket Erasure Convention proposed by Pesetsky (1979), according to which brackets are erased after every application of a WFR. 6. Cf. the Subjacency Condition in syntax, proposed by Chomsky (1973). On Adjacency in other components of the grammar, cf. van Riemsdijk (1982). 7. Muysken (1981) and Scalise (1983) show that Adjacency is not valid for Readjustment Rules. Muysken argues that RR's are constrained by the notion of c-command, while Scalise cites several cases of suffix deletion in compounds in Italian which do not support Muysken's claim.

Morphology

and

syntax

199

8. The plural ceneri actually exists, but not as the plura of cenere. Instead, it has an idiosyncratic meaning, namely "mortal remains". 9. Cf. Lepschy and Lepschy (1977). 10. The comparison is based on Scalise and Zannier (1982/3). 11. In (16) and below, we give both the tree structure and the labeled structure since it is easier to visualize the AtC with the former and the AdC with the latter. 12. There exist words such as dissolvibile "dissolvable", disponibile "available", but these words do not have the structure given in (20iii). Instead, they have the structure [ [dis + [ V ] ] + ibile]. 13. Actually, the potentiation is not limited to -mento\ it can also involve other suffixes. For example, in (25) the forms dimostrazione "demonstration" and perforazione "perforation" would also be possible. 14. This implies also that the notion of "potentiation" is a spurious notion in Williams' system. Cf. Scalise and Zannier (1982/3). 15. Exceptions to this pattern are pointed out by Bauer (1983:179). 16. The most recent bibliography includes works such as Carlson and Roeper (1979), Williams (1981b), Randall (1984). For the treatment of subcategorization frames in compounding, cf. Lieber (1983). 17. Cf. Chapter VI, sec. 2, above. 18. As Williams (1981b:83) points out, this observation was originally made by Wasow (1977). 19. Williams' proposal has been criticized by Poser (1982) who, citing examples from Bahasa Indonesia and Kinyarwanda, argues that lexical rules cannot be restricted to internalization and externalization of arguments. Instead, they must also be allowed to interchange internal arguments. A consequence of this claim is that argument structure requires a richer representation than that provided by Williams. 20. Some of the criteria offered here are based on Zwicky (1976). 21. Lefebvre and Muysken (in press) have shown that the ordering criterion holds for derivational affixes in Quechua (such as causatives and reciprocals), though there are problems for tense and object markers. 22. In (41iii) and (41v) there are some phonological readjustments which do not concern us here. 23. Derivational affixes in Italian never attach to infinitive, gerundive or imperative modes; they attach, instead, to the verbal theme (cf. Scalise 1983). 24. Cf. Lefebvre and Muysken (in press). 25. Cf. cases such as darglielo "(to) give it to him" where the attachment of the clitics glielo creates the sequence [ A ] which is not normally admitted in Italian. 26. The only exceptions are the third person plural indicative and subjonctive forms of a small group of verbs (cf. telefonano "(they) call". 27. Cf. Vogel and Scalise (1982). 28. The relevant part of this rule can be formulated as s-»z/V V. 29. We have not taken into account here another criterion proposed by Zwicky, "Rule Immunity". According to this criterion, "proper parts of words do not undergo rules of deletion under identity" (p. 30). We have seen, instead, that compounds can undergo such deletion rules (cf. Chapter IV, section 2 above). Zwicky himself, cites as exceptions cases such as German Mittag- und Abendessen "lunch and dinner", and English book- and magazine seller. Levebvre and Muysken (in press) observe, furthermore, that clitics in Quechua, like Romance clitics, can undergo certain movement rules. 30. Cf. Jackendoff (1977). 31. Other criticisms have been raised against PS rules in morphology. Cf. Allen (1978) and Scalise (1983). These criticisms were aimed, however, to a previous version of Selkirk's book, where it was mantained that the word structure rules were part of the base rules. Other criticism against a Word bar morphology can be found in Muysken (1981:285). 32. The idea of attributing a " — 1" level to the affixes is of Hoekstra et al. (1979). 33. Cf. Graffi (1980).

200

Generative

morphology

34. This observation holds also for other cases. According to Lehmann (1974:76) "The verbal compounds in a language observe the basic order patterns". As an example, Lehmann cites the case of Japanese synonims, one of which is based on borrowings from Chinese: tozan and yamanobori "mountain climbing". In the Chinese borrowing, the element for "mountain", zan, follows the element for "climb", to. This order is in accordance with the Chinese SVO basic order. In Japanese, on the other hand, the element for "mountain", yama, precedes the element for "climb", nobori, in accordance with the Japanese SOV basic order. There are, however, many difficulties with this hypothesis, especially in English, as is well known. Cf., among others, Lightfoot (1979: 160). 35. Lehmann (1974:76) observes that a direct relationship between compounds and basic syntactic patterns is found only when the compounds are primary and productive. 36. On the relationship between syntax and morphology in so far compounds are concerned, cf. Lieber (1983). 37. Cf. Thomas-Flinders (1981). 38. Cf. Chapter VI, sec. 4, above. 39. Cf., for example, Mohanan (1982), and Kiparsky (1982). 40. Cf., for example, Lieber (1980), and Selkirk (1982). 41. Cited by Lefebvre and Muysken (in press). On this notion, cf. also van Riemsdijk (1982). 42. Lefebvre and Muysken (in press) make a distinction between constraints that are extremely general, on the one hand, or stated in terms of the vocabulary of the system itself on the other hand. They attribute locality to the former. It should be noted, however, that, as we have shown, the adjacency condition can be formulated once for both derivation and inflection. 43. Lefebvre and Muysken (in press) point out that inflectional affixes and derivational affixes in Quechua share some formal properties: (a) both classes obey the same morpheme structure constraints, (b) all affixes trigger stress-shift, (c) all affixes are involved in the same type of cyclic interpretation (cf. Muysken 1981), and (d) both types of affixes can trigger vowel lowering and vowel shortening processes. 44. Answering the criticisms raised by Anderson (1982), Lapointe (1983), in fact, has proposed a modified version of the SLH. 45. Cf., among others, Bauer (1983:XIII). 46. Cf. also Beard (1981), Dowty (1975), etc.

Symbols and Abbreviations A α abstr AdC Adv AdvP Af Afr anim AP AR('s) Art AS AtC Aux BBH BIC Bnd BR C Cat com compl con conj CR('s) D dat Det Diet DR('s) DS EH Ε LH Engl EOH ER ('s) EWP fem feat

Adjective/Actor (identity symbol) Abstract Adjacency Condition Adverb Adverb Phrase Affix Afrikaans Animate Adjective Phrase Allomorphy Rule(s) Article Argument Structure Atom Condition Auxiliary Binary Branching Hypothesis Boundary Insertion Convention Boundary Blocking Rule Consonant Category Common Complement Conjugation Conjunction Compounding Rule(s) Dutch Dative Determiner Dictionary Derivation Rule(s) Deep Structure Epicycle Hypothesis Extended Lexical Hypothesis English Extended Ordering Hypothesis Evaluative Rule(s) Extended Word and Paradigm (Model) Feminine Feature

202

Generative

Fr Germ gen Gr hum infl intr IR('s) It L Lat LI masc mod morph comp MTP MUBH Ν nom NP NPC NSR obj Ρ pi Pre pres prog PP PSR('s) Q R RHR RR('s) Rum S σ SA SC SD sem SF SLH SOV Sp spec

morphology

French German Genitive Greek Human Inflection Intransitive Inflection Rule(s) Italian Lexicon/Language Latin Lexical Insertion Masculine Modifier Morphological Component Morphological Transformation Prohibition Modified Unitary Base Hypothesis Noun Nominative Noun Phrase No Phrase Constraint Nuclear Stress Rule Object Preposition Plural Prefix Present Progressive Prepositional Phrase/Past Participle Phrase Structure Rule(s) (Boundary: unspecified) Rule Right Hand Rule Readjustment Rule(s) Rumenian Sentence Syllable Structural Analysis Structural Change Structural Description Semantics Subcategorization Frame Strong Lexicalist Hypothesis Subject Object Verb Spanish Specifier

Symbols and Abbreviations ST str Suf

svo Sw sX T('s) th tr TR('s) TV UBH V V-Del voc VP

w

WBH WFR('s) WLH W, Χ, Υ, Ζ Χ, Υ, z

Standard Theory stress/strong Suffix Subject Verb Object Swedish Stem of X Transformation(s) Theme Transitive Truncation Rule(s) Thematic Vowel Unitary Base Hypothesis Verb/Vowel Vowel Deletion Vocative Verb Phrase Word Word Based Hypothesis Word Formation Rule(s) Weak Lexicalist Hypothesis (variable symbols) (major lexical category symbols)

Subject Index

Ablaut: 171, 172 Absolute exceptions: 13, 14, 19 Accidental gaps (cf. also lexical idiosyncrasies): 26, 114 Acronyms: 98 Actor: 181, 182 Adjacency Condition (AdC): 50, 53, 55, 137, 170-177, 194, 197-200 Adjacent morphemes: 50, 51 Affixation: cf. Derivation Afrikaans (Afr): 72, 136, 165 Agreement: 192 Allomorphy: 39, 58, 60, 61, 63, 65, 66-69, 97 Allomorphy Rules (AR's): 58, 60, 61, 63, 66-69, 97, 142 Ambiguity: 10 Argument Structure (AS): 78, 167, 180-183, 197, 199 Atom Condition (AtC): 137, 171-177, 197, 199 Atomic core: 98, 99 Autonomy of systems: 194 Bahasa: 199 Back formation: 118, 119 Base Component: 1, 3, 19, 20, 57, 101 Base of WFR's: 28, 29, 31, 32, 40-42, 44-51, 54, 58, 60, 62, 78-80, 105-109, 111-114, 137-140, 142, 150, 155, 157, 161, 164, 172 Base Rules: cf. Rewriting Rules Binary Branching Hypothesis (BBH): 137, 146-148, 150, 151, 163 Blendings: 98 Blocking and Blocking Rule: 51, 114, 126, 130, 135-137, 156-165 Bound forms: 8, 38, 40, 73, 74, 81, 98, 187 Boundaries: 10, 28, 42, 52, 55, 57, 68, 69, 76-82, 84, 88-90,93-95,98,99, 121,123,127 ' + ' : 28, 42, 55, 69, 77, 78, 82-84, 86-89, 94, 95,97,99, 116, 157 ' # ' : 10, 55, 76-78, 82-84, 86-90, 94, 95, 97,

99, 116, 117, 157 ' # # ' : 10, 94, 95, 97, 121 external: 55, 76-82, 84, 88-90, 95, 98, 99 internal: 28, 55,68,69,77-82, 84, 88,90,93, 95, 98, 99 strong: 94, 95, 99, 121, 127 Boundary Insertion Convention (BIC): 77, 78, 80, 81, 97-99 Boundary Weakening Rule: 94, 127 Bracket Erasure Convention: 198 Breton: 192 Case: 19, 30, 104, 123, 191, 192 C-command: 198 Chinese: 200 Class I affixes: cf. Level I affixes Class II affixes: cf. Level II affixes Clippings: 98 Clitics: 130, 183-185, 191, 199 Complex core: 98 Comparative forms: 105 Complement: 186, 190, 191 Compounding Rules (CR's): 13,62,65,66, 71, 74, 80, 88-90, 93, 95-98, 101, 115-124, 126, 134-136 Compounds: 8-13, 34, 35, 62, 74-76, 90-96, 113, 116-127, 134-136, 170, 187, 189, 190, 191, 199, 200 endocentric: 96, 136, 190 exocentric: 126, 190 left-headed: 96 lexicalized: 93, 118-122 loose: 95, 121, 122, 136 nominal: 9, 10, 118 primary: 90, 91, 200 strict: 95, 121, 122, 136 synthetic: 90 transparent: 93 and Adjacency: 170 Conditions on WFR's (cf. also Restrictions): negative: 35, 50, 51

