The Noun Phrase in Romance and Germanic
Linguistik Aktuell/Linguistics Today (LA) Linguistik AktueU!Linguistics Today (LA) provides a platform for original monograph studies into synchronic and diachronic linguistics. Studies in LA confront empirical and theoretical problems as these are currently discussed in syntax, semantics, morphology, phonology, and systematic pragmatics with the aim to establish robust empirical generalizations within a universalistic perspective.
General Editors VVerneri\braharn
Elly van Gelderen
University ofVienna I Rijksuniversiteit Groningen
Arizona State University
Advisory Editorial Board Josef Bayer University of Konstanz
Christer Platzack University of Lund
Cedric Boeckx ICRENUniversitat Autonoma de Barcelona
Ian Roberts Cambridge University
Guglielmo Cinque University of Venice
McGill University
Liliane Haegeman University of Ghent
StenV!kner University of Aarhus
Hubert Haider University of Salzburg
C. Jan-Wouter Zwart University of Groningen
Lisa deMena Travis
Terje Lohndal University of Maryland
Volume171 The Noun Phrase in Romance and Germanic. Structure, variation, and change Edited by Petra Sleeman and Harry Perridon
The Noun Phrase in Romance and Germanic Structure, variation, and change
Edited by
Petra Sleeman Harry Perridon University of Amsterdam
John Benjamins Publishing Company Amsterdam I Philadelphia
The paper used in this publication meets the minimum requirements of American National Standard for Information Sciences - Permanence of Paper for Printed Library Materials, ANSI z39.48-1984.
Library of Congress Cataloging-in-Publication Data The noun phrase in romance and germanic structnre, variation, and change I edited by Petra Sleeman, Harry Perridon. p. em. (LinguistikAktuell/Linguistics Today, ISSN 0166-o829; v.171)
Includes bibliographical references and index. Grammar, Comparative and general--NoWI phrase. 2. Grammar, Comparative and general--Syntax. 3· Linguistic change. I. Sleeman, Antonia Petronella. II. Perridon, Harry. P2;n..N6794
2011
415:5- -dC22 ISBN 978
90 272 5554 9 (Hb ; alk. paper)
ISBN 978
90 272 8729 8 (Eb)
2010045313
© 2011- John Benjamins B.V. No part of this book may be reproduced in any form, by print, photoprint, microfilm, or any other means, without written permission from the publisher.
John Benjamins Publishing Co.· P.O. Box 36224 · 1020 ME Amsterdam· The Netherlands John Benjamins North America· P.O. Box 27519 ·Philadelphia PA 19118-0519 • usA
Table of contents
Foreword The noun phrase in Germanic and Romance: Common developments and differences Harry Perridon & Petra Sleeman
VII
1
PART I. Variation
Scaling the variation in Romance and Germanic nominalizations Artemis Alexiadou, Gianina Iordachioaia & Flmian Schafer
What all happens when a universal quantifier combines with an interrogative DP
25
41
Robert Cirillo Micro-diversity in Dutch interrogative DPs. A case study on the (dis)continuous wat voor 'n N-construction Norbert Co rver & Matjo Van Koppen
57
Noun phrase structure and movement: A cross-linguistic comparison of such/scldan/solch and so/sti/so Joha.nna L Wood & Sten Vikner A unified structure for Scandinavian DPs
111
Susanne Lohmtann A semantic approach to noun phrase structure and the definite - indefinite distinction in Germanic and Romance
127
Ulla Stroh- Wollin Definite determiners in two English-based creoles: Specificity or definitenessr
Ekaterina Bobyleva
141
VI
The Noun Phrase ln Romance and Germanic: Structure, Variation, and Change
PART II.
Change
Form - function mismatches in (torrnally) definite English noun phrases: Towards a diachronic account
159
Christopher Lucas The emergence of the definite article in English: A contact-induced change?
175
Paola Crisma On the syntax of Romanian definite phrases: Changes in the patterns of definiteness checking Alexandra Cornilescu & Alexandru Nicolae
193
Coexisting structures and competing functions in genitive word order
223
Elisabetta Magni Anaphoric adjectives becoming determiners: A corpus-based account
241
Freek Van de Velde From N to D: Charting the time course of the internal rise of French n-words
257
Viviane D~prez Index
281
Foreword
Most of the papers in this volume were presented at the three-day conference Variation and Change in the structure of the noun phrase in Germanic a.nd Romance: autonomous developments or result of language contact?, which took place in January 2009 at the University of Amsterdam, and which was organized by Harry Perridon, Josep Quer, Petra Sleeman, and Fred Weerman, in collaboration with the ACLC (Amsterdam Center for Language and Communication) research group DP/NP in the Germanic and Romance languages: structure, semantics, acquisition and change. We gratefully acknowledge the financial support of the Royal Netherlands Academy of Arts and Sciences (KNAW). A large number of people have assisted us in editing this volume. We would like to thank Kees Vaes ofBenjamins Publishers and the editors of the series, Werner Abraham and Elly van Gelderen, for their help towards the publication of this volume. We are particularly indebted to the reviewers of the papers: Roberta d':A.l.essandro, Andrei Avram, Larisa Avram, Hans Bennis, Hans den Besten, Ana Maria Brito, Osten Dahl, Viviane Deprez, Marcel den Dikk.en, Jenny Doe*s, Cecilia Falk., Olga Fischer, Giuliana Giusti, Evelien Keizer, Ans van Kemenade, Alain Kihm, Elisabeth van der Linden, Muriel Norde, Phoevos Panagiotidis, Rodie Risselada, Mara van Schaik-Radulescu, Erik Schoodemmer, Mauro Scorretti, Elena Soare, Mark de Vries, and Fred Weerman. We are also very grateful to Robert Cirillo for proofreading the final manuscript of this book. At the end of the editing process we received the sad news that our beloved colleague Hans den Besten had passed away. Hans was a very inspiring member of our research group and made many significant contributions to the field of linguistics. We would like to dedicate this volume to his memory. Amsterdam, August 2010 Harry Perridon Petra Sleeman
The noun phrase in Germanic and Romance Common developments and differences Harry Perridon & Petra Sleeman ACLC, University of Amsterdam
In this introductory chapter some of the main (dis)similarities in DP-syntax between the Germanic and Romance languages, as well as between the individual languages of each group, are explored. We take a look at the following subjects: (a) the ways in which the various languages express definiteness; (b) the position of adjectives; (c) the function of the weak declension of adjectives in Germanic; (d) the evolution of genitive equivalents; and (e) the emergence of determining possessives in Germanic. In each case we try to find out whether a given construction is inherited from the parent language or is an independent development in each of the languages or language groups. Special attention is paid to common developments after the languages split up into separate entities, since they might indicate some inherent properties of human language that restrict the way in which languages may develop.
1.
Relatedness and (dis)similarities
During the last twenty years research on the internal structural of noun phrases has to a great extent been inspired by the assumption that the basic structure of such phrases is part of universal grammar (UG). The enormous variation found in the structures that surface in actual languages is in this 'maximalist' version of UG explained as the result of the various choices that languages make among the options that UG holds available. The task of a child learning the grammar of its mother tongue would then mainly consist in detecting in the linguistic input how the different parameters are set in this particular language. At least since the appearance of Hauser, Chomsky & Fitch (2002; Fitch, Hauser & Chomsky 2005), however, this strong version ofUG (or 'the faculty oflanguage in the narrow sense') seems slowly to give way to more prudent, less detailed hypotheses as to the nature of what makes human language unique. In a 'minimalist' version of UG, containing for instance just one component, viz. recursion, there would hardly be room for elaborate innate structures of the DP or CP kind, as proposed in much of the
2
Harry Perridon & Petra Sleeman
rich literature written in the generative tradition. In such a view there is no reason to assume that every category that is overtly present in a given language ~ belongs to U G, and hence lives a hidden life in a language 1y in which this category lacks any observable form. ln a recent paper, Longobardi & Guardiano (2009) claim to have found evidence that the parametric theory of UG is on the right track after all, and, hence, that the basic lay-out of all languages is much more similar than 'minimalists' and skeptics may have thought They identified 63 binary parameters, all within the domain of the noun phrase, and looked at the ways these were set in a sample of 22 Indo- European (17 modem and 5 'dead' languages) and 6 non-Indo-European languages. First the number of identities (i) and differences (d) in the parameter settings of each pair of languages was calculated, then the distance between the languages in each pair was measured by means of a method that computes what the authors call a 'normalized Hamming distance,' a figure between 0 and 1 resulting from the division of the number of differences by the sum of identities and differences: dl(i +d). Finally a (series of) genealogical tree(s) was generated by a philogenetic program (Kitsch) with the set of distances between the languages in each pair as input At first sight the trees the program produced look quite similar to the ones traditional historical linguistics has come up with. This would be good news for anybody interested in language history, since the new method (the 'parametric comparative method' or PCM for short) seems to be much more efficient and effective than the time and energy consuming methods of traditional comparative linguistics. Moreover, the debate on the nature ofUG seemed to have been settled in favor of the maximalists, since "the historical success of PCM provides evidence of an unprecedented type for Principles & Parameters models of grammatical variation" (Longobardi & Guardiano 2009: 1696). But, unfortunately, a closer look at both the primary data (the normalized Hamming distances of each pair of languages) and the trees generated with these data as input reveals some worrying anomalies and inconsistencies, which cast doubt on the usefulness of a model that only uses (dis)similarities as a basis for genetic classification. 1 According to Longobardi & Guardiano the branching within the Indo-European phylum as generated by Kitsch "is overwhelmingly the expected one': but any serious Indo-Europeanist cannot but shake his/her head in disbelief when being told that Celtic is the first subgroup to branch off from the !E-stern, that Slavic and Indian share a forefather (apart from PIE), that Rumanian does not belong to the same subgroup as Latin, etc. The distance between Latin and Classical Greek as measured by PCM is less than between Latin and Italian or Spanish, or between Spanish and Italian or French; Gothic seems to have more in common with New Testament Greek than with any Germanic language, the examples could be multiplied ad infinitum; very few of the measured distances make any sense if one's goal is to establish the genealogical relations between the languages in the 1. For anyone wanting to venture into linguistic cladistics, Holm (2007) should be obligatory reading.
The noun phrase ln Germanic and Romance
sample. With hindsight this result was to be expected: a genealogical tree is defined by shared innovations (changes), not by (dis)similarities. The Germanic languages, for instance, are defined as a separate subgroup of Indo- European by the changes their Proto-Germanic ancestor underwent, such as the change of word-initial p, t, k intof, p, x (h), the use of a dental suffix to mark past tense, and the differentiation of the weak and strong declension of adjectives. Subsequent changes in the individual Germanic languages may completely alter the way they appear, and make them look more similar to other, distantly related or even unrelated, languages, but their family ties cannot be broken, nor can any language that does not descend from Proto-Germanic ever become a member of the family.
2.
Definiteness and the definite article
Ten of the parameters Longobardi & Guardiano used in their analysis deal with the ways definiteness is (or is not) expressed. In the languages of Central and Western Europe the definite article and/or suffix is relatively recent ln the Romance languages it may have started its life at the end of the Late Latin/ Early Romance period, but it first came to maturity in the individual languages (Bauer 2007). The Germanic languages, on the other hand, did not inherit an incipient definite article from their parent language Proto-Germanic. They had become separate entities long before they developed the definite article (West-Germanic) or suffix (North Germanic). Given that the emergence of definiteness as a distinct grammatical category belongs to the history of the individual languages, and hence was not inherited from a common parent language, it is clear that none of the parameters that deal with definiteness can shed any light on the genealogical relations of the languages that were compared by means of PCM. The emergence of the article in three branches of Indo-European, Romance, Celtic and Germanic, at approximately the same time, viz. between the eighth and the twelfth century, raises a number of other questions. Is it the outcome of a completely autonomous process in each of the languages? This does not seem very likely, since it emerged in one continuous area at roughly the same time. If it was not invented over and over again, then in what language did it originate, and how did it spread? Grammar is not that easily borrowed: if it is, it usually means that contact between the two languages involved is intensive; there are, however, no indications that the language contacts were very intensive in that time and place, unless one assumes that the Christian preachers had a great impact not only on the culture, but also on the language of their audiences. If they did, what then prevented the speakers ofWest-Slavic languages who also came under the spell of Rome from developing a definite article? ln the debate on the nature of the DP-layer between maximalists, who claim that is universal, and minimalists, who argue that it is not, it has been suggested that Germanic was destined to develop a definite article, since it had lost, or was on its way of losing, its aspectual system: "definiteness and perfective aspect are [... ] just two
3
4
Harry Perridon & Petra Sleeman
instantiations of the same grammatical function. 2 So are indefiniteness and imperfective aspecf' (Leiss 2007: 73). Seen in this light the simultaneity of the development of the article in the Romance and Germanic languages could well be a mere coincidence, although contact with Romance may have functioned as a catalyst Problematic for the hypothesis that (im)pertectivity and (in)definiteness tend to show a complementary distribution in language ("Aspect languages avoid article systems, and article languages avoid aspect," Leiss 2007: 87) is the existence oflanguages like Bulgarian that combine an intricate aspectual system with a well-developed article system, on the one hand, and languages like Proto-Norse that could do without either, on the other hand: there lie at least four centuries between the loss of the prefixes that Leiss claims expressed perfectivity (approx. 5th c.) and the emergence of the definite suffix (9th or lOth c.) in this poorly attested language. A more fundamental problem for the theory, however, is that perfectivity is a binary feature while definiteness is a privative one; a verbal construction in a language that has a systematic opposition between two aspects has either perfective or imperfective aspect, it cannot be aspectless. Definite and indefinite noun phrases, on the other hand, only differ in the presence or absence of definiteness. The more or less simultaneous emergence of definiteness as a grammatical category in all the Germanic and Romance languages and dialects does not only pose some challenging questions to historical linguistics, but offers at the same time a rare opportunity to study the ways in which languages with partly different genetic backgrounds integrate a new nominal category in their grammar. In both branches this new category seems to have led to a tighter organization of the noun phrase, or perhaps even to its genesis: in the parent languages (Latin, Old English, Proto-Norse etc.) an attribute was to a certain extent still an apposition, which for reasons of emphasis or style could be placed in other positions than in the immediate vicinity of the noun it qualified, e.g.: (1) meo tu epistulam dedisti servo my.DAT.SG you letter.ACC.SG gave slave.DAT.SG "to my slave you gave a letter?" (2) Fugl vrelva slreit falvan Bird thief.Acc.sG tore pale.ACC.SG "the bird tore apart the pale thief"
(Plautus; Latin)
(Sigtuna box; runic Swedish, 11th c.)
Although it seems likely that discontinuous phrases like the ones in (1) and (2) were exceptions rather than the norm (cf. Pinkster 1990: 187 for Latin), the mere fact that they were possible indicates that word order within a nominal group was not yet as
See e.g. the papers by Abraham, Lohndal, Leiss and others in Stark, Leiss & Abraham (2007), with many useful references to previous contributions to this debate.
2.
The noun phrase in Germanic and Romance rigid as it is in the modem Romance and Germanic languages.3 All attributes could in principle occur both before and after the noun they modified, but with a difference in emphasis. The distal demonstratives grammaticalized into definite articles in a position in which they were not stressed: before the noun in West-Germanic and all the Romance languages except Rumanian, after the noun in North-Germanic and Rumanian. In this way the foundations of a more or less rigid word order in the noun phrase were laid: definite article and noun formed a phrase with a fixed order. It is possible that the demonstratives in French and the West- Germanic languages could no longer be used in postnominal position at the time the unstressed distal demonstrative developed into a definite article. In these languages, then, the article and the demonstratives occupy the same prenominal position, and belong hence to the same word class, viz. the definite determiners. In North-Germanic and Rumanian, on the other hand, the grammaticalization of the unstressed distal demonstrative led to a lexical split, as the postnominal article turned into a clitic, and eventually (at least in Scandinavian) into a nominal suffix. Rumanian demonstratives can still appear in both pre- and postnominal position, but in the latter case the definite article has to be used as well, e.g.:4. 5 (3) a.
acest an this year
b. an-ul acesta year.DEF this "this year" In the modem North-Germanic languages demonstratives may no longer tallow their head noun. In the so-called 'double-definiteness' languages (Norwegian, Swedish and Faroese) a noun that is determined by means of a demonstrative has to carry the suffixed article: (4) dette ar-et this year.DEF "this year"
(Norwegian)
Comparing the demonstrative constructions in (3b) and (4) with the corresponding ones in e.g. English and French, one notices that in the latter languages demonstratives
3· Cf. Magni (this voL) who argues that: "Latin adjectives form loose paratactic structures where the modifier-modified distinction is left unspecified, and items from the same category are juxtaposed." The opposite view is defended by Platzack (2008), who assumes that there is a difference in structure between the old and modern languages, which allows modifiers to be extracted in Latin and Old Norse, but not in e.g. French or Modern Icelandic. 4- Cornllescu & Nicolae (this voL) argue that the Rumanian enclitic article -(u)l is a suffix, too. 5· In Spanish, too, the demonstrative may appear ln postnominal position, see, e.g., Alarcos Uorach (1987: 287-306), Bernstein (2001), and Alexander (2007).
5
6
Harry Perridon & Petra Sleeman
are definite by themselves, whereas in Rumanian and Norwegian they require definiteness to be marked separately on the noun. In Greek a similar situation obtains: (5) avt6 to chr6no this the year "this year" Definiteness is hence part of the meaning of the demonstrative, which is 'spelled out' in languages like Rumanian, Greek and Norwegian. In West-Germanic and West-Romance the definite article is only used when no other definite determiner is present It is a kind of 'tool oflast resort' for expressing definiteness. Possessive adjectives followed more or less the same course of development as the demonstratives. In the parent languages the possessives could follow or precede the head noun, or even be separated from it. as in the Latin example in (1 ). Postposition of the possessive was probably the unmarked order; in front position possessives were stressed, either for emphasis or contrastively. Judging from the situation in the modem Romance languages, the markedness of both word orders seems to have reversed at some point in time. In Spanish, for example, a construction with a possessive adjective (mia "my" in 6a) following the head noun (here: casa "house") is used when one wants to emphasize the 'possessor'; (6b) with a weak form of the possessive (mi) preceding the head noun is the unmarked alternative: (6) a.
la
casa
mfa
the house mine "mjhouse" b. mi casa "my house" In the North-Germanic languages that have retained both word orders (Norwegian, Icelandic and Faroese) markedness is unaltered: postposition is unmarked, front position emphasizes the possessive (Faarlund et al. 1997: 265): (7) a.
hus-et
mitt
house.DEF my "my house" b. mitt hus Kmyhouse" It is to be noted that possessives are not definite determiners in the languages that allow both word orders.6 In Italian and (European) Portuguese possessive adjectives behave like ordinary adjectives that can be used in both definite and indefinite noun phrases. 6. Faroese is the exception here: a possessive adjective (or genitive of a pronoun functioning as a possessive) makes its head noun definite. even in postposition: hU5 mitt (*hUs-ln mitt) "my house" and mitt hU5 "my house" ('Ihnllnssonet al. 2004: 96-7; 118).
The noun phrase ln Germanic and Romance
The 'strong' possessives in Spanish function in much the same way. In Norwegian, Rumanian and, to a certain extent, Icelandic the possessives follow a noun that has been made definite by other means (cf. 7a). 7 Thus it is not the possessive itself that makes a noun phrase definite, it is rather its prenominal position that turns it into a determiner. In French, Swedish, Danish and West-Germanic this position (the determiner-position, or 'D' for short) is the only one available to possessives, which hence have become definite determiners on a par with demonstratives and the definite article. In a similar way the weak possessives in Spanish ( mi, tu, su) turned into determiners, thus parting company with the strong possessives (mio, tuyo, suyo), which remained 'genitival' adjectives, semantically akin to the PP-'genitives' (de + D P/NP in Spanish and French, of+ DP/NP in English, von + DP/NP in German, van + DP/NP in Dutch). Definiteness in the Germanic and Romance languages is not only an inherent property of some lexical elements, such as the definite article and the demonstratives, but also of a specific position in the noun phrase, viz. the D-position in the prenominal field. All elements that are placed in D ('promoted/raised to D') function as definite determiners: demonstratives, articles, genitival constructions, possessives, and even 'identi:tying adjectives like Dutch voornoemd "aforementioned" and bedoeld "intended': as described by van de Velde (this vol.), and Swedish samma "the same': niista "the next': ovannamnda "aforementioned': etc. (Agren 1912: 64-70; Perridon 1989: 207). In West-Germanic and West-Romance the histories ofthe definite article and the D-position coincide to a large extent, since it was exactly in this prenominal D-position that the definite article was born. In Rumanian and North-Germanic, on the other hand, the article emerged in postnominal position, and became eventually a suffix which only modifies the meaning of the noun but does not head a phrase. A separate D was created in these languages when demonstratives increasingly started to appear in prenominal position. In North-Germanic this leftward movement of the demonstratives ended with their no longer being able to follow their head noun, but having to precede it. At a later stage the possessives followed suit In the runic inscriptions of the lOth and 11th centuries, possessives usually followed the noun, but preceded other adjectives, e.g. on runic stone So 10 (SOdermanland, Sweden): (8)
[.•• ]paiR *litu •raisa *stain *at *iarl *fapur: sin: kupan: [...] they let raise stone at [Jarl father their good] .ACC.SG.M "they had the stone raised in memory of JarL their good father"
7· The situation in Icelandic is fairly complicated. Concrete nouns require the suffixed article when followed by a possessive, whereas abstract nouns and kinship terms disallow it, e.g.: (i) hUs-llJ mitt vs. ?1hU5 mitt house.DEF my (ii) skolJun mb1
vs. *skolJun-in min
opinion my
See Sigurllsson (2006) for a thorough discussion of these and related constructions.
7
8
Harry Perridon & Petra Sleeman
The adjective kupan (Old Norse g6lJan) has a strong, non-definite form; the weak form would have been g6lJa. When both possessive and adjective were fronted, the adjective could keep its strong form until the end of the 14th c., as e.g. in the phrase hans siukt ben "his sick leg" (Codex Bureanus Sweden, around 1350; cf. Delsing 1994), with strong siukt instead of weak siuka.. After 1400 the adjective must appear in a weak form if a determiner is present in D that is, in the form it always had when occurring within the scope of a demonstrative. Apart from a suffix 'defining' the noun to which it attaches and a D-position that takes scope over whatever is to the right of it in the D P, North -Germanic and Rumanian developed an 'adjectival article, which in principle only has scope over an adjective. In North-Germanic this article seems to be older than either the nominal suffix or the D-position. It is only used with a following weak adjective, which often functions as an epithet. e.g. in the inscription on the runic stone DR 84 (lOth c.; Sk.em in Denmark): (9) soskirijm rispi: stin: finulfs: tutiR: at: upinkaur: usbiarnaR: sun: poh: trlra:
uk: hin: turrltin:fasta: Sasgrerpr respi sten, Finulfs dottiR, at Opinkor AsbiarnaR sun, pan dyra ok hin drottinfasta. ('translated' to standardized Old Swedish) "Sasger~r. Finnulfr's daughter, raised the stone, in memory of 6~ink.arr jam's son, the valued and loyal to his lord:'
Ash-
The epithets ]?ann djra "the dear (one)" and hinn dr6ttinfasta "the (one who is) loyal to his lord" consist each of a demonstrative and a weak adjective. In this use both demonstratives have a bleached meaning, which comes close to that of the definite article in e.g. English. In Old Norse prose the construction 'adjectival article+ weak adjective' is not only used as an epithet. but also in ordinary attributive use, as e.g. in: (10) hendi inni hoegri [hand the right.WEAK].P.SG.DAT "with the/her right hand"
(Voluspa 5; poetic Edda)
Thew hole phrase could be placed before the noun, in which case it alternates with just the weak adjective. It may thus follow a demonstrative, as in (11), or even a possessive, as in (12): ( 11)
pau
hin st6rn
skip
[those the big.WEAK ships] .N.PL.NOM "those big ships" (12) minn inn hvassi hjorr [my the sharp.WEAK sword].M.SG.NOM "my sharp sword"
(Fafnismal6; poetic Edda)
The noun phrase ln Germanic and Romance
In Mainland Scandinavian it was the demonstrative sa, su, pat ~ den, det "that" that was used as a preadjectival article, in Icelandic it was (h)inn "that". Den/det is the article that appears in double-definite constructions in Swedish and Norwegian, e.g.: ( 13) den mjuka
sang-en the soft.WEAK bed.DEF "the soft bed"
Swedish
It has the same forms as the distal demonstrative denldetlde in D-position, and is therefore difficult to keep apart from it. Since article, demonstrative and pronoun all changed their plural form in exactly the same way from [di) to [cbm] in (standard) Swedish, it seems reasonable to assume that there is in fact only one lexical item. The difference in interpretation is then the result of a difference in stress and position. If this is correct, the phrases in (14) and (15) with den immediately following a demonstrative should not be analysed as the modern counterparts of the Old Norse constructions in (11) and (12), but rather as DP's in which the head noun is a DP, (14) dette (det) heje flotte hus
(Danish; Leu 2008: 31; 63)
this (the) tall stylish house (15) denna den biista av alla viirldar
this
(Swedish; Perridon 1989: 186)
the best of all worlds
Under such an interpretation they display basically the same structure as (16): (16) den sk0nne Jordens Sol the beautiful.wEAK earth.DEF.GEN sun "the beautiful sun of the war ld"
(Danish; Diderichsen 1946: 225)
which was analysed by Diderichsen (1946) as: [NP den skt<mne [NP jordens sol]], with jordens as the determiner in the inner NP (DP), and den as the determiner of the whole NP/DP. In Modern Icelandic it is no longer possible to use an adjectival article after a demonstrative or a possessive. But even in its normal position at the left edge of the DP the existence of this article is threatened: "the preposed free article is almost non-existent in common everyday language" Siguri3sson (2006: 195) writes; it is "mostly confined to abstract nouns in formal written style" (ibid.). When it occurs it appears to function as any other prenominal determiner, by requiring the noun it determines to be suffixless: (17) HiO langa kvce.Oi var frekar leiOinlegt the long. WEAK poem was rather boring
(Thrainsson 2005: 97)
But since it cannot be used in any other context than before a weak adjective it has not really changed its function of just making the following adjective definite. It became superfluous as soon as the weak form of the adjective had become an unambiguous marker of definiteness by itself. This weak torm without the 'free article' is tollowed by
9
10
Harry Perridon & Petra Sleeman
the definite form of the noun (noun+ suffix): Zanga kvrelJi-lJ long.weak poem.def"the long poem". ln Rumanian, the only Romance language to develop a suffix instead of an article, the suffix is attached to the first element in a Noun+ Adj. or Adj.+ Noun combination: (18) a.
trandafir-ul (jmmos)
(Cornilescu & Nicolae, this vol.)
rose.DEF (beautiful) "the beautiful rose"
b. frumos-ul
trandafir
beautitul.DEF rose The construction in (18b) is similar to the Icelandic one in (17): the adjectival suffix signals in much the same way as its Icelandic counterpart, the preadjectival article/ clitic, that the adjective is definite. The definite adjective functions as a determiner in D-position which like other prenominal determiners prevents the noun from taking the definite suffix. Apart from the adjectival suffix, Rumanian developed an adjectival article eel which is used with postposed adjectives: (18) c.
trandafir-ul eel frumos rose.DEF the beautiful "the beautiful rose"
Article + adjective cannot precede the noun: *eel Jrumos tra.ndafir(ul), unless the adjective (phrase) is a superlative (19), or a numeral (20):
(19) eel mai frumos tra.ndafir the more beautiful rose "the most beautiful rose" (20) eei doi trandafiri the.PL two rose.PL "the two roses" These two constructions are structurally almost identical to their Icelandic counterpart in (17). The whole phrase eel+ adjective in (19) functions as a determiner in D, not just eel, which is part of the superlative AP. From this short overview of the various ways in which definite noun phrases are construed in the Germanic and Romance languages it is clear that the similarities and differences between the individual languages are not due to a common genetic background. The main dividing line runs between D-languages that developed a definite article in D, which has scope over the rest of the D P, and suffixing languages that developed a number of articles and suffixes with only local scope: over a noun, an adjective or adjective phrase, etc. In each of the two language groups there are languages of both types: West-Germanic has D-articles, North-Germanic suffixes or
The noun phrase ln Germanic and Romance
prefixes (the South- and West Jutlandic dialects of Danish), all Romance languages are D-languages, with the exception of Rumanian, which has a rich inventory oflocal articles and suffixes.
3·
Position of adjectives
In the Romance languages attributive adjectives may either precede or follow their head noun. The question ofwhat exactly determines the choice of the order A-N or N-Ahas been hotly debated within Romance linguistics and stylistics, but has not been given a final answer yet. There is, however, some agreement on what are the most important factors that influence the choice of either order. One of these factors is the semantic nature of the adjective. In French, adjectives designating an objective property are predominantly placed after the noun, whereas evaluating, emotive adjectives regularly precede the noun: "plus un adjectif est reserve ades emplois definitionnels, techniques, excluant toute emotivite, plus regulierementcetadjectifsera place apres le substantif; (Blink.enberg 1969: 84). In the other Romance languages a similar distribution of pre- and postposed adjectives is found. Lepschy & Lepschy (1994: 165-8) stress the individuating, restrictive nature of postnominal modifiers in Italian, and ascribe a purely descriptive, epithetic function to the prenominal ones. Sole & Sole (1977: 230-239) express the same view on the position of the adjective in Spanish: "Post-nominal adjectives usually restrict, clarify or specify the meaning of the modified noun by adding an idea not expressed by the noun". In the modem Germanic languages attributive adjectives and other modifiers do not enjoy the same freedom as their Romance counterparts: they are in principle restricted to prenominal position. In English, however, postposition of an adjective is not unusua~ as in the following well-known examples: (21) a. the only river navigable b. stars visible
vs vs
the only navigable river visible stars
Bolinger (1967) argues that the adjectives in postposition in (21) express an occasional property, i.e. a property which the object designated by the noun has on some particular occasion. This might be a suitable characterization of the effect which postposition of the modifier has on the meaning of the noun phrases in (21), but it does not account for the meaning that the adjective adjacent has in (22): (22) buildings adjacent will be closed for three days
(Ferris 1993: 45)
Adjacency is hardly an occasional property of an object, as Ferris (1993) remarks in his penetrating analysis of the various uses of adjectives in English. According to Ferris both postnominal and predicative adjectives assign a property to their head nouns, whereas prenominal adjectives only modifY the meaning of the nouns they are
11
12
Harry Perridon & Petra Sleeman
subordinated to. Pre- and postnominal adjectives are part of the noun phrase, predicative adjectives of the sentence. In postnominal position adjectives are a kind of predicative attribute. 8 Some confirmation for the predicative status of postnominal adjectives comes from languages like Dutch and German, which in general only marginally allow adjectives or adjective phrases in postposition. In these languages adjectives are inflected in prenominal position (as in 23a), but remain uninflected when used predicatively. In postnominal position (2 3b) they are not inflected either:9 (23) a.
als tennisballen zo grote hagelstenen
(Dutch)
as tennis balls so big.PL hailstones "hailstones (as) big as tennis balls" b. hagelstenen zo groot als tennisballen hailstones as big as tennisballs Most of the rules that govern the placement of adjectives in the Romance languages were already in force in the parent language, Latin. As in the daughter languages, adjectives in Latin usually follow the noun "unless pragmatic factors such as Focus cause them to be preposed" (Pink.ster 1990: 186). The prenominal position of the adjective in Latin entails according to Magni (this vol.) "a tighter syntactic bond between attribute and head noun, which for the adjectives corresponds to the expression of inherent and essential properties, and to the function of description and concept formation. [... ] Conversely, the postnominal position entails a looser nexus, which is more suited to the expression of accidental properties, to object identification, and to discriminate an entity:' In the oldest phases of the Germanic languages the same freedom in position of the adjective is found. In Gothic, for instance, the adjective normally follows the noun, as in (24a), but may also precede it (24b ): (24) a.
milip
haipiwisk
(Wulfila, Marc 1, 6)
(honey wild.STRONG]N.SG.ACC "wild honey" b. unhrainjamma ahmin (unclean.STRONG spirit)M.SG.DAT "(with an) unclean spirit"
(Wulfila, Marc 1,23)
8. Writing about adjective position in Old English, Fischer (200 1: 257) describes the meaning of the postposed adjective ln the same vein: "[... ] the Old English postnominalstrong adjectives act very much like secondary predkates; they are rhematic, and as such belong to an adjectival category that is very dose to the Verb category. Their postnominal position, in other words, can be seen as iconic (because the meaning of the adjective is not incorporated into the noun." 9· Sleeman (2007) draws attention to this same difference ln morphological form, when discussing the verbal nature of postnominal (past) partictples.
The noun phrase ln Germanic and Romance
As in Latin, classifying, individuating adjectives, like hatpiwisk "wild': derived from haipi "field, heath': follow the noun, whereas qualifying adjectives like unhrains "unclean, dirty" in (24b) precede its head noun. In Old English, Old High German and Old Norse both positions were still available to the adjective, but at some point of time the default order had become A+ N, instead of N + A. as in Gothic and Latin. In the modem Germanic languages the default order has become the fixed order. Since the change from N-A to A-N, and the subsequent fixation of A-N, took place in each of the Germanic languages separately, the question forces itself upon us what they had in common in this part of their grammar that caused them to follow the same course of development. The one major feature that sets apart the Germanic adjective from its counterparts in Latin and the other Indo-European languages is its capacity to take two separate sets of endings. The 'weak' adjective with endings based on a suffix -e!on is a Germanic innovation, albeit that the construction type as such is also known from other IndoEuropean languages. In Ancient Greek nouns, esp. cognomens, could be derived from adjectives by means of this suffix, e.g.: Platon "(the) broad one" from platys "broad". In the same way nicknames were derived in Latin, e.g.: Cato "(the) smart guy" from catus "smart". It is certainly this use of the suffix that lies at the origin of the Germanic weak adjective. In Germanic, nouns derived from adjectives by means of the-e/on-suffix were always used together with the name, as a kind of extension of it, as e.g. in Old Norse Haraldr haifagri "Harald Fairhair": here the weak form of the compounded adjective hdrfagr (hdr "hair"+ fagr "fair") could in principle still be analyzed as a substantivized adjective apposed to the proper name Haraldr. In contrast to its Latin and Ancient Greek counterparts they could not be used in isolation as cognomens. Later they came to be used with common nouns as wel~ expressing an inherent property of the entity designated by the noun they were apposed to. In this way the derivational suffix -e!on- gradually grammaticalized into an inflectional suffix to be used with attributive adjectives. Already in Gothic the identifying function of the appositive weak adjective was strengthened by means of a definite article, as e.g. in (25): (25) Jon pata unhvapnando [fire the unquenchable. WEAK ).N .SG .N/ ACC "the unquenchable fire (- s of hell)"
(Wul:fila, Marc 9, 43 and 45)
Weak adjectives preceded by the adjectival article are usually postposed in Gothic. In Old Norse and Old Saxon they occur both before and after the noun, but in Old High German and Old English they almost exclusively precede the noun, as they do in all the modern Germanic languages that still distinguish between strong and weak adjectives. The unmarked order seems thus to have changed from 'N +art+ adjW' to 'art a change in word order that calls for an explanation. According to Fischer +adjW + (2001) word order in the Old English noun phrase is iconic: what is known (given information) precedes what is new. Identifying elements, such as determiners and the noun itself. are presented first, predicative elements which give new information,
N:
13
14
Harry Perridon & Petra Sleeman
follow (both in real time and in writing). Since in Fischer's view the basic meaning of weak adjectives is to give additional information that enables the hearer to identify the entity the speaker wants to say something about, it stands to reason that it stands as closely together with the noun as possible:"[ ...] adjective position is iconically motivated in that prenominal adjectives, when they are weak and definite, convey given information; they behave typically like attributive adjectives which are closer to the nominal pole of the adjective cline, they are therefore an inseparable part of the head: togetherwiththenounphrasetheyformthe'theme'oftheutterance"(Fischer2001:271). This does not explain the original word order in Germanic, exemplified by Gothic in (25). Moreover, there is no reason to assume thatadditionalinformation that is needed in order to identify the entity the speaker wants to talk about, should precede the primary identifier, the noun. The most likely scenario for the change in word order in the West- Germanic noun phrase is in our view the following: when the prenominal distal demonstrative grammaticalized into a definite article, the adjectival article was reinterpreted as an instance of that article, and hence placed in the prenominal slot for determiners, D. Since adjectival article and weak adjective formed a unity, the weak adjective moved with the article to a position before the noun. Strong adjectives could originally appear both before and after the noun. but in the course of time they lost this freedom of placement, and became restricted to prenominal position. In North Germanic, too, the adjectival article was reinterpreted as a determiner to be placed in D, and like its West Germanic counterpart it dragged along its companion, the weak adjective.
4·
Function and position of genitives and genitivals
Adjectives, genitives and genitivals (i.e. constructions that have more or less the same function as genitives in inflectional languages, e.g. the of- 'genitive' in English) have much in common: they are all subordinated to the noun and provide additional information on the entity designated by that noun. Often there is little difference in meaning between a construction with an adjective and one with a genitival, as e.g. in the examples in (26): (26) a. the French team b. the team of France
l'equipe fran~ise l'equipe de France
According to Wackernagel ( 1908: 145 ), as quoted by Magni (this vol. ), the use of adjectives for expressing a genitival relation predates in Indo-European the use of genitive case in that function. Since the kind of objective, individuating meaning mediated by these adjectives agrees well with the postnominal position in Latin (see Section 3 above), it is hardly surprising "that, when the genitive supersedes the adjective in possessive constructions, the new structure parallels the older one, and postnominal genitives replace postnominal adjectives" (Magni, this vol.:§ 3.6). In the Late Latin/Early
The noun phrase ln Germanic and Romance
Romance period the morphological genitives gave way to constructions with a preposition: head noun ('possessum') - de (preposition "of') - dependent noun ('possessor'). The Germanic languages followed a different course of development. ln the Germanic parent language the position of adjectives and genitives was presumably the same as in Latin, and expressed the same semantic distinctions: in prenominal position modifiers were closely connected with the head noun, in postposition the connection was much looser. When adjectives lost their positional freedom and became fixed to prenominal position (see Section 3), the position of genitives did not change. What changed, however, was what kind of elements could occur in either position. In postposition all kinds of genitive constructions were allowed, with or without modifiers, determiners and even relative clauses. This is still the case in those Germanic languages that have retained a morphological genitive (Icelandic and German). The prenominal position became gradually restricted to the genitives ofbare nouns, and the resulting combinations of genitive+ noun turned into noun-noun compounds. In the languages that shed their case morphology, the endings that signalled genitive case were reinterpreted as meaningless interfixes, which function as a kind of glue between the parts of a compound. In English bare nouns are placed directly betore the noun they modify, without any linker, as in: a stone bridge, an arms treaty, a holiday treat, a birthday present. The modifying noun can be modified itselfby an adjective, as in: a long distance runner or sick building syndrome. In the other Germanic languages adjectives specifying a modifying noun often keep their inflected form: (27) la.ngeaftta.ndsloper long.INFL-distance-s-runner "long distance runner"
(Dutch)
In cases like these the dividing line between morphology ("is it a compound?") and syntax ("is it a phrase?") is rather thin. In the Germanic languages that lost the genitive as a morphological case the postnominal genitive was replaced by a construction in which the genitival (or 'possessor phrase') is linked to the head noun ('possessum phrase') by means of a preposition with a reduced meaning (of in English, van in Dutch). In the two languages that kept the genitive, German and Icelandic, similar constructions emerged as alternatives to the morphological genitives. In the languages in which possessive pronouns have become determiners, a new type of genitive emerged, the determining genitive. It comes in two flavors: the element that links the two parts of the genitival construction, and functions as a
15
16
Harry Perridon & Petra Sleeman
determiner, is either (a) -s or (b) a possessive pronoun. 10 The -s genitive is found in English, Danish, Swedish, Norwegian (Bokmfil) and (marginally) Dutch, the pronominal linker in Afrikaans, Dutch, Low Saxon, German, Norwegian (Nynorsk) and the West-Jutlandic dialect of Danish. In Faroese the linking element is -sa, of unknown origin, which like the Dutch -s genitive can only be used with names. These may have phrasal structure, as in (28): (28)
Tummas a D6marakont6rinumsa
bilur (Faroese; 'Ih.rainsson et aL 2004: 64)
Thomas at legal office.DEF.DAT.SG-sa car "Thomas at the legal office's car" The remarkable fact that the Germanic languages, except Icelandic, each individually developed the same kind of construction suggests that they were all subject to the same pressure to fill a structural gap in their determiner system. The factors that made the emergence of the determining genitive possible, or even necessary, are the following: (a) there is a prenominal slot for determiners; (b) possessive pronouns are placed (or can be placed) in this slot, and function as determiners; (c) relational adjectives precede the noun. The triggering factor seems to have been the loss of the morphological genitive. In German, it is true, the morphological genitive is still alive, but its main domain is the written language, from which the construction with the pronominallinker is banned.
5·
An overview of the contributions to this volume
The discussion in the preceding sections shows that, in our view, the DP-structure in Romance and Germanic developed from rather loose relations between its constituting elements. The emergence of the definite article, which developed from a prenominal demonstrative, paved the way for a DP-structure with tighter relations between its constituents. The emergence of definiteness led to the grammaticalization of various prenominal adjectival and genitival elements as determiners. Individual languages or language groups differ in the extent to which grammaticalization has taken place. Although the Germanic languages differ in a number of respects from the Romance languages (prenominal adjectives, the determining genitive, a weak/strong adjectival inflection), individual languages do not always behave like the other languages from The first scholar to describe the origin of the -s genitive in English is Jespersen (1894), who mainly focussed on 'group genitives' like The man I saw ye.sterday~ son. Jespersen's view that -s emancipated from a mere ending to a grammatical element, an "interposition;' has in recent years led to an intense debate among historkallinguists with an interest in grammaticalization theory, since such a development towards greater and greater freedom seems to contradict the main tenet ofthat theory, viz. that grammatical change is unidirectional, from more to less freedom, see e.g.: Janda (1980; 2001), Allen (1997; 2002; 2003) and Rosenbach (2004) for English, Norde (1997; 2001; 2006) and Borjars (2003) for Swedish. 10.
The noun phrase ln Germanic and Romance
their language family. Both North-Germanic and Romanian have or had a postnominal, suffixal, definite determiner and an adjectival article. and German, Icelandic and Romanian still have morphological genitives. In this volume, similarities and differences between and within the Germanic and Romance language families are discussed. In all papers the discussion is restricted to the DP, although the discussion is not limited to the subjects introduced in this introductory chapter. The papers in this book are grouped together according to two themes: (a) variation and (b) change with respect to the DP in the Germanic and Romance languages. The papers are presented in what follows.
5.1
Variation
Alexiadou, IordAchioaia, and Schafer argue that there is no parametric difference between Germanic and Romance with respect to nominalized infinitives. Both language families have two types of nominalized infinitives: a verbal type and a nominal type. Within a Distributed Morphology approach, the authors propose that languages may only differ with respect to the distribution of verbal and nominal layers within the DP representing the nominalized infinitive. Cirillo shows that although both in Romance and Germanic universal quantifiers such as "all" can select DP as their complement and can be floated/stranded by that DR there is variation between Romance and Germanic and also among the Germanic languages with respect to the combination of a universal quantifier with an interrogative DP. In Romance, universal quantifiers cannot be combined with an interrogative DP. In Germanic, there is variation. In German, "all" can occur to the right of the wh -word and can be stranded, possibly because it also allows split DPs. American English has only the first option, and Swedish only the second. British English has neither of them. Cirillo proposes that the variation is the result of a lexical difference between the languages. Carver & van Koppens paper deals with micro-variation in a split DP-construction, the wat voor-construction, in one of the Germanic languages, Dutch. Carver and van Koppen adopt a predicate displacement analysis for all variants, but show that predicate displacement of wat is not allowed in all dialects. Carver and van Koppen relate the variation to differences in the internal syntax of the cross-dialectal variants of the wat voor-construction. Wood and Vikner discuss pre-article "so/ such"-constructions in three Germanic languages: English, Danish, and German. They analyze these pre-article constructions as predicate inversion constructions. Wood and Vikner show that there is variation with respect to the "so/ such"-constructions in Germanic. They furthermore argue that the "so"/"such" -constructions are changing their functions in English and German. Lohrmann discusses the expression of definiteness in Scandinavian, assuming that it is possible to describe the variation found in these languages by means of a single model. She argues that the three different markers of definiteness, viz. the
17
18
Harry Perridon & Petra Sleeman
preadjectival determiner, the definite suffix, and the weak adjectival inflection each express a separate aspect of the notion of definiteness. Within a Distributed Morphology approach, Lohrmann proposes a unified (double) DP structure for all Scandinavian languages and dialects she discusses in her paper, but she claims that the double definiteness languages (Norwegian, Swedish and Faroese) differ from the single definiteness languages (Danish and Icelandic) with respect to the realization of the syntactic heads representing the three different types of morphemes. Stroh-Wollin also proposes a unified (double) DP structure within a Distributed Morphology approach, both for Germanic and Romance languages. She claims that the differences in the expression of definite, indefinite, and generic noun phrases and the position of attributive adjectives in the various Scandinavian languages, in English, and in Romance are the result of a different lexical realization of both DP heads (DP and dP in her analysis) and of a difference in movement operations. Bobyleva claims that the distribution of the definite determiner in two Englishbased creoles, Jamaican and Sranan, is basically due to the amount of contact with their main lexifier, English, which has a definiteness-based determiner use, and not to the influence of Gbe, one of their most important substrate languages, which has a specificity-based determiner use. Bobyleva argues that the more extended use ofbare nouns in Jamaican and Sranan as compared to English is not due to the influence of Gbe either, but is the result of pragmatic non-referentiality. 5.2
Change
Lucas shows, just like Bobyleva (this vol.), that pragmatic non-referentiality may or may not be expressed by a definite determiner. He discusses two classes of so-called 'weak definites' in English, and outlines a diachronic explanation for the form-function mismatch of these two classes. According to Lucas such mismatches are to be expected, since definite articles in diachrony have a tendency to spread into contexts where they no longer signal semantic definiteness. Crisma argues that the definite article in Old English emerged no later than the last quarter of the 9th century. Starting out from the assumption that grammatical change does not happen without some external cause, Crisma proposes that the definite article emerged in Old English through the influence of the Celtic substratum/ adstratum. Cornilescu and Nicolae claim that the existence of the lower article in Old Romanian is evidence that the Romanian enclitic definite article originates as a post-posed demonstrative and is a suffix rather than a second position clitic. Romanian developed an inflectional genitive system (bare inflected genitives and inflected DPs preceded by the genitival article a.T) alongside the prepositional de-genitive, each with its own function and/or distribution. Comilescu & Nicolae show that there is a strong statistical correlation between the lower article and the (postnominal) inflectional bare genitive.
The noun phrase in Germanic and Romance
Magni discusses the development and the functions of prenominal and postnominal genitives in English and Latin. She argues that, both in English and in Latin, the coexistence ofboth positions can be explained through diachrony, which also accounts for the functional specialization ofboth positions. Van de Velde claims that 'anaphoric adjectiveS: i.e. adjectives that fulfil a discoursedeictic function, such as voornoemd 'aforementioned' and vermeld 'mentioned: are increasingly used as determiners in Present-day Dutch, rather than as adjectives. This suggests that anaphoric adjectives are gradually changing into Ds in Late Modern Dutch. Deprez argues that French n-words like personne 'nobody' and rien 'nothing' have undergone a change from nouns into determiners in the course of time, climbing from N, via NumP, to D. Deprez proposes that in contemporary French, n-words are merged as a strong quantifier in the highest layer of the DP, the strong determiner phrase (SDP), which accounts for their change into negative quantificational expressions.
References Abraham, Wernet: 2007. The discourse-functional crystallization of the historically original demonstrative. In Stark, Leiss & Abraham, 241-256. Agren, Karl. 1912. 0 m anvlindningen av bestiim.d slutartikel i svenskan. Uppsala: Almquist. Alarcos Llorach, Emilio. 1987. Estudios de gramatlca funcional del Espaflol. Madrid: Editorial Gredos. Alexande~ David B. 2007. The Spanish Postnomlnal Demonstrative in Synchrony and Diachrony. PhD dissertation, Ohio State University. Allen, Cynthia 1997. The origins of the 'group genitive' in English. Transactions of the Philological Society_95(1): 111-131. Allen, Cynthia 2002. The early English 'his genitives' from a Germanic perspective. In ProceedIngs of the 2002 O:mference of the Australian Linguistic Society, Peter Collins & Mengistu Amberber (eds). <WWW.als.asn.au/proceedings/als200 1.html> Allen, Cynthia 2003. Deflexion and the development of the genitive in English. E11glish Language and Linguistics 7(1): 1-28. Bauer, Brigitte. 2007. The definite article in Indo-European: Emergence of a new grammatical categorr. In Stark, Leiss & Abraham, 103-140. Bernstein, Judy. 2001. Focusing the 'right' way in Romance determiner phrases. Probus 13: 1-29. Blinkenberg,Andreas. 1969. Ihrdre des Mots en FranfaisModerne. Deu:~.iemepartie. Copenhagen: Munksgaard. Bolinger, Dwight. 1967. Adjectives in English: Attribution and predication. Li11gua 18: 1-34. Borjars, Kersti. 2003. Morphological status and (de)grammatical.isation: The Swedish possessive. Nordic Journal ofLi,~guistics 26(2): 133-163. Delsing, Lars-Olof 1994. 'Hans sjukt beri'- om Starka och Svaga Adjektiv i Fornsvenskan. In Sprdkbruk, grammaNk och sprdkfiiriiJulring. En festskrift till Ulf Thieman, Nils Jorgensen, Christer Platzack & Jan Svensson (eds), 99-108. Lund: Lund University. Diderichsen, Paul. 1946. Elementrer Dansk Grammatlk. K0benhavn: Gyldendal.
19
20
Harry Perridon & Petra Sleeman Faarlund, Jan Terje, Lie, Svein & Vannebo, Kjell. 1997. Norsk Reforanse-Grammatikk. Oslo: Universitetsforlaget Ferris, Connor. 1993. The Meani11g of Sy11tax. A Study in the Adjectives of E11glish. London: Longman. Fischer, Olga 2001. The position of the adjective in (Old) English from an iconic perspective. In 1he Motivated Sign [!conicity in Language and Literature 2], Olga Fischer & Max: Nanny (eds), 249-276. Amsterdam: John Benjamins. Fitch, W. Tecumseh, Hauser, Marc D. & Chomsky, Noam. 2005. The evolution of the language faculty: Clarifications and implications. Cogp1ition 97: 179-210. Hauser, Marc, Chomsky, Noam & Fitch, W. Tecumseh. 2002. The Language faculty: What is it, who has it, and how did it evolve? Science 298: 1569-1579. Holm, Hans J. 2007. The new arboretum oflndo-Europeum 'trees~ Can new algorithms reveal the phylogeny and even prehistory of Indo-European? Journal of Quantitative Linguistics 14: 167-214. Janda Richard. 1980. On the decline of declensional systems: The loss of OE nominal case and the ME reanalysis of -es as his. In Papers From the Fourth International Conference on Historical Linguistics [Current Issues in Linguistic Theory 14], Elizabeth Closs Traugott, Rebecca Labrum & Susan Shepard (eds), 243-252. Amsterdam: John Benjamins. Janda, Richard D. 2001. Beyond 'pathways' and 'unidirectionality': On the discontinuity oflanguage transmission and the counterability of grammaticalization. Language Sciences 23:265-340. Jespersen, Otto. 1894. Progress In Language With Special Reference to English. London: Swan Sonnenschein & Co. Leiss, Elisabeth. 2007. Covert patterns of definiteness/indefiniteness and aspectuality in Old Icelandic, Gothic, and Old High German. In Stark. Leiss & Abraham, 73-102. Lepschy, Laura & Lepschy, Giulio. 1994. La lingua Italiana. Storla, varieta dell' uso, grammatica. Milana: Bompiani. Leu, Thomas. 2008. 1he Internal Synta.r: ofDeterminers [Groninger Arbeiten zur Ger manistischen Linguistik 47]. PhD dissertation, New York University. Lohndal, Terje. 2007. On the structure and development of nominal phrases in Norwegian. In Stark, Leiss &Abraham, 287-310. Longobardi. Giuseppe & Guardiano, Christina. 2009. Evidence for syntax as a signal of historical relatedness. Lingua 119: 1679-1706. Norde, Murie11997. The History of the Genitive in Swedish: A case study in degrammaticalization PhD dissertation, University of Amsterdam. Norde, Murie12001. The history of the genitive in Swedish - The full story: A reply to Lars-Olof Delsing. Nordic Journal ofLinguistics 24( 1): 107-118. Norde, Muriel. 2006. Demarcating degrammaticalization: The Swedish s-genitive revisited Nordic Journal ofLinguistics 29(2): 201-238. Perridon, Harry. 1989. Reference, Definiteness and the Noun Phrase in Swedish. PhD dissertation, University of Amsterdam. Pinkster, Harm. 1990. Latin Syntax and Semantics. London: Routledge.
Platzack. Christer. 2008. Left branch extraction of nominal dodifiers in Old Scandinavian. In Grammatical Change and Linguistic Theory. 1he Rose11dal Papers [Linguistik Aktuell/Linguistics Today 113],1h6rhallurEyth6rsson (ed), 357-374. Amsterdam: John Benjamins.
The noun phrase in Germanic and Romance Rosenbach, Anette. 2004. The English s-genitive. In Up and Down the Cline. The Nature ofGrammaticalization, Olga Fischer, Muriel Norde & Harry Perridon (eds), 73-96. Amsterdam: John Benjamins. Sl.gtlr&son, Halld6r Armann. 2006. The Icelandic Noun Phrase: Central traits. Arkiv fiir Nordisk
Filologi 121: 193-236. Sleeman, Petra. 2007. Prenominal and postnominal reduced relative clauses: Arguments against unitary analyses. Bucharest Working Papers in Linguistics IX: 5-16. Sole, Yolanda R & Sole, Carlos A 1977. Modern Spanish Syntax. A Study in Co11trast. Lexington MA: D.C. Heath and Company. Stark. Elisabeth, Leiss, Elisabeth & Abraham, Werner (eds). 2007. Nominal Determination. 1)'pology, Co11text Constraints, and Historical Emergence [Studies in Language Companion Series 89]. Amsterdam: John Benjamins. Thniinsson, Hoskuldur. 2005.1slensk tunga. 3 bi11dL Set11ingar. Handb6k um setningafradJi. Reykjavik: Almenna b6kafelagill. Thniinsson, Hoskuldur, Petersen, Hjalmar P., Jacobsen, J6gvan f Lon & Hansen, Zakaris Svabo. 2004. Faroese. An Overview and Reference Grammar. T6rshavn: F0roya fr6llskaparfelag. Wackernagel, Jacob. 1908. Genetiv und Adjektiv. Mflanges de Li11guistique Offerts a M. Ferdi11and de Saussure, 125-152. Paris: Champion.
:n
PART I
Variation
Scaling the variation in Romance and Germanic nominalizations Artemis Alexiadou, Gianina Iordachioaia & Florian Schafer* Universitiit Stuttgart
We investigate the distribution of verbal and nominal layers in Romance and Germanic nominalizations. Specifically, we examine pairs of 'verbal' vs. 'nominal' nominalizations in two Romance (Spanish & Romanian) and two Germanic (English & German) languages. Our study proposes a large spectrum of nominal and verbal properties. While these are differently instantiated among languages, the variation we find cannot be attributed to a Germanic vs. Romance parameter; instead, we find micro-variation constrained by the compatibility between the general building blocks of verbal and nominal categories. Besides the vP-layers responsible for argument structure and Aktionsart and the DPlayer responsible for the nominal external syntax, we make a case for further functional verbal and nominal layers in nominalizations: Asp(ect)P, Class(ifier)P, and Num(ber)P. These projections are in complementary distribution in some languages and co-occur in others.
1.
Introduction
Nominalizations exhibit well-known mixed categorial properties, combining both nominal and verbal features (Grimshaw 1990, Borer 1993, Reuland & Kosmeijer 1993). Recently, this behavior has been related to their internal structure and is taken to reflect the amount of verbal structure embedded under a varied amount of nominal structure (cf. Harley & Noyer 1999, Borsley & Komfilt2000, Alexiadou 2001, van Hout & Roeper 1998, Borer 1993,2001 among others).
*
We thank the audience of the congress Variation and change in the Romance and Germanic noun phrase at the University of Amsterdam, two anonymous reviewers and the editors ofthis volwne for comments and suggestions. Our research has been supported by a DFG (German Research Foundation) grant to the project B1, The formation and interpretation ofderived nominals, as part of the Collaborative Research Center 732, Incremental specification in context, at the University of Stuttgart.
26
Artemis Alexiadou, Gianina Iordl\chioaia & Florian Schafer
In this article we address the following questions: How many and what kind of verbal and nominal layers are allowed in nominalizations and what are the possible or impossible combinations thereof? Furthermore, how are these layers distributed across Germanic and Romance languages? ln order to provide an answer to these questions, we examine pairs of a 'verbal' and a 'nominal' nominalization in two Romance (Spanish & Romanian) and two Germanic (English & German) languages. The pairs to be discussed are listed below: 1 Romanian:
supines (RS with participial morphology) vs. infinitives (RI)
Spanish:
1'erbalvs. nominal infinitives (SVI vs. SNI)
English:
verbal vs. nominal gerunds (VG vs. N G)
German:
verbal vs. nominal infinitives (GVI vs. GNI)
(Cornilescu 2001, Iordachioaia & Soare 2008); (Plann 1981, Miguel1996); (Abney 1987, Borer 1993, Kratzer 1994, Alexiadou 2005); (Esau 1973, Ehrich 1977, 1991). The paper is structured as follows. In Section 2, we introduce the four nominalization pairs. Section 3 is a summary of the general properties of nominalizations which we organize into two scales: a verbal and a nominal scale. In Section 4, we take a closer look at the aspectual information present in nominalizations and make a case tor including the inner aspect information under ClassifierR while AspectP hosts the outer aspect contribution. In Section 5, we discuss the cross-linguistic variation in the structure of nominalizations in view of the general constraints on the compatibility between verbal and nominal categories. The varying distribution of these nominal and verbal layers explains the gradual properties in nominalizations across languages (cf. Ross 1972).
2.
The verbal vs. nominal nominalization patterns: A first approximation
As explicitly stated in Borer (1993), two main properties are suggestive of the verbal nature of nominalizations: the presence of case patterns similar to those of verbal clauses (NOM/ ACC) and the licensing of adverbials. In this section we make use of these criteria in order to provide a first classification of nominalization types.
2.1
The Romance languages (Spanish vs. Romanian)
Spanish has two types ofnominalized infinitives, which we label here verbal infinitives and nominal infinitives (VI vs. NI). Miguel (1996) takes the distribution of the nominative vs. PP-subject in (1) to be the main distinction between them.
1.
Here we have nothing to say about further combinations of these two extremes.
Scaling the variation in Romance and Germanic nominalizations
(1) a.
el munnurar
la gente
the munnur.INF the people.NOM b. el munnurar de las fuentes the munnur.INF of the fountains Only VIs license accusative case (2a). Nis appear with bare Ns (2c), but not with accusative DPs (2b, see also Perez Vazquez 2002). This suggests that the bare N is incorporated (see Bosque 1989 for details). (2) a.
b.
[el can tar yo Ia. Tra.viata] the sing.INF I.NOM the.Acc Traviata
[*El cantar
estas
coplas de Lola] nos emociona.
the sing.INF these.Acc songs of Lola us
[El
cantar
coplas
moves
de Lola] nos emociona.
the sing.INF songs.Acc of Lola us
moves
Furthermore, Spanish VIs allow adverbial modification (3a, b), while Nis (3c, d) can only be modified by adjectives (Miguel1996, Ramirez 2003 ): (3) a.
el andar
errabundamente!*errabundo Juan
the go-about.INF aimlessly/
aimless
Juan
b. el (*cmtstante) escribir ella 1wvelas constantemente the constant write.INF she novels constantly c.
el an dar
errabundo/* errabundamente de Jua1t
the go-about.INF aimless/
d. el constante temer the constant fear.INF
aimlessly
of Juan
(*constantamente) de Juan constantly
of Juan
Romanian has two types of nominalizations licensing argument structure: the infinitive (RI) and the supine (RS ). Although neither of them licenses verbal case (their object or subject carries genitive), the two differ in that aspectual adverbs modifY only the supine (4a).lnfinitives combine with the corresponding adjectives (4b) (see lordikhioaia & Soare 2008). 2 (4) a.
(*constantul) citit(ul) (constant)
(constant) al ziarelor
read.suP-the constantly of journals.GEN
b. (constanta) omitere(a) (*constant) a unor infonnafii constant omit.INF-the constantly of some infos To conclude, while both Romance languages have a verbal nominalization licensing adverbs, only the Spanish one assigns verbal case to its arguments. The ungrammaticality of the pre-nominal position for consta11t indicates that the homonymous masculine-neuter adjective is excluded in (4a).
2.
27
28
Artemis Alexiadou, Gianina Iordl\chioaia & Florian Schafer
The Germanic languages (English vs. German)
2.2
The English 1'erbal gerund (VG) licenses accusative objects (Sa), while the nominal gerund (NG) takes PP-objects (6a). Furthermore, VGs take adverbial modifiers (Sa) and disallow adjectival ones (6b), while NGs display the opposite behavior (Sb vs. 6a). (S) a. Pat disapproved of]ohn's quietly leaving the room. b. *The carefully restoring of the painting took months. (6) a. his/J ohn's prompt answering of the question b. *his prompt answering the question
Gemtatt verbal infinitives (GVI) license accusative case and can be modified by adverbs (7). Nomina./ infinitives (GNI) take genitive (or PP-) objects and are modified by adjectives (8). (7)
[Hiiufig
die
Sterne Beobachten] macht Spass.
frequently the.ACC stars (8)
observe.INF makes fun
[Das haufige Beobachten der
Sterne] macht Spass.
the frequent observe.INF the.GEN stars
makes fun
A difference between German VIs and English VGs (Sa) is that the tormer cannot realize an overt subject, as (9) illustrates.3 (9) ( .. Peters) die Sterne Beobachten Peters.GEN the.Acc stars observe.INF
3·
The verbal vs. nominal scale
In this section, we put forth a set of finer-grained distinctions for the nominal and verbal properties of nominalizations. We propose two 'categorial' scales that interact with one another: a 1'erbal and a nominal one. Each scale contains a number of properties (cf. Sleeman 2010). As we will point out, languages differ as to the cut-off points they choose within these scales (cf. Ackema & Neeleman 2004). On this view, the distinction between Vs and Ns is not absolute, but gradual in nature: the V/N cut-off point of a nominalization can be located at various points in these scales (contra Panagiotidis & Grohmann 2006).4 3·
In this sense German VIs seem similar to PRO-ing gerunds, Siegel (1998): (i)
PRO smoking ctgars is fun
4- In this paper we are concerned with the criteria that help us identify the functional projections in nominalizations. A further point that needs to be clarified and which we hope to address in future research is what constraints are at work in a hierarchy of projections. We expect this to follow from a cartographic approach like in Cinque (1999) and subsequent work.
Scaling the variation in Romance and Germanic nominalizations 29 The verbal scale
3.1
Subject with nominative case: this is found only in Spanish VIs (see la, 2a). Occurrence of modal or auxiliary verbs is found in Spanish VIs (lOa), German VIs
i. ii.
(lOb) and the English verbal gerund (lOc). (10) a.
[El haber
el escrito novelas] explica su fama.
the have.INF he written novels
explains his fame
b.
[Dauernd Kuchen Essen Wollen] nervt. permanently cake eatiNF wantiNF is-annoying
c.
his ha.ving read War and Peace
iii. Accusative case on objects: this is possible with the verbal patterns in Spanish, English and German; see (2a), (Sa) and (7). iv. Projection of outer Aspect: this is evidenced by aspect shift and aspectual adverbs; see Section 4 for a detailed discussion. v. Argument Structure realization: this holds for all nominalizations considered here.
3.2
I.
The nominal scale
Genitive/PP subject: it is possible in both types of English gerunds (Sa/6a), in German (lla) and Spanish (lb) Nis and in both Romanian nominalizations (llb); it is excluded in German and Spanish VIs ( 12a, b). (11)
a.
(Toms)
Beobachten des
Kindes (durch Tom)
Tom.GEN observe.INF the.GEN child
by
b. sosirealsositul lui Ion Ia arrive.INF-the/arrive.suP-the John.GEN in (12)
Tom
timp time
a. *Toms hiiufig das Kind Beobachten Tom.GEN frequently the.Acc child observe.INF b. el escribir constantemente novelas (*de) ella the write.INF constantly novels (of) she
II. Genitive/PP-object is compatible with both Romanian nominalizations (4a-b ), and the nominal pattern in Spanish, English and German (2b, 6a, 8); excluded with the verbal pattern in Spanish, English and German (2a, Sa, 7). III. Gender features: we follow Picallds (2006) insight on the presence of a Class(ifier) Pin nouns that introduces information about declension (Gender, Case-inflection). If such features are present, an nP projection is present too, creating a
30
Artemis Alexiadou, Gianina Iordl\chioaia & Florian Schafer
nominal internal structure; on this view, n° moves to Class0 to check its nominal features (Picallo 2006).5 Spanish Nls carry gender features which - although not visible in the suffix -r- become obvious in anaphoric contexts, where an NI can be reterred to only by the masculine pronoun eland not by the default neuter pronoun ello usually employed with CPs (Miguel1996). ( 13) Accostumbrado al dulce mirar de su amada, used-to the sweet gaze.INF of his beloved,
ya
no podia vivir sin
8/*ello.
now not could live withouthim/it "Used to the sweet gaze of his loved one, he could no longer live without it:' In Romanian the infinitive establishes anaphoric relations with the feminine demonstrative aceasta (14a), while the supine rejects the masculine-neuter acesta and can only be referred to by the genderless asta (14b ), the common anap hor for CPs (14c ). This suggests that infinitives are feminine, while supines are genderless (default) (see details in Iordachioaia & Soare 2008).6 (14) a. Am vorbit despre interpretarea lui Hamlet. we spoke about interpretation.INF.the of Hamlet
Se pare ca aceasta/??asta ti apparently, this.F/??it b. Am vorbit we spoke Se pare ca apparently,
consacrtl pe
actorii
tineri
them validates Ace actors.the young
despre interpretatul
lui Hamlet
about the interpretation.suP of Hamlet *acesta/asta 1i atrage pe top actorii tineri. *this.M/it them attracts Ace all actors.the young
c. Ctl Ion a venit, asta/*aceasta/*acesta that John has come, it.N/this.F/ this.M
§tiU.
I-know
This correlates with the defectiveness of the supine with respect to case declension; the supine cannot appear in the genitive-dative form. (15)
Alunecarile de teren au
aparut
mudflows- the
occurred
have
din cauza ttlierii of cause CUt.INRGEN/
/*taiatului
ptldurilor
cutSUP.GEN woods.GEN
5· The presence of a Classifier in the syntax and semantics of nouns in English has been independently argued for in Kratzer (2005). See also Lowenstamm (2007) for an insightful view on the interaction between n, gender and nominal classes. 6. The 'long' infinitive (e.g. cd11ta-re) in Romanian is always a nominal In verbal contexts the prepositional infinitive is used instead (e.g. a c4nta). This explains the dear presence of feminine gender on the infinitive, despite its verbal properties like argument realization.
Scaling the variation in Romance and Germanic nominalizations In German, dieses is an anaphor for nouns only, while das/dies are anaphors for both nouns and CPs (see 16a, b). (16) a.
Daft Maria bereits angekommen ist, that Mary already arrived is dasl dies/ *dieses wei.B ich genau. it/ this/ this know I well
b. Hans hat ein rotes Buch. Hans has a red book
Das/ Dies! Dieses war sehr teuer. it/ this/ this was very expensive German Nis can be referred to by dieses (17a), but VIs cannot (17b).This suggests that German Nis are neuter, while VIs are genderless/default. It also correlates with the case defectiveness of VIs ( 18b ). (17)
a. Niichtliches Beobachten der Sterne gefallt ihm. at-nighLADJ observe.INF the.GEN stars pleases him Dies/Dieses/Das entspannt ihn. this/this/it relaxes him
b. Nachts
die
Sterne Beobachten gefallt ihm.
at-nighLADV the.Acc stars observe.INF pleases him *Dies/*Dieses/Das entspannt ihtt. *this/*this/it relaxes him (18) a.
wegen
des
Lesens eines Buches
because-of the.GEN read.INF a.GEN book
b. *wegen
ein
Buch Lesens
because-of a.Acc book read.INF.GEN IV. Availability ofplural: Spanish and German nominalizations never pluralize. In Romanian, plural is available only with the infinitive (19). Some English NGs allow plural, if there is no competition with other affixes (see Alexiadou, Iordachioaia & Soare 2010), but VGs do not (20). In Section 4, we explore the link between the availability of plural and the aspectual properties of nominalizations. (19)
demolarilel*demolaturile frecvente
ale ca,.tierelor
vechi
demolish.INF-PL/SUP-PL frequent-PL of quarters.GEN old "the frequent demolitions of old quarters" (20) a. the repeated killings of unarmed civilians b. *Emma's readings the poem V.
Possibility to combine with all types of detenniners: overt determiners are out in English VGs. German VIs allow definite determiners. The nominal counterparts in both languages allow all kinds of determiners.
31
32
Artemis Alexiadou, Gianina Iordl\chioaia & Florian Schafer
(21) a. *Th.at/*the/*a criticizing the book annoyed us. b. The/that/?a reading of the manuscript pleased us. c. dasldiesesl*einl*keinljedes die Marseillaise Singen the/this/a/no/every the.Acc Marseillaise sing.INF d. dasldiesesleinlkeinljedes Singen der Marseilla.ise the/this/a/no/every sing.INF the.GEN Marseillaise The Romanian infinitive (22a) and the Spanish NI (22c) freely combine with determiners. By contrast, the Romanian supine (22b) and the Spanish VI (22d) only allow the definite determiner. (22) a.
o/acea tncalcare
a drepturilor omului
de catre ministru
a/that violate.INF of rights.GEN human.GEN by
minister
al rufelor b. (*un/*acel) spalat(ul) a/that wash.suP(the) of laundry.GEN c. aquel/eseleste/un/el lamenta.r (*desesperadamente) that/this/a/the lament.INF (desperately)
de dos pastores of two shepherds d. *ese!*aquel/el haber tl escrito esa carta this/that/the have.INF he.NOM written that letter 3·3
Nominal and verbal mixed properties: Summary and conclusions
The most nominal properties involving gender and plural marking are clearly excluded in verbal nominalizations. The least nominal ones like the presence of genitive subjects are sometimes also shared by verbal nominalizations (e.g. the possessive subject in the English VG and the genitive subject in the Romanian supine). There is also an issue concerning the licensing of case: while a verbal internal structure usually involves an accusative object, this is not always so, e.g. in the Romanian supine. Moreover, only Spanish VIs license nominative case. In the next section, we focus on the aspectual properties of nominalizations, which will help us understand the interaction between nominal and verbal layers.
4
Inner and outer aspect in nominalizations
Following Verkuyl (1993), we differentiate between inner aspect (Aktionsart, lexical aspect) and outer aspect (grammatical aspect). Inner aspect is realized in the VoicePvP domain (following Alexiadou, Anagnostopoulou & Schafer (2006), Harley (2007)
Scaling the variation in Romance and Germanic nominalizations
and Marantz (2005)). 7 Outer aspect is related to an Asp(ect)P which is outside the Ak.iionsart domain (23). (23) [Tense [AspP
[VoiceP [vP [RootP])])]
outer Aspect inner Aspect We will also make use of boundedness (Jackendoff, 1991 ), a unifying notion of nominal number and verbal aspect which has long been argued for in the semantic literature (see also Mourelatos 1978, Bach 1986, Krifka 1986). In these terms, morphological plural, mass nouns, and atelic and imperfective aspect are [-b] ounded; morphological singular, count nouns, and telic and perfective aspect are [+b]. This gives us a common notion for nominal and verbal plurality to describe the mixed properties of nominalizations, where, as we will see, there is a clear interaction between the inner aspectual properties of the base verbs and the external nominal layers. The nominal layers that we assume on top of the inner aspect domain are given in (24): (24) [DP [NumberP (ClassP (±count) [nP [inner Aspect domain ... ClassP accommodates the inner aspect under a [±count] feature. Telic nominalizations, like count nouns, project Class [+count], which is the input for a further NumberP; atelic nominals, like mass nouns, project Class [-count], which blocks NumberP. While Number gives information about the form (i.e. plural/singular marking), the [±count] specification indicates the semantic 'number': [-count] means semantic plurality; [+count] means semantic singularity. The form of the noun in Class is always non-plural (cf. Kratzer 2005). These notions successfully accommodate the inner aspect specification of nominalizations, which is a semantic number that influences the availability of plural marking (i.e. the projection of NumberP): in general atelic (i.e. [-count]) nominalizations cannot pluralize, while telic ones (i.e. [+count]) can (see Borer 2005, Alexiadou, lordachioaia & Soare 2010, for details). 41
The outer aspect projection
Some nominalizations introduce aspect shift in a way similar to the outer aspect contributed, for instance, by the verbal progressive in "The train is arriving': where the telic inner aspect is overwritten by imperfective outer aspect We argue that these nominalizations project an AspectP, like verbs. The Romanian supine can be formed from most verbs and shifts their inner aspect:8 achievements (25a), accomplishments (25b) and punctual events (25c), all known as [+b]/telic, get a habitual reading in the supine:
7·
For different ways to implement this distinction, see Ramchand (2008) or Borer (2005).
8. Cornllescu (2001) argues that the main difference between the supine and the infinitive is aspectual: the supine is atelic, the infinitive is telic, see Section 4.2.
33
34
Artemis Alexiadou, Gianina Iordl\chioaia & Florian Schafer
(25) a. sositul lui Ion cu tnttrziere arrive.suP-the John.GEN with delay "John's (habit of) arriving late" b. mlncatul micului dejun pe terastl eat.suP-the breakfastGEN on terrace "(the habit of) having breakfast on the terrace"
Clipitul
.Mariei
In acest moment e enervant
blink..suP-the Mary.GEN in this moment is irritating "Mary's blinking in this moment is irritating:' Following Iordachioaia & Soare (2008), we assume that the supine, but not the infinitive, projects Aspect which hosts a plumctional operator (see Lasersohn 1995, Van Geenhoven 2004). This explains the compatibility of atelic for-PPs with inherently telic verbs (26a vs. 26b ): (26) a.
sositul
lui Ion
cu
tnttrziere timp de 3 ani
arrive.suP-the John.GEN with delay
b. #sosirea.
lui Ion
cu.
for 3 years
inttrziere timp de 3 ani
arrive.INF-the John.GEN with delay
for 3 years
The English VG is also grammatical with most verbs (27) (Borer 2005) and contributes imperfective/[-b] outer aspect (Pustejovsky 1995). The projection of AspP in Romanian supines and English V Gs is further supported by the compatibility with aspectual adverbs (4a), (28) (Alexiadou 1997, Cinque 1999). For discussion see Alexiadou et al. (2010). (27) a. b. c. d.