206

Generative

morphology

positive: 48, 50, 51, 154 Conjugation (Con): 105, 106, 109, 110 Copying rules: 198 Cycle: 170, 172, 174, 175, 198, 200 Cyclic stress assignment: 22, 83, 85 Cyclic WFR's: 136 Danish: 55 Declension: 105, 106, 109, 110 Deep Structure (DS): 5, 7-13, 17, 23, 81, 186 Defective verbs: 114 Deletion (of lexical material): 11, 12, 19 Derivation: 19-22, 24-34, 79-90, 102-115, 192-196, 199, 200 and inflection: 102-115 and compounding: 115-122 and diminutives: 131-133 Derivation Rules (DR's): 13, 14, 30-32,46,47, 53, 55, 62, 65, 71, 74, 78-80, 84-90, 95-98, 101-122, 124, 126, 127, 129, 131, 133-137, 139, 148, 149 Descriptive adequacy: 189 Dictionary (Diet): 14,15,26-35,41,43,44,59, 68, 71, 74, 76, 81, 85, 93, 97, 98, 104, 105 Diminutives (cf. also Evaluative suffixes): 131, 136, 143 Dominating feature list: 195, 196 Dummy symbol: 5-8, 11, 44, 186, 187 Dutch (D): 46,47,49,53,54,61,68,69,72-74, 87-89, 113, 116, 136, 144, 164, 179, 180 Economy (in the lexicon): 22,41,157,163,164 English (Engl): 23, 24,33,40,46, 52,62,72-74, 82, 87, 96-99, 113, 115, 116, 118-120, 124126, 129, 130, 135, 137, 138, 156, 157, 160, 164, 171, 176, 180, 182, 189, 190, 200 Epicycle Hypothesis (EH): 153 Evaluative Rules (ER's)/suffixes: 101, 127, 131-134, 136 Exceptions: 12-14, 22, 25, 32, 55, 99 Explanatory adequacy: 11, 14, 15 Extended Dictionary: 187 Extended Word and Paradigm Model (EWP): 192-196 Features: abstract: 48, 154 diacritic: 96 exception; 14, 15 inherent: 3-6, 28, 35, 42, 105 morphological: 48-50 phonological: 3

selectional: 3-5, 8, 17, 21, 22,28,35,45,105, 107, 109-111 semantic: 3, 15, 91, 92, 105, 107-112 stratal: 49, 61, 75, 78 subcategorization: 3-5, 8, 45, 78, 105-107, 109, 110, 132 syntactic: 3-5, 8, 17, 21, 22, 28, 35, 42, 45, 55, 57, 78, 96, 105-107, 109-111, 132, 133, 139, 144 syntactic category: 139, 140 [±abstract]: 3,4, 15,28,31,42,55,105,108, 110, 112, 132, 157, 161 [±animate]: 2-5,45, 105, 107, 108, 110-112 [ i c o m m o n ] : 3 , 4 , 105, 109, 110, 112 [ i c o u n t ] : 3 , 4 , 105, 107, 108, 110-112, 126, 171 [±French]: 49 [±Greek]: 49 [ i h u m a n ] : 3, 4, 28, 111 [±latinate]: 25, 38, 39,49,50,59,76,99,154 [±learned]: 61 [±native]: 61 Feminine: cf. Gender Filter: 24, 26-34, 41, 55 Filtering function: 165 Free forms: 73, 74, 81, 98, 187 French (Fr): 32, 35, 45, 49, 52-55, 61, 72, 98, 103, 104, 160, 170 Functional theory: 198 Gender: 19, 55, 104, 105, 130, 136 Feminine: 55, 104, 105, 130 Masculine: 55, 104 Generative capacity: 189 German (Germ): 53, 54, 72, 135, 136 Global rules: 15, 32, 33, 44 Goal: 181, 182 Greek (Gr): 73-76 Hapax legomena: 38 Head: 96, 97, 99, 124-127, 134, 136, 175-177, 180, 186, 190, 191 Idiomatic expressions: 113 Idiosyncrasies: 3, 12,13,15,18,21,22,25,26, 28, 31, 35, 41, 42, 107, 112, 113, 126, 168, 169, 192, 199 lexical: 13, 25, 26, 31, 114, 136 phonological: 25, 26 semantic: 12, 18, 21, 25, 31, 107, 112, 113, 126, 156 syntactic: 13, 21

207

Subject Index Incohative verbs: 29, 35 Infixes: 140 Inflection (Infi): 19, 28, 30, 35, 52, 55, 62, 7274, 81, 95, 98, 102, 106, 109, 113-115, 122-127, 129-131, 136, 191-198, 200 and derivation: 102-115, 191-194 and c o m p o u n d i n g : 122-127, 170, 171 Inflection Rules (IR's): 35, 49, 52, 55, 65, 66, 8 1 , 9 7 , 99, 101-103, 105-110, 112-115,122, 125, 128, 130-135, 148, 167, 184 Inflection (as node building rules): 136 Inflectional m o r p h e m e s / e n d i n g s : 81, 112, 114, 115, 122, 123, 133, 134, 148, 150 Inflectional Parsimony Hypothesis: 135 Internal structure: cf. Structure of complex words Internality principle: 184 " I S A " Condition: 90, 92, 93, 124 Islands: 93 Italian (It): 1 3 , 2 0 , 3 5 , 4 7 , 4 8 , 5 3 - 5 5 , 5 8 , 6 2 , 6 5 , 67, 68, 73, 74, 79, 87, 88, 99, 103-116, 119122, 124-127, 129-132, 134-136, 138-140, 144, 145, 147-150, 156, 170, 171, 173-176, 180, 183-185, 190, 191, 194, 199 Item and arrangement model: 32 Japanese: 99, 200 Latin (Lat): 53, 73, 74, 76, 88, 104, 135, 190 Learned Backing Rule: 170 Learned lexicon: 75, 76 Level I affixes: 23, 81-89, 95, 99, 116 Level II affixes: 23, 81-89, 95, 99, 116, 117, 122, 155 Level Ordering Hypothesis: cf. Ordering Hypothesis Lexemes: 115 Lexical category: cf. M a j o r syntactic category Lexical c o m p o n e n t : 1, 8, 14, 19, 20, 24, 26,30, 34, 35, 42, 43, 52, 57, 61, 63, 68, 69, 78, 83, 84, 87, 97, 98, 101, 102, 134, 135, 164, 165 Lexical core: 98 Lexical formatives: 1, 3, 71, 76, 78, 79, 99 Lexical Insertion (LI): 5-8, 15, 30, 35, 57, 68, 79, 81, 113, 187 Lexicalist Hypothesis: 17-20, 23, 24, 35, 37, 41, 52, 101-103, 133, 168, 193-195 Extended (ELH): 19, 20, 35 Generalized: 35, 101 Strong (SLH): 35, 52, 101-103, 133, 191, 194, 196, 197 Weak ( W L H ) : 101-103, 194

Lexicalized items: 15, 118, 122, 144,155,156, 164 Lexical m o r p h e m e s : 32, 44, 98 Lexical phonology: 99, 193, 194 Lexical representation: 78-81, 194 Lexicon (cf. also Lexical c o m p o n e n t ) : 1, 3-5, 8, 12-15, 17, 23, 24, 31, 41, 43, 48, 49, 61, 66, 74-76, 94, 99, 105, 122, 144, 155, 157159, 163-165, 191, 193-195, 197 Liaison: 98 Linear contiguity hypothesis: 169, 170 List of m o r p h e m e s : 26-34, 41, 44 Locality: 167-169, 178, 197, 200 Logical f o r m : 193 L o o p : 126, 134, 136 M a j o r Category Restriction: 71 M a j o r syntactic category: 3 - 5 , 8 , 2 4 , 2 5 , 2 8 , 3 0 , 35, 40, 42, 45, 51, 52, 54, 55, 58, 69, 71, 75, 76, 7 8 - 8 1 , 9 2 , 9 6 , 9 8 , 1 0 2 , 1 0 5 , 1 0 9 , 1 1 0 , 1 2 0 , 130, 134, 148 Malayalam: 116 Markedness: 48, 75 Masculine: cf. G e n d e r Maximal nesting condition: 164 Minimal sign: 37, 40, 63 Modified Unitary Base Hypothesis ( M U B H ) : 138, 141, 143, 144, 146 M o d u l a r hypothesis: 20, 167, 196 Morphemes: 24, 25, 27, 31, 32, 37-42, 44, 50, 51, 54, 57, 59, 62, 63, 66, 67, 69, 73, 98 M o r p h e m e based hypothesis/theory: 41, 44, 62 Morphological c o m p o n e n t : 20, 24, 26, 30, 31, 84, 97, 167, 170, 197 Morphological properties: 191, 192 Morphological T r a n s f o r m a t i o n Prohibition (MTP): 167 M o r p h o p h o n e m i c rules: 58 Movement rules: 168, 169, 199 Nasal assimilation: 82, 84, 88 N a t u r a l theory: 198 N o m i n a l i z a t i o n s / N o m i n a l s : 13, 14, 17-20,41, 50, 51, 65, 106, 114, 159 Non-linear theory: 198 N o Phrase Constraint (NPC): 72, 137, 154156, 164, 165 Nuclear Stress Rule (NSR): 22, 23 N u m b e r ; 19, 73, 98, 107, 108, 113, 114, 123126, 130 plural: 19,98, 107, 108, 113, 114, 119, 123-

208

Generative

morphology

126, 130, 136 singular: 73, 107, 108, 113, 125 Object: 2, 3, 5, 9, 10, 14, 45, 111, 178, 181, 199 One affix, one rule hypothesis: 137, 146, 151, 154 Order of WFR's: 151, 199 cyclic: 153 extrinsic: 152, 154, 163 intrinsic: 152 Ordering Hypothesis: 83, 85, 87, 88, 90, 95, 116, 129 Extended (HOH): 93, 95, 99, 101, 116-119, 122, 126, 134, 136 Output of WFR's: 37, 40, 44, 51-54, 59, 6166, 85, 113, 135 Overgenerating morphology: 72,99,113,114, 150, 151 Parasynthetic derivation: 21, 106, 140, 147151, 164 Past Participle (PP): 101, 127-131, 134 Percolation: 96, 97, 172, 174, 175, 180, 181, 195 Person: 19, 113 Phonological alternations: 39, 57, 61, 69 Phonological component: 1, 28, 30, 31,43,57, 66, 69 Phonological properties of complex words: 41, 42 Phonological rules: 17, 22, 23, 26, 28, 30, 31, 43, 48, 49, 57, 63, 66, 67, 82, 85, 87-90, 94, 95, 121, 122 Phrase markers: 2, 6-8 Phrase structure rules: cf. Rewriting rules Plural: cf. Number Pluralia tantum words: 114 Polycentristic theory: 198 Potentiation: 176, 199 Prefixes: 28, 32, 35, 38, 40, 45, 46, 50, 51, 55, 75, 76, 79-89,97,99, 116,119,128,135,136, 138, 147-149 Primitives: 63, 194 Productivity: 18, 50, 53, 74, 75,112,114,118, 120, 121, 125, 134, 138, 141, 144, 149, 154, 156-158, 160, 161, 163-165 Quechua: 199, 200 Raddoppiamento sintattico: 185 Readjustment Rules (RR's): 37, 52-55, 57-59,

60-69, 75-77 , 89, 97, 104, 136, 185, 194, 198, 199 Realization rules: 182 Recoverability principle: 11, 12 Recursivity: 136, 188 Redundancy rules: 3, 4, 43, 178 Representation of lexical items (cf. also lexical formatives): 3, 4, 28, 34, 44, 71, 78-81, 97 Restrictions on Allomorphy Rules: 66, 67, 69 Restrictions on WFR's: 37, 44-54, 64, 156 morphological: 44, 48-51, 157 phonological: 29-31, 35, 44, 46-48 semantic: 44, 45, 51, 53 syntactic: 44, 45, 51, 52 Rewriting Rules: 1-3, 6-8, 19, 78, 164, 168, 186-189, 194, 199 Right Head Rule (RHR): 96, 124 Rival affixes: 153, 159, 163, 164 Romance languages: 103, 104, 131, 183, 190 Roots: 20, 25, 49, 58, 60, 61, 98, 141, 165 Rule immunity: 199 Rumenian: 53, 72, 73 Russian: 99 s-voicing: 185 Sandhi processes: 94 Sanskrit: 74 Schwa deletion: 89, 90 Selectional restrictions: 3, 4, 21, 22, 28, 45 Self embedding: 188 Semantic amalgamation: 94, 121, 122 Semantic coherence: 157, 158 Semantic compositionality: 31, 42, 50, 53, 112, 127 Semantic distance: 161 Semantic drift: 54, 94 Semantic properties of complex words: 28,41, 42, 105-109, 134 Semantics of WFR's: 42, 50, 53, 54,90,91,92, 112, 137, 140, 145, 146, 149 Semitic languages: 194 Singular: cf. Number Source: 181 Spanish (Sp): 53, 72, 104 Speaker's intuitions: 11, 159 Specifier: 186, 190 Standard Theory (ST): 1-5, 17 Stems: 26, 27, 31, 35, 38-40, 59, 71, 74, 76, 79-82, 84, 86, 87, 97, 98, 121, 135, 165 bound: 38, 39 Latinate: 39 learned: 75, 76

Subject Index Stress natural affixes: cf. Level I affixes Stress shifting affixes: cf. Level II affixes Stretchable suffixes: 155, 156 Stress shift rules: 23, 81-84, 87, 89 Strong b o u n d a r y condition: 94, 121 Strong verbs: 25 Structure of complex words: 9, 13, 18-21, 24, 27, 31, 32, 41, 43, 50, 53, 77, 78, 81-86, 88, 93, 95, 98, 99, 101, 115-117, 119, 121, 124, 125, 130, 136, 145, 147, 155, 170, 171, 173175, 188, 196, 199 binary: 146-151, 168 ternary: 148, 149 S V O / S O V : 190, 200 Subcategorization Frames (SF): 3-5, 28, 42, 78,111,132,133,167,168,178,181-183,197, 199 Subtraction rules: 55, 57 Subjacency condition: 198 Subject: 2, 9, 10, 111, 178, 181 Suffixes: 20, 23, 25, 28, 31, 32, 35, 39, 42, 44, 47, 50, 52, 55, 58-63, 65, 68, 75, 76, 79-89, 97,99, 105, 108, 114,116,118-121,133,135, 136, 139-144, 147-150, 152-156, 158-162, 173-177, 179, 182, 187, 188 Suppletion: 105, 195 Surface structure: 10, 11, 13, 30, 52, 57, 62, 73 Swedish: 72, 73 Syllabification rules: 89 Syntactic c o m p o n e n t : 1, 3, 8, 9, 17, 19, 20, 32, 57, 101, 167 Syntactic properties of complex words: 28,41, 42, 105-109, 134 Tagalog: 52, 99, 115 Tense: 19, 106, 113 Thematic derivatives: 74 Thematic relations: 181 Thematic Vowel (TV): 74, 106, 141 Theme (th): 74, 141, 199 Theme: 181, 182 T r a n s f o r m a t i o n a l c o m p o n e n t : 1, 8, 9, 17, 20, 32, 194 Transformational approach to word f o r m a tion: 8-15, 17-24 Transformations: 3, 6, 9-15, 17, 19, 20, 23, 35, 64, 69, 77, 78, 101, 102, 167, 168, 197 aux-shift: 10 boundary insertion: 10 Chomsky-adjunction: 69 dative-shift: 19 deletion: 3, 10, 11, 14, 102

209 morpholexical insertion: 187 movement: 19, 102 nominalization: 13, 17 noun-shift: 10 relative: 10 Transitive/intransitive: 2-4, 8 , 4 2 , 4 5 , 63, 110, 128, 132, 196 Trisyllabic laxing rule: 26 Truncation Rules (TR's): 58-62, 64,66-68, 97, 158 Uninflected base hypothesis: 103, 105, 126 Unitary Base Hypothesis ( U B H ) : 4 4 , 1 3 7 , 1 3 9 , 141, 163, 164 Unitary O u t p u t Hypothesis: 137, 138, 143, 144 Variable R Condition: 90-92 Velar softening rule: 48, 49 Vowel a r m o n y : 184 Vowel Deletion (rule) (V-Del): 59, 65, 66, 68, 69, 121, 142 Well-formedness conditions: 71, 80, 95-98, 132 Welsh: 95, 116, 121 Words: 71-75 abstract: 52, 72 complex: 8, 9, 17, 19, 21, 22, 27-29, 31, 42, 43, 48, 50, 51, 53, 59, 62, 74, 75, 77, 80, 81, 85, 86, 90, 96, 97, 113, 114, 123, 124, 135, 136, 188 existent: 29, 30,40-43, 71, 72, 113, 119, 132, 149-151, 163, 164 n a t i v e / n o n native: 30,40-43, 51, 71,90, 150 inflected: 30, 35, 52, 72-74, 102-105, 113, 115, 124, 126, 130, 131, 135 morphological: 187 new: 30, 40-43, 51, 71, 90, 150 phonological: 191 possible: 12, 1 3 , 2 6 , 3 1 , 3 2 , 3 4 , 3 5 , 4 0 , 4 1 , 4 4 , 72, 102, 114, 119, 122, 149-151, 164 simple: 8, 21, 76-78, 81, 85, 86, 93, 113, 188 syntactic: 187 uninflected: 52, 115, 122 word forms: 115 Word b a r theory: 186, 199 Word Based Hypothesis ( W B H ) / m o r p h o logy: 27, 34, 37, 40-44, 54, 61-63,69, 71, 72, 74, 98, 151, 154 Word F o r m a t i o n Rules ( W F R ' s ) : 8, 11, 17, 19, 20, 24-35, 37,40,42-54, 57, 59-61,63,68,

210

Generative

morphology

71, 73, 77-86, 90, 93, 97, 99, 101, 112, 115, 128, 130, 132, 135, 137, 140, 146, 151, 154156, 158, 160, 162, 167-169, 181, 197, 198 as "once only" rules: 43 Word order: 190, 197, 200

Word structure autonomy condition: 102 Word structure rules: 187 χ theory: 19, 138, 187, 188, 190, 191, 197 Zero affixation: 54, 129, 130, 131

Affix Index

a- (It): 185 -a (It): 142, 143 -a (It): 62, 79, 106 -aal (D): 61, 89 - a a n (D): 89 -aar (D): 61 -abile/-ibile (It): 88, 103, 107, 111-113, 174 -able (Engl): 3 9 , 4 2 , 4 4 , 4 5 , 5 0 , 5 2 , 5 3 , 8 5 , 137, 145, 157, 182 -acchiare/icchiare (It): 110 -accio (It): 131 -achtig (D): 89, 164 -age (D): 89 -aggio (It): 106 -aio (It): 55, 108, 110, 112, 119, 156 -aia (It): 110, 112 -air (D): 61, 89 -al (Engl): 1 7 , 2 3 , 2 7 , 2 9 , 3 1 , 3 3 , 3 5 , 5 0 , 5 1 , 52, 79, 82, 116, 152, 153, 165, 196 all- (Engl): 136 -ame (It): 112 anglo- (Engl): 75 -ant (Engl): 27, 159 -ante (It): 112 anti- (Engl): 75 -anza (It): 139 -are (It): 62, 114, 145, 147 -ario (It): 113, 114, 120, 147 -astro (It): 59 - a t a / - i t a / - u t a (It): 103, 106, 114, 140-143,164 -ate (Engl): 62-64, 82, 84, 159 -(a)(t)ion (Engl): 13-15, 23, 27, 29, 31, 32, 39, 52, 57, 60, 61, 62, 65, 82, 84, 152-154 - a t o (It): 112, 174 be- (D): 179 be- (Engl): 154 bio-(Engl): bio- (It): -cide (Engl):

circum- (Engl): 164 cis- (It): 185 co- (Engl): 138, 164 con- (Engl): 39 con- (It): 87 c o n t r o - (It): 138 counter- (Engl): 164 -erat (Engl): 75 -cy (Engl): 160 de- (Engl): 39, 75, 79, 80, 82, 117 d e - ( I t ) : 110, di- (It): 147, 164 dis- (Engl): 75, 80, 138, 164, 169, 170, 177 dis- (It): 174, 175, 185 - d o m (Engl): 52 -e (Gr): 74 -e (It): 65, 125, 126, 130 -ed (Engl): 84, 99, 127, 128, 129, 131, 171, 172 -ee (Engl): 45, 63, 64 -eel (D): 61 -eggia(re) (It): 173 -el (D): 179 electro- (Engl): 75 -ello (It): 131 en- (Engl): 50, 51, 80, 99, 154, 176 -en (Engl): 27-29, 30, 46, 52 -en (Engl, PP): 30 -ent (Engl): 159 -er (Engl): 14, 53, 118 -erellare (It): 132 -erello (It): 132 -ery (Afr): 165 -es (Gr): 73 -ese (It): 109, 112 -esimo (It): 139 -etä (It): 135 -eto (It): 108 -etto (It): 131 -eur (D): 89

212

Generative

morphology

-eus (D): 61 -eux (Fr): 160, 161 -evole (It): 114, 174 ex- (D): 179 ex- (Engl): 99, 164 extra- (Engl): 136 -ezza (It): 65, 87, -fer (Engl): 76 - f o n o (It): 79, 80 fore- (Engl): 164 f r a n c o - (Engl): 75 -ful (Engl): 50, 84 ge- (D): 179 - g r a m m a (It): 80 -graphy (Engl): 76 -heid ( D ) : 136, 164, 179 -heit (Germ): 136 - h o o d (Engl): 28, 31, 42, 49 -i (It): 73, 103, 113, 115, 125, 126 -i (Lat): 73, 98 -i (Rum): 73 -ia (It): 59, 139 -iamo (It): 103 -ic (Engl): 50, 75, 79, 82, 84, 85, 116, 165 -ic- (It): 139, 140 -ico (It): 58, 59, 68, 80, 87, 99, 103, 113, 114, 120, 121, 150, 183 -id (Engl): 50 -ien (Fr): 163 -iera (It): 107, 108, 111, 119, 120 -iere (It): 67, 112, 133 -ifica(re) (It): 20, 21 -ify (Engl): 52 -ile (Engl): 157 -ile (It): 112 -ime (It): 139 in- (Engl): 75, 79, 80, 82-85, 116, 164 in- (Engl): 39 in- (Fr): 46 in- (It): 87, 88, 119, 120, 149, 174, 175 in- (It): 147-149 -ing (Engl): 18, 50, 82, 84 -ingo (It): 140 -ino (It): 120, 131, 132, 143 -ino (It): 112, 140, 143-146 inter- (Engl): 138, 164 inter- (It): 138 iper- (It): 97