John's arriving at 5 pm is unlikely. John's eating breakfast Marfs blinking is annoying. John's knowing the answer
(28) John's constantly reading the morning newspaper
Spanish verbal infinitives have no special aspectual contribution, but they can appear in the perfective with haber (lOa), so they exhibit aspect shift, and they are not sensitive to the inner aspect of the root ((29) vs. (32)). (29) a.
el a.ndar
el nino tan ta.rde por esa zona
the go-about.INF the child so
late
in
b. el lamentar la familia lo sucedido the regretiNF the family the happened "the family regretting what happened" c.
el llegar
tan tarde el ni1io
the arrive.INF so
late
the child
that district
Scaling the variation in Romance and Germanic nominalizations
d. el comprar una casa the buy.INF a
Juan
house Juan
Both German infinitives induce impeifectivity (Ehrich 1991); this is shown by the fact that even N Is of telic verbs do not tolerate resultative VPs, but allow atelic process-VPs.9 (30) Das Abholzen des Waldes the deforesliNF the.GEN forest a. *muss bis morgen friih erreicht sein must till tomorrow morning achieved be b. wird zwei Jahre lang fortgesetzt is two years long continued 4.2
Inner aspect: The [±count] feature on ClassP
Some other nominali:zations are sensitive to the inner aspect of the base. We argue that they do not project AspP, but accommodate the inner aspect under Class. 10 The Romanian infinitive is not compatible with atelic bases like the unergatives in (31) (Cornilescu 2001). (31) *muncirea l*alergarea lui Ion "Work.INF-the/ *run.INF-the of John.GEN
The Spanish NI and the English NG are both incompatible with telic bases (Miguel 1996, Borer 2005). Spanish Nls are fine with activities (32a), but out with achievements (32b) and accomplishments (32c). (32) a.
el trabajar de Juan en el campo the work.INF of John in the garden
de Pedro a 1a habitaci6n b. *el intenso llegar the intense arrive.INF of Pedro to the room c. *el rapido construir la casa. de los albaiiiles the fast build.INF the house by the workers Borer (2005: 2391f.) shows that English NGs are possible with non-culminating events (activities and semelfactives in (33)), but out with achievements (34), creating an anti-
telicity effect. 9·
The difference between GNis and GVIs is that the latter refer only to generic events.
On this analysis, the ClassP must 'see' the features within the VP domain. This can be formalized in terms of Agree: the specification of ClassP concerning singularity, division and homogeneity (Borer 2005, vol 1: cb.. 4) has to Agree with the specification of the predicate via probing into the VP. This is possible irrespectively of whether nP is a phase. as Agree is not subject to the Phase-Impenetrability Condition and intervention effects (see Boskovic 2007). 10.
35
36
Artemis Alexiadou, Gianina Iordl\chioaia & Florian Schafer
(33) a. the sinking of the ships b. the falling of the stock prices c. the jumping of the cows (34) a. *the arriving of the train b. *the erupting of Vesuvius c. *the exploding of the balloon Spanish N Is, Romanian infinitives and English N Gs are incompatible with aspectual adverbs, which indicates the unavailability of AspectP (35a-c). (35) a. El (constante) murmurar (*constatamente) del mar the constant whisper.INF constantly of the sea b. omiterea constanta!*constant a detaliilor omit.INF constant/constantly of details.GEN
c. John's constant omitting of details/*constantly A correlation emerges concerning the aspectual structure of nominalizations and their ability to pluralize (Alexiadou, Iordachioaia & Soare to appear): Plural marking is only available with bounded (i.e. perfective or telic) aspect and in a nominal environment (i.e. presence of gender, case inflection, adjectival modification). For a structure to be nominal, nP and ClassP must be projected (cf. Picallo 2006, Borer 2005). NumberP can be further projected on top of a Class [+count].
5·
The building blocks of nominalizations and their cross-linguistic distribution
In general, nominalizations come either with the verbal internal structure in (36a) or with the mixed internal structure in (36b). In our terms, a verbal internal structure is associated with verbal functional projections, while a mixed internal structure is associated with the additional presence of nominal layers (see Borsley & Kornfilt 2000): (36) a. [DP [Verbal FP ... [...]] b. [DP [Nominal FP ... [Verbal FP ... [...]]]] We now address this distribution with respect to our constructions. The most verbal nominalization type is Spanish VIs (37). The licensing of nominative Case indicates that Tense is projected. The presence of Tense is evidenced by the presence of reflexive clitics in Spanish VIs (Pesetsky & Torrego 2004), assuming that clitics in Romance attach toT" (38a vs. 38b). (37) [DP [TP [Aspect [VoiceP [vP [Root]]]]]]
Scaling the variation in Romance and Germanic nominalizations ( 38)
a.
el afeitar-se
la barba Juan
the shaving-clitic the beard Juan
b. *el afeitar-se de Ia ba.rba
Romanian supines, English VGs and Germa.n VIs have the structure in (39). The difference between them only concerns the features under Aspect0 which are distributed as in (40). 11 (39)
[DP [AspectP [VoiceP [vP [Root]]]]]
(40) Romanian supine English verbal gerund German verbal infinitives
~ ~ ~
pluractionality imperfectivity genericity
(41) represents the constructions which have a rich nominal internal structure in addition to the verbal layers. German Nls have the structure in (41a), Spanish Nls the one in (41b), Romanian infinitives and English NGs the one in (41c): (41) a. b. c.
[DP [ClassP [nP [AspectP [VoiceP [vP ... [DP [ClassP[ -count] [nP [VoiceP [vP [DP [(NumberP) [ClassP[±count] [nP [VoiceP [vP
Adjectival modification and genitive Case assignment are located within the nP. Plural is available under NumberP, provided that ClassP is [+count]. Low adverbs will be licit if AspectP is present (Borer 1993, Alexiadou 2001). This means that German Nls will license both adjectives and adverbs: ( 42) das dauernde laut Singen der Marseillaise the constant loudly sing.INF the.GEN Marseillaise
Romanian supines constitute an odd case, as the internal argument bears genitive case, although the structure is clearly verbal But note that the Romanian supine is the only structure introduced by a suffixed determiner. We suggest that there is a link between this and the genitive Case in this nominalization. Specifically, the affixed article in the supine creates a nominal environment, albeit a defective one, hence the case that appears is the one found in nominal environments. Following Giusti (2002), the Romanian article is nothing more than a grammatical morpheme responsible for realizing nominal features (cf. Abney 1987). As Giusti shows, it lacks semantic import, as the cooccurrence of two definite articles in one DP does not produce a two-reterentinterpretation effect This is certainly not the case for the English or the German determiners. A further point of variation concerns elements that appear in Spec,DP. Romance languages and German VIs disallow genitive subjects in pre-nominal position. For German we would like to suggest that PRO is present and for this reason an overt DP
n. One could argue that different Aspect projections are involved in each case, following Cinque (1999).
37
38
Artemis Alexiadou, Gianina Iordl\chioaia & Florian Schafer
is blocked. In Romance this relates again to the X-status ofSpec,DP in these languages as opposed to English (Abney 1987).
6. Conclusion
The variation found in nominalizations within and across languages cannot be attributed to a Germanic vs. Romance parameter; we find micro-variation constrained by the compatibility of the general building blocks ofverbs and nouns. The layers responsible for argument structure/Aktionsart properties (VoiceP/vP) and the nominal external syntax (DP) are always present. For the fine-grained differences, we argued for the inclusion of further functional layers from the verbal and nominal domain: Asp P, ClassP, NumberP. The varied distribution of these nominal and verbal layers explains the gradual properties in nominalizations across languages (cf. Ross 1972).
References Abney. Steven 1987. The English Noun Phrase in its Sentential Aspect PhD dissertation, MIT. Ackema, Peter & Neeleman, Ad 2004. Beyond Morphology. b1terJace Conditions on Word Formation. Oxford: OUP. Alexiadou, Artemis. 1997. Adverb Placement: A Case Study in Antisymmetric Syntax [Lingulstik Aktuell/Linguistics Today 18]. Amsterdam: John Benjamins. Alexiadou, Artemis. 2001. Functional Structure i11 Nominals: Nomi11alizatlon and Ergativity [Linguistik Aktuell/Linguistics Today 48]. Amsterdam: John Benjamins. Alexiadou, Artemis. 2005. Gerund types, the present participle and patterns of derivation In Event Arguments in Syntax, Semantics and Discourse, Claudia Maienborn & Angelika WOl.lstein-Leisten (eds), 139-152. Tlibingen: Niemeyer. Alexladou, Artemis, Iordl\chioaia, Gianina & Soare, Elena. 2010. Number/aspect interactions in the syntax of nominalizations: A distributed morphology approach. Journal of Linguistics 46(3): 537-574. Alexiadou, Artemis, Anagnostopoulou, Elena & Schafer, Florian. 2006. The properties of anticausatives crosslinguistically. In Phases of Interpretatio11, Mara Frascarelli (ed), 187-212. Berlin: Mouton de Gruyter. Bach, Emmon 1986. The algebra of events. Linguistics and Philosophy 9: 5-16. Borer, Hagit. 1993. Parallel mmphology. Ms, University of Utrecht. Borer, Hagit. 2001. The forming, the formation and the form ofnominals. Handout USC. Borer, Hagit. 2005. Structuring Sense. Oxford: OUP. Borsley, Robert & Kornfilt, Jaklin 2000. Mixed extended projections. In The Nature and Function ofSyntactlc Categories, Robert Borsley (ed), 101-131. San Diego CA: Academic Press. BoSkoviC, 2eljko. 2007. Agree, phases and intervention effects. Linguistic Analysis 33(1-2): 54-96. Bosque, Ignacio. 1989. Las Categorfas Gramaticales. Madrid: Sfntesis. Cinque, Guglielmo. 1999. Adverbs and Functional Heads: A Crosslinguistic Perspecti1'e. Oxford: OUP.
Scaling the variation in Romance and Germanic nominalizations Cornllescu, Alexandra. 2001. Romanian nominalizations: Case and aspectualstructure. Journal of Linguistics 37(3): 467-501. Ehrich, Veronika. 1977. Zur Sy11tax u11d Semantik vo11 Substa11tivierungen im Deutsche11. Kronberg: Scrlptor. Ehrich, Veronika. 1991. Nominalisierungen. In Semantik. Ei11 Handbuch der internationalen Forschung, Arnim von Stechow & Dieter Wunderlich (eds), 441-458. Berlin: Mouton de Gruyter. Esau, Helmut 1973. Nominalizatio11 and Complementation 111 Modern German. Amsterdam: North-Holland Giusti, Giuliana. 2002. The functional structure of noun phrases. A bare phrase structure approach. In Fu11ctlonal Structure in DP and IP: The Cartography ofSyntactic Structures, Guglielmo Cinque (ed), 54-90. Oxford: OUP. Grimshaw, Jane. 1990. Argument Structure. Cambridge MA: The MIT Press. Harley, Heidi. 2007. External Arguments: On the independence of voice and v. Paper presented at the XXX GLOW Colloquium, University ofTroms0. Harley, Heidi & Noyer, Rolf 1999. Mixed nominallzations, short verb movement and object shift in English. Proceedings ofNELS 28. van Hout, Angeliek & Roeper, Tom. 1998. Events and aspectual structure in derivational morphology. In.l\.-IIT Working Papers 111 Li11gulstics: Papers from the UPem1/MIT Roundtable on Argument Structure and Aspect, Vol. 32, Heidi Harley (ed), 17 5-200. Cambridge MA: The MIT Press. Iordikhioaia, Gianina & Soare, Elena. 2008. Two kinds of event plurals: Evidence from Romanian nominalizations. In Empirical Issues In Synta.r: and Semantics 7, Olivier Bonami & Patricia Cabredo-Hofherr (eds), 193-216. <WWW.cssp.cnrs.fr/eiss7/i.ndex_en.html> Jackendoff, Ray. 1991. Parts and Boundaries. Cognition 41: 9-45. Kratzer, Angelika. 1994. On external arguments. In Functional Projections [University of Massachusetts Occasional Papers 17], Elena Benedicto & Jeffrey Runner (eds ), 103-30. Amherst MA: GLSA. Kratzer, Angelika. 2005. On the plurality of verbs. In Event Structures In Linguistic Form and Interpretation, Johannes Dolling, Tatjana Heyde-Zybatow & Martin Schafer (eds). Berlin: Mouton de Gruyter. KrUka, Manfred 1986. Nomi.nalreferenz und Zeitkonstitution. Zur Semantik von Massentermen. Pluraltermen und Aspekt:kl.assen. PhD dissertation. University of Munich. (Published 1989, Munich: Fink). Lasersohn, Peter. 1995. Plurality, Conjunction and Events. Dordrecht: Kluwer. Lowenstamm, Jean 2007. On little n, -../,and types of nouns. In Sounds of Silence, Jutta Hartmann, Henk van Rtemsdijk & Veronika Hegedus (eds). Amsterdam: Elsevier. Marantz. Alex. 2005. Objects out of the lexicon: Objects as event. Ms, NYU. de Miguel, Elena 1996. Nominal infinitives in Spanish: An aspectual constraint. Canadian Journal ofLinguistics 41(1): 29-53. Mourelatos, Alexander P.D. 1978. Events, processes and states. Linguistics and Philosophy2(3): 415-434. Panagiotidis, Phoevos & Grohmann, Kleanthes. 2006. On positioning categorial switches. Paper presented at the conference Universality and Particularity in Parts-of-Speech Systems, University of Amsterdam. Perez Vazquez. Enrtqueta. 2002. A mixed extended projection: The nominalized infinitive in Spanish and Italian. Quaderni del Laboratorio di LingiUstica 14: 143-159. Pisa: Scuola Normale Superiore.
39
40
Artemis Alexiadou, Gianina Iordl\chioaia & Florian Schafer Pesetsky, David & Torrego, Ester. 2004. Tense, case and the nature of syntactic categories. In The Syntax of Time, Jacqueline Gueron & Jacqueline Lecarme (eds), 495-537. Cambridge MA: The MIT Press. Picallo, M. Carme. 2006. Some notes on grammatical gender and I- pronouns. In Proceedings of the Workshop Specificity and the Evolution!emerge11ce ofNominal Determination Systems i11 Romance [Arbeitspapier Nr. 119], Klaus von Heusinger, Georg A Kaiser & Elisabeth Stark (eds), 107-121. Konstanz: Fachbereich Sprachwissenschaft der Universitiit Konstanz. Plann, Susan. 1981. The two el + infinitive constructions in Spanish. Li11gulstic A11arysis 7(3): 207-241.
Pustejovsky, James. 1995. The Generative Lexico11. Cambridge MA: The MIT Press. Ramchand, Gll.lian. 2008. First Phase Syntax. Cambridge: CUP. Ramirez, Carlos Julio. 2003. The Spanish nominall.zed infinitives: A proposal for a classification. Toro11to Working Papers 111 Linguistics 21: 117-133. Reuland, Eric & Kosmeijer, Wim. 1993. Projecting inflected verbs. In The Parameterization of lT11iversal Grammar [Lingulstlk Aktuell/Linguistks Today 8], Gisbert Fanselow (ed), 3772. Amsterdam: John Benjamins. Ross, John. 1972. The category squish: Endstation hauptwort. In Proceedings of the Eighth Regional Meeti11g of the Chicago Li11guistic Society, Paul M. Peranteau, Judith N. Levi & Gloria C. Phares (eds), 316-338. Chicago IL: CLS. Siegel, Laura 1998. Gerundive nominals and the role of aspect. InProceedingsofESCOL 1997, Jennifer Austin & Aaron Lawson (eds). Ithaca NY: CLC Publications. Sleeman, Petra. 2010. The nominalized infinitive in French: Structure and change. Lingufstica, Revista de Estudos Lingufsticos da Universidade do Porto 5. Van Geenhoven, Veerle. 2004. For-adverbials, frequentattve aspect, and pluractionality. Natural Language Semantics 12: 135-190. Verkuyl, HenkJ. 1993. A Theory of Aspectuality. The Interaction between Temporal and Atemporal Structure [Cambridge Studies in Linguistics 64]. Cambridge: CUP.
What all happens when a universal quantifier combines with an interrogative DP* Robert Cirillo Universal quantifiers such as all select a DP as their complement and can be 'floated' or 'stranded' by that DP, and in certain Germanic languages they can also co-occur with an interrogative DP. The pwpose of this article is to investigate whether interrogative and non-interrogative DPs that co-occur with a universal quantifier in the Germanic languages have the same relationship to that quantifier and have gone through the same selection process. I begin with evidence from German that universal quantifiers can select and be stranded by interrogative as well as non-interrogative DPs, but I ultimately argue, going back to an analysis in Giusti ( 1990b), that a universal quantifier co -occurring with an interrogative is base-generated to the right of that interrogative, not to its left. I also propose that the formation of interrogative expressions involving universal quantifiers may take place in the syntax or in the lexicon, depending on the language.
1.
Introduction
It is well known that universal quantifiers in the Germanic and Romance languages select a DP as their complement and that they can also appear in a position left-adjoined to vP: (1) a.
Alle die Studenten haben das Buch gelesen. all the students
have
(German)
the book read
b. Die Studenten haben alle das Buch gelesen. the students (2) a.
the students have
the book read
all
libro.
(Italian)
read the book
Gli studenti hanna tutti letta il the students have
*
all
Tutti gli studenti hanna letta il all
b.
have
libra.
read the book
My thanks to two anonymous reviewers for their many very significant contributions to this article.
42
Robert Cirillo The first one to propose that floating quantifiers such as those in (lb) and (2b) are base-generated as left adjuncts to DP rather than to vP and are then stranded by the DP when it moves to a higher position was Sportiche (1988). This so-called Stranding Analysis of floating quantifiers was refined by Giusti (1990a), who treated the stranded quantifier not simply as a left adjunct to DP but as the head of a quantifier phrase that dominates DP. Shlonsky (1991) put the finishing touches on the Stranding Analysis, and Cirillo (2009) updated it for more recent developments in linguistic theory. It is widely but not universally accepted. Its validity is assumed in this article. In deriving (lb) and (2b) under the Stranding Analysis, one begins with the following structure: (3)
QP
~
Q'
SPEC
~
Q
DP
all
~ SPEC
D'
~
D
NP
the
~ SPEC
N'
I
N
students
Under this approach, the entire QP in (3) can move to a higher position, producing (1a) and (2a), or the DP selected by the universal quantifier can move up by itself, via [SPEC, QP], stranding the quantifier and producing (1 b) and (2b ). Note that when the D P moves out of Q P via [SPEC, Q P] it must continue on its own. It cannot pied-pipe the quantifier. Put in another way, once the DP has moved to [SPEC, QP], the QP as a whole can no longer move the way it did in (la) and (2a). The following sentences, based on those in (1) and (2), are ungrammatical: (4) a. *Die Studenten alle haben das Buch gelesen. the students all have the book read
b. *Gli studenti tutti hanno letto il the students all
libro.
(Germ.) (Italian)
have read the book
The significance of the data in (4) will be discussed in Section 2 .2. 1 1. Cardlnaletti & Giusti (2006) point out that the word order in (4b) is possible in languages like Portuguese, Romanian and Spanish and conclude that in these languages tt is possible for an entire QP to move even after a DP has moved to [SPEC, QP]. Due to space limitations it is
What all happens when a universal quantlfier combines with an interrogative DP
Given the data in (1) and (2), it appears that universal quantifiers have the same selectional properties, semantics and syntax across the Germanic and Romance language families. They select a DP, they have universal semantics, and they have the syntactic property of being able to be stranded, due to their position as head of QP. This article deals with the question of what happens when a universal quantifier combines with an interrogative DP. The remainder of the article is divided into three sections. In Section 2, I show how universal quantifiers and wh-phrases interact in German and offer a tentative hypothesis. I then point out the weaknesses in that hypothesis and propose an alternative approach. Section 3 supplements the findings from German with data from other languages. In Section 4 I present the conclusions that can be drawn from the data and discussion in Sections 2 and 3.
2.
What all happens when universal quantifiers combine with wh -words in German
This section is organized as follows: Section 2.1 presents examples from German of universal quantifiers in combination with interrogative phrases and offers a tentative analysis and model based on the assumption that interrogative D Ps can be selected by a universal quantifier. Section 2.2 contains some sample derivations within the framework of this proposed model. Section 2.3 points out the weaknesses in the model and proposes an alternative.
2.1
Data and a possible analysis
We begin with the following sentences: (5) a.
Welche Studenten alle sind gekommen? which students
b.
come
Welche Studenten sind alle gekommen? which students
(6) a.
all are are
all come
Was alles liegt auf dem Tisch? what all
lies on the
table
not possible to debate this issue here, but a counter-suggestion is simply that in these languages the universal quantlfier has adjectival qualities that allow it to appear in normal post-nominal adjectival position. An English sentence such as The Stude11ts all passed the test might also seem to be evidence that an entire QP can move after a DP has moved to [SPEC, QP], however in English it is widely assumed that verbs do not move to AgrS/T /I. This would mean that in The Students all passed the test the subject DP The students is in the SPEC position of AgrSP /TP /IP and the quantlfier has been stranded in [SPEC, vP], as argued in Cirillo (2009).
43
44
Robert Cirillo
b. Was liegt alles auf dem Tisch? what lies all on the table In these sentences the universal quantifier has the function of turning the question into a request tor an enumeration In (5), one is asking, "Tell me the names of all the students who came:' In (6) one is asking, "Tell me all the things that are lying on the table." I will refer to the semantic feature that allows a universal quantifier to combine with an interrogative word in this way as the list readingfeature. As we will see, this feature is not borne by the universal quantifier in all languages, and even if it is present in a particular language, it might be applied arbitrarily and inconsistently in regional variants of the same language. This will be discussed in Section 3. For now, we will concentrate on German.2 The first question is how one would analyze (5) and (6). Since welche Studenten in (5) and was in (6) are DPs, and since we know that universal quantifiers select DPs and can be stranded by them, it seems logical to assume that the structure in (3) applies to (5) and (6) as well as to (1) and (2). Therefore, I will tentatively propose the diagrams in (7) and (8) as the base-structures for the subjects of the sentences in (5) and (6), respectively. In (7) we see a QP in which the universal quantifier alle has selected a full DP headed by the wh-determiner welche. In (8) we see a QP in which the universal quantifier, which bears the [s] ending corresponding to neuter singular morphology, has selected the bare interrogative DP was. (7)
QP
~
Spec
Q'
~
Q
aile (all)
DP
~ SPEC
D'
~
D
NP
welche ~ (which) SPEC N' I
N
Studenten
(students) 1. Thls listreadlngfeature'J$ not to be confused with list readings such as those discussed ln Chlerchla (1992 and 1993) and elsewhere. For example, a sentence such as Whom do all the roys love? can have a pair-list interpretation (John loves Mary, Hans lo'l'es Gudrun, and Yoshi loves Yruhiko) or a single pair/indi'l'idual interpretation (All the boys love Lisa). These interpretations are dependent upon whether the wh- word whom has wlde or narrow scope with respect to the DP all the boys. The list reading feature that I am referring to ls not related to scope in the same way. It ls an extension of the semantics ofthe universal quantifier and implies that the speaker is asking for an emuneratioiL
What all happens when a universal quantlfier combines with an interrogative DP (8)
QP
~
Q'
Spec
~
Q
alles (all)
DP
~ SPEC
D'
~
D
NP
was~ (which) SPEC
N'
I
N
0
Before moving on to the actual derivations of the sentences in (5) and (6), I will demonstrate why one must assume that welche in (7) is in the D position, I will point out an interesting feature of welche as a determiner, and I will explain why I place the pronoun was in Din (8). It is characteristic of determiners occupying Din German that they do not bear the weak ending [n] but rather the strong null ending. The sentences in (9) illustrate this. In (9a) and (9b) the demonstrative diese and the possessive pronoun meine do not bear the weak [n] ending.3 In (9c), however, the adjective must bear the weak ending, since it does not occupy D. Given that in (5) the interrogative word welche cannot bear the weak [n] ending, one can conclude that it is located in D. (9) a. All(e) diese!*diesen Studenten sind intelligent. all these students are intelligent
b. All(e) meine!*meinen Studenten sind intelligent. all my students are intelligent c. Alle guten!*gute Studenten si1td Melligent. all good students are intelligent It is important to point out that the determiner welche triggers weak [n] inflection on the adjectives that follow it and thus seems to behave more like a definite than an indefinite determiner/quantifier/numeral. This is seen in (10). The significance of these examples will be discussed shortly. (10) a. Alle!Die/Diese/Meine/Welche guten Studenten all the these my which good students
3· For analyses in which possessive pronouns in the Germanic languages and some Romance languages move to D, see Schoorlemmer (1998) and den Besten (2006).
45
46
Robert Cirillo b. Einige/Viele/Wenige/Mehrere/Funf gute Studenten some manyfew several five good students Regarding the base-position of was in (8), I follow the arguments in Abney (1987), Longobardi (1994) and Sleeman (1996) and assume that pronouns are base-generated in D. I also offer additional evidence that pronouns in German are base-generated in D. namely, the fact that they have the syntactic effect of the definite article. The sentences in (11) show that the weak ending, indicated by the [n] on Deutschen, is obligatory for adjectives and substantivized adjectives that follow the definite article. The sentences in (12) show a very interesting phenomenon. Example (12a), with the strong null ending on the adjectival noun Deutsche, represents what is considered to be 'Standard German' and what is taught in the schools. However, a very large percentage of German speakers, perhaps even a majority of them, prefer the weak ending shown in (12b). This shows that the pronoun wir has the effect of the definite article, implying that pronouns occupy D. (11)
a. *Die Deutsche essen keinen Gorgonzola.. the Germans eat no Gorgonzola b. Die Deu.tschen essen keinen Gorgonzola.. the Germans eat no Gorgonzola
(12) a.
Wir Deutsche essen keinen Gorgonzola. we Germans eat
b.
Gorgonzola
Wir Deutschen essen keinen Gorgonzola. we Germans eat
2.2
no no
Gorgonzola
Derivations
In deriving (5) and (6) from (7) and (8), I assume that in the Germanic languages a wh-DP moves to [SPEC, CP], following den Besten (1983) and others, and that whmovement is locaVcyclical. 4 This means that before moving to [SPEC, CP] the wh-DP must move to [SPEC, QP]. After the wh-DP has moved to [SPEC, QP], if we follow the Stranding Analysis of floating quantifiers, one of two things can happen. The entire QP can move to [SPEC, CP], producing (Sa) and (6a), or the wh-DP can move to [SPEC, CP] by itself, stranding the quantifier and producing (Sb) and (6b). Note that there is an important difference between the movement ofwh-DPs and non-wh-DPs. As shown in (4), once a non-wh DP has reached [SPEC, QP], it must continue on its own. The QP as a whole can no longer move. However, as shown in (Sa) and (6a), in the case of a wh-DP the entire QP can move even after the DP has reached [SPEC, QP], with the effect that the 4· Arguments that A-bar movement is subject to locality can be found in Chomsky (1977, 1995), McCloskey (1990), Torrego (1983), Rizzi (1990, 2006) and elsewhere.
What all happens when a universal quantlfier combines with an interrogative DP
wh-DP pied-pipes the quantifier. Giusti (1991) pointed out a similar pattern with pronouns. Like a wh-DP in QP, a pronoun in QP must also move to [SPEC, QP] in German and English, and once it does it can strand the quantifier, as expected, or the entire QP can move upward: (13) a.
Sie
aile haben gegessen.
they all have
eaten
b. Sie haben alle gegessen. they have all eaten c.
*Alle sie all
haben gegessen.
they have
eaten
To explain why a QP containing a wh-word or a pronoun may move in its entirety, with the effect that the quantifier is pied-piped as suggested in (Sa), (6a) and (13a), one could propose that the features that trigger the obligatory movement of pronouns and wh-phrases are projected through the entire QP. As pointed out by an anonymous reviewer, however, this would not explain the optionality of the movement of the entire QP. If a feature must be checked and it permeates an entire phrase, then the entire phrase should have to move. One would thus have to postulate that it is the upward projection of the features that is optional, which would be an ad hoc stipulation. Nonetheless, optionality does seem to be relevant. I therefore tentatively propose that the features that force a wh-phrase or a pronoun to move leftward create a kind of momentum that allows but does not force the entire QP to move, as seen in (5), (6) and (13). It is relevant that speakers generally prefer the (b) sentences in (5), (6) and (13). In other words, they prefer not to move the entire QP, which implies that economic principles are at work. In (4), in which there is no feature forcing the DP to move outofQP, principles of economy need not compete with the momentum created by obligatory movement, and pied-piping of the QP does not take place. Some readers might object to analyzing (6b) as a case of stranding, based on the following examples: (14) a.
Wer alles ist gekommen? who all
is come
b. Wer ist alles gekommen? who is all
come
In these sentences the universal quantifier has to be an adverbial adjunct because its neuter inflection does not agree with the non-neuter gender of the interrogative pronoun wer. One could therefore argue that what we see in (6a) and (6b) is the same as what we observe in (14a) and (14b). Whereas I will not deny that one could analyze (6) as being analogous to (14), I must point out that one cannot rule out the possibility that (6b) is derived from (8) and is an instance of stranding. Observe the following examples, in which a bare demonstrative pronoun that presumably occupies D and is
47
48
Robert Cirillo
thus very comparable to the interrogative word was can be selected by and strand the universal quantifier: (15)
a. Alles das ist nur meine personliche .Meinung. all that is only my personal opinion
b. Das ist alles nur meine personliche Meinung. that is all only my personal opinion The following examples from Giusti (1990b) are also relevant: (16) a. Mit wem allem hast du gesprochen? with whom all have you spoken b. Mit wem hast du allem gesprochen? with whom have you all spoken In the sentences in (16) the quantifier, whether floating or not, exhibits the singular, masculine dative morphology of the interrogative pronoun wem, just as the quantifier in both examples in (6) agrees in number, case and gender with the singular, nominative, neuter pronoun was. One can therefore not exclude the possibility that (6b) is derived from (8) and is an example of stranding. 2.3
Weaknesses in the model and a possible alternative
In Section 2.1 I proceeded from the similarity between (1) and (5) and proposed a unified analysis in which both sentences are instances of a universal quantifier that selects a DP and can be stranded. Let's call this proposal the Selector Approach. A unified analysis is of course always desirable, but this one has some serious weaknesses. One of these is the need for the ad hoc stipulation regarding the optionality of quantifier piedpiping by wh-D Ps but not by non-interrogative D Ps. This contrast is shown in (4a) and (Sa) above. This weakness is perhaps mitigated by independent motivation provided by the fact that a similar phenomenon occurs with pronouns, but, as I will argue shortly, there are reasons not to pursue a unified approach for pronouns and wh-words. A second weakness in the Selector Approach is that it has been well argued in Giusti (1991) and Cardinaletti & Giusti (2006) that universal quantifiers can select only definite or generic D Ps. I had mentioned that the data in (10) suggest that at least in German there might be some definiteness or specificity associated with a word like welche because it triggers weak morphology. However, the evidence in (10) is unconvincing for two reasons. First of all, there are two German quantifiers, solche and manche, that would seem to have indefinite semantics but which trigger weak [n] morphology on adjectives, meaning that the data in (10) are inconclusive: (17)
Manche!solche guten Studenten many
such
good students
What all happens when a universal quantlfier combines with an interrogative DP
Secondly, there is morphological evidence thatwh-words are not definite. In Hungarian, for example, interrogative words, including milyen "which': fail to trigger the verbal morphology required by definite direct objects: (18) a.
Ez a konyvet kered. this the book
b. Kit I whom
(you) want
Mit what
I Milyen konyvet which
book
kersz? (you )want
A third weakness in the Selector Approach, which 1present in Section 3, is that there are languages in which universal quantifiers and interrogatives do not co-occur at all, or, if they do co-occur, do so arbitrarily and not always with quantifier stranding. The fact that universal quantifiers select and can be stranded by definite and generic DPs in all the Romance and Germanic languages but do not universally show the same behavior with interrogative DPs casts doubt on the Selector Approach for interrogatives. Given the above three weaknesses in the Selector Approach (and a fourth one in English that will be presented in Section 3) an alternative analysis must be considered. If the universal quantifier in (5) and (6) is not the selector of the interrogative DP, then it can only be its complement This idea of base-generating the universal quantifier to the right rather than to the left of an interrogative D P is very similar to a proposal in Giusti ( 1990b ), who treats a universal quantifier that occurs with an interrogative as a "right-adjoined partitive constituent': Giusti was concerned mainly with occurrences of the German neuter alles exemplified in (14), but her idea can be extended to all occurrences of the universal quantifier with interrogatives, not only in German but in other languages, and it is consistent with her claim that a universal quantifier only selects definite and generic DPs. (In this article I will use the term complement rather than right adjunct.) This Complement Approach eliminates the need for the stipulation on optionality for QPs containing wh-DPs. As I will argue in Section 3, it also simplifies the syntax because it implies that inter- and intra-linguistic variation result from a lexical difference, namely, the presence or absence of the list reading feature. This will be laid out in detail in the next section. Before moving to Section 3, although this article is not about pronouns as such, I would like to discuss the possibility of applying the Complement Approach to pronouns as well as to wh-expressions, thereby explaining data such as (5) and (13) in a unified manner. There are two problems with this idea. First of all. unlike interrogatives, pronouns co-occur with universal quantifiers all across the Germanic and Romance language families. This suggests that there is no language-specific 'pronoun selection' feature comparable to the list reading feature. Secondly, in Italian and Spanish pronouns may but need not move to the left of their selecting quantifier. The examples in (19) would be difficult to account for by a complement approach and provide strong evidence in favor of an analysis in which pronouns are selected by universal quantifiers and then move leftward:
49
50
Robert Cirillo (19) a.
Loro tutti/Tutti loro hanno letto questo. they all
all
they have
(Italian)
read this
b. Ellos todos!Todos ellos han leido esto. they all all they have read this
(Spanish)
We can now look at relevant data in other languages.
3·
What all happens with universal quantifiers and interrogatives in English, Swedish and the Romance languages
In Section 2 we saw that while the Selector Approach works well in German mechanically, it has theoretical issues. The Complement Approach was thus proposed as analternative. We will now see that in other languages there is even more compelling evidence for the Complement Approach. The sentences in (20) and (21) are the American English equivalents of the German examples in (5) and (6). Whereas all four sentences in ( 5) and (6) are grammatical, American English only allows the equivalent of (6a ): (20) a. ..Which students all have come? b. *Which students have all come? (21) a. What all is lying on the table? b. *What is all lying on the table? The following sentences are also relevant: (22) a. Where all have you been? b. *Where have you all been? c. Who all has arrived? d. *Who has all arrive? Note that speakers of British English find all of the sentences in (20) to (22) ungrammatical, even (21a), (22a) and (22c). This fact, combined with the arbitrariness with which the universal quantifier can co-occur with a wh-word in American English and the lack of stranding, strongly suggests that wh-words are not selected by the universal quantifier in American English. There is additional evidence that the Selector Approach is not the correct explanation for sentences like (21a) and (22c). Keep in mind that under the Selector Approach the wh-word in these two sentences is the complement of the quantifier and has piedpiped it to subject position after moving through [SPEC, QP]. A wh-word can also be the complement of a preposition, and the preposition can be stranded or pied-piped. One would expect to see similarity between a PP and a QP if a wh-word is present, but instead we see two major differences. First, when a preposition is pied piped with a wh-word that is its complement. the wh-word has not moved to [SPEC, PP]:
What all happens when a universal quantlfier combines with an interrogative DP
(23) a. *Whom with did he go to the party? b. With whom did he go to the party? Secondly, prepositions, unlike quantifiers, do not have to be pied-piped: (24) Whom did he go to the party with? The implication is that if the wh-word in (21a) and (22c) were really the complement of the quantifier and if pied-piping were really involved, we would see a pattern more like that in (23b) and (24). Given the contrary, the Selector Approach for interrogatives becomes still more doubtful. Turning now to Swedish, we see that it seems to exhibit a mixed pattern. All four of my native informants gave me the following judgments: (25) a. *Vilka studenter alla har last boken? which students all have read book-the b. lVilka studenter har alla liist boken? which students have all read book-the c.