-ique (Fr): 162 -ire (It): 147-149 -ish (Engl): 27, 52, 82, 84, 155 -ism (Engl): 33, 35, 152 -isme (Fr): 161, 162 -ismo (It): 140, 141, 156 -issimo (It): 114 -ist (D): 144, 145 -ista (It): 108,109, 112-114,140,141,147,150, 156, 183 -iste (Fr): 161, 162 -istico (It): 119, 120, 150 -itä (It): 65, 87, 88, 113, 114, 135, 147, 174 -ite (Fr): 160 -iteit (D): 49 -itude (Engl): 84 -ity (Engl): 27, 49, 50, 77, 78, 80-84, 99, 137, 154, 157-159 -ive (Engl): 39, 69, 82, 84, 157 -ivo (It): 67 -ize (Engl): 17, 153, 182, 196 -izeer (D): 179 -izza(re) (It): 87, 173 -je (D): 136 -lieh (Germ): 136 -like (Engl): 50, 84 -ling (D): 87 -logy (Engl): 75, 76 -loos (D): 89, 136, 164 -ly (Engl): 50, 52, 82, 84 m a l - ( E n g l ) : 117 -ment (Engl): 18, 23, 50, 51, 82, 176 -mente (It): 104, 105, 113, 147 - m e n t o (It): 106, 114, 173, 176, 199 mid- (Engl): 164 -ness (Engl): 18, 23, 49, 52, 82, 84, 96, 137, 155-160, 188 non- (Engl): 117, 118 -o (Gr): 74 -o (It): 62, 106 -ό (It): 113 -on (Gr): 74 -oor (D): 61 -oos (D): 61 -or (D): 49 -or (Engl): 14, 15, 39 -or (Sw): 73

Affix Index -ory (Engl): 39, 153 -os (Gr): 74 -oso (It): 59, 65, 120, 174 -ou (Gr): 74 - o u s (Engl): 75, 80, 157-159 over- (Engl): 136 para- (D): 136 -phile (Engl): 75, 76 philo- (Engl): 75 p h o n o - (Engl): 79 pre- (Engl): 117, 138 pre- (It): 138 re- (Engl): 32, 39, 45, 82, 96, 117, 164, 188 ri- (It): 120, 164, 185 s- (It): 47, 48, 137, 147, 164 -s (Engl): 30 , 72, 125 -schap (D): 47 -sei (D): 46 - s o m e (Engl): 82, 84 sopra- (It): 138 -ster (D): 47

213 sub- (Engl): 82, 164 super- (Engl): 82, 136 tele- (It): 80 -trice (It): 62, 120 -ucolo (It): 131, 133 -ucchia(re) (It): 132 ultra- (D): 136 un- (Engl): 46, 50, 53, 82-85, 99, 117, 129,169, 170, 177 -ura (It): 103, 106, 112, 114, 119 -uto (It): 65 -uzzo (It): 131, 132 -ver- (D): 179 -vore (Engl): 76 -worthy (Engl): 84 -y (Engl): 50, 82, 84, 155 -zaam (D): 47 -(z)ione (It): 67, 87, 106, 114, 173

Word Index

abbattere: 176 a b b a t t i m e n t o : 176 abbellire: 21 abitato: 48, 174 a b s o r b : 60 absorption: 60 academique: 162 academisme: 162 accentramento: 176 accentrare: 176 acclamare: 176 acclamazione: 176 acceptable: 137 accogliere: 176 accoglimento: 176 acconciare: 164 accrescere: 176 accrescimento: 176 accusation: 60 acidity: 83, 99 acquafrescaio: 119 acquasantiera: 119 acqueux: 161 acquittal: 25 acquosite: 161 action: 18 a d a t t a t o : 174 addolcire: 21 addurre: 176 adduzione: 176 admission office: 136 admit: 35 advance: 23 advancement: 23 affectation: 60 affossamento: 106 affossare: 106 affumicare: 139, 140 afono: 79 agricola: 190 aggress: 14, 15

aggression: 14, 15 aggressor: 14, 15 agrodolce: 120 agter-die-muur-rook-er: air condition: 118 air conditioner: 118 aiuto macchinista: 121 alcolise: 163 alcoolisme: 162 algebra: 99 algebrico: 99 alienabile: 174 alienare: 114 alienazione: 114 allargamento: 106 allargare: 106 allarme: 108 allarmista: 108 alleggerimento: 106 alleggerire: 106 a l l o n t a n a m e n t o : 114 allontanare: 114 allunaggio: 106 allunare: 106 alpe: 143 alpino: 143 altezza: 108 altopiano: 119, 120 amali: 184 amandoli: 184 amareggiavate: 194 amarezza: 65 amarli: 184 amateur: 162 amateurisme: 162 a m b a s s a d e u r : 89 a m m o n i t o : 67 ammonizione: 67 a m ö : 113 analphabete: 162 analphabetisme: 162

216

Generative

morphology

ancorare: 114 ancorato: 174 andirivieni: 120 anglo italo soviet production: 75 anglophile: 76 a n n a t a : 141 a n n o : 141 answer: 5 anxieux: 160 anxiete: 160 apostolico: 87 apprendere: 141 apprendista: 141 apprezzamento: 176 apprezzare: 176 a p p r o b a t i o n : 26, 154 a p p r o f o n d i m e n t o : 106 a p p r o f o n d i r e : 106 approval: 25, 26, 31, 154 approve: 31 aranciata: 141 arancio: 141 argentiere: 67 argento: 67 a r m a : 148 armare: 148 arms: 125 arms conscious: 136 armsrace: 125 A r n o : 109 arresisi: 184 arresosi: 185 arret: 55 arreter: 55 arricchimento: 106 arricchire: 106 arrival: 25-27 arrivare: 141 arrivismo/ista: 141 artigianato: 112 artigiano: 112 artistico: 183 ascoltami: 183 asinino: 143 asino: 143 assimilare: 106 assimilazione: 106 assume: 35 astronaut: 76, 79 a s t r o n a u t a : 120 astronautico: 120 athlete: 162

athletisme: 162 at homeish: 155 atipico: 185 atomico: 58, 68, 103 a t o m o : 58, 68, 103 attaccare: 143 attacchino: 143, 144 attacco: 144 atterraggio: 106 atterrare: 106 attivare: 148 attivo: 148 aunt: 1, 2 autorizzare: 87 a v a n z a m e n t o : 13 avere: 106 avvocatura: 103 avvocato: 103 baas: 47 baasschap: 47 b a d : 55 bade: 55 bagnare: 144 bagnino: 143, 144 bagno: 143 balena: 107 baleniera: 107, 108 baptismal: 33 battere: 176 beautifully: 52 believe: 27 bello: 21 benev: 131 benino: 131 Beobachter: 53 berry: 37 beve: 141 bevuta: 141, 142 bevuto: 142 bewilderment: 50, 154 bezwanger: 179 bilingue: 162 bilinguisme: 162 biochimica: 80 biografico: 120 biografo: 120 black: 49 black and blueness: 155 blackberry: 38 blackboard: 92, 124, 125 blacken: 29

Word Index blackish: 49 blood and thunderish: 155 blu: 59 bluastro: 59 blueberry: 38 blue jay: 92 boek-in-die-bed-lees-ery: 165 bois: 161 b o o k : 1, 2, 4, 5, 78 b o o k - and magazine seller: 199 bottega: 110 bottegaio: 110 boy: 4 , 5 , 2 8 , 3 1 , 4 2 , 4 3 , 6 2 , 7 2 , 7 3 , 9 8 , 1 1 5 , 1 2 4 b o y h o o d : 28, 31, 42, 98 boysenberry: 37 brandweerladderwagenknipperlichtinstallatiemonteurs: 34 bread and buttery: 155 brighten; 46 brother: 27 b r o t h e r h o o d : 49 b r u t t o : 147, 149 buig: 47 buigzaam: 47 buoni: 65 burned: 136 b u r n t : 136 cable television: 92 cagna: 108 cal: 161 clacio: 151 calcistico: 151 calleux: 161 callosite: 161 calmare: 148 c a l m o : 148 calvinisme: 161 c a m m i n a : 141 camminare: 62 c a m m i n a t a : 141, 142 camminato: 142 c a m p o s a n t o : 120 cancellata: 141 cancello: 141 cane: 73, 112, 113 canile: 112 capitalism: 33 capostazione: 62 cappelli: 107 cappelliera: 108 cappello: 107, 108

car: 11 carino: 62 carrettiere: 67 carretto: 67 carto-chiromante: 69 carto-libreria: 69 car thief: 11 casa: 62, 65, 140 casalingo: 140 cattivo: 21 cenere: 170 ceneri: 126, 199 ceppo: 113 cerino: 143 cero: 143 certa: 104 certamente: 104 certo: 104 cessation: 60 chase: 4, 5 cheval: 98 chiarificare: 22 chiaro: 22 childish: 27 chioccia: 145 chiocciare: 145 cicala: 145 cicalare: 145 cisalpino: 185 citta d o r m i t o r i o : 121 civetta: 145 civettare: 145 civile: 48 clansman: 123 climb: 4, 5 closed: 128 c o a u t o r e : 87 c o a u t h o r : 138 coexist: 138 coextensive: 138 cogliere: 176 colombaia: 110, 112 Colombo: 110, 112 color blind: 92 colpevolezza: 87 coltura: 99 c o m a n d a n t e : 112 c o m a n d o : 112 cominciare: 106 c o m f o r t : 85 c o m f o r t a b l e : 85 commence: 178

218

Generative

commendation: 60 commit: 38, 39 commitment: 154 committal: 154 commonness: 49 commune: 60 communion: 60 comparire: 107 compassion: 83 compel: 39 componist: 144, 145 conceive: 38, 39, 65 conception: 65 concretizzarle: 184 condanna: 114 condannare: 114 condensation: 27 conditional: 50 conduce: 38, 39 confer: 38, 39 conformisme: 161 conformiste: 162 conifer: 76 conservador: 104 conservadoramente: 104 consolation: 154 consolement: 154 constitutional: 33 constitutionalism: 33 construct: 19 construction: 18, 19 consultation: 60 consume: 38, 39, 60 consumption: 60 conta: 103 contabile: 103, 107 contare: 107 contamelo: 185 contiamo: 103 contrabbandierucolo: 133 contrabbando: 133 contraddittorio: 21 controargomentare: 138 controintuitivo: 138 controllabile: 174 controrivoluzione: 138 conversion: 20 corista: 150 correttivo: 67 corretto: 48, 67 correzione: 67 cortesia: 139

morphology cosmogonico: 12i cotton tail: 93 courage: 80 courageous: 80 craftsman: 128 cranberry: 37, 93 crescere: 107, 176 cretin: 162 cretinata: 141 cretinisme: 162 cretino: 141 cristallino: 143 cristallo: 143 critical: 49 criticism: 17, 19, 49, 178 criticize: 17, 178 crocerossina: 120 crocevia: 120 crowded: 127 cucchiaiata: 103, 140, 141 cucchiaio: 103, 141 curieux: 160 curiosite: 160 curiosity: 81, 157, 159 curious: 81, 157, 159 curiousness: 81, 159 cut-down: 156 cut-downess: 156 daltonien: 163 daltonisme: 162 dammi: 185 dampen: 29, 46 Dante: 109 darglielo: 199 darken: 27 Darwin: 109 darvinista: 109 dattiloscritto: 120 deaf mute: 92 decente: 119 decentrare: 176 deceive: 38, 39 decide: 60 decision: 60 declamare: 176 declamazione: 176 declaration: 60 dedicatorial: 153 deduce: 38, 39, 80, 82 dedurre: 176 deduzione: 176