"Vad allt ha.r hiint
efter det?
what all has happened after that
d. *f? Va.d ha.r allt hiint efter det? what has all happened after that Swedish, if it allows the combination of a universal quantifier and an interrogative expression at all, insists that the quantifier be stranded. This makes Swedish somewhat analogous to German, which prefers but does not demand stranding. Given the marginality of (2Sb) and (2Sd), however, one could also say that Swedish is approaching British English in its disfavoring of universal quantifiers in combination with interrogatives. For the sake of completeness, note that in the Romance languages the universal quantifier never appears with an interrogative: (26) a. *Quali studenti tutti sono venuti? which students all are come
(Italian)
b. *Quali studenti sono tutti venuti? which students are all come c. *Che tutto ha.i fatto? what all have-you done
fa.tto tutto? d. *Che hai what have-you done all The data in this section show that the list reading feature on universal quantifiers is not universal. British English and the Romance languages do not have it, Swedish seems to be evolving away from it, and American English applies it arbitrarily and inconsistently, allowing a.ll to occur only with singular interrogative DPs. In American English there
51
52
Robert Cirillo
is also the syntactic question of why stranding is not allowed in the case of wh-DPs but freely allowed with non-wh-DPs selected by all. The evidence strongly suggests that universal quantifiers do not select wh-DPs but can occur as their complement/restrictor as long as they bear the semantic (lexical) list reading feature. This is consistent with the arguments in Giusti (1991) and Cardinaletti & Giusti (2006) regarding the selectional properties of universal quantifiers, and it is consistent with the proposal in Giusti (1990b) that in a German phrase such as was alles (what all) the universal quantifier is base-generated to the right of the interrogative DP. It is also consistent with Chomsky & Lasnik (1993) and Baker (2008), who claim that variation between closely related languages or regional varieties of a language is due mainly to lexical differences. The presence or absence of the list reading feature would be an example of a lexical difference. Having presented my data, I can now tie together some loose ends and reformulate an analysis for each language, with the goal of presenting a unified approach in which universal quantifiers do not select interrogatives. In British English and the Romance languages, the universal quantifier does not bear the list reading feature, which means that it cannot combine with interrogative words. American English does possess the list reading teature, but the arbitrariness that it shows in applying it, combined with the lack of stranding/floating of the quantifier, suggests that interrogative expressions involving the universal quantifier in American English are formed as indivisible phrases in the lexicon. The following examples illustrate the arbitrariness in the formation of interrogatives with all in American English: (27) a. Whom all have you talked to? b. What all have you done today? c. Where all have you been this year? d. *Why all do people not like George? e. *How all could we solve this problem? The fact that all can co-occur with the adverb where but not with why and how, combined with the lack of stranding illustrated in (20) to (22), strongly reinforces the idea that the formation of this type of expression is lexical rather than syntactic in American English and that the combinations in (27) are comparable to lexicalized, 'frozen' expressions. German also possesses the list reading feature, but, unlike American English, it is very productive in its combination of the universal quantifier with any sort of interrogative expression and with quantifier stranding, as seen in (5) and (6). In other words, German does not exhibit the restrictions and arbitrariness seen in (20) and (21). This suggests that German interrogative expressions such as those in (5) and (6) are not 'frozen phrases' formed in the lexicon like their American English equivalents, but are normal DPs formed in the syntax. The universal quantifier is thus comparable to other kinds of complements that can modify or restrict an interrogative phrase. This
What all happens when a universal quantlfier combines with an interrogative DP
explains why the universal quantifier can be stranded in a manner typical of split DPs in German questions:5 (28)
a.
Was fur ein Buch hast du gekauft? what for a
book have you bought
b. Was hast du for ein Buch gekauft? what have you for a book bought (29)
a.
Was alles hast du gekauft? what all
b.
have you bought
Was hast du alles gekauft? what have you all
bought
Regarding Swedish, given the tact that the universal quantifier can combine, albeit marginally, with both full and bare interrogative DPs, and given that interrogative phrases are split, as seen in (25b) and (25d), it would seem that Swedish follows the German pattern and merges the quantifier and the interrogative word or phrase in the syntax. The marginality of (2Sb) and (2Sd) would be due to the fact that Swedish is losing the list reading feature, becoming more like British English and the Romance languages. The mandatory splitting of the interrogative DP could well be due to the economic principle of moving as little 'weighf as possible. This can be related to German speakers' general preterence for splitting interrogative DPs, as shown by the fact that the (b) sentences in (5), (6) and (13) are preferred to the (a) sentences. It is not clear why German and Swedish allow the syntactic merging of a full or bare interrogative DP with a universal quantifier while American English handles this in an arbitrary way in the lexicon, but we must accept this difference between the languages if we do not want to claim that in German and Swedish the universal quantifier can select interrogative DPs.
4·
Summary and conclusions
German shows interesting symmetry between non-interrogative and interrogative D Ps that co-occur with a universal quantifier, especially if analyzed within the framework of the Stranding Analysis of floating quantifiers. This was a compelling reason to attempt a unified analysis of phrases such as alle die Studenten Kall the students" and welche Studenten alle ~hich students all". Whereas the mechanics of this unified approach work well in German, there are four problems with it that are both theoretical and empirical in nature. First of all, it is in conflict with arguments in the literature that the universal quantifier only selects definite and generic DPs. Secondly, it requires an
5·
The pattern illustrated in (28) and (29) is also noted and discussed in Giusti (1990b).
53
54
Robert Cirillo
ad hoc stipulation regarding the optional movement to [SPEC, CP] of an entire QP that contains an interrogative. Whereas this stipulation is somewhat independently motivated by the fact that pronouns behave similarly to wh-words when they combine with a universal quantifier, there are also reasons for not applying a uniform analysis to pronouns and interrogatives, as 1 argued at the end of Section 2.3. Further research is required in order to determine exactly what is happening with pronouns in (13 ). The third weakness in the unified approach is that there are languages in which either universal quantifiers and interrogatives do not co-occur at all, or they occur in an inconsistent, arbitrary manner and sometimes without stranding. The tourth weakness is that in English there are discrepancies between a wh-word in a PP and a whword co-occurring with a universal quantifier. These discrepancies, which concern movement to SPEC position, stranding and pied-piping, are evidence that the whword in an expres sian such as what all is not the complement of the quantifier. In view of the evidence against the Selector Approach, an alternative analysis is desirable. The best way to analyze the co-occurrence of universal quantifiers and interrogatives is to apply an idea first presented in Giusti (1990b) whereby the universal quantifier can be base-generated to the right of a wh-D P. I suggested that this depends upon whether in a particular language the universal quantifier bears what I call the list reading feature. Because the combination of interrogatives and the universal quantifier in German is a very unrestricted, productive process, I suggested that it occurs through syntactic merging and that the stranding of the quantifier can be explained as a typical example of the splitting of an interrogative DP in German. In American English, on the other hand, given the arbitrariness seen in (27) and the lack of stranding, I concluded that interrogative expressions involving the universal quantifier are created in the lexicon. Swedish seems to be like German in that it merges the universal quantifier and interrogatives in the syntax, but it is stricter than German in requiring rather than preferring that the interrogative DP be split It seems to be losing the list reading feature, becoming more like British English. What would be worth investigating further is the question of why English, as opposed to German and Swedish, combines universal quantifiers and interrogatives in the lexicon rather than allowing them to be merged and split in the syntax. ln summary, the Complement Approach that I have proposed enables one to explain the inter- and intra-linguistic variation that one finds in the co-occurrence of universal quantifiers and interrogative expressions.
References Abney, Steven Paul. 1987. The English Noun Phrase in its Sentential Aspect. PhD dissertation, MI't Baker, Mark. 2008. The Macroparameter in a Microparametric World In The Limits of Syntactic Variation [Lingulstlk AktuellJLinguisics Today 132], Theresa Biberauer (ed.), 351-374. Amsterdam: John Benjamins.
What all happens when a universal quantlfier combines with an interrogative DP den Besten, Hans. 1983. 0 n the interaction of root transformations and lexical deletive rules. In On the Formal Syntax of Westgerma11la [Linguistlk Aktuell/Linguistics Today 3], Werner Abraham (ed.), 47-131. Amsterdam: John Benjamins. den Besten, Hans. 2006. The origins of the Afrikaans pre-nominal possessive system(s). In StudIes in Co11tact Lb1guistics: Essays in Honor of Glenn G. Gilbert, Linda L. Thornburg & Janet M. Fuller (eds), 103-124. Bern: Peter Lang. Cardinalettl, Anna & Giusti, Giullana. 2006. The syntax of quantlfied phrases and quantified ditics. In The Blackwell Companion to Syntax 15, Martin Everaert & Henk van Riemsdijk (eds), 23-73. London: Blackwell Chierch!a, Gennaro. 1992. Functional WH and weak crossover. Proceedi11gs of the Tenth We.st Coast Conference on Formal Linguistics. Stanford CA: Stanford Linguistics Association. Chierch!a, Gennaro. 1993. Questions with quantifiers. Natural Language Semantics 1: 181-234. Chomsky,. Noam. 1977. On Wh-movement. InFormal Syntax, Peter William Culicover, Thomas Wasow & Adrian Akmajlan (eds), 78-132. New York NY: Academic Press. Chomsky,. Noam. 1995. The Minimalist Program. Cambridge MA: The MIT Press. Chomsky,. Noam & Lasnlk, Howard. 1993. The theory of principles and parameters. In Syntax: A11 Inter11ational Ha11dbook ofContemporary Re.search, Joachim Jacobs, Arnim von Stechow, Wolfgang Sternefeld & Theo Vennemann (eds), 506-569. Berlin: Mouton de Gruyter. Cirillo, Robert. 2009. The Syntax ofFloating Quantifiers: Stranding Revisited. Utrecht: Landelijke Onderzoekschool Taalwetenschap. Giusti, Giullana. 1990a. Floating quantifiers, scrambling and configurattonality. Linguistic Inquiry 21: 633-641. Giusti, Giullana. 1990b. The syntax of floating 'alles' in German. In Issues In Germanic Syntax, Werner Abraham, Wim Kosmeijer & Eric Reuland (eds), 327-350. Berlin: Mouton de Gruyter. Giusti, Giuli.ana 1991. The categorial status of quantlfied nominals. Linguistische Berichte 136: 438-452. Longobardi, Giuseppe. 1994. Reference and proper names: A theory ofN-movement in Syntax and Logical Form. Linguistic Inquiry 25(4): 609-665. McCloskey, James. 1990. Resumptive pronouns, A:-binding, and levels of representation in Irish. In Syntax and Semantics 23: The Synta.r: of the Modern Celtic Languages, Randall Hendricks (ed.), 191-248. New York NY: Academic Press. Rizzi, Luigi. 1990. RelaNvizedMinlmallty. Cambridge MS: The MIT Press. Rizzi, Luigi. 2006. On the form of chains: Criteria! positions and ECP effects. In WH-Movement: Moving on, Lisa Lai-Shen Chen & Norbert Corver (eds), 97-133. Cambridge MA: The MIT Press. Schoorlemmer, Maaike. 1998. Possess tves, articles and definiteness. In Possessors, Predicates and Movement in the Determiner Phrase [Lingutstlk Aktuell/Linguisttcs Today 22], Artemis Alextadou & Chris Wilder (eds), 55-86. Amsterdam: John Benjamins. Shlonsky, Ur. 1991. Quantlfiers as functional heads: A study of quantifier fioat in Hebrew. Lingua 84: 159-180. Sleeman, Petra. 1996. Licensing Empty Nouns in French. PhD dissertation, University of Amsterdam. (The Hague: Holland Academic Graphics). Sportiche, Dominique. 1988. A theory of floating quantifiers and its corollaries for constituent structure. Linguistic Inquiry 19: 425-449. Torrego, Esthe.t 1983. More effects of successive cyclic movement. Linguistic b1quiry 14: 561-565.
55
Micro-diversity in Dutch interrogative DPs* A case study on the (dis)continuous wat voor 'n N-construction Norbert Corver & Marjo Van Koppen UiL-OTS, Utrecht University
This paper presents a micro-comparative perspective on the Dutch wat voor 'n N-construction (meaning: "what kind ofN"). Besides a description of the various formal manifestations of this nominal construction as found intradialectally and cross-dialectally, an analysis is given of the dimensions of variation at the level of internal syntax and external syntax. As regards the external syntax, an analysis will be given of patterns that permit a split wat voo r 'n N-pattem and those that do not. An important outcome of our analysis will be that a uniform structural basis 'underlies' the different manifestations of the wat ~·oor 'n N-construction: more specifically, predication, configurationally defined as a DP-internal small clause structure, and predicate displacement.
1.
Micro-variability within the Dutch wat voor 'n N-construction
The syntax of the Dutch wat voor 'n N-construction, as exemplified in ( la), has been an important topic on the Dutch generative syntactic research agenda ever since Hans den Besten's discussion of this nominal expression in the 1980s (see Den Besten 1981, 1985 ). The research reported on in this article is part of a larger research project entitled Diversity in Dutch DP Design (DiDDD), which is financially supported by The Netherlands Organization for Scientific Research (NWO). We are grateful to the informants of the DiDDD-project for providing us with the relevant data of their Dutch dialect. The content of this article was presented at the 2009 conference Variatilm and Change in the Romance and Germanic Nou11 Phrase, which took place at the University of Amsterdam. We thank the audience for their comments and remarks, and we thank the three anonymous reviewers and the editors of this volume for their useful comments. The contents of this article were also presented at a 2008 UCLA-minicourse entitled 'The study of syntactic microvariation: perspectives and tools: which was sponsored by a joint collaborative grant from the Universities of California and the University of Utrecht We would like to thank the audience, and especially Hilda Koopman and Ed Stabler, for very useful and stimulating discussion.
58
Norbert Corver & Marjo Van Koppen
Its most striking property is, arguably, the possibility of subextracting the wh-word wat out of the nominal expression, which yields a discontinuous pattern like ( lb ). (1) a.
[Wat voor 'n auto] heb je
gekocht?
whatfor a car have you bought 'What kind of car did you buy?" b.
Wati heb je
[ti voor 'n auto] gekocht?
what have you for a car bought 'What kind of car did you buy?" Also its internal syntactic structure has triggered much discussion. It raises questions such as: What is the internal constituency of the nominal expression, and what is the nature of the words voor and 'n? In the present paper, we hope to contribute to our understanding of this nominal construction by investigating its internal and external syntax from a micro-comparative (i.e. cross-dialectal) perspective. 1 As will become clear in the course of this paper, the various manifestations of this nominal construction type, as attested both intradialectally and interdialectally, display interesting dimensions of variation, both regarding their internal syntax and their external syntax. For example, it turns out that certain variants of the wat voor 'n N-construction do not permit the split pattern. That is, the wh-word cannot be removed from within the nominal expression. The question obviously arises as to how this asymmetry in subextraction behavior can be accounted for. Ever since Den Besten's seminal discussion of this nominal construction type, a variety of syntactic analyses of it has been given in the generative literature. 2 In this paper, we will take the analysis as given in Bennis et al. (1998) as our theoretical basis. There are two important ingredients in this analysis: firstly, there is a DP-internal predication relationship between the noun auto and the wh-word wat in (1). The former is the external argument (i.e. 'the subject') of the predication relation, and the latter is the (nominal) predicate. Secondly, the surface order is derived by means of DP-internal predicate displacement That is, the predicate wat undergoes movement to a position which precedes its subject. In this paper, we will try to show in what ways the various surface manifestations of the Dutch wat voor 'n N-construction relate to differences in the "formal implementation" of the predication relation and the predicate displacement process. This paper is organized as follows: in Section 2 we provide the reader with some backgroundinformationaboutthephenomenonofDP-internalpredicatedisplacement
t.
We will use the orthographic convention 'n to represent the indefinite article segment of the
wat voor 'n N-construction. In the literature, one also finds the orthographically non-reduced form ee11 ("a"), as in wat voor een auto (what for a car, "what kind of car"). Importantly, the two orthographic forms have the same pronunciation: f-;,nf. 2. See, for example, Bennis (1983), Corver (1990, 1991), Pafel (1996), Leu (2008a,b). For reasons of space, we abstract away from a discussion ot;. and comparison with, analyses different from the one presented here, which is based on Bennis et al (1998).
Micro-diversity in Dutch interrogative DPs
Following Bennis et al. (1998), we take this computational operation to be part of the internal syntax of the wat voor 'n N-construction. In Section 3, we present a description of the various manifestations of the wat voor 'n N construction. Furthermore, we will give a formal analysis of the internal syntax of each variant. In Section 4, we examine the subextraction behavior of the different variants of the wat voor 'n N-construction: that is, which variants permit a discontinuous pattern and which do not? Section 5 concludes the article.
2.
Predicate displacement within the wat voor •, N-construction
In this section we will outline the structural analysis of the wat voor 'n N-construction in Standard Dutch as proposed by Bennis et al. (1998 ). We will take this analysis as the starting point for our investigation of the micro-diversity attested within this nominal construction across different dialects of Dutch. As will become clear in this section, Bennis et al make two major claims regarding the internal syntax of the wat voor 'n N-construction: first of all this nominal construction involves a predication relationship between the pronoun wa.t ('the predicate') and the noun ('the subject'), which is configurationally represented as a DP-internal small clause representation XP. Secondly, the surface position of wat in the left periphery of the noun phrase is a derived position; that is, DP-internal (predicate) displacement moves wa.t from the small clause predicate position to Spec,DP. In what follows, we will first place these claims about DP-internal predication and predicate displacement in a somewhat broader context, so that the reader can see that the empirical basis for these claims is not restricted to the wat voor 'n N-construction. We will start our discussion with the nominal construction which perhaps shows the phenomenon of DP-internal predication and predicate displacement most clearly, namely the so-called N ofN-construction.
2.1
Predicate Inversion and the spurious indefinite article
In recent generative studies, a number of nominal construction types have been (re) analyzed in terms of predicate displacement, most notably the so-called N of/van Nconstruction (cf. (2)-(3)). (2) die idioot van 'n dokter (3) that idiot of a doctor
(Dutch) (English)
Den Dikken (1995, 1998, 2006), for example, proposes that in constructions like (2)(3) the displaced predicate originates in a DP-internal small clause configuration, which is represented as XP in ( 5); see also Bennis etal. (1998) .3 ThisXP is asymmetrically 3· See Kayne (1994) for an analysis in terms of predicate displacement, in which the inverted predicate originates in a predicate position of a clausal IP.
59
6o
Norbert Corver & Marjo Van Koppen constituted such that the external argument of the predicate, located in the complement position of the small clause head X, occupies the specifier position of XP, which following Den Dikken (2006) may also be referred to as R(elator)P(hrase); see (4). Predicate displacement involves movement of the predicate (i.e., the complement of the small clause head) to the specifier position of a higher functional head FP, as depicted in (5):4 (4) [DP die [pp Spec [F F lxP dokter [X' lx 'n] idioot]]]]] (5) [0 p die [pp idiootj [F F (= va.n) +Xi(= 'n) fxp dokter [X' ti t}JJJJ
As indicated in (4)-( 5), we will take X to be the indefinite article 'n. As noted in Bennis et al. (1998), 'n is spurious in this nominal environment in the sense that it does not seem to 'belong to' the noun that follows it, nor in fact to the noun that precedes it Normally, the indefinite article is compatible with singular NPs only (see (6a,b)). Furthermore, it does not co-occur with proper names and mass nouns (see (6c,d)). (6) a. Ik heb 'n boek gelezen. I have a book read b. *Ik heb 'n boeken gelezen. I have a books read c.
*Ik heb 'n Westertoren gezien. I have a Westertoren seen
d. ''1k heb 'n spinazie gegeten. I have a spinach eaten As illustrated in (7a), the second noun of the N van 'n N construction can be plural. Furthermore, 'n can precede proper names (7b) and mass nouns (7c). That 'n does not belong to the preceding noun either (i.e. the displaced predicate) is shown by the existence of examples like (7d), in which the first noun (and also the second one) is plural (data drawn from Bennis et al. 1998). (7) a.
1die
ramp
van 'n getalscongruentiefeiten
that disaster of
a number agreement facts
b. die pracht van 'n Westertoren that beauty of
a Westertoren
'n pracht van 'n spinazie a beauty of
a spinach
d. die schatten van 'n kinderen those darlings of a children
4·
In Den Dlkken (2006), FP is characterized as a L(inker)P(hrase).
Micro-diversity in Dutch interrogative DPs
Consider, finally, the prepositional element van in (7). With Den Dikken (1995, 2006), we will assume that van in the N van 'n N construction is a nominal copula, which surfaces at PF in the functional head position F. In fact, this nominal copula is considered to be the nominal equivalent of the verbal copula (English to be). As shown in the English example (Sa), the infinitival copula can be freely omitted in copular sentences with a straight subject-predicate order. It cannot be left out, however, in the Predicate Inversion counterpart of (8a), given in (8b) (see Mora 1988 for discussion): (8) a. I consider John (to be) the best candidate. b. I consider the best candidate (*to be) John. Predicate displacement in the wat voor 'n N-construction
2.2
Let us next consider the derivation of the interrogative noun phrase wat voor 'n
boek(en) in (9). (9) Wat voor 'n boek(en) heb jij gelezen? what for a book(s) have you read "What kind ofbook(s) did you read?"
Following Bennis et al. (1998), we will assume that this nominal pattern also has a 'small clause basi, as in (10). Thus, wat is a predicate nominal in the complement position of a small clause head X. which is lexicalized by the spurious indefinite article 'n. 5 The noun boek(en) functions as the external argument and occupies [Spec,XP]. (10)
lxP boeken lx 'n [wat]]]
The spurious status of the indefinite article 'n is suggested by the fact that it can precede the plural noun boeken in (9). Further support for its spurious status comes from the fact that it can precede a quantifying noun like iemand "someone': as in (lla). As shown in (11b ), iemand normally cannot be preceded by an indefinite article. 6 5· M shown in (i), the lexical item wat also occurs as a nominal predicate in clausal copular constructions:
(i)
a
b.
Dat is (me) wat. that is (me) what "I think that's quite something." Wat Is dat? what ls that "Wh.afs that?"
M opposed to Leu (2008a,b) and van Riemsdijk (2005), we will not assume that the indefinite article 'n precedes a sUent noun soRT /KIND. Given the possibUity of having a lexical noun soort"sort/kind" in the wat voor 'n N construction. as in watvoor 'n soort boeke11 (what for a sort books, "what kind ofbooks"), it is, of course. tempting to say that there is a sUent noun soRT/ KIND present in wat voor '11 boeken, which then has the representation: wat voor 'n sooRT boeken. 6.
61
62
Norbert Corver & Marjo Van Koppen (11)
a.
Wat voor 'n iemand heb jij
ontmoet?
what for a someone have you met "What kind of person did you meet?" b. Jan heeft gisteren (*'n) iemand ontmoet. Jan has yesterday (a) someone met "Jan met someone yesterday:' The derived structure of wat voor 'n boeken is given in (12):7
Importantly, however, we also find patterns like the following, in which the article 'n precedes the plural noun soorten (see also Corver 1990) (i)
Wat voor 'n soorten vtrussen zijn er? what for a sorts viruses are there "What kinds of viruses are there?"
An analysis in which wat voor '11 soorten virussen receives the analysis wat voor '11 sooRT soorten virussen is not so obvious semantically. Its 'non-silent' counterpart, wat voor een soort soorten virussen, seems infelicitous.
7· The question may be raised as to whether there is any independent support for the application ofDP-internal displacement of wat. It is quite hard to use islandhood effects (e.g. the complex: NP-constraint, the wh -island constraint, et cetera) for showing that movement is involved in the derivation of the wat voor 'n N-construction. A potentially interesting contrast is the one given in (i) and (ii): (i)
a.
wat voor iemand/iets
heb jlj gezie11?
what for someone/something have you seen "You saw someone who is like what (Le. what kind of person)?" "You saw something which is like what (Le. what kind of thing)?" b.
*wat voor niemandlniets
heb }ij gezien?
what for no one/nothing have you seen "You saw no one who is llke what?" "You saw nothing which is like what?" (ii) a
wat voor 'n dieren/wat
voor twee dleren
heb }ij gezien?
what for a animals/what for two animals have you seen "What kind of (two) animals did you see?" b.
*wat voor geen dieren
heb ji} gezien?
what for no animals have you seen "You saw no animals which were like what!" The observation here is that wat cannot co-occur with a negative nominal expression within the wat voor N-construction. IfDP-internal movement of watto [Spec,DP] takes place, as in (12), the wh-word crosses the negative nominal expression (11iemand, 11iets, gee11 dieren), yielding a DP-structure in which the displaced wh-word is separated from the trace in the base position (Le. complement of X) by an intervening negative word in [Spec,FP]. As is well-known from clausal constructions, a non-argwnental wh-word cannot cross a negative expression (see Ross's Inner Island effects (1983); see also Corver 1990 and Honcoop 1998 for this observation in
Micro-diversity in Dutch interrogative DPs
(12) [DP wati [D' voorf+WHJ
[pp
tj [p lx 'n]i + F lxP boeken lx ti tj]]]]]]
According to this representation, the nominal predicate wat undergoes Predicate Inversion and moves into [Spec,FP]. Furthermore, the spurious article 'n raises and adjoins to the functional head F. Then there is a final movement step, called Predicate Fronting by Bennis et al., which places the inverted nominal predicate wat into [Spec,DP].8 Observe that the D-head is occupied by the preposition-like element voor. With Bennis et al. (1998), we will assume that this element is the lexicalization of a [+WH] operator D-head. It is this prepositional D-head which defines interrogative force on the nominal expression, and it is the Spec-position of this D-head from where the DP-intemally displaced predicate watcan leave the nominal construction, yielding a discontinuous pattern like (lb ).9. 1o Recall from our discussion of theN of/van N-construction in Section 2.1 that van surfaces as a nominal copula in contexts of DP-internal Predicate Inversion, i.e., when relation to the wat voor-split phenomenon). Thus, one might interpret the ill-formedness of (ib) and (lib) as some sort ofDP-internal inner island effect. (iv) a
b.
*Wat1 heeft niemand [t1 1•oor 'n boeken] gekocht? what has noone for a books bought "What kind of books did no one buy?" [Wat voor 'n boeken] 1 heeft niemand t 1 gekocht?
One might, of course, object that negative expressions like nlets cannot be qualified for kind, and consequently cannot be questioned for this. M shown in (v), however, we do find nominal expressions containing a negative word and a qualifying element: (v) Ik heb [lets dergelljks]l[niets dergelljks] gezien I have something such-llke-s/nothing such-like-s seen "I saw something like that/nothing like that." 8. In Bennis et al. (1998), the operation of Predicate Inversion is taken to be anA-movement type operation. The operation of Predicate Fronting is analyzed as an A-bar movement operation. In what follows, we will use the term 'predicate displacement' when we abstract away from the exact type of predicate movement that is involved 9· Bennis et al. (1998) note that, in many southern varieties of Dutch and also in substandard Dutch, voor is used as the infinitival complementizer in constructions that feature operatormovement to [Spec,CP], as, for example, in the infinitival relative clause in (i). (i)
['n boek lcp OP1 [ c voor [PRO [pp t1 in] te kljken]]]] a book for into to look "a book to look into"
See Postma (1995), Bennis (1995), Bennis et al. (1998) for discussion of the structural meaning of the wh-word wat; i.e., the idea that the pronoun wat receives part of its meaning derivationally by moving to a specific syntactic position in the syntactic structure - for example, a Spec-position of a functional head In (ia), for example, the pronoun wat has an indefinite meaning (l.e., "something") when it occupies its base position. In (ib ), wat receives its interrogative meaning by occupying the Spec-position of a functional category C, which is specified for interrogativity (represented here as [+WH]).
10.
63
64
Norbert Corver & Marjo Van Koppen
the nominal predicate undergoes movement to [Spec.FP]. If the derivation of wat voor 'n boeken in (10) also involves A-movement of wat to [Spec,FP], we expect the appearance of the nominal copula va.n, which yields the pattern in (18). Schematically: (13) [0 p wati [0 , voor[+WHJ [pp tj [p F (= *van) + 'n; [XP boeken
lx t; tiJJJJJ
However, presence of van yields an ill-formed structure: *wat voor van 'n boeken. Den Dikk.en, who also observes the impossibility of this pattern, states the following: 11 The suggestion that presents itself is that the presence of an overt meaningless functional element ( voor) under D causes the emptiness of the LINKER in a context in which it would otherwise be obligatorily overt. This can be thought of as a kind of "non-proliferation treaty"- a desire to keep the amount of meaningless material to a minimum. (Den Dikken 2006: 226)
As an alternative account of the non-co-occurrence of voor and 1'an, one might propose that the two elements compete for the same structural position. Suppose, for example, that the nominal copula van (i.e. the F- head) raises to D when the specifier position of D becomes occupied by the wh-element wat. Compare, in this respect, T-to-C movement of a finite auxiliary or the dummy verb do in English wh-interrogative main clauses (Who will/ do you invite!). IfF(= van) has to raise to D when a wh-elementraises to [Spec,DP], then the complementary distribution of voor and van is directly accounted for: they compete for the same structural slot, viz. D. In what follows, we will show that certain dialects permit the pattern wat11a.n 'n boeken (what of a books; "what kind of books"). We will argue that in this variant of the wat 1'00r N construction, the nominal copula van raises to D. Having provided the reader with some background on the phenomenon of DPinternal predicate displacement, we will now turn to a discussion of the cross-dialectal variation attested for the wat voor 'n N construction.
(i)
a
Ik denk fcp dat Jan
wat gelezen heeft]
I think that Jan has what read "I think that John read something." (wat =indefinite pronominal reading) b.
has
me af[ep wa~ fcC[+-whJ [IP Jan ti gelezen heeft]]]? I wonder me PRT wat John read has (wat =interrogative pronominal reading)
Ik vraag
Wat acts as an (interrogative) operator in (ib), since it binds a variable (Le. the 'wh-trace' left behind after wh-movement). Analogously to (ib), we may assume that the wh-word wat in (13) acts as an operator-like element, since it binds a wh-trace (variable) which is contained within the nominal expression. More specifically, the wh-word picks up its interrogative meaning in [Spec,DP] under spec-head agreement with the [+WH] prepositional determiner voor. It then moves to [Spec,CP], where it takes scope at the clausal level, yielding the reading: For what~ [x,: thing/property] John read [Xj book].
n. The notion 'linker' corresponds to F in our representation.
Micro-diversity in Dutch interrogative DPs
3· Dimensions of diversity within the wat voor 'n N-noun phrase In this section we will present a descriptive overview and an analysis of the variants of the wat voor 'n N-construction as attested across dialects of Dutch. The dialectal data presented are collected as part of the DiDDD-project (Diversity in Dutch DP Design), which is executed at the University ofUtrecht (see Corver et al. 2007). For this project, the nominal system ofS 3 dialects, evenly distributed over the Nether lands and Flanders (the Dutch speaking part of Belgium), has been investigated in depth.l2 From a descriptive point of view, a first global distinction within the class of wat voor 'n N patterns can be made along the following dimensions: (a) presence of the nominal element soort "sort"; (b) presence of a 'doubling' pro-form, e.g. zulk "such"; (c) 'bare form' (i.e. absence of soort and zulk). As will become clear in the course of this article, further distinctions can be made within this descriptive tripartition on the basis of formal characteristics such as (d) presence versus absence of preposition-like elements; (e) presence versus absence of the indefinite article 'n. In (14), an example from each class of the descriptive tripartition is given. (14) a.
wat voor 'n soort boeken what for
b. wat voor zulke boeken what for such books c.
wa.t voor 'n boeken what for
(the soort-pattem)
a sort books (the doubling pattern) ('bare' pattern)
a books
In what follows, we will discuss each of these descriptive patterns. For each pattern, we will first give a descriptive overview of the variants of the pattern as attested in the DiDDD database, and subsequently provide an analysis of these variants. We will start our discussion with the bare pattern in (14c). 3.1
The 'bare' pattern: Wat voor 'n boeken
Starting with the 'bare pattern - i.e. the pattern in which neither soort nor a doubling pronoun like zulke is present - we can distinguish the following variants:
wat voor 'n boeken b. wat voor boeken (16) a. waffer 'n boeken b. waffer boeken (15) a.
what for a books what for books what+ for a books what+ for books
See Vangsnes (2008) for an in-depth micro-variation (i.e. dialectal) study of the watvoor '11 N-construction in the Scandinavian languages and Leu (2007, 2008a,b) for a comparative study of the Germanic languages in general, including variants of German (e.g., Swiss German). See also Van Riemsdijk (2005).
12.
65
66
Norbert Corver & Marjo Van Koppen (17) a. wat van 'n boeken b. wat 1'an boeken
what of a books what ofbooks
(18) a. wat 'n boeken b. wat boeken
what a books what books
The a-examples differ from the b-examples in the presence versus absence of the spurious indefinite article 'n. We will assume that the two variants have the same structure and only differ in the lexical (i.e. phonological) contents of the spurious indefinite article: 'n (i.e./n/) versus 121 (i.e.//; absence of sound). Schematically, represented here for (15): (19) [DP wati [D. voorf+WHJ [pp tj [F" [F~ +Xi(= 'nle)] fxP boeken [X' ti tj]JJJ] The pair in (16) differs from the one in (15) as regards the phonological strength of the preposition-like determiner: In (15) we have the phonologically strong form voor whereas in (16) we have the weak form fer. We will assume that the two forms occupy the same syntactic position, viz. D, and that the weak form, as opposed to the strong form, phonologically cliticizes onto the left-adjacent wh-word, yielding waffer.U (20) [DP wat1 [D.voor!fer[+WHJ [pp tj [F"[Fjl + x; (= 'nle)] lxphoeken fx, ti t)JJJJJ Consider next the pair in (17), where we find the preposition-like element van instead of voor. 14 The question, obviously, arises as to what kind of element van is. Two options are available: (a) van is an instance of the nominal copula van, as in (21); (b) van is a preposition-like D( eterminer ), just like voor (see (22)): (21) [DP wati [D. Df+WHJ [pp t'1 lrlF (=van)+ x; (= 'nl€1)] lxP boeken fx, ti t)JJJJJ (22) [DP wati [D,van[+WHJ [pp tj [p,[F~ +XJ= 'nle)] lxphoeken [X' ti 911111 It is admittedly quite hard to decide between these two structural analyses. A nice characteristic of the analysis in (21) is that it displays the interaction between the preposition-like D(eterminer) voor and the nominal copula va.n. In none of the dialects investigated did we find a pattern in which voor and van co-occur: *·wat voor va.n 'n boeken. The absence of voorwould make it possible for the nominal copula to surface. Turning next to the potential analysis in (22), we should ask ourselves whether van as a preposition-like determiner, analogously to voor, makes any sense. In Bennis et al.
13. From an orthographic point of view, we might also have represented fer as ver, as this expresses the parallelism with voor even more clearly. We have chosen fer as it (orthographically)
expresses the [-voiced] property of the fricative more dearly. 14. The wat van 'n N-pattern is found in the following dialects that are part of the dialectal
database ofthe DtDDD-project: Hooghalen Dutch, Klazienaveen Dutch, Groenlo Dutch, Noord Deurntngen Dutch, Borgloon Dutch, Oosteeklo Dutch. These dialects are not all part of the same geographical region in the Netherlands or Belgium. For example, Klazlenaveen Dutch is spoken in the province ofDrenthe in the northern part of the Netherlands. Oosteeklo Dutch, on the contrary, is spoken in East Flanders, Belgium.
Micro-diversity in Dutch interrogative DPs (1998), the analysis of voor as a preposition-like determiner was given support by drawing a parallel with the clausal system. That is, voor occurs as a (prepositional) complementizer in certain dialects. Given the often assumed parallelism between the clausal system and nominal system, one might expect there to be prepositional determiners as well (see note 9); the element voor in the wat voor 'n N-construction is taken to be such an element. Following this same line of reasoning, the question arises as to whether van ever occurs as a prepositional complementizer in varieties of Dutch. As suggested by the Aarschot Dutch example in (23), van can fulfill this role of'prepositional complementizei in certain varieties of Dutch (see Pauwels 1958: 400).
(23) Hij had schrik [van te vallen]. of to fall he had fear "He was afraid to falL"
(Aarschot Dutch)
Taking the perspective of cross-categorial parallelism, one might propose then that 1'an can occur as a prepositional determiner as well. There is a subtle difference, though, between van in (23) and the prepositional complementizer voor in the example in note 9: the latter occurs in a clausal construction in which an (empty) operator has been moved to [Spec,CP]. In (23), it is not obvious that (empty) operator movement to [Spec, CP] has taken place. We seem to have an infinitival clause here, the subject position of which, [Spec,TP], is occupied by an empty pronoun, i.e. PRO. In other words, van in (23) is not really similar to prepositional voor in example (i) of note 9, in that it does not require any operator-like element in its Spec-position. Given this asymmetry between the prepositional complementizers voor and van and also adopting the view of cross-categorial symmetry, one might propose that the element 1'an which features in the nominal expression wat van ('n) boeken is not a prepositional D(eterminer) either, but rather a nominal copula. The fact that van does not occur in any of the wat voor- variants of Aarschot Dutch may also be suggestive for rejecting the prepositional D-analysis of van in (17). 15 Acknowledging that the evidence against aD-analysis is not very strong, we will tentatively assume in what follows that van is a nominal copula, as in representation (21).16 Consider, finally, the variants in (18), which may rightfully be called 'bare patterns: since there even is no preposition-like element (voorha.n) present in the structure. We propose here that these variants have the same structure as the slightly 'more lexicalized' patterns in (15)-(17). As indicated, we assume that wat raises to [Spec,DP] via [Spec,FP]. (24)
[DP
wati [D,Df+WHJ [pp tj [p[F (=e) +Xi(= 'n!e)] fxP boeken [-r:' ti t}JJJJJ
15. Aarschot Dutch has the following variants: wat 'n N, wat N, wajfer N, wat voor N, hoe'n N, oke N (see Pauwels 1958: 350). 16. It will be argued in Section 4 that va11 (=F), together with the adjoined small clause head X. raises to D.
67
68
Norbert Corver & Marjo Van Koppen 3.2
The soort-pattern: wat voor 'n soort boeken
Within the descriptive pattern featuring the noun soort we have identified the following variants in our database: 17 (25) a.
wat voor ('n) soort van boeken what for (a) sort of books "what kind ofbooks"
b. wat voor ('1t) soort boeken what for (a) sort books (26) a.
waffer
('n) soort van boeken
what+ for (a) sort of
b. waffer
books
('n) soort boeken
what+ for (a) sort books
(27) wat van soort boeken 18 what of sort books (28) a.
wat ('n) soort van boeken what (a) sort of
books
b. wat ('n) soort boeken what (a) sort books (29) wat soortige boeken what sort-ig + e books
Except for the presence of soort and modulo example (29), the patterns in (25)-(28) are quite similar to the ones discussed in Section 3.1.19 One important dimension of variation concerns the presence versus absence of van after soort. Furthermore, the pattern in (29), where wat soort is followed by the bound morpheme -ig(e), is also quite remarkable. Obviously, the first question which we will have to address concerns the status of the noun soort. That is, what role does it fulfill within the nominal construction. 20 We 17. Besides soort, also other nouns can function as a predicate nominal in this variant of the wat voor N-construction: e.g. type, merk, as in wat voor type/merk auto~ (what for type/sort cars; "what kind of cars").
is the only pattern of the type wat van soort N attested in our database. Possibly, other variants of this type exist, such as wat van 'n soort N or wat van soort van N. We have only represented here what we found in our database. 19. In the examples (25)-(28), we have represented the variation regarding the presence versus absence of 'nina single example. In (15)-(18), this dimension of variation was represented in different examples (the a-examples versus the b-examples). 20. In certain dialects, we also find the noun soortement besides soort. Presumably, -ment is the same morpheme as the one found in nouns like: mankement ("defect"), amusement 18. (27)
Micro-diversity in Dutch interrogative DPs
propose that soort is a noun which acts as a nominal predicate and takes boeken as its external argument within a (DP-internal) small clause configuration (see 30a)_21 The linear order soort (=the predicate)+ boeken (=external argument) is the result of Predicate Inversion; see (30b). We further propose that the nominal expression soort (van) boeken functions as the external argument of the nominal predicate wat, as in (31a). The inverted surface order is obtained by predicate inversion of ·wat to a position preceding the external argument soort (van) boeken (see (31b)).22 In other words, predicate inversion applies more than once in this nominal construction. 23 (30) a.
fxP boeken
fxX~ soort]]
b. [ppsoorti [p,[F (=van) +Xi(= 0)] [li:Pboeken fx, ti tj11124 ("entertainment"), traktement ("salary, pay"). The non-productive, non-Germanic suffix -ment typically attaches to the root of verbs ending in -e(e)r-en; e.g. mankeren ("'to lack"), amuseren ("to amuse"), trakteren ("to treat"). So we also find: sorteren ("to sort': "assort"). 21.
See also Zamparelli (1998) for discussion of the nominal construction this kind of car.
As an alternative to the analysis in (30)-(31), one might propose that the pattern wat voor 'n soort van boeken has the following internal sytax: [[wat voor een soort] van boeken]. That is, wat voor 'n soort constitutes a complex nominal expression which is derived by predicate displacement of wat to a position preceding soort. This complex expression acts as a predicate nominal which predicates over boeken, as in (i). (i) a. [:xp boeken l;rX, [watvoor '11 soort]]] b. [pp[watvoor '11 soort]; [p[F(= van) +XJ lxpboeken fxt1 tJ]]] We do not adopt this analysis for the following reasons: firstly, the coordination patterns in (ii) are incompatible with the structural analysis in (ib ). Secondly, subextraction of wat out of boeke11 wat voor 'n soort van N is possible. In (ib), wat is part of a left branch spedfier phrase. Normally. subextraction is not possible from within such a configuration; see also Section 4. (ii) a. wat [voor '11 soort van boeken] en [voor 'n soort van kra11ten] what for a sort of books and for a sort of newspapers "what kind of books and newspapers" b. wat voor 'n [soort van boeken en soort van kranten] 23. That a nominal expression derived by predicate inversion can be involved in a more complex nominal involving predicate inversion is dear from an example like that asshole ofan idiot ofa doctor (see Den Dikken 2006 for discussion). 14. AB pointed out by two reviewers, our analysis predicts that the spurious indefinite article 'n should also occur as a small clause head in (30), yielding- in combination with the derivation in (31)- the pattern: wat voor 'n soortvan 'n boeken (what for a sort of a books; "what kind of books"). Although we have not found this pattern in our database, arguably because this logically possible form was not presented in the questionnaires which were filled out by our informants, both two of the reviewers and we as authors think this pattern is quite acceptable. Notice by the way that the occurrence of this pattern makes the analysis according to which 'n is followed by a silent noun SOORT (see e.g. Leu (2008a,b) and see also note 6) implausible; it is quite hard to see what it means: watvoor 'n soort van 'n SOORT 22.
69
70
Norbert Corver & Marjo Van Koppen (31) a.
lxPsoort van boeken lxX (= 'n!e) wat]]
b. lpp watm lp· lF + X 1 (= 'nle)] lxpsoort van boeken lx t1 tnJ]]] c.
lop watm lo· voorl+WHJ lpp f mlp lF + X1 (= 'n/0)] lxpsomt van boeken lx t1 tnJ]]]]]
Taking this derivation as our starting point, we will discuss in what follows different variants of this wat voor 'n soort N pattern. Before doing that, however, we will first concentrate on some of the ingredients of the derivational analysis in (30)-(31), starting with the first part, i.e. (30a,b ). Let us first of all point out that the pattern soort van boeken is not restricted to the wat voor-environment. As shown in (32), it can also, for example, be part of a nominal expression (DP) introduced by an indefinite article or a demonstrative determiner: (32) a. Jan heeft leen soort va.n gedicht] voorgelezen. Jan has a sort of poem out-read KJan read out a sort of poem."
b. Jan leest ldit soorl van boeken]. Jan reads this soort of books "Jan reads this kind of books:' Following the derivation in (30a,b), a nominal expression like een soort 1'a.n gedicht will have the derived representation in (33): 25 (33) lop een lppsoortj lF' lF (=van)+ .x; (=e)] lxpgedicht lx ti t)JJJJ An analysis according to which van is a nominal copula, analogously to the element van in the N van N-construction, leads to the expectation that they display similar grammatical behavior. This is indeed the case. First of all, it is impossible to coordinate two sequences each of which is introduced by van; see (34). In other words, van +NP cannot act as a conjunct. Secondly, as exemplified in (35), extraposition of the van +NP is impossible. Thirdly, as illustrated in (36), va.n +NP cannot be pronominalized as daar +van (there-of). (34) a. *Jan heeft leen soort van gedicht en van opstel] voorgelezen. Jan has a sort of poem and of essay out-read "Jan read out a kind of poem and a kind of essay."
boeken (what for a sort of a SORT books; *"what kind of a kind of books"). Notice also that the same pattern with a second, non-silent noun soort is infelicitous: *wat voor 'n soort van 'n soort boeken. 25. Spurious 'n is possible here: 'n soort van 'n gedicht (a sort of a poem; "a sort of poem"), dit soort van 'n gedichten (this sort of a poems; "this kind of poems").
Micro-diversity in Dutch interrogative DPs b. *Jan heeft [een pracht van een gedicht en van een opstel] Jan has a beauty of a poem and of an essay
1'oorgelezen. out-read "Jan read out a beauty of a poem and a beauty of an essay:' (35) a. *Jan heeft een soort voorgelezen van gedicht Jan has a sort out-read of poem KJan read out a sort of poem:' b. *Ja.n heeft een pracht voorgelezen van een gedicht Jan has a beauty out-read of a poem "Jan read out a beauty of a poem:' (36) a.
*Jan heeft [een soort daarvan] voorgelezen. Jan has a sort there-of out-read KJan read out a (poem-like) thing:'
b. *Ja.n heeft [een pracht daarvan] voorgelezen. Jan has a beauty there-of out-read "Jan read out a beauty of a (poem-like) thing:' Observe that van as found in a nominal expression like een gedicht va1t Jan "a poem of Jan's" behaves differently in all three respects, which suggests that it has an internal syntax different from the N van N-construction and the soort van N-construction: (37) a.
Jan heeft [een (gezamenlijk) gedicht van mij en van Kees] Jan has
a
(joint)
poem of
me and of
Kees
voorgelezen. out-read "Jan read out a poem jointly written by me and Kees." b. Jan heeft ee1t gedicht voorgeleze1t van Kees. Jan has a poem out-read of Kees "Jan read out a poem written by Kees:'
c. Jan heeft [een gedicht daarvan] voorgelezen. Jan has a poem there-of out-read "Jan read out a poem written by him/her." Having provided support for the parallelism between the N van N-construction and the soort van N-construction, let us next tum to a variant of the latter construction, viz. the pattern in which soort and N are juxtaposed; i.e. there is no intervening nominal copula van. (38) ]an heeft [een soort gedicht] voorgelezen. Jan has a sort poem out-read "Jan read out a kind of poem:'
71
72
Norbert Corver & Marjo Van Koppen
In the spirit of Carver's (2002) analysis of Dutch pseudopartitive constructions like 'n
doos sigaren (a box cigars, "a box of cigars"), where we also find a juxtaposition of two nominal elements, we will assume that soort, being a bare noun, is able to undergo predicate displacement of the head movement type; that is, soort raises to X, forming the complex head lx X+ soort], which subsequently head-moves to F, yielding [F F +
lx X+ soort]].26 1his yields the derived structure in (39):27 ( 39)
[DP
een [pp [[p F + [X+ soortjil lxpgedicht lx ti tj]]]
We propose that, since the complex head F becomes 'lexicalized' (i.e., it has sound properties) after adjunction of the complex head [X + soort] to F, there is no reason anymore to spell out F as the nominal copula van. In other words, there is a sort of economy principle which blocks copular spell out of the F-head when the complex head already has obtained lexical contents. So far, we have given an analysis of the lower part of the wat voor 'n soort (1'a.n) Npattern: the variant soort 1'an 'n N results from phrasal predicate displacement. whereas the variant soort N involves predicate displacement of the head movement type. Turning now to the various instantiations of the pattern wat voor 'n soort N in (25)-(28), we see that the analyses of the various realizations are quite similar to the ones given for the ·wat voor 'n N-pattern. The major difference is that in the latter pattern we have a simple phrase (e. g. the N P boeken) as the external argument of wat (see (19)), whereas in the former pattern we have a complex phrase (e.g. soort (van) boeken) as the external argument of wat; see (31). The variant wat voor (n) soort (van) 26. An alternative line of analysis would be one according to which the predicate soort is basegenerated as a specifier of the small dause and gedicht as the complement of X. Under such an analysis, the predkation relationship in syntax is configurational yet nond!rectional (see Den Dikken 2006: 43). In this paper, we will adopt an analysis in which the predkate originates in the complement position of X.
27. A question which may arise is whether the N van N-construction has a juxtaposed N-N variant as well. As a matter of fact, we do find the pattern in (ib) besides (ia). Importantly, this N-N variant can have a phrasal stress pattern; i.e. stress falls on the second noun ( 'n pracht geDICHT), whkh suggests that it is not a nominal compound It should be noted, however, that this pattern is exceptional. In general, the juxtaposed pattern is not permitted as a variant of the N van N-construction. as exemplified in (ii). (i)
a
Jan heeft ['11 pracht van 'n gedicht] voorgelezen Jan has a
b. (ii) a
beauty of
Jan has a
b.
a poem
out-read
Jan heeft ['n pracht gedicht] voorgelezen Jan heeft ['n juweel van 'n gedicht] voorgelezen jewel of a poem
out-read
""Jan heeft ['n juweel gedicht] voorgelezen Jan has
a
jewel poem
out-read
We have nothing of Interest to say here about the restricted occurrence of the N-N counterpart of theN van N-construction.
Micro-diversity in Dutch interrogative DPs
boeken now has the structure in (40a), where we abstract away from the internal structure of the external argument soort (van) boeken. The variant waffer ('n) soort (van) boeken has the structure in (40b). Finsally, the variant wat ('n) soort (van) boeken is assigned the structure in (40c): (40) a.
[DP wati [D,voor[+WHJ [pp tj [F'[F +XJ= 'nle)] fxPsoort (van) N [X' ti t)JJJJJ
b. [DP watj [D,for[+WHJ [pp tj [F' [F +X;(= 'nle)] fxpsoort (van) N fx, ti tJIJJJl
[DP watj [D' Dl+WHJ [pp tj [F' [F +X;(= 'nle)] fxP soort (van) N fx, ti tilJJJJ ~ for the variant wat van soort boeken, we will assume, in line with our discussion of van in Section 3.1, that van is a nominal copula.
(41)
[DP watj [D,D[+WHJ [pp tj [F'[F (=van)+~(= e)] fxpsoort boeken fx, ti tiJJJJJ
Summarizing, our analysis of the wat voor soort N- pattern is similar to our analysis of the wat voor N-pattem. The crucial difference is that in the former pattern, the external argument of the nominal predicate wat is complex (soort (van) boeken) rather than simplex (boeken). Importantly, the complex nominal expression soort (van) boeken involves a predication configuration and is derived by means of predicate displacement. There is one variant of the soort- pattern which we have not discussed so far, viz. wat soortige boeken (see (29)). We will wait with the discussion of this variant until Section 3.4. 3·3
The doubling pattern: Wat voor zulk.e boeken
Having provided an analysis of the bare pattern (wat voor 'n N) and the soort- pattern (wat voor 'n soort (1'an) N), let us now tum to the third descriptive pattern, i.e. the doubling pattern, as in wat voor zulke boeken (what for such books, "what kind of books"). Recall that we called it the 'doubling' pattern because of the co-occurrence of two pronominal elements, namely the interrogative pronoun wat and the indefinite demonstrative element zulke (or, as we will see below, the interrogative pronoun ·welke). We will claim that zulk is not an atomic word, but rather consists of various subparts, one being -lk. We will propose that -lk is an adjectival small clause head (i.e. a Relator in the sense of Den Dikken (2006)), which establishes a relation between a predicate (i.e. the complement of -lk) and a subject (the specifier of -lk). But before getting into this more deeply, we would like to start with the pair (42)-(43): (42) [Wat voor 'n boek] heeft Jangelezen? whatfor a book has Jan read (43) [Zo'n hoek] heeft Jan gelezen. such a book has Jan read
73
74
Norbert Corver & Marjo Van Koppen
(42) and (43) are related: in a way, (42) is the interrogative counterpart of (43), which features what is called the indefinite demonstrative indefinite pronominal zo'n in Dutch traditional grammar. Modulo the element 1'oor, the two constructions are very similar: a pro-form precedes a noun and the two are separated from each other by an intervening indefinite article 'n. In Bennis et al (1998), this parallelism is made visible in the structural analysis of the two nominal constructions: both constructions involve a predication configuration and, in both constructions, the surface word order is derived by means of Predicate Inversion. As we already know, in (42), watis the inverted predicate. In zo'n boek, the pro-form zo is the inverted predicate. In other words, what is traditionally analyzed as a single word actually consists of two parts: zo and 'n. Schematically:28 (44) a.
f.lf:Pboek lxlx 'n]zo]]
b. [ppzoJp,[F+X;(= 'n)] lxpboekfx,tJJJJJ
Thus, put informally, zo'n boek expresses the meaning: "book which is so"; i.e. "a book like that': That zo can function as a pro-predicate is also clear from clausal patterns like the following: 18. Also here the question arises as to whether there is any independent evldence for the displacement of the pro-predicate zo. Interestingly, zo cannot combine with a negative word (see ia,b) but can be combined with the 'positive' counterpart (see iia,b ). This may be interpreted as an inner island effect: the predicate zo cannot cross the intervening negative word, which occupies [Spec,XP]; see (iii).
(i)
a b.
(ii) a
b.
*zo niets so nothing; "nothing like that" *zo 11iemand zo noone; "noone like that" zo iets so something "something like that" zo iemand so someone "someone like that"
(iii) *[ppzoi [p[F + XJ [XP11iets fx t1 tJlll There does not seem to be any semantic reason for the ill- formedness of the patterns in (I) and (ii). Notice, for example, the following constructions, which semantically correspond to the intended meanings in (ia) and (iia): (ill) a
b.
niets dergelijks nothing such-like-s "nothing like that" lets dergelijks something such-like-s "something like that"
Micro-diversity in Dutch interrogative DPs
(45) a.
Jan is normaal nooit zo.
(e.g. zo =arrogant)
Jan is normally never so "Jan normally never behaves that way." b. Jij wordt later ook zo. you become later also so "Later, you will also be like that:'
(e.g. zo =bald)
In (45a,b) zo functions as the predicate in a clausalcopular construction, with zijn ("to be") as the copular verb in (45a), and worden ("to become") as the copular verb in (45b ). If the analysis in (45b) is on the right track, we expect 'n in (45b) to be an instance of the spurious indefinite article. Although in Standard Dutch we do not find a pattern in which 'n precedes a plural noun, in colloquial Dutch and Dutch dialects a pattern like that is attested (see, for example, Pauwels 1958): 29 (46)
[Zo'n boeken] heeft Jan gelezen. so a books has Jan read "Jan read such books."
ru an alternative to the interrogative pattern in (42), we find the pattern in (47a) in a number of Dutch dialects. This pattern is also found with plural nouns: hoe'n boeken. We propose that, analogously to the non-interrogative zo'n hoek, this variant has the structure in (47b): (47) a.
[Hoe'n boek] heeft Jan gelezen? how a book has Jan read 'What kind of book did Jan read?"
b.
[0 p hoei [0 , Df+WHJ
[pp
tj [p [F + ~ (= 'n)] [XP boek lx t; t)lllll
Although in certain dialects it is possible to use a nominal expression like (46), in which 'n is followed by a plural noun, in standard Dutch (and many other dialects) one has to use the variant in (48), featuring the pronominal elementzulk(e): (48) [Zulke boeken] heeft Jan gelezen. such-e books has Jan read "Jan read such books." Also here, Dutch traditional grammars characterize zulke as a single word. However, extending our decompositional analysis for zo'n in ( 44a), we would like to propose that zulke contains the following sub-parts: zo + -lk +-e. Thus, the first part corresponds to the pro-torm zo. But what about -lk and -e? As for -lk we propose that this boundmorphemic element is the equivalent of the English word like, which typically establishes a predication relationship between an argument and a predicate, as for example 29. In the DiDDD-d.atabase, the following dialects display. for example, this pattern: Lier Dutch, Maasbree Dutch, D!lbeek Dutch.
75
76
Norbert Corver & Marjo Van Koppen in (49), where eyes functions as the subject and diamonds as the predicate within the nominal expression eyes like diamonds. (49) She has [eyes like diamonds]. In Dutch, we have the equivalent pattern in (50), featuring the elementgelijk "like': The use ofgelijk is somewhat archaic; normally the word als "like/as" is used in this context:
(SO) a. Zij heeft [ogen gelijk diamanten]. b. Zij heeft [ogen als diamanten]. With Bennis et al (1998) and Den Dikken (2006 ), we assume that like is a small clause head; i.e. a Relator in the sense of Den Dikken (2006). We further propose that -lk is a bound morphemic instance of the small clause head. This relator head establishes a predication relationship between a predicate and an argument In the case of zu.lke boeken in (48), we take -lk to establish a predication relation between the predicate zo and the argument boeken. This is represented in (51a). The surface order results from the application of Predicate Inversion. as in (51b ): (51) a.
f.lf:Pboekenfxfx-lk]zo]]
b. [ppzo 1 [p,[[Xi (= -lk) +F)+ -e]} fxP boeken [X' ti t)JJJ (= zulke boeken) Importantly, zu.lke is morphologically inflected: the inflection -e - which is the same inflection as the one found on attributively used adjectives (mooi-e boeken; beautiful-e books)- agrees with the plural noun boeken. The property of being inflected like an adjective suggests that -lk is an adjectival (i.e. [+N,+V]) relator head. In this respect it differs from the spurious indefinite article, which we take to be nominal (i.e. [+N,-V]) in nature. In the context of this article we will not present a full-fledged theory on the nature of adjectival inflection. We will assume that the bound-morphemic adjectival relator-head enters into a spec-head agreement relationship with the noun boeken, which manitests itself morphologically as the adjectival inflection-eon -lk (i.e. -lk-e).30 Interestingly, besides the form zu.lke we also find the form zu(l)kse in certain Dutch dialects. We propose that the -s in zu(l)kse is a bound-morphemic instance of the nominal copula.31 Schematically (example from Uithuizen Dutch): (52) Ik heb J..-rekt zukke/zukse schounen as mien zuster. I have exactly such-e/such-s-e shoes as my sister "I have exactly the same shoes as my sister." (53) [ppzoj[F'[XJ=-(l)k) +F(=s/0) +-e) fxPschounen [X'tJi]]]] (F = -s---,). zukse schounen; F = 0---,). zukke schounen) 30. Under an Agree-based analysis, one might propose that -lk enters into an Agree-relation with the noun boeken in [Spec,XP] after -lk has raised to F. 31. See Den Dillin (1998) and Corver (2008, 2009) for other nominal environments in which the nominal copula -s surlaces.
Micro-diversity in Dutch interrogative DPs In certain dialects (not in Standard Dutch) we find the interrogative pattern hoelke boeken (how-Ike books, "what kind ofbooks").32 If (51b) is the analysis for zulke hoeken, then arguably (54) is the analysis for hoelke boeken: (54) a. b.
fxP boeken fx fx -lk] hoe]] [pphoei [p[F (=~)+Xi(= -lk) +-e) fxPboeken fxtJ)JJJ
Having given an analysis of wat 1'oor 'n N variants such as hoe'n boek and hoe(l)ke boeken, let us next turn to the slightly more complex patterns in (55): (55) a.
·wat voor zulke what tor
b.
wat voor welke what tor
boeken
(Uithuizen Dutch)
such-INFL books
boeken
which-INFL books
In these variants, we find two pronominal elements: the wh-word wat and a second pronominal element: (a) the indefinite pronominal zu.lke!zukke in (55a) and (b) the interrogative pronominal welke in (55b).33 Two potential analyses come to mind for these constructions: one analysis would be to say that the additional pro-form and the noun boeken constitute a complex nominal expression which functions as the external argument of the predicate nominal wat (as in (56a)). Importantly, the small clause subject zulke boeken is itself derived via predicate displacement, as is depicted in (51). The surface word order wat voor zulke boeken is derived by application of predicate movement, as schematically represented in (56b). (56) a.
b.
fxP [zulke boeken] fxX wat]] [DP wati [D' voor [pp tj [p~ + F fxP [zulke boeken] f.lf:' ti t)JJJJJ
According to the second, alternative analysis, it is not the small clause subject which is complex, but rather the predicate nominal More specifically, the 'base structure' is the one in (57 a), with boeken as the small clause subject, -lkas the small clause head (i.e. the Relator in Den Dikken's sense), and wat voor zo as the complex predicate. This predicate has the 'underlying structure' in (58a): zo is the small clause subject and wat is the interrogative predicate. Thus, the meaning corresponding to this small clause configuration can informally be paraphrased as: "so is what?". The pattern wat voor zo is the result of predicate movement of wat to the specifier position of voor, as in (58b). The complex predicate wat voor zo undergoes predicate displacement to a position
32· In the DiDDD-database, Urk Dutch and Gilze Dutch exhibit this pattern, for example. 33· In Uithuizen Dutch, we also find the pattern wat van zukke boeken (what of such books; "what kind of books"). It does not seem unlikely that van is the nominal copula here. Another interesting pattern is watvukke boeke11 (what -v-(s )uch books; "what kind ofbooks") from NoordDeurningen Dutch. The form watvukke, possibly. is a contracted form of wat + van + zukke.
77
78
Norbert Corver & Marjo Van Koppen preceding boeken, with concomitant movement of the small clause head -lk(e). This yields the sequence wat 1'oor zo + -lke boeken.34 (57) a.
[XP [boeken] fx-lk [wat voor zo]]]
b. [pp [wa.t 1'oor zo ]i [F' [~ ( = -lk) + F] + e [XP boeken fx, ti t)]]] (58) a.
lxpzo fxX wa.t]]
b. [DP wati [D' voor [pp tj [~ + F] lxpzo fx tJ)1111 The question, obviously, arises as to whether there is any empirical support for one or the other analysis. One potential argument comes from subextraction. Getting ahead of our analysis of the discontinuous wat voor N- patterns, we already point out here an important contrast between the patterns wat voor 'n boeken and wat voor 'n soort boeken, on the one hand, and the wat voor N-patterns in (55), on the other hand: the former permit subextraction of ·wat out of the nominal expression (see (59a)), whereas the latter do not (see (59b )). (59) a.
Wati heb jij
voor [ti voor 'n (soort) boeken] gekocht?
what have you for a (sort) books "What kind of books did you buy?"
b. *Wati heb jij
bought
voor [ti voor zulke!welke boeken] gekocht?
what have you for sort such/which books "What kind of books did you buy?"
bought
This asymmetry in subextraction behavior is quite remarkable, since the patterns are superficially quite similar. Under a structural analysis like (56b ), it is not immediately obvious why wat cannot be moved out of the nominal expression, since, after DP-internal predicate displacement has taken place, wat occupies [Spec,DP], which is the escape hatch for subextraction. Under a structural analysis like (57), on the contrary, the ill-formedness of the subextraction pattern may follow from the fact that wat is too deeply embedded within the wat voor N-construction. Importantly, it does not occupy the edge position of the nominal expression wat voor zulke boeken, but rather the edge position of the inverted predicate wat voor zo, which is a left branch specifier contained within the larger nominal expression wat voor zulke boeken. This implies that removal of wat out of the wat voor N-construction involves subextraction out of a left branch specifier of the nominal expression. Such subextractions are generally ruled out for reasons of locality. For example, subextraction of wat is impossible out of a wat voor N-construction which acts as left branch possessor within a complex nominal expression (see (60a)). As shown in (60b), pied piping of the entire nominal expression is required.
34· Certain dialects permit the pattern wat voor zukse boeken, featuring the element -s, which we Interpret as an instance of the nominal copula: (i) [pp[watvoor zo]1 [F' [)G (= -lk) + F (=s)] + e [x.p boeken [X'~ tJ]]]
Micro-diversity in Dutch interrogative DPs
(60) Ik vraag me af ... I wonder REFL PRT ...
a. * .... [wat]; ze [[t; voor 'n jongen] z'n fiets] hebben gestolen . ....what they for
a boy
his bike have
stolen
b. ... [[wat voor 'n jongen] z'n fiets]; ze t; hebben gestolen . ...what for a boy his bike they have "I wonder what kind of boy's bike they stole."
stolen
We will assume that the pattern wat voor welke boeken has the same derivation as wat voor zulke boeken, the only difference being the small clause that constitutes the predicate of boeken. Instead of lxP zo fx,X wat]] in (58), we have the predicate lxP wat fx,X wa.t]], which can informally be paraphrased as: "something is what?".35 We propose that the 'doubling' patterns in (61) essentially have the same derivation as the patterns in (55). (61) a.
hoe zo'n boek how so a book "what kind ofbook"
b. hoe zulke boeken how such books "what kind ofbooks" The sequence hoe zo forms a complex phrase derived by predicate inversion of the interrogative word hoe, as in (63 ): (62) a.
b.
lxpzo fx,X hoe]] [pp
hoei [p ~ +F) lxP zo fx, t; tjJJJ (hoe zo)
The complex phrase hoe zo constitutes a predicate of boeken. The surface word order is derived by applying predicate inversion to the complex predicate hoe zo, as in (63) and (64). In (63), 'n is the small clause head, and in (64) -lk is the small clause head. (63) a.
lxP [boeken] fx, 'n [hoe zo]]]
[hoe zo]i [p ~ (= 'n) +F) lxP boek [X' t; tjJJJ (hoe zo'n boek) (64) a. lxP [boeken] fx,-lk-e [hoe zo]]] b.
[pp
b. [pp [hoe zo]i [p [X;(= -lk-e) +F) f.lf:P boeken fx, t; tjJJJ (hoe zulke boeken)3 6
35· wat in [Spec,XP] is the indefinite pronoun, meaning "something': 36· In certain dialects of Dutch, we also find the pattern hou zukse bouken: (i) Hou zukse bouken hes de koft? how such-s-e books have you bought "What kind of books did you buy?"
(Uithuizen Dutch)
79
so
Norbert Corver & Marjo Van Koppen Our analysis of -lk as an adjectival (i.e. [+N, +V]) small clause head opens the way to an analysis of the pattern wat soortige boeken in (29). In the next section, we will briefly discuss this pattern.
3·4
Another soort-variant: wat soortige N
The central question about this variant of the wat 1'oor 'n N-construction is, of course, the following: what is the nature of the bound morpheme -ig, which appears attached to the noun soort? We propose that -ig is an adjectival suffix; see also Leu (2008a,b) for German. More specifically, we assume that, just like adjectival-lk, it is a small clause head that mediates between an external argument (i.e. the subject) and a predicate. Schematically:
(65)
fxpNPrubject fx,ig yppredicate]]
Potential support for this Relator-status of -ig may come from the following pairs, which suggest a certain parallelism between the N van N construction and the construction featuring -ig. (66) a.
'n schat
van n JOngen
a darling of
a boy
b. 'n schatt-ig-e
jongen
a darling-ig-e boy Ka darling of a boy" /"a cute boy"
(67) a.
'n etter van 'n ventje a puss of
a guy
b. 'n etter-ig ventje a puss-ig guy "a jerk of a boy" /"a nasty boy" It does not seem implausible to assign an analysis to the examples (66b)-(67b) which parallels the analysis of theN van N-constructions in the a-examples. This structurally parallel analysis is depicted in (68): (68) a.
[DP
'n [ppschatj [p F (= va.n) +xi(= 'n) lxpjongen (l(' ti t)JJJJ
"a darling of a boy"
b.
[DP
'n [ppschati [p [F + lx ig +-eM lxpjongen fx, ti tJIJJJ
Taking -ig to be an adjectival Relator-head, we can now assign the following analysis to the expression wat soortige boeken. First of all, we propose that wat soort constitutes a phrase whose internal syntax involves predicate movement. More specifically, (69a) is taken to be the base structure. and (69b) the derived structure.
Micro-diversity in Dutch interrogative DPs (69) a.
fxPsoort fxX wat]]
b. [DP wati [D' D [FP tj [Xi+ F) fxP soort fx ti t)JJJJ The phrase wat soort, in turn, functions as a predicate over the NP boeken, with the adjectival head -ig functioning as the mediating Relator-head: (70) a.
fxP [boeken] fx,-ig [wat soort]]]
b. [FP [wat soort]j [F [X;(= -ig) +F)+ e fxP boeken fx ti t)JJJ Summarizing, we have analyzed three descriptive patterns of the wat voor 'n N-construction: the bare pattern (wat voor boeken ), the soort- pattern (wat voor 'n soort (van) boeken) and the doubling pattern (wat voor zulke boeken). In all three patterns, wat starts out as a predicate nominal in a predication configuration and undergoes predicate displacement to a position preceding the 'subject' of the predication relationship. The nature of the subject differs in the three descriptive patterns: in the bare pattern, a bare NP (boeken) is the subject; in the soort-pattem, soort + NP (soort boeken) constitutes the subject of wa.t; in the doubling pattern, finally, a pro-form (e.g. zo) acts as the subject of wat.
4
Micro-diversity in the split wat voor 'n N-construction
Having provided an analysis of the internal syntax of various wa.t voor N- patterns, we will now tum to the external syntax of this interrogative nominal construction. More specifically, we will examine the phenomenon of wat voor-split from a micro-comparative perspective. As will become clear, this phenomenon is not attested in all dialectal varieties of Dutch. The question obviously arises then whether this variation can somehow be related to differences in the internal syntax of the cross-dialectal variants of the wat voor 'n N-construction. Before turning to a more detailed discussion of each of the split patterns, let us formulate some major empirical findings that emerged from our cross-dialectal study: There are Dutch dialects that do not permit the split pattern. If Dutch dialect L allows for the split wat voor-pattem, it also allows for the nonsplit pattern. There are no dialects which have the split wat voor-pattem but do not have the non-split pattern. These findings are represented in the map in (71). The yellow dots represent the dialects that do not permit subextraction of the wh-element The small squares with a yellow dot are the dialects which have both the split pattern and the non-split pattern. The absence of dotless squares shows that there are no dialects which only permit the split pattern.
81
82
Norbert Corver & Marjo Van Koppen (71) • wat voor-construction +extraction (78) •
wat voor-construction without extraction (79)
Another generalization that can be formulated on the basis of our cross-dialectal study is that subextraction of the wh-word is only possible when the prepositional element voor or van is present The examples in (72) exemplify these split patterns. (72) a.
(dialectal Dutch)
Wai het je [ti veur appels] gekauch? what have you- for apples bought
b. Wai het je [ti
(dialectal Dutch)
va beuk] gekauch? what have you- for books bought
Let us now turn to the various subextraction patterns, starting with the variants that do permit subextraction of wat: (73) a. wat... voor ('n) boeken what .. for (a) books
(see (15))
b. wat... voor ('n) soort (van) boeken what.. for (a) sort (of) books
(see (25))
c. wat... van ('n) boeken what.. of (a) books
(see (17))
d. wat... van soort boeken what .. of sort books
(see (27))
Micro-diversity in Dutch interrogative DPs Following standard assumptions (see e.g. Szabolcsi (1983, 1994)), we will assume that a wh-phrase can leave a nominal expression via the edge (i.e. [Spec,DP]) of this expression. That is, [Spec,DP] functions as an escape hatch. In (73a,b ), wat has been moved to the Spec-position of the prepositional determiner voor before it leaves its 'nominal home' (see (19) and (31)). Before turning next to the examples in (73c,d), we would first like to point out that subextraction is impossible from the patterns in (7 4), which lack the element van: 37 (74) a. *wat ... ('n) boeken
b. *wat... ('n) soort (1'a.n) boeken
(see (18)) (see (28))
The question obviously arises as to how to account for this contrast between (73c,d) and (7 4a,b ). Remember that we analyzed van in (7 3c,d) as an instance of the nominal copula, realizing the functional head F. This functional head F (i.e. van) forms a complex head together with the raised small clause head X (i.e. the spurious indefinite article 'n ). In other words, the sequence van 'n has the structure: [p F (= van) [p X (= 'n) ]] . Suppose now that the nominal copula van (dragging along the spurious indefinite article) is able to raise to the [+interrogative] D-head, quite analogously to the raising of a (copular) finite verb to C in an interrogative main clause in Dutch.38 (75) a.
Wat zijn dat? what are that "What are those things?"
b.
fcP wati [C'zijni [TP datA: fr t'i [yp tk. fv, ti tjlJJJJ
We tentatively propose that it is the raising of the nominal copula (i.e., the complex head [F + X]) which makes it possible to sub-extract wat out of the nominal wa.t voor N-phrase. In a certain way, the D-head needs "support" of the nominal copula van for subextraction of wat to be possible. When the nominal copula van is absent (i.e. F has no phonological content) no raising of [F + X] to D takes place, even if the small clause head X has lexical content by means of the spurious indefinite article 'n. In other words, it is the lexical content of the head of the complex head [p F [X]] which is crucial for movement to the interrogative D-head to be possible. With [p F (= van) [X]] raising to 37· In standard Dutch, we do find the split pattern in (ia) in exdamative constructions: (i) a. Wat heb jlj 'n boeken gekocht! what have you a books bought "How many books you bought!" b. Wat 'n boeken heb jij gekocht! In Corver (1990), its is argued, though, that this split pattern does not result from subextraction of wat from within the nominal expression wat 'n boeken, which, as shown in (ib) is also possible in exclamative clausal contexts. Rather, he proposes that wat in (Ia) is base-generated as an exdamative clausal operator in [Spec,CP]. 38. We abstract away from the question as to whether Dutch is SVO or SOV underlyingly.
83
84
Norbert Corver & Marjo Van Koppen
D, the structure in (21), where the nominal copula is still in situ, can now be more precisely represented as (21'), where the nominal copula (i.e. the complex head [F + X]) has been raised to the interrogative D-position. (21') [DPwat1 [D.[van + 'nladk [FPtj [F'tk[xpboeken [X' ti t)JJJJJ When van is absent (i.e. F is not spelled out phonologically), the complex F-head does not raise to D. Thus, wat ('n) boeken has the representation in (24). We will assume that itis the absence ofKlexical support" forD which blocks subextraction of the wh-word wat in (Spec,DP]. In short, for subextraction of wat from [Spec,DP] to be possible, the D-head needs to have lexical content, either via direct Merge (as in the case of voor) or via head movement (i.e. raising of the nominal copula (i.e. [p F (=van) +X ( = 'n/a)] to D).39 Consider next the ill-formed patterns in (76). Given the fact that subextraction is permitted when voor is phonologically strong, the conclusion seems inescapable that the ill-formedness of these patterns is due to the phonologically weak status of the prepositional determiner. Subextraction of wat will yield a 'dangling weak determiner, which has no host to attach to; see (20) and (40b) for the relevant representations. (76) a. *wat .. .fer ('n) boeken b. *wat ... fer ('n) soort (van) boeken
(see (16)) (see (26))
We have already discussed in Section 3.3 the impossibility of subextracting the whword wat from within the wat voor zulke N construction, repeated here as (77a). According to our analysis, ·wat voor zo constitutes a phrase, which after predicate displacement occupies a left branch specifier position, i.e. [Spec,FP]; see (57b). Importantly, wat does not occupy the edge-position of DP. Subextraction from within a left branch specifier position is generally taken to be impossible. This account of the non-extractability ofwatalso applies to the wa.tvoorwelke N-pattern in (77b), the only difference being that wat voor wat constitutes the wh-phrase. Finally, removal of wat out of the nominal expression wat soortige boeken in (77c) is blocked for the same 39· As pointed out by a reviewer, also in the Dutch clausal domain, lexicallzation of C Is required for subextraction of a wh-phrase to be possible. As shown in (ia), the C-head does not have to be lexicalized when a wh-phrase occupies [Spec,CP]. However, when a wh-phrase is removed from an embedded CP, the C-position cannot be empty; i.e. it must have lexical contents: (i) a
Ik weet nlet [CP wat; [hij t1 doet]] I know not
b.
what he
does
Ik weet nietfcp wat1 [hij denkt[CP t'1 fc *(dat) [hij t1 doet]llll I know not
what
he thinks
(that) he
does
An interesting illustration oflexicalizatton ofC via (head) movement of a finite verb to C comes from Belfast English. As pointed out by Henry ( 1995), intermediate wh-movement to the Specposition of an embedded CP triggers I-to-C movement in this English variant: (ii) Wha~ did Mary claim b f 1 [c· did, [theytJ steal t 1]]]?