Word Index deersman: 123 defend: 69 defensive: 69 defer: 38, 39 deforest: 117 degenerate: 79, 83 degradation: 60 demagnetize: 117 demast: 117 demit: 38, 39 deportation: 60 deprezzamento: 176 deprezzare: 176 deridere: 110, 181 derivation: 23, 26 derive: 23 derubare: 110 description: 26 descrittivo: 67 descritto: 67 descrizione: 67 desensitize: 117 desiderabile: 112 determinare: 141 determinate: 130, 131 determined: 128, 129 determinismo: 140, 141 determinista: 140, 141 dialectology: 75 dictate: 82 difterico: 59 difterite: 59 diglielo: 183 dillo: 183, 185 dimettere: 164 dimostramento: 177 dimostrare: 177 dimostrazione: 199 diocesano: 109 diocesi: 109 diploma: 58 diplomatico: 58 directeur: 61 directoraat: 61 dirigere: 113 dirigibile: 113 dirimpettaio: 156 dirimpetto: 156 dirozzare: 147 disabitato: 48, 175 disadattato: 175 disancorato: 175

discern: 51 discernment: 51 discharge: 80 disciplina: 114 disciplinare: 114 disciplinato: 174 discontent: 138 discontinue: 138 discuss: 25 diseducato: 48 disfavor: 80 dishonest: 80 disonesto: 48, 185 displeasure: 138 disponibile: 199 dissolvibile: 199 distruggere: 21 distruzione: 21 disumano: 48 diver: 52 divine: 25 divinity: 25 dogana: 112 doganiere: 112 dolceamaro: 120 dolcificare: 21 domabile: 174 d o m u s : 135 Donaudampschiffahrgeselleschaft: 34 dopolavoristico: 119, 120 d o p o l a v o r o : 120 doppiolavorista: 156 dormi: 141 dormicchiare: 110 dormire: 110 d o r m i t a : 141, 142 dormito: 142 downer: 71 down towny: 155 drink: 55 drinkable: 52 dryclean: 52 dubiousness: 157 durare: 106 eat: 2 eau: 161 educate: 60 education: 60 educational: 52 educato: 48 egoi'sme: 161

220

Generative

morphology

egoiste: 161 Einheitlichkeit: 136 eis: 46 elettrico: 150 elicitation: 60 elitair: 89 employ: 45, 51, 63, 64 employee: 45, 63, 64, 124 employment: 51 enclose: 63, 80 encroachment: 50 enlarge: 28 enrich: 28, 188 enrichment: 154, 176 entirity: 63 entrust: 80 epicurien: 162 epicurisme: 162 eponge: 161 equivocation: 62 espace: 160 essayist: 145 essere: 106 evacuate: 63 evacuee: 63 every dayness: 155 evocation: 60 evoluzionistico: 59 examination: 60 excited: 128 e x h u m a t i o n : 60 existentialisme: 161 existentialiste: 162 ex-minnaar: 179 faggeto: 108 faggio: 108 falsificare: 21 fama: 59 fame: 80 famoso: 59 famous: 80 famousness: 157 fashionable: 137 fashionableness: 137 fasten: 29 fastidiousness: 52 favoleggiamento: 173 fedupness: 155 felicity: 49 ferroviario: 120 fiets: 47

fietsster: 47 filantropico: 120 finire: 106 finite: 84 fiorista: 150 fire box: 92 fire m a n : 90-92 fire t r u c k : 92 firma: 148 firmare: 148 fish: 135 fleur: 61 flika: 73 florale: 61 flour mill: 11 focherellino: 132 focherello: 132 foglia: 112 fogliame: 112 f o n d a assistenza: 125 fonico: 80 f o n o g r a m m a : 80 forare: 177 formale: 104 formalmente: 104 f o r m a t i o n : 60 formazione base: 121 formlessness: 163 f o r t u n a t o : 47 fotocopiatrice: 120 fragment: 51 fragmental: 51 francese: 109 Francia: 109 francophile: 75 Freiheitheitlichkeit: 136 freshen: 46 friction: 82 friend: 31 frighten: 4, 5, 46 frullare: 143 frullino: 143 frutticoltura: 99 f u m o : 140 fundamentalisme: 61 fundamenteel: 61 fuoco: 132 fuoriuscito: 65 furious: 159 furiousness: 159 fury: 159 f u t u r o : 55

Word Index galvanometer: 75 gatto: 108 g e h o o r z a a m : 179 gehoorzaamheid: 179 gehunker: 179 geophysics: 76 germanophile: 75, 76 giallastro: 59 giallino: 131 giallo: 59, 131 gift: 19 giocare: 132 giocherellare: 132 giornalaio: 55, 112 giornale: 55, 112 giusto: 48 giving: 19 glisten: 46 globality: 50 globe-trot: 118 globe-trotter: 118 gloire: 160 gloria: 55 glorieux: 160 glorify: 52 glorioso: 65 glorious: 159 gloriousness: 159 glory: 159 go: 195 goddelijkheidloos: 164 goddeloosheid: 164 gomme: 161 gommeux: 161 goose: 138 gooseberry: 38 gracious: 159 graciousness: 159 gradevole: 48 grande: 105, 191 grandezza: 105 grandioso: 105 graphomaniac: 76 grassime: 139 grateful: 84 greenish: 59 groei: 47 groeizaam: 47 groenteachtig: 89 groundsman: 123 gruesome: 82 guardare: 140, 141

guardarobiera: 120 guardata: 141 guardingo: 140 guerrafondaio: 156 halfling: 89 handsome: 27 handtowel: 122 hapless: 82 happy: 99 happyness: 49, 83, 188 harden: 29, 46 harmonie: 160 harmonieux: 160 heated: 127 heavily: 83 herbivore: 76 heroic s o u n d i n g : 116 highschool: 93, 96, 189 hippisme: 162 hippos: 74 historic: 79 historiography: 75, 76 history: 79 hit: 57, 181 hoedloos: 89 home: 25 honest: 78 house: 62 huckleberry: 37 hunker: 179 icy cold: 189 idea: 132 ideuzza: 132 ignore: 35 illegal: 82, 84 illegale: 88 imbeccare: 106 imbeccata: 106 imbiancare: 143 imbianchino: 140, 143 imboccare: 106 imboccatura: 106 imbruttire: 147-149 immaculate: 79, 80, 82, 84 immangiabile: 88 i m m a t u r o : 88 immerse: 65 immersion: 65 immorale: 88 impel: 39

222

Generative

morphology

impious: 84 implacable: 84 imply: 82 impopulaire: 46 impoverimento: 106 impoverire: 106 impressionisme: 161 impressioniste: 161 i m p r o b a b l e : 79, 80 i n a d a t t o : 149 inalienabile: 175 incantesimo: 139 incapable: 80 incattivire: 21 incertitude: 84 inciampare: 140 inciampicare: 140 incivile: 48 incollare: 106 incollatura: 106 incolpevole: 87 i n c o m p i u t o : 149 incontrollabile: 175 indecente: 87, 119 indecision: 84 indecisive: 84 indeformable: 75 indeformabilitä: 146, 147 indisciplinato: 175 individuale: 150 individualista: 150 individualistico: 150 indomabile: 175 indorsation: 60 indovinare: 114 induce: 38, 39 indurimento: 106 indurire: 106 industrialization: 153, 154 industrialize: 153 inelegante: 149 inequality: 83 infer: 38, 39 infinite: 84 inflazione: 119 inflazionistico: 119 i n f o r m a t i o n office: 12 infuriare: 106 infuriato: 106 ingannare: 114 inganno: 114 ingiuntivo: 67

ingiunto: 67 ingiusto: 48 ingiunzione: 67 ingrandire: 106 inhospitable looking: 116 innes: 71 inorganic: 84, 85 in-preparato: 175 insano: 48 insecticide: 76 insensible: 48 insicuro: 48 insipido: 87 inspirational: 33 intellettualistico: 59, 150 interdisciplinare: 138 interfaccia: 138 international: 138 interporre: 138 interpretabile: 111 interpretare: 111 interstate: 138 interweave: 138 intristire: 21 inumidire: 21 inutilitä: 88 invecchiare: 21, 147 invendicato: 175 inverosimile: 119, 120 inversion: 52 invert: 52 ipersaturo: 87 Irene: 109 irregular: 82 Italo anglo soviet production: 75 jardin: 73 jobs committee: 136 junction: 154 juncture: 154 junior high school teachers association curriculum planning committee: 34 juste: 45 justesse: 45 justice: 45 juteachtig: 89 kaats: 47 kaatsster: 47 kerk-toe-stap-ery: 165 Kind: 72, 73 kindness: 82, 83

Word Index kindjietjie: 136 king: 14 kingdom: 52 kinsman: 123 kleurloosheidloos: 136 kneus: 46 know: 4, 5 kras: 46 laat-in-die-bed-kom-er: 165 lab: 98 lachen: 74 laid: 55 laide: 55 laine: 161 laineux: 161 large: 45 largesse: 45 largeur: 45 laugh: 62 lava: 113 lavapiatti: 120, 125 lavava: 113 lavoro: 119 lawful: 82, 84 leale: 47 lees: 46 left-handed: 99 legal: 82, 84, 182 legal: 179 legalize: 182 legalizeer: 179 leggerlo: 183 letta: 143 lettere: 125, 136 lettino: 132 letto: 132 lezen: 24 library committee meeting: 151 libro: 73, 109, 115, 191 libros: 72 ligneux: 161 lignosite: 161 limitation: 83 lionman: 92 lodevolezza: 173 lonelyness: 96 lontananza: 139 lovable: 53 lucid: 81 lucidity: 50, 81 lucidness: 81

lume: 132 lumino: 132 lungarno: 121 lupus: 73, 98 lustrascarpe: 13 macchina: 108 macchinista: 108 macinare: 143 macinino: 143 maculate: 79 maestria: 139 maggioranza: 105 maggiorato: 105 maggiore: 104, 105 maggiorenne: 105 maggiormente: 104 magistrato: 112 magistratura: 112 mailman: 92 maladroit: 104 maladroitement: 104 malavitoso: 120 m a l f o r m a t i o n : 117 malnutrition: 117 man: 9, 10 mangiabile: 21, 107, 174 mangiare: 107, 114, 132 mangiata: 114, 140 mangime: 139 mangiucchiare: 132 manifestation: 57 manservant: 9, 10, 13 manualistico: 150 m a r e m m a : 109 m a r e m m a n o : 109 Maria: 109 marriage: 18 materials center: 136 medicate: 49 medicine: 49 mescolare: 177 mescolamela:185 messa: 143 metamorfico: 59 metamorfosi: 59 met-die-hand-in-die-sak-staan-ery: 165 metrico: 150 metro: 150 microscopio: 120 mietitrebbiatrice: 62 mietitrice: 61

224

Generative

morphology

milanese: 109 Milano: 109 miljardair: 89 militair: 61 militarist: 61 militarista: 150 mindful: 50 m i n n a a r : 179 Mittag- u n d Abendessen: 199 m o d a n a r e : 114 m o d a n a t u r a : 114 moisten: 29, 30 moneta: 67 monetiere: 67 m o n o - and trisyllabic: 99 monosyllabe: 83 monstre: 161 monstreux: 161 mostruosite: 161 moralitä: 87 Moravia: 109 mordacious: 158 mordacity: 158 mordiller: 52 mordre: 52 morire: 107 m o r t o : 129 mostrare: 177 motel: 98 m o t h e r h o o d : 49 mountain: 31 mouse trap: 122 munte: 73 muscle: 161 muscolosite: 161 musculeux: 161 musical: 79, 83 musical sounding: 116 musicologico: 120 musicology: 75, 76 nationalize: 195, 196 nationwide: 189 naturale: 150 naturalista: 150 naturalistico: 150 nautisme: 162 nave traghetto: 125 need: 9, 10 neve: 140 nevicare: 139, 140 nice: 62

nicety: 26 nobori: 200 nodosite: 161 noeud: 161 nominate: 63, 64 nominee: 63, 64 non class-conscious: 117 non color-blind: 117 non h o m e - m a d e : 117 noueux: 161 nudista: 150 nudistico: 150 nut: 113 obese: 80 obesity: 26, 27, 80 obscene: 25 obscenity: 25 observador: 53 observateur: 53 observator: 53 odiositä: 173 oftalmico: 87 oliare: 148 olio: 148 onesto: 48 open-airy: 155 organ: 85 organic: 85 organization: 152 organizational: 152, 153 organizzazione: 173 ornament: 23, 51 o r n a m e n t a l : 23, 51 orologiaio: 108, 112 orologio: 108 ortofrutticoltura: 99 osservabile: 174 osservatore: 153 overdo: 93 overdose: 189 overseas investor: 136 overshoot: 92 overwide: 189 paesano: 109 paese: 109 paglia: 111 pagliaio: 111 paint: 54 paleface: 12, 93 pant leg: 123