Micro-diversity in Dutch interrogative DPs
reason: wat is part of a left branch phrase occupying [Spec,FP]; see (70). Crucially, it does not occupy [Spec,DP]. & a result of that, subextraction is impossible. (77) a. *wa.t ... voor zulke boeken
(see (55))
b. *wat ... voor welke boeken c. *wat ... soortige boeken
(see (29))
Consider, finally, the ill-formed split patterns in (78), where the wh-word hoe "how" has been removed from a nominal expression. (78) a. *hoe ....zulke boeken
b. *hoe ... -lke boeken *hoe ...'n boek
(see (61)) (see (54)) (see (47))
The ill-formedness of (78) is related to the ill-formedness of the split patterns in (77): hoe forms a phrase together with zo. This phrase is a left branch specifier of the functional head R see (64). Subextraction from within this specifier position is blocked. The ill-formedness of (78b) arguably relates to the bound morphemic status of the stranded adjectival Relator-head -lk(e): a bound morpheme cannot be stranded; see (54) for the relevant structure. Consider, finally, (78c). It seems likely that this split pattern is out for the same reason that (7 4a) is out: there is no raising of a complex Fhead, containing the nominal copula van, to D. Consequently, D is not lexicalized. Even though hoe occupies the edge position within DP (see (47b)), subextraction of hoe is blocked because of the absence of"lexical content>' in the D-position.
5·
Conclusion
In this paper, we have tried to give a description and an analysis of the intra- and interdialectal variants of the Dutch wat voor N-construction. Building on Bennis et al:s (1998) structural analysis of the standard Dutch wat voor N-construction, we have argued that both a DP-internal predication configuration and a process of predicate displacement are at the basis of the different variants. The variants differ from each other in a number of respects, such as (i) the nature of the 'subject' of the predication relation (bare NP, soort+ NP, a pro-form likezo); (ii) the realization of the small clause head ('n, e, -lk); (iii) the (non)realization ofF as a nominal copula (1'an or the bound morpheme-s). Importantly, we have tried to show that there is a common structural core (predication and predicate displacement) behind these different variants, which is in line with Chomsky's (2001) Uniformity Principle, which states that "In the absence of compelling evidence to the contrary, assume languages to be uniform, with variety restricted to easily detectable properties of utterances': In a way, then, the different intra- and interdialectal wat voor N-manifestations are all variations on a common structural theme.
85
86
Norbert Corver & Marjo Van Koppen
References Bennis, Hans. 1983. A case of restructuring. In Linguistics in the Netherlands 1983, Hans Bennis & W. U.S. van Lessen Kloeke (eds ), 9-19. Dordrecht: Forls. Bennis, Hans 1995. The meaning of structure: The wat voor construction revisited In Linguistics in the Netherlands 1995, Marcel den Dikken & Kees Hengeveld (eds). Amsterdam: John Benjamins. Bennis, Hans, Corver, Norbert & den Dlkken, Marcel 1998. Predication in nominal phrases. Journal of Comparative Germanic Lb1guistics 1: 85-117. den Besten, Hans. 1981. Government, syntaktlsche Struktur und Kasus. In Sprache: Formen und Strukturen. Alden des 15. Lingulstischen Kolloquiwns, Miinster 1980, Vol 1 [Linguistlsche Arbeiten 98], Manfred Kohrt & Jiirgen Lenerz (eds ), 97-107. 'llibingen: Niemeyer. den Besten. Hans. 1985. The ergative hypothesis and free word order in Dutch and German. In Studies in German Grammar, Jindrich Toman (ed), 23-62. Dordrecht: Foris. Chomsky, Noam. 2001. Derivation by phase. In Ken Hale: A Life in Language, Michael Kenstowicz (ed), 1-52. Cambridge MA: The MIT Press. Corver, Norbert. 1990. The Syntax of Left Branch Extractions. PhD dissertation. T:dburg University. Corver, Norbert. 1991. The internal syntax and movement behavior of the Dutch 'wat voor'construction. Linguistlsche Berichte 133: 190-228. Corver, Norbert. 2002. On three types of movement within the Dutch nominal domain. In From NP to DP, Vol. 1: The Syntax and Semantics of Noun Phrases [Lingulstik Aktuell/Linguistics Today 55], Martine Coene & Yves d'Hulst (eds), 297-328. Amsterdam: John Benjamins. Corver, Norbert. 2008. Uniformity and diversity in the syntax of evaluative vocatives. Journal of Comparative Germanic Linguistics 11 (1): 43-93. Corver, Norbert 2009. Getting the (syntactic) measure of measure phrases. The Linguistic Review 26(1): 67-134. Corver, Norbert, van Koppen, Marjo, Kranendonk. Huib & Rigterink. Mirjam. 2007. The noun phrase: Diversity in Dutch DP design. Nordlyd 34(1): 73-85. den Dlkken. Marcel 1995. Copulas. Paper presented at GLOW Troms0; Ms., Vrije Universiteit Amsterdam/HI L den Dikken, Marcel 1998. Predicate inversion in DP. In Possessors, Predicates and Movement in the Determiner Phrase [Linguistik Aktuell!Linguistics Today 22], Artemis Alexiadou & Chris Wilder (eds), 177-214. Amsterdam: JohnBenjamins. den Dikken, Marcel. 2006. Relators and Linkers. The Syntax of Predication, Predicate Inversion, and Copulas. Cambridge MA: The MIT Press Henry, Alison 1995. Belfast English and Standard English: Dialect Variation and Parameter Setting. Oxford: OUP. Honcoop, Martin. 1998. Dynamic Rr:cursions on Weak Islands. The Hague: HAG. Kayne, Richard 1994. The AnNsymmetry ofSyntax. Cambridge MA: The MIT Press. Leu, Thomas. 2007. A Note on 'what for' Split. NYU Working Papers In Linguistics 1: 1-20. Leu, Thomas. 2008a. 'What for' internally. Syntax 11(1): 1-25 Leu, Thomas. 2008b. The Internal Syntax of Determiners. PhD dissertation. New York University. Moro, Andrea. 1988. Per una teoria unificata delle frasi copulari. Rivista di Grammatica Generatlva 13(8): 1-110. Pafel, Jiirgen. 1996. Die syntaktische und semantlsche struktur von wasforphrasen. Linguistische Berichte 161: 37-67.
Micro-diversity in Dutch interrogative DPs Pauwels,Jan 1958. HetDialectvan Aarschoten Omstreken, Vol I. Tongeren: Belgisch Interuniversitatr Centrwn voor Neerlandistiek. Postma, Gertjan. 1995. Zero semantics- The semantic encoding of quantlficational meaning. In Linguistics In the Netherlands 1995, Marcel den Dikken & Kees Hengeveld (eds ), 175-190. Amsterdam: John Benjamins. van Riemsdijk. Henk. 2005. Silent nouns and the spurious indefinite article in Dutch. In Grammar & Beyond: Essays In honour of Lars Hellan, Mila Vulchanova & Tor A. Afarli (eds), 163-178. Oslo: Novus Press. Ross, John-Robert. 1983. Inner islands. Ms, MIT. Szabolcsi, Anna 1983. The possessor that ran away from home. The Linguistic Review 3: 89-102. Szabolcsi., Anna. 1994. The noun phrase. In Syntax and Semantic 27: The Syntactic Structure of Hungarian, Istvan Kenesei (ed.), 179-274. New York NY: Academic Press Vangsnes, 0ystein A. 2008. What kind of Scandinavian? On interrogative noun phrases across North Germanic. Nordic Journal ofLinguistics 31(2): 227-251 Zamparelli, Roberto. 1998. A theory of kinds, partitives and of!z possessives. In Possessors, Predicates and Movement In the Determiner Phrase [Ltnguistik Aktuell/Linguistics Today 22], Artemis Alexiadou & Chris Wilder (eds), 259-301. Amsterdam: John Benja.mins.
87
Noun phrase structure and movement A cross-linguistic comparison of such!sadan!solch and so/sa/so* Johanna L. Wood & Sten Vikner University of Aarhus
We investigate the etymologically related words so and such (English); sd and sM.an (Danish); and so and solch (German). Similarities and differences that have to be accounted for cross-linguistically are i. position (pre- or post- indefinite article), ii. agreement morphology (in Danish and German), and iii. semantics (whether an AqiP or a DP/NP is modified). English and Danish solsd may only modify an AdjP, while German so may also modify the DP/NP. English such may only modify the DP/NP (Bolinger 1972, Wood 2002) and may only precede the indefinite article. Danish and German allow inflected sadan/solch to follow the article. We discuss two possible syntactic derivations, predicate raising (e.g. Corver 1998, Bennis, Corver & den Dikken 1998) and XP movement from an attributive adjective position within the nominal (e.g. Matushansky 2002). The analysis links up with the morphological agreement facts of predicate and of attributive adjectives in Danish and German (Vikner 2001).
1.
Introduction
The focus in this paper is on two etymologically related words, cognates of which are used to express degree in three Germanic languages: so and such in English, sa and stldan in Danish, and so and solch in German. The syntax and semantics of degree
*
We are grateful to participants at ESSE-9 (Aarhus), SyntaxLab (Cambridge) and the Conference on Variation and Change in the Romance and Germanic Noun Phrase (Amsterdam), and to Eva Engels, Steffen Krogh. two anonymous reviewers, and the editors of this volume for comments and suggestions. This work is carried out as part of the project Similarities and Differences between Clauses and Nominals financed by ForskningsrAdet for Kultur og Kommunikation (Danish Research Council for Culture and Communication).
90
Johanna L. Wood & Sten Vikner expressions have been of interest to linguists for at least forty years, as seen in e.g. Bolinger (1972) and Bresnan (1973). Abney (1987: 301-321) suggested not only that noun phrases occur inside determiner phrases but also that adjective phrases occur inside degree phrases. English expressions involving so, too, as, how, this, that plus an adjective are particularly relevant for Abney's suggestions, as these expressions may occur on the left periphery of nominals, preceding the indefinite article: (1) a. ...which are so big a part of the present system b. . )?which are a so big part of the present system
(COCA)
The fact that these degree expressions may precede the indefinite article has led to further research into the fine-grained structure of DPs. Degree phrases such as so big are either taken as evidence for functional structure above DP (e.g. Kennedy & Merchant 2000: 125 ), if the indefinite article is assumed to be D 0 , or for functional structure below D P (e.g. Matushansky 2002, Wood 2002 ), if the indefinite article is assumed to be the head of a phrase that occurs inside DP, e.g. Num(ber)P or Card(inality)P. Similarly, within both functional and generative frameworks, the analysis of the word such has long been discussed, as also such has a degree reading, intensifying in Bolinger's (1972: 60) terms. As soon as there is a gradable element in the context, either an adjective as in (2), or a gradable noun as in (3), such may have the meaning a reaction as violent as this reaction or a person as foolish as this person. (2) I did not expect such a violent reaction. (3) I did not expect to meet such a fool. In the terms of Quirk et al. (1985: 257), such is a predeterminer, albeit a special predeterminer which, in English, may only precede the indefinite article, (4) and not the definite article, (5). (4) a. ... which are a major part of the present system b. ... which are such a major part of the present system (5) a. b.
which are the major part of the present system ... which are such the major part of the present system
The syntax and semantics of such are complicated by the fact that it also has what we will refer to as a kind reading, a reading that Bolinger (1972: 60) terms identifying. In (6), such a machine means a machine of this kind because there is no possibility of construing machine as a gradable noun: (6) I have never seen such a machine before. Whereas English such may have both kind and degree readings, and sometimes be ambiguous between the two, English so (inside a nominal expression) is only a degree adverb, as in (la) above. Although Wood (2002) tocussed on explaining the syntax of such in terms of kind and degree, the starting point in this paper is on what is modified, the DP/NP, or the
Noun phrase structure and movement
adjective phrase. In (6), the only possibility is for such to have a kind reading, but (2) and (3) are both ambiguous between kind and degree. In Sections 3 and 4 the focus is on the link between word order, morphology, and the possible derivations. We will return to kind, degree and modification in Section 6 when we discuss language change. With respect to the syntax, two possible derivations of pre-article such and so in English, ( la) and (4b ), have been suggested in the recent literature. 1 In one derivation there is movement from the prototypical Germanic adjective position preceding the noun to a position preceding the indefinite article, as in (7): (7)
[such/so bad]; at; hotel
4
I
The other suggestion is that such and so expressions originate as predicates in a small clause that has a DP subject as in (8): (8)
[such/so bad]k a hotel [ec ~l
4
I
In this paper we compare so and such to their equivalents in Danish, satsddan, and in German, so!solch, in order to decide which of the two derivations above is most plausible. The significance of introducing these languages lies in the two ways in which they differ syntactically and morphologically from English. First, unlike English solsuch, it is possible to find sa!scldan and so/solch following as well as preceding the indefinite article. Secondly, unlike English, both languages have morphological agreement on adjectives. Danish, however, differs from German in that both predicative and attributive adjectives agree, whereas German adjectives only agree when they are attributive. It is thus possible to tell whether an adjective that has moved comes from an attributive position as predicted in (7) or a predicate position as predicted in (8). We would like to argue, on the basis of adjectival agreement morphology, that post-article expressions are represented by the base order of (7), a so bad hotel and pre-article ones are derived from predicate raising as in (8), from a hotel so bad to so bad a hotel. Another way in which a comparison between these three languages can be informative is when considering language change and the grammaticalization of degree expressions. Bolinger (1972: 92) suggests a semantic change, from kind to degree, whereas we will show, specifically with reference to German, that the significant change is in syntactic scope. i.e. whether the lexical items under consideration modifY the entire nominal or just the adjective phrase. Before the discussion of the derivation of nominals involving solsatso and those involving such!scldan!solch in Sections 3 and 4 below, Section 2 sets out our assumptions about the structure of the DP.
Note that we use pre-article so/such to refer to so/such preceding the indefinite article, as in (la) and (4b).
1.
91
92
Johanna L. Wood & Sten Vikner
2.
The structure of nominals
As was mentioned above, we will consider two alternative derivations for pre-article solsd/so and su.ch!sd.da.n!solch expressions, movement from the prenominal position or movement from the postnominal position, and we will argue for the latter. We asswne a structure of the DP in which the definite and indefinite articles occupy different positions as the head ofDP and the head ofNumP respectively, and we argue that so and such expressions that precede the indefinite article move to Spec-NumP. More detailed structures may be found in Sections 3 and 4. First, however, we will consider a third option and our reasons for not adopting it It has been argued that nominals with fronted degree modifiers and with such involve movement to Spec-DP. Alexiadou, Haegeman & Stavrou (2007: 108-9) observe the parallels between such and demonstratives, this and tha.t (below), and suggest that as a first approximation it can be proposed that such occupies Spec-D P: (9) I did not expect this reaction. I did not expect such a reaction.
(Alexiadou et al. 2007: 108, (63a)).
Demonstratives and such are supposed to have the same (derived) position as shown in ( 10) below: (10)
DP
~
Spec
D'
such
~
this
D"
NP
that
a
reaction
Similarly, Haegeman & Gueron (1999: 419-420) suggest that expressions with fronted degree modifiers are derived as in (7) above. but that the movement is to Spec-DP. However, there are two problems with such an analysis. The first problem is that expressions like such and so big can co-occur with another determiner, unlike this and that. Therefore if this and that are in Spec-D P, expressions like such and so big clearly are not: ( 11) *Any that a proposal ... (12) *No that a change ... ( 13) D'Cey feared he would not cut halfso good a figure. (The Iron wood TI-ee Tony DiTerlizzi. COCA) (14) I can at any rate promise you that I will not be the medium of any so a.bsu.rd a. requisition. (Anthony Trollope. Barchester Towers, 1857) (15) For the time being at least. no such a change in Congressional attitudes would occur. (Wood 2002: 110)
Noun phrase structure and movement
Admittedly, a structure such as (10) can account for (13-15) if an additional higher functional category, QP, is introduced (as pointed out to us by an anonymous reviewer). However, because (10) places so bad!su.ch in Spec-DP, this structure cannot possibly accommodate the definite article in (17)-(20): (16) so bad a hotel (17) the so bad hotel (18) An example based on the so impressive work of]ose Roca (www.powerbasic.com/ supporVforums) (19) Day at leisure to explore the so impressive city. (www.aegeantours.com.au/ south-america-highlights-tour) (20) A judge may receive compensation and reimbursement of expenses ... if the source of the such payments does not give.... (COCA)
In Sections 3 and 4 below we will discuss what we take the structures of the preceding examples to be, arguing that pre-article expressions are derived as in (8), rather than as in (7). We will propose that the indefinite article is the head ofNumP, as illustrated in (26) below, whereas the definite article is the head of DP. 2
3·
So constructions in English, Danish and German
In this section we describe the possible positions of so, sa and so with respect to the indefinite article and discuss two possible derivations. English so belongs to a class of degree adverbs that are part of the leftmost functional structure of the AdjP (Bresnan 1973, Carver 1997). The class includes the degree adverbs: how, as, too, this, that. This also holds for Danish: sa, for, hvor and German so, zu., wie. The tables below set out the logical possibilities in the three languages: (21)
pre-article modifying the whole DP/NP modifYing only the AdjP
a c
post-article
b d
2. In other words, we use pre-article so/such to refer to SlJ/such preceding the indefinite article in Nwno, whereas we use post-article SlJ/such to refer to so/such following the indefinite article in Num0 • This again means that when SlJ/such follows a definite article (which is in Do), as in (17)-(20), it may either precede or follow Num 0 •
93
94
Johanna L. Wood & Sten Vikner (22) English pre-article DP/NP AdjP
post-article
a. so
b.
'16so
c. d
*so a hotel *a so hotel so bad a hotel 11a SO bad hotel
(23) Danish pre-article DP/NP AdjP
si
post-article
sA
I
a. *sa et hotel b. *et sa hotel c. sa darligt et hotel d et sa darligt hotel
(24) German pre-article DP/NP AdjP
post-article
so
so
a. so ein Hotel b. *ein so Hotel c. *so schlecht ein Hotel d ein so schlechtes Hotel
As may be seen, English and Danish are similar, in that solsa only modifies an adjective phrase and has to be immediately adjacent to that phrase. The difference between English and Danish is in the possible word orders. In Danish it is grammatical for sa plus adjective to either precede or follow the indefinite article, whereas, in English, if native speakers accept (22d) at all it is with heavy stress on so. German so is similar to Danish sa. in that so plus adjective may follow the article. An important difference that separates English from Danish and German is the obligatory agreement between the ad jective and the (neuter) noun. The significance of this agreement morphology will become apparent in the following section when we discuss such!sadan!solch. The most striking semantic difference that separates German from the other two languages is that pre-article so is possible without a following adjective. This means that it is possible for so in German to modify the entire DP as well as the AdjP or NP, and hence German so can be ambiguous between kind and degree unlike English and Danish so and sa nominals which only have degree readings.
3.1
Deriving pre-article so
As has already been mentioned, there are two possible derivations that give the required surface structure. In one possible derivation of pre-article so, which we will not be adopting,
Noun phrase structure and movement
the adjective phrase so bad would be base-generated in the prototypical adjective position (ie. following the article and preceding the noun) and then move into Spec-NumP, as was described in (7) above, shown in (25) below (Matushansky 2002; Wood 2002, 2004): (25)
DP
~
D'
~
NumP
D"
~ AdjPk
Num'
DN~P sobig
a
~
~~
lp N'
I
N" hotel
An alternative derivation, which we preter, would be through predicate raising from the small clause, which has the same position that a relative clause would have, i.e. rightadjoined to the NP as in (8) above and shown in (26) below (e.g. Zamparelli 1995; Bennis, Corver & den Dikk.en 1998; Corver 1998; Wood 2002: 106): (26)
3.2
DP
~
Support for the predicate raising analysis
There is more evidence to support the derivation in (26) than the one in (25). Firstly, constructions with the surface word order of (26) are possible:
95
96
Johanna L. Wood & Sten Vikner (BNC) (27) She told me she never saw a man so pleased by a glass of wine. (28) It is rare to see a house so little altered. (BNC) (29) Maybe he was a man so obsessed with hidden things that he simply couldn't see open criminality. (BNC) (30) The imagination had become a power so terrible that he could not risk (BNC) indulging it These are possibly reduced relative clauses. The second piece of evidence comes from constructions such as ( 31) below, from Zamparelli (1995: 132-33), and (32), from COCA. Many dialects of English have an optional of in so +adjective constructions. (31) so tall (of) a man (that he had to lean on the basket) ( 32) I was so cold of a person at one time in my life; closed off all my own feelings. (COCA) This ofis not a case-assigning preposition, nor a possessive ofor a partitive of It is similar to the meaningless element, Dutch van, argued by Bennis, Carver & den Dikken (1998: 86) to be a functional head parallel to the clausal copula as in the Dutch examples below: ( 33) De grootste beer is die kerel the biggest bear is that guy
(34) een beer 1'an een kerel a bear of a guy Similar constructions are possible in English and Danish: ( 35) The island is a jewel. (36) a jewel of an island
(37) Sta.tsministeren i tegneserien 1'a.r en lille na.r. Prime-minister-the in cartoon-the was a little fool (38) en lille na.r af en statsminister a little fool of a prime-minister
(KorpusDK)
A third piece of evidence is that in English, adjectives that are usually only predicative (e.g.
alike, awake, ashamed, upset, afloat, alike, alone) appear to be better in these constructions than when they occur in the canonical attributive position. Comparing the b and c examples in (39) and (40) below it may be seen that the c examples are grammatical (39) a. The shoplifter is ashamed. b. *the ashamed shoplifter c. the so ashamed shoplifter d. So ashamed a shoplifter I have never seen. (40) a. The twins are alike. b. *the alike twins c. the so alike twins
Noun phrase structure and movement
If the c examples are derived through predicate raising it would explain why they are grammatical, whereas the b examples are not A fourth piece of evidence is provided by Lenerz & Lohnstein (2004: 83), who point out that although (41b) is ungrammaticaL native speakers nevertheless have clear intuitions of what the form should be, not only uninflected (as predicate adjectives are in German, see e.g. Vikner 2001) but also in the predicate form, hoch. Therefore, although both are ungrammatical, (41b) is better than (41a). (41) a. *so *hohein Haus so high.ATTR a house b. *so hoch ein Haus so high.PRED a house We argued above that the predicate raising analysis is the preferred analysis for expressions involving pre-article so, sa. and so. In the following section, we go on to compare in detail the possible positions of such, sd.dan and solch with respect to the indefinite article and whether the D P/NP or the Adj P is modified in each of the three languages. We then argue that also for pre-article such, sadan and solch expressions, the predicate raising analysis is preferred over fronting from attributive position.
4· Such constructions in English, Danish and German Also for such/stlda.nlsolch there are four possible combinations ofword order and modification. Such obligatorily precedes the article in present-day English (although, as pointed out in Wood (2004: 31 5), in earlier English and in some dialects post-article such is found). However, in both Danish and German, both pre- and post-article orders are found: (42) pre-article modifying the whole DP/NP modifying only the AdjP
post-article
b d
a c
(43) English pre-article DP/NP AdjP
such
post-article
a. b.
c. d
such a hotel *a such hotel *such bad a hotel ..a such bad hotel
97
98
Johanna L. Wood & Sten Vikner (44) Danish pre-article DP/NP AdjP
sidan(t)
post-article sl\dant
a.
sadan(t) et hotel
b. et sadant hotel c. *sadan(t) darligt et hotel d *et sddan(t) darligt hotel
(45) German pre-article DP/NP AdjP
solch
post-article solches solch
solch ein Hotel b. ein solches Hotel c. *solch(es) schlecht(es) ein Hotel d ein solch schlechtes Hotel a.
As may be seen in ( 43) above, English is the most restrictive as far as word order goes, because such must precede the article, ( 43a). As in German and Danish, pre-article such in English modifies the entire DP; it cannot just modify an adjective. Danish allows both a pre-article sadan, which modifies the DP, (44a), and a post-article sadan, (44b), which modifies the NP. Post-article sadan is always inflected, whereas pre-article sadan may or may not be inflected. German is the least restrictive. Like English and Danish, it has a pre-article solch, which modifies the DP, (45a), but in addition, it has two different forms of post-article solch. Inflected post-article solch modifies the NP, (45b), whereas uninflected post-article solch modifies an AdjP, (45d).
4.1
German post-article solch
Before we focus on the derivation of pre-article such constructions, a few more words about German post-article solch. For ( 45d), there might seem to be two possibilities, as solch could also have been inflected in this position (i.e. following an article and preceding an adjective). However, when solch is inflected in this position, it modifies the NP, and so this would really be a subcase of (45b ). Therefore ( 45d), solch in the postarticle position modifying an AdjP, only exists in one version, uninflected, which, we suggest, is an adverb, not an adjective (see Section 5 below). In other words, when post-article solch modifies the NP, it inflects like an adjective: (46) ... was eituolcher Kunstler for eine Gage bekommt whata such artist for a fee receives
However, when post-article solch modifies an adjective, it is never inflected:
Noun phrase structure and movement
(47)
... mit
einetn solch groflet' Aufwand
with a.DAT such big.DAT effort
(Fabridus- Hansen et al. 2005: 330)
solch dickes Buch meine (48) Selten hat ein Seldom has a.NOM such thick.nom book my Aufmerksamkeit mehr gefesselt als dieses. attention
more captured than this. (Fabridus- Hansen et al. 2005: 332)
This difference is parallel to the difference between attributive adjectives (inflected) and adverbs (uninflected) in German: (49)
Bin unglaubliches
Buch
An incredible.NEUT book
(SO) Bin unglaublich_gutes Buch An incredibly good book Deriving pre-article such
4.2
One possible derivation of pre-article such is for it to be base-generated in the prototypical adjective position (i.e. following the article and preceding the noun) and then to be moved to a position preceding the indefinite article. One implementation based on Ritter (1992) was suggested by Wood (2002) (see also Matushansky 2002: 48 ). Here, such moves from the prototypical attributive adjective position, which is left-adjoined to the NP~ (51)
DP
~
D'
~
D"
NumP
(no)
~
such
t
a
4.
I
NP
I I N" N'
hotel
3·
We are abstracting away from the internal structure of the adjective phrase (e.g. Bresnan
1973, Corver 1997, Wood 2002).
99
100
Johanna L. Wood & Sten Vikner Another possible derivation is predicate raising, as in Wood (2002: 112 ), who takes the idea from Zamparelli (1995) and from Bennis, Carver & den Dillin's (1998) derivation of wat voor constructions. The main idea is that such originates as a predicate in a verbless clause (a so-called small clause). 4 1his small clause has the same position that a relative clause would have, i.e. right-adjoined to the NP.
(52) [small aause [se-subje..1: teachers] [se-predica2 su.ch as these]] is what every student fears. (53)
DP
~
D'
~
o•
NumP
tiO
~
ati.V
AdjPk
Num'
6N~P such
a
~
NP
SC
6~k hotel
4·3
German pre-article solch
Deriving German pre-article solch from a predicate as illustrated in (53) would account for why it is never inflected (similar to German manch ein "many a" and welch ein "what (a)" given that predicate adjectives are never inflected in German, as opposed to Danish, cf. e.g. Vikner (2001): 5
4· We have to admit that although we derive pre-article such from a predicate, it would appear that such is only posslble as a predicate in small clauses and not in finite clauses: (i) teachers such as these (are what every student fears). (ii) ~No teachers are such as these. 5· M pointed out by Harry Perridon (p.c.), German may leave pre-article all uninflected. Thls is not posslble in Danish, where infiection on alie is obligatory. (i) a mit all diesen Buchern GERMAN b. *med al disse wger DANISH with all_ these books (ii) a ... mit allen diesen Buchen~ GERMAN b. med alle dlsse wger DANISH ... with all.PL these books
Noun phrase structure and movement
(54) Dann wird es klar, wie klug und peinlich genau solch ei" Kunstler wie Wagner
war. Then it becomes clear how intelligent and pedantically exact such an artist as Wagner was. (Fabricius- Hansen et al. 2005: 331). (55) a.
Wo findet man solch ein Hotel?
German
b. * Wo findet man solches ein Hotel? where finds one such a hotel? (56) a.
Bin Bus ist grii.n_, die anderen sind gelb_.
b. En bus er gr€1n_, de Andre er gule. one bus.MASC.COM is green, the others.PL are yellow. (57) a.
Bin Haus ist grii.n_, die anderen sind gelb_.
b. Et hus er grent, de andre er gule. one house.NEUT is green, the others are yellow. 4·4
German Danish German Danish
Danish pre-article sadan
The derivation of Danish pre-article sadan from a predicate as illustrated in (53) is at first glance less obvious than the parallel derivation for German pre-article solch. Given that ( 57b) indicates that Danish predicative adjectives always agree, predicative sddan does not appear to behave as a prototypical adjective, because adding the agreement morpheme-tin the neuter is only a rarely used option, not obligatory. This is in sharp contrast with Danish attributive adjectives, where adding the agreement -t in the neuter is very much the preferred option. It is therefore not surprising that prearticle sadan, derived, we argue, from a predicate, may add -tin the neuter, but most often it does not, (44a). This is completely in line with it being derived from a predicate since, if pre-article sddan were derived from an attributive, it would be expected to always add the agreement -tin the neuter. This analysis receives further support from the following corpus data. First, consider the following figures for pre-article sadan in Danish (KorpusDK, November 2009): (58)
a. sddant et + NOUN b. sddan_ e_! + NOT.JN
~ ~
( such.NTR a.NTR + NOUN) (such.COM a.NTR +NOUN)
Although the possibility ofleaving all uninflected in German is reminiscent of pre-article solch, it is not clear to us whether (ia) could or should be derived from a predicative structure, given that all is not a very likely candidate for a predicate.
101
102.
Johanna L. Wood & Sten Vikner
(59) a. sti.da.nt et + ADJ +NOUN
b. sadan_ e[ + ADJ +NOUN
~ (such.NTR a.NTR + ADJ +NOUN) ~ (such.COM a.NTR + ADJ + NOUN)
Pre-article sadan always modifies a DP, and regardless of whether this D P contains an AdjP, (59), or not, (58), adding the agreement -t to sada.n in the neuter is possible, but clearly not preferred (the b-figures are very much bigger than the a-figures). Compare these figures to the following figures from a search in Itifomedia, a newspaper database, of Danish sadan as a predicate (these strings are not sufficiently frequent to show up in KorpusDK): 6 (60) a. deter sadant- at .. -
D
(it.NTR is such.NTR that ... )
b. de!_ er sadan_ at ..
~
(it.NTR is such.COM that ... )
This shows that predicative sadan may add the agreement -t, but like pre-article sadan, the non-inflected form, (60b), is very much more frequent than the inflected one, (60a). There is thus a sharp contrast with the figures for post-article stldan, (KorpusDK, November 2009): (61)
a. etsadant+NOUN -
b. e! sadan_ +NOUN
~
G
(a.NTR such.NTR +NOUN) (a.NTR such.COM +NOUN)
(62)
a. e! sti.da.n!
+ ADJ +NOUN
~
(a.NTR such.NTR + ADJ +NOUN)
b. e! sadan_
+ ADJ + NOUN
[3
(a.NTR such.coM + ADJ + NOUN)
These figures show that adding the agreement -t to post-article sadan in the neuter is not only possible, but clearly also preferred (the b-figures are very much smaller than the a-figures). In sum, the figures cited here support our analysis that pre-article sadan is derived from a small clause predicate (both only add the agreement - t in the neuter in a small minority of the total cases) and not from post-articlesada.n (which adds the agreement -t in the neuter in a large majority of the total cases).
6. The search was carried out In July 2010. The results were checked for mistakes and repetitions. For (60b), the figures have been estimated based on the first 100 returns. Note that a Google search Ouly 2010) returned 6,030 cases for (60a) (compared to the 488,000 cases returned by (60b)). This is support for the 3 occurrences In (60a) representing a real option.
Noun phrase structure and movement 103 Having established our arguments for deriving the pre-article expressions in Section 3 and 4 above we now summarize, and compare the morphology of the pre-article expressions with post-article expressions in Section 5 below.
5·
Summary of the derivations
Shown below is a summary of the possible word order and agreement morphology in the three languages: (63) pre-article
post-article
b d
a c
modifying the whole DP/NP modifying only the AdjP (64) English pre-article DP/NP AdjP
post-article
such
pre-article
I I
so
post-article so
(65) Danish pre-article sidan(t)
DP/NP AdjP
post-article
pre-article
post-article
sl\dant
I I
sl\
'*'sA
(66) German
DP/NP AdjP
I
pre-article solch
post-article solches solch
pre-article
I
I
post-article
so so
Although it might be expected that pre-article expressions are somehow derived from the post-article ones, we suggest that this is not the case. In the post-article versions the so, sa, so and such, sddan, solch expressions are generated as attributes as in (67) and in the pre-article ones they are generated as predicates as in (68).
104
Johanna L. Wood & Sten Vikner (67)
DP
~
D'
~
o•
NumP
(no)
~ Nwn'
(ar1y)
~
Num•
NP
~ : NP
a ••••• AdjP
6 such
~
::
so good i •.........•
(68)
J, I N" hotel
DP
~
D'
~
o•
NumP
(no> •••• (any):
~
AdjPk :
Num'
:/\:~ =~=Num" NP : :a ~ • such • i ~.~: m ~ I
I
'·T····
o 1·
As (67) shows, (63b,d) (i.e. post-article so/such) are base-generated in situ, and inflected as are all attributive adjectives in most Germanic languages. (67) does not give rise to (63a,c), presumably because such a movement would be a left branch constraint violation. As (68) shows, we instead derive (63a,c) (i.e. pre-article so/such) by a movement out of a postnominal predicate position, which is why pre-article so/such are inflected only in languages where predicative adjectives are inflected (in Germanic, these are only the YO-languages). Thus, the reason why German pre-article solch!so-phrases (i.e., 63a,c) are never inflected is that they are derived from a predicate. The reason why the post-article solch that modifies an AdjP (i.e., 63d) is not inflected is that it is an adverb, not an adjective. In Danish, there is no post-article sdda1t that modifies an
Noun phrase structure and movement adjective (i.e., 63d). Danish pre-article sddan (i.e., 63a,c) is peculiar in that it may, but most often doesn't, show agreement with the NP. This is completely in line with it being derived from a predicate, because also with predicative sadan, agreement is only a rarely used option, not a must (cf. 58-60 above).The structure in (68) is fully compatible with complicated examples such as (69), where such follows any but precedes a, as any is in D 0 and a is inN um0 : (69) on the basis of any such a proposal or application form Having discussed the structure and derivation of so/such in the three languages we now examine, mainly with reference to German, changes that have occurred and are occurring with so, with su.ch, and in the interaction between so and such.
6. Language change As was mentioned in Section 1, Bolinger (1972: 92) noted, that for English, the direction of change is from kind to degree. This is borne out by, for example, considering the semantics of English this and that, where the earliest attested examples with a degree reading are from the 17th century, whereas this and that have a much longer history as demonstrative determiners. However, as will be shown below, the significant factor is syntactic, whether or not an element can modify the DP/NP (as with German so) and whether or not it can modify an adjective phrase. The kind and degree readings then fall out from whether the modified nouns and adjectives are gradable, as seen below.
6.1
German so and grammaticalization
As was noted above, German is the only one of the three languages in which so may precede the indefinite article without an adjective, i.e. so ein Hotel in (24a). It is also the only one of the languages in which so may have a kind reading as well as a degree reading. As long as the noun is not gradable, (70a), or the noun plus adjective is not gradable, (70b ), the kind reading is the only one available. (70) a. Mit so einem Ergebnis hat niemand von uns gerechnet with so a result has no one of us counted (Die Zeit, 08.03.2010, www.zeitde) b. Jetzt steht so ei, medizi,isches Groflgerat zur now stands such a medical large-instrument for the
Untersuchung von Patienten itt fast examination of
jeder grofleren Klinik.
patients in almost every larger clinic (Die Zeit, 02.05.2002, www.zeitde)
However, as soon as there is a gradable noun, (71a), or a gradable adjective, (71b), in the DP/NP, it is possible for the reading to be ambiguous between kind and degree.
105
106
Johanna L. Wood & Sten Vikner
(71) a. Es ist so eine Verschwe,du,g. it is such a waste
(Die Zeit, 16.07.1998, www.zeit.de)
b. So ein breites Repertoire finde ich nicht in jeder Epoche. so a wide selection find I not in every period (Die Zeit, 02.02.2010, www.zeit.de) It appears that a change has taken place in German in which so has added to its functions. The grammaticalization of German so to a eli ticized determiner is evident from the observation that (72b) is even possible in the plural, so that as ein becomes more grammaticalized, it loses its singular feature (Hole & Klump 2000, Lenerz & Lohnstein 2004). (72) a.
So'n
Buch wurde ich nie
so-a.sG book would I
lesen
never read
b. So'ne Bacher wurde ich nie lesen So-a.PL books would I never read Although in (22) above we have indicated that pre-article so, as in (22a), is ungrammatical in English, it may be acceptable in colloquial PDE as seen in (73) and (74). Although both examples involve a gradable adjective, the position of so makes a kind reading possible? (73) I am delighted that there has been so a good response to the park and ride scheme in its first week. (http://www.carmarthenshire.gov.uk/English/news) (74) And that's why Hulu has been so a. huge success. (http:/!blog.brandexperiencelab.orglexperience_manitesto/2009/011) Something similar with pre-article so may be seen in Middle English: (75) 'Iher roose so a grete torment in the see. there arose such a great torment in the sea (OED: 1471 Caxton 'The Recuyell Of 'The Historyes Of Troy)
The change that appears to be happening with German so (and marginally in colloquial English) is from a degree adverb to a cliticized determiner, resulting in a kind reading as well as degree. This is not the direction of change predicted by Bolinger but it is completely in line with recent theories of grammaticalization within the generative framework in which grammaticalized elements move up the tree, i.e. widen their scope (van Gelderen 2004: 29). In addition, if further investigation of the Middle English data reveals that earlier stages of the language permitted modification of the entire DP, we could be looking at a cyclic change (cf. van Gelderen 2009 ).