Word Index

225

pants: 123

plat: 55

paramedical: 83

platte: 55

para-paranormaal: 136

plavei: 4 6

parks commissioner: 136

plaveisel: 4 6

parts distributor: 136

play: 78

passare: 143

poison: 161

passino: 143

pollame: 112

patience: 4, 5

polio: 112

pateras: 73

polsino: 143

pattume: 107, 111

polso: 143

pattumiera: 107, 111

polvere: 107

pay: 63

polveriera: 107

payee: 63

porno: 127

peaceful: 82, 83

pomodore: 127

pecorino: 140

populaire: 4 6

pedata: 141

porta: 170

pedestre: 104

portabandiera: 190

pedestremente: 104

portabile: 107

penna: 143

portaborse: 13

pennino: 143

portacenere: 126, 170, 171

pentitosi: 185

portaerei: 65

perceive: 38

portalettere: 125, 126

perdite: 13

portare: 107

perdition: 13

portascarpe: 190

perforamento: 177

possible: 99

perforare: 177

posta: 143, 144

perforazione:

postino: 140, 143, 144, 145

permission: 39

pouvoir: 5 4

permissive: 39

power: 11

permit: 38, 39

prank: 30

permutation: 18

praticabile: 174

perturbation: 60

pre-air-condition: 117

Perugia: 112

precieux: 160

perugino: 112

preciosite: 160

Petrarca: 109

preciosity: 159

petrarchismo/ista: 109, 140

precious: 159

pettinare: 114

preciousness: 159

pettinatura: 114

precipitevolissimevolmente: 114

philanthropist: 75, 76

precocious: 158

phonic: 79, 82

precocity: 158

phono: 79

prefer: 38, 39

pianista: 109

preistoria: 138

piano: 109

prehistory: 138

piatti: 125 piede: 141

premature: 138

pig: 4 9

preparato: 174

pigish: 49

pre-proof-read: 117

preordinare: 138

Piemonte: 112

prerinascimentale: 138

piemontese: 112

presa: 143

Piero: 109

pressapochismo: 156

pious: 84

pressapoco: 156

piscicapi: 190

prete operaio: 121

226

Generative

morphology

presume: 38 presuppose: 138 prevedere: 107 prevedibile: 107 priest: 28 priesthood: 28 prikkelen: 179 prikken: 179 probable: 79 probity: 82 profanare: 114 profanazione: 114 profane: 25 profanity: 25, 83 programs c o o r d i n a t o r s : 193 prohibit: 23, 39 prohibition: 23, 39 prohibitive: 39 prolongation: 60 promessa: 143 proposal: 25, 26 proposition: 26 prosodia: 58, 59 prosodico: 58, 59 proven: 136 proved: 136 psicologico: 120 pugilistico: 150 purgation: 60 purificare: 20-22 puro: 20 purosangue: 119 questura: 143 questurino: 143 quick: 62 quintessenza: 121 r a a d : 46 raadsel: 46 radar: 98 radiosomaggismo: 156 ragazzo: 62 ramingo: 140 r a m o : 140 rattlesnake: 189 read: 6, 24, 42, 182 readability: 50 readable: 42, 182 re-air-condition: 117 real: 82 realization: 60

realize: 60 rebel: 60 rebellion: 60 receive: 38, 39, 57 reception: 57 recital: 25, 26, 31 recitation: 26 reclean: 117 recommence: 178 reduce: 38, 39 refer: 38, 39 refract: 82 refusal: 17, 19, 25, 28, 31, 83 refuse: 17, 28, 31 refusing: 19 regular: 82 rehearsal: 25, 26 religieus: 61 religieux: 161 religion: 161 religiosite: 161 religiositeit: 61 remail: 117 remission: 39 remissive: 39 remit: 38, 39 repeat: 60 repel: 39 repetition: 60 representation: 60 re-proof-read: 117 resolution: 60 resolve: 60 resolved: 128 restare: 107 restructuration: 32 restructurer: 32 resume: 38, 39, 60 resumption: 60 retroterra: 120 revelation: 60 revolution: 18 revolve: 18 rewash: 82, 83, 117 rewrite: 96 rich: 28 ridacchiare: 110 ridere: 62, 110, 180 rifare: 185 rights organizzation: 136 rigid: 49 rigor: 49

227

Word Index rimescolamento: 177 rimescolare: 177 rincasare: 106 rincaro: 106 risalire: 164 roast beef sandwich platter: 151 roodachtig: 89 roodachtigheid: 164 roller-skate: 118 roller-skating: 118 ros: 73 rosor: 72 rovina: 114 rovinare: 114 rozzo: 147 rubare: 110 Ruf: 54 rule: 14 ruler: 14 ruis: 47 sabino: 48 sale: 143 sales: 125 sales manager: 125 sales oriented: 136 sales receipt: 136 salino: 143 saliscendi: 120 sano: 48 satirico: 150 saturazione: 87 sbarcare: 106 sbarco: 106 sbloccare: 137 s b r a n a m e n t o : 114 sbranare: 114 scaldare: 143 scaldino: 143 scandalize: 30 scarcerare: 106 scarcerazione: 106 scarsity: 26 schoolteacher: 189 scimmiottare: 145 scimmiotto: 145 scissor handle: 123 scissors: 123 sconto: 48 scorretto: 48 scribacchiare: 143 scribacchino: 143

sea breeze: 11 secretaris: 47 secretarisschap: 47 seducente: 48 seemly: 50 segnalibri: 190 semplice: 150 semplicista: 150 semplicistico: 150 sensation: 60 sensibile: 48 serene: 80, 98 serenity: 80, 98 servant: 9, 10 s f o r t u n a t o : 47 sforzo: 48 sgradevole: 48 sgravio: 48 sgocciolare: 137 shorten: 52 shoe d e p a r t m e n t sales manager: 151 Siam: 109 siamese: 109 sicuro: 48 signifer: 190 simbolista: 108, 112 simbolo: 108, 112 simultaneity: 158 simultaneous: 158 sincere: 76-78 sincerity: 77, 78 sing: 55 slapen: 179 sleale: 47 sleeplessness: 163 slegare: 164 slitta: 48 snake poison: 10 snow removal: 90 socialismo/ista: 140 soften: 29 solidificare: 20, 21 solido: 20 solingo: 140 solo: 140 sonante: 48 soprabito: 65, 121 sopranazionale: 138 soprappeso: 138 sopravvalutare: 138 spacieux: 160 spazzare: 143, 144

228

Generative

spazzino: 141, 143, 145 specificity: 50 speciosity: 158 specious: 158, 159 speciousness: 159 speranza: 108 stage-manage: 118 stage-manager: 118 stand: 171 stand-offish: 155 steal: 11 stijf: 46 stijfsel: 46 stivale: 109 stood: 171 stoppa: 143 stoppino: 143 storia: 99, 109 storicistico: 114 storicista: 183 storico: 99 storiografico: 120 strangeness: 49 straw: 38 strawberry: 38 strijkage: 89 structure: 32 student: 30, 73 studenti: 72 stukkietje: 136 stupidata: 164 stupido: 164 s u b h u m a n : 82 submit: 38, 39, 82 subsume: 38 subvert: 65 subvertion: 65 sunshine: 93 superintelligent: 80 superman: 80, 83 supersaturate: 80 supervisory: 83 synodaal: 89 syracusaan: 89 systems analysis: 136 swordsman: 123 tafel: 72, 73, 113 take: 1, 2, 57, 171 tangerineade: 26 tarlare: 145 tarlo: 145

morphology t a r m a : 145 tarmare: 145 tavolino: 131 tavolo: 131 teacher: 30 telefonano: 199 telefonare: 114 telefonata: 114 'elefonista: 108 telefono: 80, 108 thanklessness: 163 t h o r o u g h : 23 thoroughness: 23 tisico: 150 to: 200 tolleranza: 139 took: 171 total: 27 toughen: 46 tozan: 200 transfer: 38, 39 t r a n s f o r m a t i o n : 29 t r a n s f o r m a t i o n a l : 29 transformationalist: 29 transmission: 41, 54 transmit: 38 transmittal: 25 trasmissione radio: 121 trattamento treatment: 82 trebbiatrice: 62 triste: 62 trouser leg: 123 trousers: 123 truck driver: 90 type-write: 118 type-writer: 118 uccelliera: 108 uccello: 107, 108 ultra- u l t r a m o d e r n : 136 u m a n e s i m o : 139 u m a n o : 48 umidificare: 20, 22 umido: 20 umore: 109, 150 umorista: 109, 150 unchildlike: 84 unconfortable: 85 u n d e r a r m : 92 underfeed: 189 understand: 171, 172

Word Index understood: 172 undiscoverable: 170 undistinguished: 170 undrinkable: 24, 27 unfriendly: 84 unfruitful: 84 ungrateful: 84 unhappiness: 84 unhappy: 53 unitarietä: 114 unlawful: 82, 84 unlucky: 84 unorganic: 85 unprecedented: 84 unreal: 82 unsea-worthy: 84 unselfish: 84 unsurprising: 84 unwholesome: 84 uppity: 71 up-to-date: 156 up-to-date-ness: 156 usuraio: 55 usurpation: 60 utile: 113 utilitä: 113 utilitario: 113 utilitarista: 113 utilitaristicamente: 113, 146, 147 utilitaristico: 113 vacant: 27 vacate: 82 vain: 61 vanite: 61 valigia a r m a d i o : 121 variety: 159 various: 159 vecchio: 147 vedendoti: 183 veloce: 104 velocemente: 104 velocitä: 65 vendicato: 174 venduto: 65 veneneux: 161 venenosite: 161 vent: 161 venteux: 161 verdura: 108 verduraio: 108 verhuis: 47

229 verificare: 21 verslapen: 179 vertu: 161 vertueux: 161 vettura: 143 vetturino: 143 vigna: 108 vigneto: 108 vinaio: 55 vino: 55 virtu: 59 virtuosite: 161 virtuoso: 59 vivacity: 49 vomit: 39 waarnemer: 53 walk: 1, 2, 62, 195 walked: 195 warlike: 50 was: 47 wash: 45 washable: 45 water mill: 90, 91 weaken: 46 went: 195 werk: 47 werkzaam: 46 whinchat: 93 whiten: 46 whiteness: 49 wide-awake: 93 wind: 46 windmill: 11, 12, 76, 125 windsel: 46 with-it: 156 with-itness: 156 wijsheidachtig: 164 W o r t : 135 worthy: 50 write: 25 xerox: 98 yama: 200 y a m a n o b o r i : 200 zan: 200 zijdeachtig: 89 zittire: 148 zitto: 148 zoppicare: 139, 140

230

Generative

ζορρο: 140 zuccheriera: 107 zucchero: 107 zwanger: 179

morphology

Index of Names

Alinei, Μ. 140 Allen, R.M. 11, 12, 46, 51, 55, 68, 72, 83-85, 90-96, 99, 102, 116-118, 121-124, 126-128, 136, 137, 150, 154, 157, 165, 169, 170, 199 Anderson, S.R. 69, 102, 191-197, 200 Aoun, Y. 130, 131 Aronoff, M. 23, 27, 34, 35, 37-46, 49-55, 58, 60-64, 66, 68, 69, 71-74, 99, 102, 107, 114, 115, 126, 137, 138, 141, 145, 146, 150, 153, 154, 156-160, 163, 165, 169, 198 Bach, E. 6, 15 Basbell, H. 55 Bauer, L. 68, 71, 99, 156, 164, 199, 200 Beard, R. 200 Bloomfield, L. 49, 55, 74, 98 de Boer, M. and S. Scalise 35 Booij, G. 10, 11, 34, 44,46, 47, 49,55,61,64, 68, 69, 72, 76-78, 88, 89, 99, 102, 105, 113, 116, 136, 150, 164, 198 Botha, R.P. 9, 15, 72, 136, 164, 165 Brame, M.K. 102 Bresnan, J.W. 22, 23, 35 Carlson G. and T. Roeper 199 Carrier, J.L. 52, 65, 69, 99, 102, 115, 135 Carstairs, A. 135 Chapin, P.G. 137, 151-153 Chomsky, N. 2-4,11,15,17-20,22,28,34,35, 37, 40, 63, 101, 102, 139, 187, 198 Chomsky, N. and M. Halle (SPE): 23, 48, 57, 77, 102 Corbin, D. 22, 35, 150-164 Dardano, M. 58 Dell, F. 45, 68 Dell, F. and E. Selkirk 61, 170 Dik, S. 198 Dowty, D.R. 200 Dressler, W. 198 Drigo, M. 190

Freidin, R. 127 Graffi, G. 199 Greenberg, J.H. 103 Gruber, G. 181 Hale, K. 194 Halle, M. 17, 23-35, 37, 41,43,44,54,68, 102 Harris, J.W. 104 Herczeg, 142 Hocke», C.F. 35, 37 Hoekstra, T.H. et al. 19, 20, 35, 72, 74, 187, 199 v.d. Hulst, H. 144 Hyman, L. 94 Jackendoff, R. 13-15, 19, 35, 37,43,54,68,72, 102, 169, 199 Kageyama, T. 99 Katz, J.J. and J.A. Fodor 11 Kayne, R. 198 Keyser, S.J. and P.M. Postal 8 Kiparsky, P. 72, 99, 102, 126, 130, 164, 200 Kuhn, T.S. 198 Lakoff, G. 13-15 Lapointe, S. 101, 102, 195, 200 Lees, R.B. 1, 9-12, 14, 17, 34 Lefebvre, C. and P. Muysken 185, 191, 193195, 199, 200 Lehman, W.P. 200 Lepschy, G. 142 Lepschy, A.L. and G. Lepschy 190, 199 Lieber, R. 8, 54, 68, 72,102,127-130,136,168, 198-200 Lightfoot, D. 200 Marchand, Η. 46, 136 Matthews, P.H. 35, 63, 127 Mayerthaler, W. 136, 198

232

Generative

morphology

McCarthy, J. 102, 167, 168, 194, 198 Meys, W.J. 11 Miller, G.A. 55 Mohanan, K.P. 85, 99, 116, 126, 200 Moody, M.D. 103 Moortgat, M. 144, 178, 179 Muljacic, Z. 48 Muysken, P. 71, 72, 102, 198-200 Pesetsky, D. 80, 85, 96, 99, 132, 198 Platt, D. 102 Postal, P.M. 37, 49 Poser, W. 199 Randall, J.M. 199 v. Riemsdijk, H. 198, 200 Roeper T. and E.A. Siegel 98, 169 Saciuk, B. 48, 49 Scalise, S. 15, 5 5 , 6 5 , 6 8 , 6 9 , 7 2 , 7 4 , 8 8 , 9 6 , 9 9 , 102, 103, 124, 132, 135, 136, 142, 164, 170, 198, 199 Scalise, S. and I. Zannier 199 Scalise, S. et al. 165 Schane, S.A. 55 Schultink, H. 46, 49, 136 Selkirk, L. 68, 72, 77,96, 98,99, 102,123,124, 136, 169, 187-190, 197, 199, 200

Siegel, D. 23, 34, 35, 46, 50, 54, 55,63, 72,77, 78, 83, 85, 98, 99, 102, 113, 116, 127, 137, 165, 169, 170 Strauss, S.L. 87, 88, 102, 136, 164 Tekaviic, P. 139, 140, 149 Thomas-Flinder, T. 102, 200 Toman, J. 187 Tuller, L.A. 102 Vergnaud, J.R. 35, 111 de Villiers, R. 98 Vogel, I. 48 Vogel, I. and S. Scalise 136, 199 Wasow, T. 127, 168, 198, 199 Williams, E. 14, 39, 45, 55, 72, 96, 102, 103, 124, 127, 137, 169, 171, 172, 176, 178, 182, 183, 199 Wurzel, W. 136 Zingarelli, N. 54 Zuffi, S. 136 Zwanenburg, W. 51, 55, 68, 72, 157, 161-163, 165 Zwarts, F. 46 Zwicky, A.M. 184, 199

Bibliography Alinei, Μ. (1962), Dizionario inverse italiano, Mouton, The Hague. Allen, M. R. (1975), Vowel Mutation and Stress In Welsh, Linguistic Imquiry, 6:2, 181-201. Allen, M. R. (1978), Morphological Investigations, Unpublished doctoral dissertation, University of Connecticut. Allen, M. R. (1980), Semantic and Phonological Consequences of Boundaries: A Morphological Analysis of Compounds, in M. Aronoff and M. L. Kean (eds.), Juncture, Amna Libri, 9-27. Anderson, S. R. (1975), On the Interaction of Phonological Rules of Various Types, Journal of Linguistics, 11, 39-63. Anderson, S. R. (1982), Where's Morphology?, Linguistic Inquiry, 13, 571-612. Aoun, Y. (1981), Parts of Speech, Unpublished paper, MIT. Aronoff, M. (1976), Word Formation in Generative Grammar, The M I T Press, Cambridge (Mass.). Bach, E. (1974), Syntactic Theory, Holt, Rinehart and Winston, New York. Basb011, H. (1970), Notes on the Phonology of the Imperative in Danish, ARIPUC, 4, 15-42. Bauer, L. (1979), Against Word-Based Morphology, Linguistic Inquiry, 10:3, 508-9. Bauer, L. (1980), Deux problemes au sujet des noms composes comprenant un premier element verbal en frangais moderne, Lefanqais moderne, 48:3, 219-224. Bauer, L. (1983), English Word-Formation, Cambridge University Press, Cambridge. Beard, R. (1981), The Indo-European Lexicon, North Holland, Amsterdam. Bloomfield, L. (1933), Language, Holt, New York. de Boer, M. G. and S. Scalise (1978), Problemi di morfologia generativa, Lingua e Stile, 4, 551-71. Booij, G. (1977), Dutch Morphology. A Study of Word Formation in Generative Grammar, The Peter de Ridder Presse, Lisse. Booij, G. (1979), Semantic Regularities in Word Formation, Linguistics, 17, 985-1001. Booij, G. (1980), Syllabus morfologie, Unpublished paper, Amsterdam. Botha, R. P. (1968), The Function of the Lexicon in Transformational Generative Grammar, Mouton, The Hague. Botha, R. P. (1981), A Base Rule Theory of Afrikaans Synthetic Compounding, in M. Moortgat and T. Hoekstra (eds.), The Scope of Lexical Rules, Foris, Dordrecht, 1-77. Brame, Μ. K. (1978), The Base Hypothesis and the Spelling Prohibition, Linguistic Analysis, 4, 1-30. Bresnan, J. W. (1971), Sentence Stress and Syntactic Transformations, Language, 47, 257-81. Carlson, G. and T. Roeper (1979), Morphology and Subcategorization: Case and the Unmarked Complex Verb, in T. Hoekstra and M. Moortgat (eds.), Lexical Grammar, Glot, 3/4, 123-64. Carrier, J. L. (1979), The Interaction of Morphological and Phonological Rules in Tagalog: a Study in the Relationship between Rules Components in Grammar, Unpublished doctoral dissertation, MIT Cambridge (Mass.). Carstairs, A. (1981), Notes on Affixes, Clitics and Paradigms, IULC. Chapin, P. G. (1967), On the Syntax of Word-Derivation in English, Unpublished doctoral dissertation, MIT, Cambridge (Mass.). Chapin, P. G. (1970), On Affixation in English, in M. Bierwisch and Κ. E. Heidolph (eds.), Progress in Linguistics, Mouton, The Hague, 51-63. Chomsky, N. (1957), Syntactic Structures, Mouton, The Hague. Chomsky, N. (1965), Aspects of the Theory of Syntax, The MIT Press, Cambridge (Mass.).

234

Bibliography

Chomsky, Ν. (1970), Remarks on Nominalization, in A. Jacobs and P. S. Rosembaun (eds.), Readings in English Transformational Grammar, Ginn and co, Waltham (Mass.), 184-221. Chomsky, N. (1972), Studies on Semantics in Generative Grammar, Mouton, The Hague. Chomsky, N. (1973), Conditions on Transformations, in S. Anderson and P. Kiparsky (eds.), A Festschrift for Morris Halle, Holt, New York, 232-86. Chomsky, N. and M. Halle (1968), The Sound Pattern of English, Harper and Row, New York. Corbin, D. (1976a), Le Statut des exceptions dans le lexique, Langue FranQaise, 30, 90-110. Corbin, D. (1976b), Peut-on faire l'hypothese d'une derivation en morphologie? in J. CI. Chevalier (ed.), Grammaire transformationeile: Syntaxe et lexique. Publications de l'Universite de Lille III, Villeneuve d'Ascq, 49-91. Corbin, D. (1980), Contradictions et inadequations de I'analyse parasyntetique en morphologie derivationelle, in Theories Linguistiques et traditions grammaticales, Presses Universitaires de Lille, Villeneuve d'Ascq, 181-224. Dardano, M. (1978), La formazione delle parole neWitaliano di oggi, Bulzoni, Roma. Dell, F. (1979), La morphologie derivationelle du fran^ais et l'organization de la composante lexicale en grammaire generative, Revue Romane, 2, 185-216. Dell, F. C. and E. Selkirk (1978), O n a Morphologically Governed Vowel Alternation in French, in S. J. Keyser (ed.), Recent Transformational Studies in European Linguistics, Linguistic Inquiry Monograph, 2, 1-51. Dik, S. (1979), Funktionele morfologie, in T. Hoekstra and H.v.d. Hulst (eds.), Morfologie in Nederland, Glot special, 73-100. Dowty, D. R. (1975), Towards a Theory of Word Formation for Montague G r a m m a r , Texas Linguistic Forum, 2, 69-96. Dressier, W. (1977), Elements of a Polycenristic Theory of Word Formation, Wiener Linguistische Gazete, 15, 12-32. Drigo, M. (1983), Per una teoria dei composti. Unpublished dissertation, University of Venice. Freidin, R. (1975), The Analysis of Passives, Language, 51:2, 384-405. Graffi, G. (1980), "Universali di Greenberg" e grammatica generative, Lingua e Stile, 3, 37187. Greenberg, J. H. (1963), Some Universale of G r a m m a r with Particular Reference to the Order of Meaningful Elements, in J. H. Greenberg (ed.), Universals of Language, The M I T Press, Cambridge (Mass.), 58-90. Gruber, G. (1976), Lexical Structure in Syntax and Semantics, N o r t h Holland, Amsterdam. Hale, K. (1976), Linguistic Autonomy and the Linguistics of Carl Vogelin, Anthropological Linguistics, March, 120-8. Halle, M. (1973), Prolegomena to a Theory of Word Formation, Linguistic Inquiry, 4, 3-16. Harris, J. W. (1977), Spanish Vowel Alternation, Diacritic Features and the Structure of the Lexicon, Ν ELS, 7. Herczeg, G. (1972), La funzione del suffiso -ata: sostantivi astratti deverbali, Studi di Grammatica Italiana, 2, 191-260. Hockett, C. F. (1954), Two Models of Grammatical Description, Word, 10, 210-31. Hockett, C. F. (1958), A Course in Modern Linguistics, The Macmillan Company, Toronto. Hoekstra, Τ., H. v.d. Hulst, M. Moortgat (eds.) (1979), Introduction to Lexical Grammar, Glot, 3/4, 1-48. Hyman, L. (1978), Word Demarcation, in J. H. Greenberg (ed.), Universals of Human Language, vol. 2., Stanford University Press, 443-70. JackendofT, R. (1972), Semantic Interpretation in Generative Grammar, The M I T Press, Cambridge (Mass.). JackendofT, R. (1975), Morphological and Semantic Regularities in the Lexicon, Language, 51, 639-71. JackendofT, R. (1977), X Syntax: A Study of Phrase Structure, The M I T Press, Cambridge (Mass.). Kageyama, T. (1982), Word Formation in Japanese, Lingua, 57, 215-58. Katz, J. J. and J. A. F o d o r (1964), The Structure of a Semantic Theory, Language, 39, 170210.