7· It must be admitted that these examples are from the Internet and perhaps not the most solid of data.
Noun phrase structure and movement German solch and grammaticalization
6.2
In addition to the changes affecting German so, Germansolch is also changing. According to Zifonun et al. (1997: 1936), solch is changing from a determiner to an adjective. In German, the morphology gives an indication of the word class, and the fact that solch can be followed by a weak adjective indicates that it is a determiner, whereas the fact that it can also be tollowed by a strong adjective indicates that it may be an adjective: (76) a. fUr solche neuen Wege for such.STR new.WK approaches b. fur solche neue Wege for such.STR new.STR approaches
(Fabricius- Hansen et al. 2005: 330)
A second change affecting German solch involves pre-article solch. Recall that pre-article solch modifies the entire DP and is not inflected as in (54), repeated below:
(77) Dann wird es klar, wie klug u.nd peinlich genau solch Then it becomes clear how intelligent and pedantically exact such ein Kunstler wie Wagner war. an artist as Wagner was. (Fabricius- Hansen et al. 200 5: 331 ). This use ofsolch is archaic, and is being replaced in more colloquial registers by so and so'n, cf. (72a) and (72b) above. However, this is not the only place where solch is being replaced by so. In examples such as (45d), post-article solch in ein solch schlechtes Hotel, i.e. the adverbial use of solch is giving way to (24d), ein so schlechtes Hotel, the degree adverb so, i.e., here also solch is being replaced by so. 8 A widely accepted fact regarding language change is that change starts in colloquial registers and more formal registers are the most resistant to change. In German it is apparent that both the pre-article adjective solch and the post-article adverb solch are used in more formal registers and so is used in colloquial registers, indicating the replacement of solch by so. A similar observation is made regarding English such. Both Bolinger (1972: 62) and Altenberg (1994: 238-9) point out that English such with a kind reading is more archaic than such with a degree reading. The above discussion reveals that changes involving these two words are interdependent and Bolinger's observation that kind readings tend to change to degree readings is only a small part of the story. Further investigations into grammaticalization 8.
It should be noted that German uses so and not such ln predicates, whereas Danlsh has such: a Es 1st so, dass ... German
(i)
b.
*Det er sd at it
(ii) a
b.
is
Danish
so that ...
*Es 1st solch, dass ... Det is sddan at it is such that ...
German Danish
107
108
Johanna L. Wood & Sten Vikner
and the interaction between these two words are likely to be informative in the investigation of possible directions and processes in language change.
7·
Conclusion
We have aimed to clarify the possible readings of these two etymologically related words in three languages and found that English such respects a more restricted word order than its Danish and German counterparts. Although it might be expected that of the two derivations, there might be a difference between the derivation of kind and degree readings (as suggested by Wood 2002), we have shown that the morphological evidence gives support for pre-article word orders being predicates. We have seen that German solch modifying a DP is archaic (and formal in English), which lends support to Bolinger's suggestion that the direction of change is from the modification of DP/ NP to the modification of AdjP (in his terms: from kind to degree). The interaction between the such forms and the so forms in the three languages is complex and much work remains to be done.
Sources British National Corpus (BNC): http://corpus.byu.edu/bnc/ Corpus of Contemporary American English (COCA): www.americancorpus.org/ Infomedia: www.infomedla.dk KorpusDK: http://ordnet.dk/korpusdk
References Abney, Stephen 1987. The English Noun Phrase In its Sentential Aspect PhD dissertation, MIT. Alexiadou, Artemis, Haegeman. Liliane & Stavrou, Melita. 2007. Noun Phrase In the Generative Perspective. Berlin: Mouton de Gruyter. Altenberg, Bengt. 1994. On the functions of such in spoken and written English. In Corpus Based Research into Language, Nelleke Oostdijk & Pieter de Haan (eds), 223-240. Amsterdam: Rodopi. Bennis, Hans, Corver, Norbert & den Dikken, Marcel1998. Predication in nominal phrases. Journal of Comparative Germanic Linguistics 1: 85-117. Bolinger, Dwight. 1972. Degree Words. The Hague: Mouton. Bresnan, Joan. 1973. Syntax of the comparative clause construction. Linguistic b1quiry 4(3): 273-349. Corver, Norbert 1997. Much-support as a last resort. Linguistic Inquiry 28: 119-164.
Noun phrase structure and movement Corver, Norbert. 1998. Predicate movement in pseudopartitive constructions. In Possessors, Predicates andMovementin the Determiner Phrase [LinguistikAktuellJLinguistics Today22], ArtemisAlexiadou &Chris Wilder (eds), 215-257. Amsterdam: John Benjamins. Fabridus-Hansen, Cathrine, Gallmann, Peter, Eisenberg, Peter, Fiehler, Jorg Peters, Reinhard, Niibling, Damaris, Barz, Irmhlld & Fritz, Thomas A. 2005. Grammatik der Deutschen GegeJ~wartssprache, Duden, Vol. 4, 7th edn. Mannheim: Bibliographisches Institut. van Gelderen, Elly. 2004. Grammaticalisation as Economy [Linguistlk Aktuell/Linguistics Today 71]. Amsterdam: John Benjamins. van Gelderen,Elly. 2009. Cyclical Change [LinguistikAktuellJLinguistics Today 146]. Amsterdam: John Benjamins. Haegeman, Liliane & Gueron, Jacqueline. 1999. English Grammar: A Generative Perspective. Oxford: Blackwell Hole, Daniel & Klump, Gerson. 2000. Definite type and indefinite token: The article son in colloquial German. Li11gulstische Berichte 182: 231-144. Kennedy, Chris & Merchant, Jason. 2000. Attributive comparative deletion. Natural La11guage a11dLinguistic Theory 18:89-146. Lenerz, Jiirgen & Lohnstein, Horst. 2004. Nur so - Strukturaspekte der Vergleichskonstruktion. In Deutsche Sy11ta:~.~ Empirie u,ld Theorie. Symposium Giiteborg. 13-15 Mal 200 [Acta Universitatis Gothoburgensis 4], Franz Josef
Wood, Johanna L. 2002. Much about such. Studia Linguistlca 56: 91-115. Wood, Johanna L. 2004. Number phrase and fronted pre- modifiers in Middle English. Proceedlngsofthe 32nd Western Conference on Linguistics, 305-318. Fresno,CA: DeptofLinguistics, California State University. Zamparelli, Roberto. 1995. Layers in the Determiner Phrase. PhD dissertation, University of Rochester NY. Zifonun, Gisela, Hoffmann, Ludger & Strecker, Bruno (eds). 1997. GrammaNk der Deutschen sprache. Berlin: Mouton de Gruyter.
109
A unified structure for Scandinavian DPs* Susanne Lohrmann Universitiit Stuttgart
This paper discusses doubling patterns of determiners in Scandinavian, the interaction between definite markers and adjectival inflection, and its effects on the structure of the DP. When an adjective precedes a definite noun an additional article is added preceding the adjective. This phenomenon has been referred to as 'double definiteness: Moreover, the adjective receives definite inflection so that the noun is actually defined three times. Questions arise as to the function of this multiple exponence of definiteness and the syntactic/ semantic role of the morphemes involved. This article develops an account that argues for three components of definiteness in Scandinavian expressed by three distinct morphemes (the two articles and the adjectival inflection). The result is an analysis, based on the framework of Distributed Morphology, which captures the different realizations of definiteness in the Standard Scandinavian languages and which shows that multiple exponence in Scandinavian DPs evidently contributes to interpretation.
1.
Introduction
Adjectives in definite DPs in Swedish, Norwegian, and Faroese trigger an additional lexical determiner. If the variation in terms of usage of the articles is considered, it shows that the different determiners encode different aspects of definiteness. These are commonly referred to as 'inclusiveness' and 'specificity' (Julien 2005, among others). In a number of cases, one of the determiners is absent and different readings are obtained. The presence or absence of weak adjectival inflection can also yield different readings, that is, inflection interacts with interpretation. Questions that arise are: (i) What is the function of the doubling pattern of determiners? Is this a mere agreement phenomenon or is there a semantic correlate? (ii) What is the function of the adjectival inflection? and (iii) What does this tell us about the structure of the DP?
*
I would like to thank Artemis Alexladou, the participants of the conference Variation and Change and two anonymous reviewers for their valuable help, comments and discussion. All errors are mine.
112
Susanne Lohrmann
After a brief introduction to double definiteness I will show how the different morphemes involved constitute the notion of definiteness in Scandinavian. I suggest that definiteness in Scandinavian DPs is made up of three components, 'discourse reference!, 'specific reference, and 'identitY, which are expressed by three distinct morphemes: the preadjectival article, the suffixed article, and the adjectival inflection respectively.1 In the last section I will then point to a possible syntactic analysis of the patterns found in the different Scandinavian languages. 2
1.1
Basic data
Standard Swedish, Norwegian and Faroese very much pattern alike with regard to double definiteness. (1) a. film-en
film-DEF "the film" b. den rolig-a film-en DEF funny-W film-DEF "the funny film"
(Swedish)
In non-modified DPs, the definite article is attached to the noun (la). A second article appears when an attributive adjective modifies the DP (lb). The adjective is marked with the so-called weak inflection. This type of double definiteness is the default structure, irrespective of the type of adjective. The preadjectival article is triggered only once, further adjectives do not trigger additional determiners. Neither Danish nor Icelandic show structures involving double definiteness. The article in non-modified definite DPs is attached to the noun (2), as in the other Scandinavian languages. With respect to adjectival modification, the languages differ: in Danish (3a), a separate article is introduced preceding the adjective and the suffixed article is omitted. In Icelandic, the suffixed article is retained and no further article is
The morphological status of the bound article has been a matter of debate. Since this ls not relevant to the discussion here, I refer to, for example, Faarlund (2007) and Borjars & Harries (2008).
t.
2. Note that there is not always a one-to-one mapping regarding the functions of the functional mmphemes in all Scandinavian languages. For instance, there is not necessarily the same variation in adjectival inflection to the same extent in Swedish as in Norwegian. However, there are instances of adjectival variation in other contexts in Swedlsh that point into the same direction as in Norwegian. A detailed analysis of the respective languages has shown that it ls tenable to deal with the Scandinavian languages in one go (for an extended discussion see Lohrmann 2010).
A unlfied structure for Scandinavian DPs 113
introduced (3b ), i.e. in both languages, the suffixed and the preadjectival articles occur in complementary distribution.3 (2) a.
hus-et house-DEF "the house"
(Danish)
b. ht:ts-iJ house-DEF "the house"
(Icelandic)
(3) a.
det stor-e hus DEF big-w house "the big house"
(Danish)
b. gaml-a ht:ts-ilJ old-w house-DEF "the old house"
(Icelandic)
In all Standard Scandinavian languages adjectives are inflected. Attributive adjectives show weak or strong inflection, the form of which is determined by semantic aspects: the weak form is chosen if the modified DP is definite, the strong form if the DP is indefinite. (4) a.
den gran-a
bil-en
DEF green-w car-DEF "the green car" b. en gron-e bil a green-s car "a green car" c.
de
stor-a hus-en
DEF.PL big-W house-DEF.PL "the big houses"
(Swedish)
If dialectal variation is taken into account, the phenomenon becomes much more complex. In Icelandic, for instance, strong adjectival inflection can be combined with definite contexts to achieve a non-restrictive reading. The variation in the realization of adjectival inflection leads to questions regarding the meaning and function of ad jectival inflection, even more so if it is considered that adjectival inflection can interact with meaning (see Section 4.3).
3· In Icelandic, the definite article may also be placed preceding the adjective: (h)il1 gamla hU5 "DEF old-W house': However, the Icelandic native speakers participating in this research preferred that structure in only one particular example and otherwise exclusively used N + DEF.
114 Susanne Lohrmann
Swedish, Norwegian, and Faroese
2.
Optionality or elimination of one of the articles
2.1
In order to get a better understanding of the function of the respective articles, those cases are of interest in which either the preadjectival article or the suffixed article is optional or even obsolete. If double definiteness is not a mere agreement phenomenon but is of interpretive value, then a difference in meaning is predicted for DPs that do not exhibit the default structure. This prediction is borne out. as the following sections show. 2.LI
The suffixed article
If the suffixed article is omitted in Norwegian, the result is an abstract reading (Sa). Here, reference is not made to a particular school but to a teacher who is one of the old school for example, in his way of teaching. However, if the intention is to refer to a particular building, that is, if a concrete reading is intended, then the suffixed article is obligatory (Sb). Examples (5) and (6) are from Julien (2005). (5) a. Han er en lrerer av den gaml-e skole(-n). he is a teacher of DEF old-w school-DEF "He is a teacher of the old school."
b. Vi sa
paden gaml-e skole*(-n)
we saw at DEF old-w school-DEF "We looked at the old school." (6) a.
Dei oppforer seg
som dei vet-st-e
brall-a.r
they behave REFL as DEF worst-w brute-PL "They behave like the worst brutes"
b. Dei oppforer seg
som dei verst-e
b"ll-a-ne
they behave REFL as DEF worst-w brute-PL-DEF "They behave like the worst brutes" (7) Han uppfOr sig som den varst-e buse he behaves REFL as DEF worst-w brute "He behaves like the worst brute" A similar contrast is shown in the Norwegian phrase in (6), and for Swedish in (7).4 Both examples contain so-called 'absolute superlatives~ which do not express comparison but a very high degree of the quality contributed by the adjective. In the example without the suffixed article (6a), the reading is non-referential that is, the speaker does not know who those people are. In (6b), where the suffixed article is present, the speaker refers to specific people and a referential reading is obtained.
4·
Example (7) has been contributed by an anonymous reviewer.
A unlfied structure for Scandinavian DPs 115
2.L2
The preadjectival article
Julien (2005) notes that even if the referent of the DP in (Sa) has not been mentioned before, it is clear that there must be "a new car in the universe of discourse" (Julien 2005: 33). The structure without the preadjectival article (8b) would rather be used if the referent is very fiuniliar, that is to say if the people involved in the discourse already know about the new car. (8)
a.
Du kan ta
den ny-e
bil-en.
you can take DEF new-w car-DEF "You can take the new car."
b. Du. kan ta
ny-e
bil-en.
you can take new-w car-DEF "You can take the new car."
(Norwegian, Julien 2005)
3· Danish and Icelandic The examples in (9) and (10) show that different readings depending on the use of the suffixed article (abstract vs. concrete, as obtained in the equivalent sentences in Swedish and Norwegian) can only be achieved by context (9) a.
Han er kennari af gaml-a sk6la-num. he is teacher of old-w school-DEF "He is a teacher of the old school."
b. VilJ horfoum
a
gaml-a sk6la-nn.
we looked at old-w school-DEF "We looked at the old school." (10)
a.
(Icelandic)
Han er en lcerere af den gaml-e skole. He is a teacher of DEF old-w school. "He is a teacher of the old school."
b. Vi sa
paden gaml-e skole.
we saw at DEF old-w school "We looked at the old school."
(Danish)
The same holds for differences in referentiality. The equivalent sentences in Danish and Icelandic are ambiguous (11). (11) a.
De
opferte sig
som de vcerst-e ooll-er.
they behaved REFL like the worst-w brute-PLU "They behaved like the worst brutes." b. Peir hoglJu ser eins og hin verst-u 6hrces-i. they behaved REFL like the worst-w brute-PLU "They behaved like the worst brutes."
(Danish)
(Icelandic)
116
Susanne Lohrmann
With respect to different readings in Swedish and Norwegian depending on the use of the preadjectival article, Danish and Icelandic remain ambiguous, too: (12)
a.
Du kan tage den ny-e
bil
you can take DEF new-w car "You can take the new car."
(Danish)
b. Pu getur tekilJ njj-a bil-inn you can take new-w car-DEF "You can take the new car."
(Icelandic)
Whether the car is familiar or not, the definite article in (12) is obligatory in Danish as well as in Icelandic.
The semantics of the articles and the adjectival inflection
4· 4.1
Swedish, Norwegian, and Faroese
4.1.1
The suffixed article
Julien (2005) suggests that the semantic content of the suffixed article is 'specificity': the suffixed article can be omitted if a non-specific reading is intended, and a specific reading is only possible if the suffixed article is spelled out. I agree with Julien in large part but consider the term specificity not unproblematic (indefinite DPs can also be specific) and suggest extending the term to 'specific reference: thus including that the denotation ofN + DEF yields a referential reading and that the denotation is identifiable and locatable for the hearer. Example (6), repeated here as (13), illustrates this point: in the example without the suffixed article (13a), the reading is non-referential. In ( 13b ), the speaker refers to particular people and a referential reading is obtained. This observation is supported by restrictive relative clauses. Here, too, the suffixed article is redundant (14). 5 Since restrictive relative clauses limit and specify the denotation ofN and always yield concrete readings, the suffixed article is superfluous. (13) a.
Dei oppforer seg
som dei verst-e
they behave REFL as
bell-ar
DEF worst-w brute-PL
b. Dei oppforer seg som dei verst-e bell-a-ne they behave REFL as DEF worst-w brute-PL-DEF "They behave like the worst brutes"
5· The so-called 'determinative pronoun' de is introduced preceding the noun. Since the determinative is always stressed unless followed by an adjective, I assume that it functions like a demonstrative when stressed, but that in the case of adjectival modification the preadjectival article is triggered rather than the determinative kept
A unlfied structure for Scandinavian DPs 117 (14) De turist-er som akte till Island fick mycket sol. DET tourist-PL who drove to Iceland got a lot of sun. "The tourists who went to Iceland got a lot of sun:' (Swedish, Holmes & Hinchliffe 2003) 4.1.2 The preadjectival article The share the preadjectival article has in the notion of definiteness is commonly called 'inclusiveness' (cf. Hawkins 1978, 4"ons 1999, Julien 2005). Inclusiveness assimilates 'uniqueness' and is meant to express reference "to the totality of the entities that satisfy the description" (Lyons 1999: 11), i.e., mass and plural nouns are included. In my opinion, this definition does not cover the function of the preadjectival article. I suggest that the preadjectival article introduces a 'discourse referent' that contains a new 'discourse variable'. In other words, what the preadjectival article does is signal that a new modified definite N is entering the discourse. Example (8), repeated as (15), supports this view: since the new car in (15b) is a familiar entity for those involved in the discourse, there is no need to introduce it as a new discourse variable.
(15) a.
Du kan ta
den ny-e
bil-en.
you can take DEF new-w car-DEF "You can take the new car."
b. Du kan ta ny-e bil-en. you can take new-w car-DEF "You can take the new car." 4.2 Danish and Icelandic
As opposed to Swedish and Norwegian. there is no difference between an abstract/ concrete or referentiaVnon-reterential reading in the equivalent Danish and Icelandic examples. This is not surprising, since the two definite articles in Swedish and Norwegian seem to contribute different aspects of definiteness. These aspects appear to fall together in one morpheme in Danish and Icelandic.
4·3 The adjectival inflection The two sentences in (16) are identical apart from the adjectival inflection. (16a) shows the strong ending, (16b) the weak one. However, the meaning differs: (16a) is not presuppositional so that it is not clear whether there are any unripe apples at all, whereas in ( 16b) the reading is presuppositional, i.e., there is at least one unripe apple. The presuppositional reading is rendered by the weak adjectival ending only and this suggests that the weak adjectival ending individuates the relevant members in the A+ N denotation.
us
Susanne Lohrmann (16) a.
Legg hvert umoden-t eple i
denne kassen.
put every unripe-s apple in this "Put every unripe apple in this box:'
box-DEF
b. Legg hvert umodn-e eple i denne kassen. put every unripe-w apple in this box-DEF "Put each unripe apple in this box". (Norwegian, Vangsnes 2007) If it is correct that the weak adjectival ending states the existence of the A + N denotation, then we should not find this ending if existence is stated otherwise. The adjective egen "own" might be such a case. (17) a.
den egn-a
tm11-a.n
DEF own-w garden-DEF "one's own garden"
b. hans egen-e hemlighet his own-s secret "his own secret" c.
deras ege-t fin-a
hus
their own-s fine- w house "their own fine house"
d. hans egn-a
upptriidande
his peculiar-w behaviour "his peculiar behavior"
(Swedish, Holmes & Hinchliffe 2003)
( 17a) displays the default structure: double definiteness plus weak adjectival inflection. In ( 17b ), egen follows a possessive pronoun and shows strong inflection although this context normally requires the weak ending. This is shown in (17c), where eget has the strong ending and the other adjective fina the weak one. If, on the other hand, egen is used after a possessive but carries weak adjectival inflection ( 17 d), the meaning changes. Two things are noteworthy here. First, the adjectival inflection clear1y interacts with interpretation, as the change in meaning shows. Second, the question arises as to why egen carries the strong adjectival inflection when following possessive structures. The use of egen in (17b) and (17c) differs from (17d) in that in the former examples the adjective does not have to help identify the referent of the noun, since it clearly refers back to the possessor, whereas in the latter, egen solely refers to the noun it describes. In other words, the use of the possessive presupposes the existence of the possessor so it is not necessary to state this existence otherwise, therefore the weak adjectival inflection does not have to be chosen. This supports the hypothesis that the weak adjectival inflection identifies the member( s) of the subset in the A+ N denotation.
A unlfied structure for Scandinavian DPs 119
5· Diachronic developments The above data suggest that double marking of definiteness in the Scandinavian double definiteness languages is not superfluous but mirrors the semantic contribution of the respective morphemes. Support for this observation comes from diachronic data, namely from the development of the definite article and of the different realizations of definiteness. 5.1
The development of the definite article(s)
There have been several theories that tried to explain the development of the suffixed article (cf. Delbriick 1916, discussed in Syrett 2002). Afl is assumed in the literature, definite articles commonly develop out of demonstrative pronouns (ct~. for example, Braunmuller 1982). Whatis of interest in the scope of this paper is that there were different forms of demonstratives making use of two different inflectional paradigms: (18) a. sa/}:Jessi "this" b. hinn, inn, enna "that"
(Kristoffersen 2002)
Satpessi had pronominal inflection, and hinn, inn, enna inflected like indefinite adjectives. The crucial point is that the definite article developed out of a different demonstrative pronoun than the independent article did: according to Delsing (2002 ), it is plausible to assume that the independent article developed from the demonstrative pronoun in (18a) whereas the suffixed article developed from that in (18b) (Delsing 2002: 930-931). 5.2
The development of the different realizations of definiteness
The development of the different realizations of definiteness - single definiteness in Danish and Icelandic as opposed to double definiteness in Swedish, Norwegian and Faroese - provides further support tor the hypothesis that the suffixed article and the independent article are not identical. In Ancient Nordic (1st-7th century), the beginnings of the development of the suffixed article can be seen in the inscription on the whet-stone ofStr0m. (around 600 AD). (19) hali hino stone this "this stone"
(Braunmuller 2002)
The demonstrative pronoun hino is postposed. This is not very surprising since NP modifiers are normally found in postposition in the early West Nordic languages. Following Braunmuller (1982 ), suffixation of postposed elements after a process of
120
Susanne Lohrmann
weakening of their original (here: deictic) function is not very surprising, either. Another outstanding phenomenon of that kind in the Northern Germanic languages is, for instance, the fusion of the originally independent reflexive pronouns sik, ser with verbal endings to form the Modern Scandinavian middles. The suffixed article -en has a West and South Jutlandic equivalent. ce, which also developed from the demonstrative pronoun (h)inn (Perridon 2002: 1019), but which is, as opposed to the suffixed article in East Jutlandic, placed before the noun. (20) a. ce by/by-en "the town" b. den lille by/den lille by
"the small town" c. ce hele hus/hele huset
"the whole house"
(South Jutlandic/Danish)
Two things are notable here: (i) In adjectival modification den is used, whereas in nonmodified contexts the equivalent of the Standard Danish suffixed article, ce, is used. (ii) In the case of (20c ), hele does not trigger the independent article but the equivalent of the suffixed article both in South Jutlandic and Standard Danish. Hele "whole" is different from, for instance, lille "little" in that it does not introduce a new discourse referent but modifies an existing one. This supports the hypothesis that the preadjectival article introduces a discourse referent that contains a new discourse variable. According to Perridon (2002 ), the use of the suffixed article had spread to the whole of the Scandinavian language area by the first half of the 12th century. However, West and South Jutlandic developed a prefix and not a suffix. I further follow Perridon (2002) and Braunmiiller (1982) in the assumption that the reason for the different realizations of definiteness is due to major word order changes within the NP: attributes were placed before instead of after nouns. This change took place in the Viking Age (800-1000 AD) and must have originated in Jutland.6 In the rest of Scandinavia. the change in the position of attributes must have occurred after noun plus (h)inn had turned into definite noun forms.
6. Towards an analysis
Adjectival inflection in Scandinavian comprises five different endings (strong -0, -t, -a, and weak -a., -e). Depending on the context (almost) every adjective can occur with either a weak or a strong ending. If it is assumed that the ending has a particular function, and if it is further assumed that lexical items are not stored as complex heads, the 6. By comparison of runic inscriptions it was found out that the majority of attributes (24 out of 28) was placed before the noun in Jutlandic but only some in the rest of Danish inscriptions. In Swedish inscriptions from the 11th century postposition of attributes still seems to be the rule (Perridon 2002).
A unlfied structure for Scandinavian DPs t:n most economical strategy would be to regard both the ending and the stem as independent items that are inserted depending on their morphosyntactic features. This is why 1 will adopt the framework of Distributed Morphology (DM). Contrary to traditional approaches dealing with the question of how language is represented in the mind, and which postulate that there is a lexicon which inserts words into syntactic structures, in DM word formation follows syntactic structure-forming operations. Therefore, features that will come to be realized as a subpart of a phonological word will be treated the same as features that will be realized as words (see, for example, Harley & Noyer 1999). There have been attempts to account tor the structure of Scandinavian DPs in the framework of Distributed Morphology. A relatively recent account is that by Embick & Noyer (2001). I will not go into the details of their analysis but will only briefly mention some facts Embick & Noyer cannot account tor: (i) Cases where the suffixed article is not present; (ii) The difference in content of the preadjectival article and the suffixed article (Embick & Noyer's account implies that they are identical); (iii) Diachronic facts; (iv) Differences in meaning due to the weak/strong alternation of adjectival inflection. As shown above, these facts need to be taken into consideration.
6.1
The structure of Scandinavian DPs
6.u Assumptions i.
ii.
Borer's (2005) DP structure includes a classifier phrase (CLP), which has a dividing function, and a number phrase (# P), which is the quantity phrase. The absence of CLP gives rise to mass interpretation. If no quantity interpretation is intended, #Pis absent. The existence of CLP is a precondition for #P, whereas the existence of#P is nota precondition for CLP, i.e. nouns can be divided but not counted (bare plurals, such as cats in "Cats like mice"). Following Borer (2005) in part, I assume that nouns enter the derivation without being classified as count or mass nouns. Thus nouns need to be individuated, irrespective of their being modified or not. There has been some debate as to whether prenominal adjectives should be analyzed as heads. However, since adjectives in Scandinavian can take complements and phrasal APs can appear prenominal (21 ), I assume that prenominal adjectives in Scandinavian are APs. (21) a.
alTa i stadsmiijo boende medborgare all in downtown living citizens "all citizens living downtown"
b. en for rockkonserter oliimpig
lokal
a for rock concerts unsuitable venue Kan unsuitable venue for a rock concert" (Swedish, Holmes & Hinchliffe 2003) iii. In plural formation, Swedish nouns are divided into essentially five declensions. I assume that the form of the declensional affix does not carry any semantic content
122
Susanne Lohrmann
apart from plural information. Reasons for this assumption: (1) There is no clear-cut distinction with respect to the allocation of nouns to declension classes, i.e. the declensional affix seems phonologically motivated. (2) Some nouns have alternative plural endings and can be used with either of the declensional affixes, for example en ka.tt, katt-eti-or "cat, cats~ en kollega, kolleg-eti-or "a colleague, colleagues" (Holmes & Hinchllife2003: 13).7 (3) Often there is no distinction between the two declension classes -or and -er in spoken Swedish (Holmes & Hinchliffe 2003). (4) The declension class can change, e.g. with loans that become familiar and are then used with an indigenous plural (reporters> repmtra.r, Holmes & Hinchliffe 2003: 23). 6.L2 Full structure
The full syntactic structure I assume for Scandinavian DPs is the following. (22) [0 p 2 [disc]
[#P
[ppAP [F' [ident] [0 p 1 [sref] lciassP [ind] NP]]]])]
6.L3 Swedish, Norwegian, Faroese
The following Swedish phrase (23) serves as an example for the other double definiteness languages in Scandinavian. The syntactic analysis is illustrated in (23'). (23) de tre ny-a bil-ar-na DEF three new-w car-PL-DEF "the three new cars" (23')
DP2
~
[disc]
lfP
de~ tre FP
AP~
~
F'
~ [ident] DP1
ny-
a-~ [sref]
~
[ind]J
~
bilt
[sref] -na
[ind]
-ar 7· An anonymous reviewer commented that the form katt-or ls the plural of katt-a "remale cat': whereas katt-er ls used as plural for katt, "cat, female or male~ However, gender ls not relevant here. -or ls not a feminine plural ending (Swedish only has neuter and non-neuter gender) but ls chosen because in katt-a, the final syllable ls open. as opposed to katt, which ends in a dosed syllable.
A unlfied structure for Scandinavian DPs The features [sref], [disc], and [ident] each head their own phrase. DP 1, DP 2, and FP respectively. The syntactic structure also includes a classifier phrase, ClassP, because the feature [ind], individuation, functions as a classifier which individuates nouns. In case of plural marking, the declensional affix is inserted here, in singular DPs, €1 is inserted. The Vocabulary Item bil enters the derivation as a non-classified noun (i.e. neither count nor mass) and is individuated by its movement to the head of ClassP. Nouns that are to be interpreted as mass nouns remain in situ. Bare plurals move to the head of ClassP, too, as do mass nouns that carry plural inflection and/ or are combined with numerals. The head of ClassP then moves further and adjoins to [sref] under Dl' the preadjectival article is realized in DP 2' 6.L4 Icelandic and Danish The distribution of determiners in Icelandic and Danish shows that both languages may show both patterns of definiteness, that is, either the determiner precedes the adjective or it is attached to the noun.
(24) a. goJ-i maiJur-inn good-w man-DEF "the good man"
(Icelandic)
b. hel-e hus-et whole-w house-DEF "the whole house"
(Danish)
(25) a.
hinn golJ-i
malJur
DEF good-w man "the good man"
(Icelandic)
b. det stor-e hus DEF big-w house "the big house"
(Danish)
Both Icelandic and Danish have access to the full structure, too, differing in that the features [sretl and [disc] are not heading one DP each but are united eitherinDP 1 (26) or in DP 2 (27). However, closer examination of the feature content of the D-nodes shows that [disc] is not present in the lower D-node in Danish (Lohrmann 2010). In (26) this is indicated by the use of parentheses. (26)
[DP2[#P [ppAP [p [ident] [np 1 [sref],([disc]) laassP[ind] NP]]]]]]
(27)
[np2 [sref],[disc] [#P [ppAP [p [ident] [np 1 laassP [ind] NP]]]]]]
123
124
Susanne Lohrmann 7·
Conclusion
I have argued that the notion of definiteness in Scandinavian D Ps is made up of three particular components, which are expressed by three distinct morphemes: discourse reference [disc), identity [ident], and specific reference [sref]. The suffixed article brings about specific reference [sref] and is merged under D 1, the adjectival inflection identifies the member( s) in the A + N denotation and is merged in FP to identify the [ident] teature, and the preadjectivalarticle introduces a new, modified discourse variant and matches the [disc] feature under D2 Subsequent movement operations then provide the correct order of morphemes. All Scandinavian languages have access to the same structure with the difference that Danish and Icelandic use either DP 1 or DP 2 as opposed to DP 1 and DP 2 in Swedish, Norwegian, and Faroese. The Modern Scandinavian data shown here suggest that double marking inside the DP in Scandinavian is not a mere agreement phenomenon but is of interpretive value, as is the case with adjectival inflection.
References Borjars, Kersti & Harries, Pauline. 2008. The clitic-affix distinction, historical change, and scandinavian bound definiteness marking. Joun1al of Germanic Linguistics 20(4): 289-350. Borer, Hagit. 2005. b1 Name Only. Oxford: OUP. Braunmiiller, Kurt. 1982. Sy ntaxtypologische Stud len zum Germanischen. Tlibingen: Narr. Braunmiiller, Kurt. 2002. The Andent-Nordic linguistic system from a typological point ofview. In The Nordic Languages. A11 International Handbook of the History of the North Germa11ic Languages, Vol. I, Oskar Bandle, Kurt Braunmiiller, Ernst Hakon Jahr, Allan Karker, HansPeter Naumann & tnfTeleman (eds), 649-656. Berlin: Walter de Gruyter. Delbri.ick, Berthold 1916. Germanische Syntax. Leipzig: Teubner. Delsing, Lars-Olaf. 2002. The morphology of Old Nordic II: Old Swedish and Old Danish. In The Nordic Languages. An International Handbook of the History of the North Germanic Languages, Vol.l, Oskar Bandle, Kurt Braunmii.ller, Ernst Hakon]ahr, Allan Karker, HansPeter Naumann & tnfTeleman (eds), 925-939. Berlin: Walter de Gruyter. Embick, David & Noyer, Rolf 2001. Movement operations after syntax. Linguistic Inquiry 32(4): 555-595.
Faarlund, Jan T. 2007. From clitic to affix: The Norwegian definite article. Working Papers in Scandina1•ian Syntax 79:21-46. Harley, Heidi & Noyer, Rolf 1999. State-of-the-article: Distributed morphology. GLOT 4(4): 3-9.
Hawkins, John 1978. Definiteness and Indefiniteness: A Study in Reference and Grammaticality Prediction. London: Croom Helm. Holmes, Philip & Hinchliffe, Ian 2003. Swedish - A Cotnprehensi1•e Grammar. London: Routledge. Julien, Marit 2005. Nominal Phrases from a Scandinavian Perspective [Lingulstlk Ald:uell!Lingulstics Today 87]. Amsterdam: John Benjamins.
A unlfied structure for Scandinavian DPs 125
Kr!stoffersen, Kr!stian Emil. 2002. The morphology of Old Nordic I: Old Icelandic and Old Norwegian. In The Nordic La11guages. A11 International Handbook ofthe History ofthe North Germa11ic La11guages, Voll, Oskar Bandle. Kurt Braunmilller, Ernst Hakon Jahr, Allan Karker, Hans-Peter Naumann & Ulf Teleman (eds ), 911-925. Berlin: Walter de Gruyter. Lohrmann, Susanne. 2010. The Structure of the DP and its Reflex: in Scandinavian. PhD dissertation, University of Stuttgart. Lyons, Christopher. 1999. De.finitene.ss. Cambridge: CUP. Perridon, Harry. 2002. Dialects and written language in Old Norse II: Old Danish and Old Swedish. In The Nordic Languages. An International Ha11dbook of the History of the North Germa11ic La11guages, Voll, Oskar Bandle. Kurt Braunmilller, Ernst Hakon Jahr, Allan Karker, Hans-Peter Naumann & UlfTeleman (eds), 1018-1027. Berlin: Walter de Gruyter. Syrett, Martin. 2002. Morphological developments from Ancient Nordic to Old Nordk. In The
Nordic Languages. An Internatio11al Ha11dbook of the History of the North Germanic La11guages, Vol I, Oskar Bandle, Kurt Braunmilller, Ernst Hakon Jahr, Allan Karker, Hans-Peter Naumann & UlfTeleman (eds), 719-729. Berlin: Walter de Gruyter. Vangsnes, 0ystein. 2007. On the mmphosyntax of Scandinavian adjectives. Paper presented at the guest lecture series of the SFB 732/GK 609.
A semantic approach to noun phrase structure and the definite - indefinite distinction in Germanic and Romance Ulla Stroh-Wollin Department of Scandinavian Languages Uppsala University, Sweden
This article deals with the relation between the structure and the semantics of noun phrases. Two complexes of meaning are taken to be grammatically derived in noun phrases. One complex concerns the conceptualization of the entity denoted by the noun phrase, whereas the other concerns the restriction of the set of referents. The latter complex is related to the distinction between definite and indefinite noun phrases. The article also deals with the variation within and between Germanic and Romance noun phrases, showing how this variation may be explained on the basis of the preceding analysis. In particular, the different uses of definite and indefinite noun phrases are highlighted.
1.
Introduction
The purpose of this article is twofold: to demonstrate and argue for a semanticallymotivated structural model of noun phrases and to demonstrate how this model can account for variation within and between Germanic and Romance as concerns the structure and use of different kinds of noun phrases. The analysis is based on a semantic view on grammatical derivation. An important part of the study is to examine what meanings are, or may be, grammatically encoded in noun phrases and how the derivation influences the semantics of the phrases. The overall task is to find a model that accounts for the semantics as well as for the empirical evidence from different languages. The structural description is basically worked out within the general framework of generative grammar as developed in Chomsky (2000) and (2001). There are two complexes of meaning that are considered to be grammatically derived in noun phrases. One complex, treated very briefly in the article, concerns the conceptualization of the entity denoted by the noun phrase as a set of countable objects, as a mass or as a single object The other complex, which will be more thoroughly
128
Ulla Stroh- Wollin discussed here. concerns the restriction of the set of referents, which is linked to the definite - indefinite distinction. Empirically, the analysis is chiefly based on evidence from Germanic, not the least from the Scandinavian languages. However, as will be demonstrated below, the model also works for Romance languages. In fact, I believe it has a much wider applicability. It might also work for languages without articles, though such languages will not be considered here (but see Stroh-Wollin 2009 for an account of Old Scandinavian). Only regular 'argumentaf noun phrases characterized by the use of determiners and reterential power are considered, whereas e.g. predicative noun phrases and vocatives are not taken into account
2.