Bibliography

235

Kayne, R. (1981), Unambiguous Paths, in R. May and J. Koster (eds.), Levels of Syntactic Representation, Foris, Dordrecht, 145-83. Keyser, S. J. and P. M. Postal (1976), Beginning English Grammar, Harper and Row, New York. Kiparsky, P. (1973), Phonological Representations, in O. Fujimura (ed.), Three Dimensions of Linguistic Theory, TEC, Tokyo, 1-136. Kiparsky, P. 1982, Lexical Morphology and Phonology, ms. Kuhn, Τ. S. (1962), The Structure of Scientific Revolutions, The University of Chicago Press, Chicago. Lakoff, G. (1970), Irregularity in Syntax, Holt, New York. Lapointe, S. (1978), A Theory of Grammatical Agreement, Unpublished doctoral dissertation, University of Massachussetts, Amherst, MA. Lapointe, S. (1983), A Comparison of two Recent Theories of Agreement, in Papers from the Parasession on the Interplay of Phonology, Morphology and Syntax, Chicago Linguistic Society, Chicago, 122-34. Lees, R. B. (1960), The Grammar of English Nominalizations, Mouton, The Hague. Lefebvre, C. and P. Muysken (in press), Quechua Nominalizations, Reidel, Amsterdam. Lehmann, W. P. (1974), Proto-Indo-European Syntax, University of Texas Press, Austin. Lepschy, A. L. and G. Lepschy (1979), The Italian Language Today, Hutchinson, London. Lepschy, G. (1981), Le parole dello sciacallo. Journal of the Association of Teachers of London, 32, 65-71. Lieber, R. (1980), On the Organization of the Lexicon, Unpublished doctoral dissertation, MIT, Cambridge (Mass.). Lieber, R. (1983), Argument Linking and Compounds in English, Linguistic Inquiry, 2, 251— 85. Lightfoot, D. (1979), Principles of Diachronic Syntax, Cambridge University Press, Cambridge. Marchand, Η. (1969), The categories and Types of Present-Day English Word-Format ion, Snd edn., München, Beck. Matthews, P. H. (1970), Recent Developments in Morphology, in J. Lyons (ed.) New Orizons in Linguistic. Penguin, Harmondsworth, 96-114. Matthews, P. Η. (1974), Morphology: an Introduction to the Theory of Word-Structure, Cambridge University Press. Cambridge. Mayerthaler, W. (1981), Morphologische Natürlichkeit, Athenaion, Wiesbaden. McCarthy, J. (1979) Formal Problems in Semitic Phonology and Morphology, Unpublished doctoral dissertation, MIT, Cambridge (Mass.). Meys, W. J. (1975), Compound Adjectives in English and the Ideal Speaker-Listener, North Holland, Amsterdam. Miller, G. A. (1978), Semantic Relations among Words, in M. Halle, J. Bresnan and G. A. Miller (eds.). Linguistic Theory and Psychological Reality, The MIT Press, Cambridge (Mass.), 60-118. Mohanan, K. P. (1982), Lexical Phonology, IULC. Moody, M. D. (1978), Some Preliminaries to a Theory of Morphology, Glossa, 12, 16-38. Moortgat, M. (1981), Subcategorization and the Notion "Lexical Head", in S. Daalder and M. Gerritsen (eds.), Linguistics in the Netherlands 1981, North Holland, Amsterdam, 4554. Muljacic, Z. (1969), Fonologia generale e fonologia delta lingua italiana, II Mulino, Bologna. Muysken, P. (1981), Quechua Word Structure, in F. Heny (ed.), Binding and Filters, Longmans, London. Pesetsky, D. (1979), Russian Morphology and Lexical Theory, unpublished manuscript, MIT, Cambridge (Mass.). Platt, D. (1981), The Balance between Regularity and Irregularity in Verbal Inflection, in T. Thomas-Flinders (ed.), Inflectional Morphology: Introduction to the Extended Word-andParadigm Theory, UCLA occasional papers, 4, 41-71. Poser, W. (1981), Lexical Rules may Exchange Internal Arguments, The Linguistic Review, 2, 97-100.

236

Bibliography

Postal, P. M. (1969), Anaphoric Islands, in Papers from the Fifth Regional Meeting, Chicago Linguistic Society, 205-39. Randall, J. M. (1984), Thematic Structure and Inheritance, Quaderni di Semantica, 1, 92-110. van Riemsdijk, H. (1982), Locality Principles in Syntax and P h o n o l o g y , in The Linguistic Society of K o r e a (ed.), Linguistics in the Morning Calm, H a n s h i n Publishing Co., Seoul, 693-708. Roeper, Th. and E. A. MufTy Siegel (1978), A Lexical T r a n s f o r m a t i o n for Verbal C o m p o u n d s , Linguistic Inquiry, 9, 199-260. Saciuk, B. (1969), T h e Stratal Division of the Lexicon, Papers in Linguistics, 1, 464-532. Scalise, S. (1980), T o w a r d s an "Extended" Italian M o r p h o l o g y , Journal of Italian Linguistics, 1/2, 197-244. Scalise, S. (1983), Morfologia lessicale, Clesp, Padova. Scalise, S. (in prep.), II ruolo delle Regole di Riaggiustamento. Scalise, S. and I. Z a n n i e r (1982/83), Restrizioni sulle regole di formazione di parola: la condizione di adiacenza e la condizione atomo, Quaderni Patavini di Linguistica, 3, 159210. Scalise, S. et al. (1983), Sulla nozione di "Blocking" in morfologia derivazionale, Lingua e Stile, 2, 243-69. Schane, S. A. (1968), French Phonology and Morphology, T h e M I T Press, C a m b r i d g e (Mass.). Schultink, H. (1962) De morphologische valentie van het ongelede adjectief in modern Nedelands, T h e Hague. Schultink, H. (1974), Dubbele diminutiven in het Afrikaans: en trasformationeel-generative taalbeschrijving, in Taalkunde: η lewe, studies opgedracht aan prof. dr. W. Kempen, K a a p s t a d - J o h a n n e s b u r g , 105-9. Selkirk, L. (1972), The Phrase Phonology of English and French, doctoral dissertation, MIT, Cambridge (Mass.), reproduced by I U L C 1981. Selkirk, L. (1982), The Syntax of Words, The M I T Press, C a m b r i d g e (Mass.). Siegel, D. (1971), Some Lexical Transderivational Constraints in English, unpublished paper, M I T , C a m b r i d g e (Mass.). Siegel, D. (1974), Topics in English Morphology, Unpublished doctoral dissertation, MIT, C a m b r i d g e (Mass.). Siegel, D. (1977), T h e Adjacency Condition and the Theory of Morphology, N E L S , 189- 97 Strauss, S. L. (1979), Against Boundary Distinctions in English Morphology, Linguistic Analysis, 4, 387-419. Strauss, S. L. (1982), Lexicalist Phonology of English and German, Foris, Dordrecht. Tekavcic, P. (1972), Grammatica storica delFitaliano, Vol. III: Lessico, II Mulino, Bologna. Thomas-Flinders, T. (1981), Evidence for Inflection as a Phonological Process, in T. T h o m a s Flinders (ed.), Inflectional morphology: Introduction to the Extended Word-and-Paradigm Theory, U C L A occasional papers, 4, 72-83. T o m a n , J. (1980), Wortsyntax, Unpublished dissertation, University of Köln. Tuller, L. A. (1981), O n Nominal Inflection in Hausa, in T. Thomas-Flinders (ed.). Inflectional Morphology: Introduction to the Extend Word-and-Paradigm Theory, U C L A occasional papers, 4, 117-57. Vergnaud, J. R. (1973), F o r m a l Properties of Lexical Derivations, Quarterly of the Research Laboratory in Electronics, M I T , 108, 280-7.

Progress

Report

de Villiers, R. (1979), Phrasal Categories in Word F o r m a t i o n Rules, Stellenbosh Papers in Linguistics, 2, 39-69. Vogel, I. (1982), La sillaba come unitä fonologica, Zanichelli, Bologna. Vogel, I. and S. Scalise (1982), Secondary Stress in Italian, Lingua, 58, 213-242. Wasow, Th. (1977), Transformations and the Lexicon, in P. Culicover, Th. Wasow and A. Akmajian (eds.). Formal Syntax, Academic Press, New York, 327-60. Williams, E. (1981a). O n the N o t i o n s "Lexically related" and " H e a d of a Word", Linguistic Inquiry, 2, 245-74. Williams, E. (1981b). Argument Structure and Morphology. The Linguistic Review. 1, 81-114. Wurzel, W. (1970). Studien zur deutschen Lautstruktur, Studio Gramatica, 8.

Bibliography

237

Zingarelli, Ν. (1970), Vocabolario delta lingua italiana, Zanichelli, Bologna. Zuffi, S. (1981), The Nominal Composition in Italian. Topics in Generative Morphology, Journal of Italian Linguistics, 2, 1-54. Zwanenburg, W. (1980a), La forme de la composante morphologique, in S. Daalder and M. Gerritsen (eds.), Linguistics in the Netherlands 1980, North-Holland, Amsterdam, 47-57. Zwanenburg, W. (1980b), Form and Meaning in Morphology, Quaderni di Semantica, 2, 327 38. Zwanenburg, W. (1981), Le principe du blocage dans la morphologie derivationelle, in S. Daalder and M. Gerritsen (eds.), Linguistics in the Netherlands 1981, North Holland, Amsterdam, 65-72. Zwarts, F. (1973), -Aar, -arij, en -te, Taalkundig Bulletin van het Nederlands Instituut der Rijksuniversiteit Groningen, 6, 9-23. Zwicky, A. M. (1977), On Clitics, in W. Dressier and Ο. E. Pfeiffer (eds.), Phonologica 1976, Akten der dritten internationalen Phonologie-Tagung, Innsbrucker Beiträge ruz Sprachwissenshaft, Insbruck, 29-39.