Theoretical assumptions
Theoretically, I follow, as already mentioned, the approach outlined in Chomsky (2000, 2001), where the old X-bar principle is combined with basically three operations to conduct the derivation: Merge, Agree and Move. The derivation is abstract, and 'spell-out' is only possible at some points of the derivation, namely when a 'phase' is completed. Below, I will provide some clarifications concerning my view on the operations Agree and Move. When it comes to Agree, I follow the account of Pesetsky & Torre go (2007). In contrast to Chomsky (2000, 2001 ), Pesetsky & Torrego make a distinction between the interpretability and the value of features. In this way, Agree appears above all as an operation which establishes relations between categories. It can briefly be described as follows. If we take for example an adjective which agrees in number with the noun head, it has a value, let's say PLURAL. But PLURAL has no meaning, i.e. is not interpretable, on the adjective as such; PLURAL is a meaningful teature, i.e. an interpretable feature, only on a noun, or in relation to a noun. Consequently, an adjective can probe for a goal with an interpretable number feature, i.e. a noun, with which it can share the relevant feature. Thus, a formal relation, an Agree-relation, between the attribute and the noun is established. It is, however, not necessarily the case that an adjective has a value for number; not all languages inflect their adjectives in number - English, for instance, does not Nor is it necessarily the case that a noun has a value for number; not all languages inflect their nouns for number; e.g. (spoken) French normally does not. This means that both an interpretable feature and an uninterpretable one can be valued as well as unvalued. Whenever a feature is unvalued and/or uninterpretable, it probes its c-command domain for value and/or interpretation by Agree with another instance of the same feature. In case neither the probe nor the goal has a value for the feature, the feature remains unvalued (and the words will be spelled out in the default form). However, an uninterpretable feature must always be turned interpretable through Agree with a corresponding interpretable feature, whether valued or not
A semantic approach to noun phrase structure 129
The consequence for the noun phrase structure from Agree. as described above, is that the kind of attributes in the noun phrase that regularly show agreement in number and case must be merged to the structure at some position where they can probe for interpretability; such features are not interpretable on the attributes themselves. Since a probe probes a goal in its c-command domain, we know that agreeing attributes cannot be merged below the head where the noun is inflected for number and case. When it comes to Move, no so called EPP-features (i.e. features without independent semantic impact that simply trigger movement of constituents so that they are spelled out in the right position) are assumed. Thus, every head is solely defined by its abstract (unvalued) functional features, which attract lexical counterparts during the course of a derivation. The host of the attracted lexical feature can be a head constituent or a phrase constituent I take elements with purely formal features, such as articles and inflectional affixes, to be merged in head positions, and elements which, besides the formal feature, have descriptive content, including deictic elements and quantifiers, to be merged in specifier positions (basically in line with Giusti 1997). In case an affix is first merged in a head position, it will in turn attract the nearest head down the tree. Finally, it should be emphasized that formal morphemes may be phonologically null. It is not unusual that languages lack tor example visible affixes to indicate present tense or singular number. But if the language has contrasting forms, such as past tense and plural, then the present tense verb and the singular noun should rather be considered inflected with null morphemes, than not being inflected at all. However, I believe that null morphemes also have to be assumed in some other cases, i.e. when a contrast is not at hand. But of course a null morpheme must be 'discemable' in language acquisition, if not by contrast, then in some other way. 3·
A broad outline of the proposal
Nowadays it is generally assumed by generative linguists working according to the Chomsk.yan approach that (argumental) noun phrases are DPs. The proposal I will advocate here is a slightly expanded variant of the DP-NP-model (dating back to Abney 1987), consisting of four projections: a little nP, a big NP, a little dP and a big DP; see (1). (1) DP
~ dP
~ NP
~
nP
Although the division of labor is not that clear-cut, I take the lower projections, NP and nP, to form an N-domain and the higher projections, DP and dP, to form a D-domain, and will treat the respective domains in tum below.
130
Ulla Stroh-Wollin 3.1
The N -domain
The expansion of the NP into a little nP and a big NP is due to a shift in the view of what is the input to a grammatical derivation. Many agree now that it is not a fully inflected word. Instead a grammatical derivation is assumed to begin with a lexical unit without word class features entering into a local relation with a category defining head in the lowest X-bar-module, whereas further inflection is localized in superstructure projections. This is the view I adopt here. (So far the description is in accordance with the theory of Distributed Morphology; see e.g. Marantz 1997, Embick & Noyer 2007. This does not mean, however, that I adhere to this theory in full.) The lexical unit mentioned, a lexical unit without word class features, is normally referred to as a root However, the complex outcomes of lexical compounding and lexical affixation may also be entries to a grammatical derivation. Thus, the derivation of a noun phrase starts with the merging of a root (or a more complex 'word base') as the head of little nP. The root itself (or the complex unit) has no word -class features, but amalgamates with a word class-defining affix inn, from which follows a conceptualization as ENTITY. In this mode~ the complex of a root and affix is not a noun, but a noun stem, i.e. a base for further nominal inflection, such as inflection in number and case, which is supposed to take place in the head ofbig NP, on top oflittle nP. I will not discuss case inflection further, since case does not affect the meaning of the noun phrase as such. But inflection in number is semantically important. I take number inflection to account for a primary conceptual specification of the entity. A noun in the plural denotes an entity comprised of separate objects, whereas a noun in the singular denotes an entity not distinguishable as separate objects. No further conceptual specification of the noun is assumed at the N P-level. While it is true that singular noun phrases refer either to single objects or masses, such a disambiguation does not seem to be a matter pertaining to number marking. This means that plural and singular nouns as such can be assumed to have simply opposite values on some semantic feature; I think+ COUNTABLE on plural nouns and- COUNTABLE on singular nouns is a suitable labelling. However, some determiners (inserted above NP) have disambiguating power, cf. glass (as a DP) and a glass, which means that a full DP may be further specified; let us say a DP such as glass is +DIVISIBLE and the DP a glass is- DIVISIBLE. Traditionally, the lexicon is said to contain nouns which (in principle) denote either indivisible (in the sense of 'individual') objects, countables, or divisible masses, uncountables. However, a consequence of the model explored so far is that the lexicon provides the speaker with noun stems, rather than with nouns, and further, that (un) countability (of a sort) and (in)divisibility are syntactically-derived semantic features, not lexically inherent qualities in the noun stems. This view is both theoretically and empirically motivated. Since many nouns, glass, stone, ham etc., can appear in noun phrases referring to masses as well as to plural entities and single objects, it is an advantage not to presume any inherent countability or divisibility features on the word stern, but to take these kinds of specifications on the noun phrase to be accounted for
A semantic approach to noun phrase structure 131
by the derivation above nP. The countability specification may also be visible in the singular case, e.g. in Swedish skinka "ham" with an -a marking the singular, cf. skinkor "hams" with -or for the plural Thus, it is not the case that a plural noun is a singular noun+ a plural affix; a plural noun is a noun stem+ a plural affix and a singular noun is a noun stem + a singular affix. The derivation within theN -domain is demonstrated in (2a-b ), showing the analyses of Swedish skinka and skinkor. Following e.g. Josefsson (1997), I take the first element of a compound, such as skink in Swedish skinksmorgas "ham sandwich': to be a root Consequently, the stem affix in the stem skink is invisible. NP
(2)
~
~
nP
N
~ n
N
~
-a
root skink
n
-0
NP
b.
~
nP
~
N
n
~-or root skink
n -0
The rootskink is not, contrary to the noun stem skink, formally conceptualized as ENTITY, even though such a meaning could seem a natural inherent component already in the root. However, it is less evident that roots such as travel and jump should be inherently conceptualized as ENTITY, although there are nouns like travel and jump. Thus, I presume there is a formal difference between roots and word stems, with this difference being semantically linked to the basic conceptualizations of the word classes. Roots like tra11el and jump can of course also merge as a verb-defining head, with which follows a conceptualization as ACTION. This is, in principle, possible also with ham, even though the verb (to) ham so far has no conventionalized meaning. 3.2
The D-domain
The expansion of the DP into a little dP and a big DP is mainly motivated by the semantic nature of definiteness. In the proposal, the two projections are taken to collaborate in the restriction of the set of referents. This restriction is accounted tor by articles, quantifiers, demonstratives and other determiners. Definite suffixes, occurring in e.g. the Scandinavian languages, are also part of the restriction system. The semantics of the D-domain will be discussed in more detail in Section 4. However, a split DP is also empirically supported by so-called double definiteness in Scandinavian languages. This support will be demonstrated below, simultaneously giving the reader an idea of what the derivation in the D-domain may be. Double definiteness means that two definite morphemes, a definite suffix on the noun and a pre-posed free definite article, occur in one and the same noun phrase.
132
Ulla Stroh- Wollin
This double marking is (in most cases) obligatory in Swedish, Notwegian and Faroese if the noun is preceded by some attribute; see the Swedish example in (3a). Without a preceding attribute, the inflected noun expresses definiteness on its own; cf. ( 3b ). (3) a.
den lilla flickan
"the little girl"
DEP little girl.DEF b. flickan girl.DEP
"the girl"
The derivation of the two noun phrases in (3a-b) is demonstrated in (4a-b). In both examples, the derivation starts with the merging of the definite suffix as little d. Then the suffix immediately attracts the (complex) noun-head inN. As adjectival attributes always precede their head nouns in Swedish (as in Germanic on the whole), we can take the adjective lilla "little" of the a-example to be inserted in the specifier-position of little dP. As the noun phrase in this case cannot stand without the pre-posed article, there is actually the need for the big DP on top of little dP; the pre-posed article lexicalizes the higher head. When, on the other hand, there is no preceding attribute, as in the b-example, the inflected noun can itself move to the head of the DP. DP
(4) a.
~ D den
DP
b.
~
dP
D
~
d'
spec
dP
~
~
li11a
NP
d
A
N
flicka
d -n
A
The evidence from double definiteness in Scandinavian indicates that definiteness demands two collaborating projections. Below, we will return to the question why. 3·3
The positions of attributes
The structure in (1) provides several possibilities to insert attributes in its specifiers. Attributes that do not have to probe for agreement on number and case, e.g. PPs, adverbs and clausal attributes, are supposed to attach to the nR but adjectives, possessives, quantifiers and demonstratives for example, which must enter into an Agree-relation with the noun, cannot be inserted lower than the NP, where the noun is inflected. The proposal does not assume universally fixed positions or earmarked projections tor the pre-posed modifiers of the noun phrase, basically on empirical grounds. Notwegian possessives for example may be used as definite determiners in first
A semantic approach to noun phrase structure 133
position or as post-noun attributes (in spec-NP), on condition that definiteness is expressed by other means in the left periphery, cf. (Sa-b). (5) a.
min lille bil
"my little car"
my little car b. den lille bilen min DEF little car.DEF my
"my little car"
Presuming non-fixed positions is not denying the fact that certain relative orders are cross-linguistically well attested (thus, hopefully possible to explain in some way or the other) or pretending that any modifier can be inserted anywhere. One general principle is that determiners precede adjectives, but there is no distinct division on projections. Adjectives can be inserted in N P or dP, and determiners can be inserted in dP or DP. The double use of the dP may actually explain why the border between adjectives and determiners is not always clear-cut.
4· The semantic implications of definite and indefinite noun phrases The D-domain of the noun phrase structure is the locus of different determiners. Determiners define the noun phrase as structurally definite or indefinite. The same kind of power can also be attributed to definite affixes in languages, such as the Scandinavian languages, which inflect their nouns for definiteness. Furthermore, we can note that determiners also restrict the set of referents in relation to the largest possible set, given the descriptive core of the noun phrase. In the following, I will take as my standpoint that the definite - indefinite distinction, from the formal perspective, is solely a question of how the restriction of the set of referents is derived grammatically. In this respect, my account differs considerably from others, where semantic or pragmatic notions, such as specificity, uniqueness, identifiability or inclusiveness, are taken to be inherently grammatically encoded features of definite noun phrases (see e.g. Lyons 1999, Vangsnes 1999, 2001; Julien 2005). At first glance, my standpoint might seem odd, as e.g. the noun phrases three dogs and the dogs in (6a-b ), for example, could actually be used to refer to exactly the same set of dogs. The choice of expression is in such a case a pragmatic question of who is the speaker and who is the hearer. As the pragmatic definition of definiteness tells us: a definite noun phrase is appropriate when the speaker can assume that the hearer, within the context, can uniquely identify the intended referents. (6) a. There are three dogs in the garden. b. The dogs are in the garden. Nevertheless, I will detend the standpoint that the difference, formally, is a question of how the restriction of the set of referents is settled. First, the restriction in (6a) looks very straightforward; the noun phrase thJ-ee dogs denotes three arbitrary individuals
134
Ulla Stroh-Wollin
taken from the universal set of dogs. But the case of (6b) is trickier. Even so, if we imagine a situation where a person we are visiting has three dogs and utters the sentence (6b ), we will make two conclusions with quantitative implications. The first conclusion is that the speaker (if nothing in the situation speaks to the contrary) is talking about her own dogs and no other dogs. The other conclusion is that all three dogs are in the garden. And this is in principle how I believe the semantics of definite noun phrases should be understood in the prototypical case: a definite noun phrase denotes the totality of a (contextually) restricted selection of the largest possible set of referents. The grammatical consequence of this view is that 'the definite meaning' is derived in two steps. The first step could be described as something like a restriction in relation to the largest possible set, given by the descriptive core of the noun phrase. Formally, I take the descriptive content ofNP to define this 'largest possible se~ disregarding the fact that elements merged in the structure above NP may also sometimes convey descriptive content. The set defined by the NP is in the following labelled the universal set, U for short This first selection results formally in a set of selection, S for short In my model, the set of selection is settled by the derivation of little dP. Then the set of referents, R for short, may finally be settled in big DP.ln the definite case, R equates S. Thus, using the meta-language of set theory, the prototypical definite meaning can be expressed by the formula in (7), where c stands for "is a (true) subset of'. (7) Prototypical definiteness: R = S c U Notice that the formula in (7), contrary to the pragmatic definition cited above, holds true even in those cases where a definite noun phrase actually can be used, although the referent/ -s is/are not uniquely identifiable; see e.g. (8a-b ), where Ia jam be "the leg" in the French example corresponds to her leg in the English counterpart. Most people have two legs, but in neither language is it a problem that we cannot identify which one is broken. However, 'S is a subset of U' may be taken just as a very general instruction to the hearer to find out (in anyway) how Scan be reduced in relation to U, but it does not postulate that the hearer must find out the exact limits of S. (8) a. Elle
she
s~t
REFL
casst Ia jambe. is broken DEF leg
b. She has broken her leg. We may then take it as a pragmatic, non-grammatical question to sort out under what conditions a definite noun phrase is acceptable in a non-uniquely identifiable context. Now, one can object that definite noun phrases also can be used with generic reference, as the Brazilians in (9), which implies that the formal meaning of definite noun phrases is less precise than is shown by the formula in (7). Generic reference means no restrictions whatsoever between U and R, so in fact, we have to assume that the formal meaning of definite noun phrases is R = S ~ U, which allows both the prototypical definite interpretation, R = S c U, and a generic interpretation, R = S = U as in (9).
A semantic approach to noun phrase structure 135
(9)
The Brazilians are crazy about football.
In Romance, generic reference is always expressed with definite noun phrases. 1 In Germanic on the other hand, generic reference is most often expressed by noun phrases with no visible determiner, so-called bare nouns. It is, for instance, fully possible to omit the definite article in (9) and still have a grammatical sentence with approximately the same meaning. Hence, it follows that the interpretation R = S = U must also be a possible interpretation of noun phrases with no visible determiner in Germanic. But Germanic bare noun-phrases are ambiguous; they may have both generic and more restricted reference, cf. dogs in (lOa-b). I will use the label 'unquantified partreference' for the kind of reference that (1Ob) represents, as opposed to the 'quantified part-reference' expressed by noun phrases with a q uantitying determiner, e.g. a cardinal (as in three dogs) or a word like some or many. In cases such as (lOb) we can presume that the interpretation implies a restriction of R in relation to S, whereas S still equals U (which, of course, is also the case with noun phrases expressing quantified part-reference); thus, R c S = U. In Romance, unquantified part-reference can be expressed by noun phrases with partitive articles; see the French counterpart to (lOb) in (lOc). (10) a. She doesn't like dogs. Interpr.: R = S = U b. There are dogs in the garden. Interpr.: R c S = U c. Il y a des chiens da.ns le jardin. there-are PART dogs in the garden In French, the partitive construction is the only way to express unquantified part-reference, since the language completely lacks bare noun DPs. Romance languages other than French, however, have at their disposal both bare noun DPs and DPs headed by partitive articles and can use them in the same contexts. From the interpretations of the different uses of dogs in (1Oa-b ), we may suggest R ~ S = U to be the formal meaning ofbare noun-phrases in Germanic. However, the case of bare noun-D Ps in Romance is less clear, since they are not used with generic reference, but show the same distribution as the partitive construction. We will return to this below.
5· 5.1
Noun phrases in Gennanic and Romance- a comparison The derivation of noun phrases in Germanic
The derivation of definite noun phrases was demonstrated above with examples from Swedish; see (4a-b). A common and characterizing feature of the Scandinavian Longobardi (2001: 352 f) distinguishes between 'kind-denoting' and 'object-denoting' generictty. The latter type is, contrary to the former, possible with bare noun-DPs as well as with various kinds of determiner-headed indefinite noun-phrases. Generic reference in this article comprises, however, only (the noun phrase-inherent) kind-denoting generictty.
1.
136
Ulla Stroh-Wollin languages is the inflection for definiteness on nouns. Here this inflection is assumed to take place in little d. We may take this head to host an abstract feature, a, which attracts a corresponding lexical feature in the suffix, whereby S, formally a, is valued (S ~ U). The suffix in tum attracts the noun from N. The final settling of the set of referents (as R = S) takes place in the DP. However, this second step is accounted for in different ways in the Scandinavian languages. One possibility is the lexical identification of R, which takes place if the noun itself with its lexical content, representing S, moves to D, where S equates with R. This procedure is, however, not possible if e.g. an adjective precedes the noun, because an adjective in spec-dP contributes to the restriction of the set of selection by its descriptive content and cannot be left in dP, if R is to be identified lexically. The Scandinavian languages actually handle this situation in three different ways; see (lla-c).
(11) a.
den lilla flickan
"the little girl"
Swedish
"the little girl"
Danish
"the little girl"
Icelandic
DEP little girl.DEF b. den lille pige DEF little girl c.
litla. stUlka.n little girl.DEF
The double definiteness languages (Swedish, Norwegian and Faroese) and Danish merge a free definite article in D, which means that R is formally, i.e. not lexically, specified. The Danish article is a 'true' definite article with specification on the a- feature; it merges first to little d before it moves to big D. Thus, the Danish noun in ( llb) is not inflected for definiteness. The article in the double definiteness languages, on the other hand, is an expletive with an unvalued a- feature. In these languages, the valued definite suffix is merged in little d (and attracts the noun) and the free article directly in D, where it probes the definite suffix in d for a value on a. Icelandic, finally, has no free definite article and applies lexical identification of R also when an adjective precedes the noun, however, by moving the entire dP to spec-DP. A similar raising analysis, for this and other cases, is also put forward by Julien (2005: 54 ff.). 2 English and continental Germanic practice only formal specification of R with their free and pre-posed definite articles. The derivation of an English noun phrase such as the dogs is demonstrated in (12a). The corresponding bare noun-DP dogs is demonstrated in ( 12b ). In the indefinite case, we can take both little d and big D to be lexicalized
It is an intriguing question why the Scandinavian languages make use of different strategies in phrases such as the ones in (11). It may be a question of'economf. It is not too far-fetched to argue that movement of the entire d.P to DP is the most 'costly' strategy and therefore that it applies only in Icelandic, which lacks a free definite article. It is also worth considering that a phrase like de11 pige in Danish is possible, but exclusively with demonstrative reading: "that girl': cf pigen, the only way to express "the girl" in Danish.
2.
A semantic approach to noun phrase structure 137
by invisible 'articles~ the lower one leaves the set of selection unrestricted in relation to U (S = U), and the higher one leaves R with an open value in relation to S (R ~ S). (12) a.
b.
DP
~
~
dP
~
d
~
DP
D 0
dP
~
NP
d
~
0
N dogs
NP
~ N dogs
As mentioned above, I take adjectives to be inserted either in the specifier ofNP or in the specifier of dP. In the case of English and continental Germanic, it is difficult to tell whether adjectives are merged with NP or dP or ifboth strategies are applied. In Scandinavian, on the other hand, we can deduce from adjectives preceding nouns in the definite form, cf. (lla), that adjectives must be in dP in definite phrases. Adjectives in definite noun phrases appear in their so-called weak form, whereas adjectives in indefinite noun phrases, where the nouns stay in N, appear in the strong form. This may indicate that the strong adjectives, contrary to the weak ones, are inserted in NP. Continental Germanic languages also have strong and weak declensions of adjectives but apply slightly different rules than Scandinavian for the use of the strong and weak forms. I am less sure that the strong weak distinction has anything to say about the position of adjectives in these languages. 5.2
The derivation of noun phrases in Romance
Romance languages (at least the varieties of which I have some knowledge) differ from Germanic by taking adjective attributes both before and after the head noun. Partly following Cinque (1994), I account for this difference by assuming that different kinds of adjectives are distributed on different projections in Romance, i.e. NP and dP in my model, and that the noun obligatorily raises to (what in my model is) the d-head in the same languages. (I do not follow Cinque when assuming universally-fixed positions for adjectives, nor do I accept the generalizing statements about Germanic having a fixed low position for their nouns; ct~ the previous subsection.) Now, the question remains what triggers the movement of nouns to din Romance. My answer is that nouns are attracted by an invisible affix in d, which imposes a possible restriction on the set of selection, i.e. S ~ U. Above, this restriction was associated with definite noun phrases, but should, if I am on the right track, be present in definite as well as in indefinite noun phrases in Romance. This means that the formal semantic difference between definite and indefinite noun phrases in Romance is solely attached to the value on R. A grammatical consequence of this hypothesis is that the definite article can merge directly to big D. The derivation of a definite noun phrase in Romance, represented by
138
Ulla Stroh-Wollin French les chiens "the dogs~ is shown in (13a). The derivation of a corresponding indefinite phrase with a partitive article, des chiens "(some) dogs': is shown in ( 13b ).
(13) a.
b.
DP
~
~
dP
D
~
les
d
N chiens
5·3
d
dP
D
~
des NP
A
A
DP
N
-fll
D
A
A N chiet1s
NP
d
N
-fll
The use of definite and indefinite noun phrases in Germanic and Romance
It may seem odd to presume, as I did above for Romance, that an invisible, alwayspresent, morpheme in d attracts all nouns, but does not contribute at all to the contrast between definite and indefinite noun phrases. It is probably due to the more regulated distribution of adjectives of different kinds in Romance, forcing some adjectives ('thematic adjectives' and 'manner-adjectives' according to Cinque 1994) to appear to the right of the noun, that the system is at all learnable. However, the semantic implications do make sense in a comparison between Germanic and Romance in the use of definite and indefinite noun phrases. In principle, we can distinguish four basic kinds of reterence that a noun phrase can express, described above in Section 4, viz. prototypical definite reference, generic reference, unquantified part-reference and quantified part-reference. There are also, in principle, four kinds of noun phrases to express the different references: definite DPs, bare noun-DPs, DPs headed by partitive articles and DPs headed by quantifying determiners. However, four kinds of reference and four kinds of noun phrases do not in this case mean a simple one-to-one-relation between form and meaning. The different pairings of reference-type and D P-type are illustrated in ( 14-17), with English and corresponding French examples. The DPs are marked with bold face and the choices of DP-types are, for the sake of clarity, noted to the right (14) PROTOTYPICAL DEFINITE REFERENCE a. She doesn't like the dogs. b. Elle n'aime pas les chiens.
definite DP definite DP
(15) GENERIC REFERENCE a. She doesn't like dogs. b. Elle n'aime pas les chiens.
bare noun-DP definite DP
(16) UNQUANTIPIED PART-REFERENCE a. There are dogs in the garden. b. Il y a des chiens dans le jardin.
bare noun-DP DP with partitive article
A
(17)
semantic approach to noun phrase structure 139
QUANTIFIED PART-REFERENCE
a.
There are three dogs in the garden.
b. Il y a. trois chiens dans le jardin.
DP with quantifying determiner DP with quantifying determiner
The examples in (14-17) do not exhaust the possibilities of Germanic and Romance, but rather show very typical examples. We shall, however, keep in mind that even Germanic sometimes uses definite noun phrases for generic reference, cf. (9) above, and that the Romance languages, other than French, use bare noun-DPs to some extent To understand the possible and preferred pairings of reference-type and type of DP in different languages, we have to consider three facts. The set of DP-types is not the same in all languages. The encoded meaning of a certain kind of DP may be ambiguous. And a certain kind of DP in one language may be semantically different from a formally-close counterpart in another language. If we want to express a generic reference, we need a DP that allows an interpretation R = S = U. In Germanic, the meaning of definite DPs, i.e. R = S ~ U, and the meaning of bare noun-DPs, i.e. R ~ S = U, are equally good for this purpose. The preference for expressing generic reference with bare noun-DPs in Germanic may then be due to the fact that it is the formally unmarked variant There is, however, room for variation, and the choice ofDP-type is not in every situation identical in all Germanic languages. In French, no choice exists when there is a question of generic reference; a definite DP is the only alternative. But a language such as Italian, which permits bare nounDPs, also uses exclusively definite DPs to express generic reference and restricts bare noun-DPs to contexts where even DPs with partitive articles are possible; see (18), where the definite article i can not be omitted.
(18) Non le piacciono *(i) cani. not her please DEF dogs KShe doesn't like dogs."
(Italian)
Following the hypothesis that all noun phrases express S ~ U in Romance, we may assume that the encoded meaning of bare noun- DPs in Italian is R ~ S ~ U, whereas definite DPs encode R = S ~ U. If so, the definite DP is obviously closest to the generic meaning, and we can take the bare noun-DP to be ruled out by contrast The similar distribution on bare noun-D Ps and D Ps headed by partitive articles in Italian points to a part-reference in both cases, even though the encoded meaning is more distinct in the latter (also formally marked) DP-type (R c S ~ U). There seems to be a subtle division oflabor: "the bare noun tends to be used where the tocus is on the concept expressed by the noun, [... the] partitive construction, in contrast, focuses more on entities denoted by the noun" (Maiden & Robustelli 2000: 76). In Germanic, an imprecise quantifier, such as English some, can have a similar specifying effect as the Italian partitive article.
140
Ulla Stroh-Wollin
References Abney, Steven Paul. 1987. The English Noun Phrase and Its Sentential Aspect. PhD dissertation, MIT. Chomsky, Noam. 2000. Minimalist Inquiries. In Step by Step: Essays on Minimalist Syntax In Honor of Howard Lasnik, Roger Martin, David Michaels & Juan Uriagereka (eds), 89-156. Cambridge MA: The MIT Press. Chomsky, Noam. 2001. Derivation by Phase. In Ken Hale: A Life In Language, Michael Konstowitz (ed.), 1-52. Cambridge MA: The MIT Press. Cinque, Guglielmo. 1994. On the evidence for partial N-movement in the Romance DP. In Paths Thwards Universal Grammar. Studies i11 Honor of RichardS. Kayne, Guglielmo Cinque, Jan Koster, Jean-Yves Pollock, Luigi Rizzi & Raffaela Zanuttini (eds), 85-110. Washington DC: Georgetown University Press. Embick, David & Noyer, Rolf 2007. Distributed morphology and the syntax- morphology interface. In Linguistic Interfaces, Gillian Ramchand & Charles Reiss (eds ), 289-324. Oxford: OUP. Giusti, Giuliana. 1997. The categorical status of determiners. In The New Comparative Syntax, Liliane Haegeman (ed), 95-123. London: Longman. Josemson, Gunlog. 1997. 011 the Principle.s of Word Formation 111 Swedish. Lund: Lund University Press. Julien, Marit. 2005. Nomb1al Phrases from a Scandinavian Perspective [Linguistik AktuellJLinguistics Today 87], Amsterdam: John Benjamins. Longobardi, Guiseppe. 2001. How comparative is semantics? A unified parametric theory of bare nouns and proper names. Natural Language Semantics 9: 335-369. Lyons, Christopher. 1999. Definiteness. Cambridge: CUP. Maiden, Martin & Robustell!, Cecilia. 2000. A Rejere11ce Grammar of Modern Italian. London: Arnold. Marantz. Alec. 1997. No escape from syntax: Don't try morphological analysis in the privacy of your own Lexicon. UPe,w Worki11g Papers in Linguistics 4: 17 5-192. Pesetsky, David & Torrego, Esther. 2007. The syntax: of valuation and the interpretability of features. In Phrasal a11d Oausal Architecture: Syntactic Derivation and b1terpretation [Linguistik AktuellJLinguistics Today 101], Simin Karimi, Vida Samiian & Wendy K. W!lkkins (eds), 262-294. Amsterdam: John Benjamins. Stroh-Wollin, Ulla. 2009. On the development of definiteness markers in Scandinavian. Working Papers In Scandinavian Syntax 83.1-25. Vangsnes, 0ystein Alexander. 1999. The Identification of Functional Architecture. PhD dissertation, University of Bergen. Vangsnes, 0ystein Alexander. 2001. 0 n noun phrase architecture, referentiality, and article systems. Studia IJnguistlca 55(3): 249-299.
Definite determiners in two English-based creoles Specificity or definiteness? Ekaterina Bobyleva University of Amsterdam
It has been repeatedly observed in the literatme that unlike their European lexifiers, creoles do not mark definiteness categorically. The question addressed here is whether the distribution of overt definite determiners and bare definite NPs in two Atlantic English-based creoles, Jamaican and Sranan, can be accounted for in terms of the specificity constraint that underlies the determiner use in Gbe, one of their most important substrate languages. The data examined shows that definite determiners in the two creoles tend to follow the definiteness-based pattern. This suggests that English rather than Gbe played the most crucial role in the formation of the discourse-semantic and distributional properties of Jamaican and Sranan definite determiners. Bare definite NPs in the two creoles may occur when the referent is either unimportant or self-evident. Thus, determiner use in Jamaican and Sranan is sensitive not only to languagespecific grammar rules but also to universally prominent pragmatic factors.
1.
Introduction
It has frequently been observed that creole languages use determiners in a way different from the way determiners are used in their European lexifiers. Where Germanic and Romance languages would invariably require definite or indefinite determiners, creoles allow for bare, determinerless NPs. The present study investigates the use of definite determiners and bare definites in two English-based creoles, Jamaican Creole (JC) and Sranan, and attempts to identify the sources of their discourse-semantic and distributional properties. The literature on determiner use in creoles primarily seeks to account for the use of bare NPs. Some researchers regard the occurrence of bare NPs in creoles as an instance of system irregularity. Others, in contrast, view the use of bare NPs as a property of the creole grammar, and argue that the use of overt determiners as
142
Ekaterina Bobyleva opposed to bare NPs in creoles is governed by rules different from the ones applied in European languages. With regard to some Atlantic creoles, it has been proposed that they follow the specificity-based pattern of determiner use, which is also attested in their West African substrate languages. Remarkable parallels in the distribution of definite determiners have been observed between Haitian Creole, Saramaccan and Gbe (cf. Lefebvre 1998; Aboh 2006). Since Gbe also constitutes an important substrate component of both JC and Sranan, the verification of the substratist hypothesis of determiner use advocated by Lefebvre and Aboh is one of the central objectives of this study. Next to the accounts of determiner use in creoles that appeal to substrate influence, there are also accounts that appeal to universal aspects of human language capacity. Bickerton (1981) relates the behavior of determiners in creoles to the principles of UG, while Giv6n (1984) argues that it has to do with universal properties of discourse organization.
2.
2.1
Setting the stage Jamaican Creole and Sranan: Their social and linguistic history
As already mentioned in the introduction, the two case studies to be considered in this paper are JC and Sranan. Both English-based creoles, they developed in the latter part of the 17th century in the British colonies of Jamaica and Surinam, as a result of contact between the British settlers and the West African slaves. The slaves came predominantly from Kwa-, in particular, Gbe- speaking areas, but also from the areas where Bantu and Benue-Congo languages were spoken. Despite their common socio-historical sources, JC and Sranan have a very different linguistic history, the main difference being the amount of contact with their main lexifier, English. JC developed in a continuous contact with English. which has always remained the official language of the colony. Combined with the pressure from English as a language of prestige and socio-economic growth, the contact between English and JC gave rise to a situation where very few (if any) Jamaicans use the 'deep' creole, referred to as basilect. The majority of Jamaicans, especially in the urban areas, speak a variety (or varieties) in between the basilect and English, so-called mesolect, creating a creole-to-English speech continuum. The linguistic situation of Surinam is very different. The direct influence from the varieties of English spoken by the British only lasted thirty years. Surinam started out as a British colony in 1651 but in 1667 it was taken over by the Dutch. By 1680 almost all British slave-owners had left the colony with their slaves. The slave population of Surinam, however, continued speaking an English-based creole, and the Dutch influence hardly penetrated beyond the lexical level. The different amount of contact with the lexifier has had a significant impact on the linguistic properties of the creole languages that developed in Jamaica and Surinam.
Definite determiners ln two English-based creoles 143 J C, in particular its most widely spoken mesolectal variety, shows much more affinity with English than Sranan. 2.2
The noun phrase in Jamaican Creole and Sranan
Similarly to English, JC and Sranan have definite and indefinite determiners. The major difference between English and the two creoles is in the use of bare nouns. 2.2.1
Definite deta·mina·s
The JC definite determiner is di (also occurring as de and the). The form most likely derives from the English definite determiner the. Similarly to its English etymon, di occurs prenominally and is neutral with regard to number. The combination of di with a count noun in the absence of plural marking generally gives a singular reading. In order to obtain the plural interpretation of a definite NP, the use of the postnominal plural marker dem, or the plural inflection -(e)s is required. Sranan has two definite determiners: the singular a (da in early Sranan) and the plural den (also occurring as dem in early Sranan). The two definite determiners have inherited their number specifications from their English etyma, the demonstratives that and them (as it is used in colloquial English, e.g. them boys). Similarly to demonstratives in English, Sranan a and den occur prenominally. Sranan has no number agreement in the noun phrase. The noun itself always occurs in the bare form. 2.2.2
Indefinite determiners
The form of the indefinite determiner, and of the homophonous lowest numeral, in both JC and Sranan is wan. This form transparently derives from the English lowest numeral one. JC, on a par with wan, frequently uses the English-like form a (which only occasionally occurs as a.n ). Creole wan, which has inherited the [-plural] specification from its English etymon, can only occur with [-plural] NPs. Plural indefinites in both JC and Sranan surface without a determiner. In JC, they can be overtly marked for number by means of the inflection -(e)s. In Sranan, semantically indefinite plural NPs never receive overt number marking. 2.2.3
Bare NPs
The use of bare N Ps, i.e., NPs with no determiner and no inflection, is known as one of the features typical of creole languages. In ]C and Sranan, bare NPs are also broadly used and have a rich interpretative variability. In addition to the plural indefinite bare NPs mentioned in the previous section, bare nouns in JC and Sranan may receive generic, singular and plural indefinite, and singular definite interpretation. 2.3
Working definitions
Given the lack of unanimity with regard to the definitions of the notions of definiteness and specificity, before I proceed to the discussion of the distribution of determiners in JC and Sranan, I shall provide the working definitions of these concepts.
144 Ekaterina Boby1eva
Definiteness
2.,3.1
Following Lyons (1999), I assume that identifiability (subsuming tamiliarity) and uniqueness are the core semantic features associated with grammatical definiteness. 2.,3.2
Specificity
Following Ihsane & Puskas (2001), I consider specificity to be an independent category, which can be applied not only to indefinite NPs, as some scholars (e.g., En~ 1991) suggested, but also to definite NPs. I assume that unlike definiteness, specificity is a matter not just of identifiability but of a particular type of identifiability, namely, identifiability via discourse-linking (ct~ Pesetsk:y 1989; En~ 1991; Von Heusinger 2002). While definite NPs have a presupposition of existence in the world of discourse, they are not necessarily linked to elements previously introduced into the discourse. In addition to discourse-linking, which gives rise to directly anaphoric definites, identifiability can also rely on the association with previously introduced referents (associative anaphora), shared situational context, general knowledge, and particular types of NP modifiers and complements (certain adjectives, prepositional complements and relative clauses) (cf. Hawkins 1978). Specific definite NPs, on the other hand, always relate to pre-established elements in discourse. Thus, only directly anaphoric definite NPs are specific.
3· 3.1
Distribution of definite detenniners in Gbe, JC and Sranan Specificity-based determiner use in Gbe
According to Aboh (2004, 2006, 2010), Gbe languages have a specificity-based pattern of determiner use. Overt determiners only occur with specific NPs. Non-specific N Ps, generic, indefinite and definite, occur with a zero-determiner. Examples below illustrate determiner use with [+specific] and [-specific] definite NPs and with generics. The examples are from Gungbe, a Gbe language spoken in Benin.
tavo
(1) K!>ku m:lm tavo ce (Aboh 2004: 76) bO a!> eml na x!> fJ. Koku see table lSG.POSS and say 3SG FUT buy table DET KKoku saw my table and then he said he would buy the table:'
(2) De im jt xwegbe, un m:Jn As lsG.NOM get home lsG.NOM see x:Jnt:Jn ce lt te to agbtl jf friend 1sG.Poss PL stand be.locate veranda on "When I got home, I saw my friends waiting on the veranda:'
(Aboh, p.c.)
(3) A.O lu wt n:J du tevf titan. king FOC HAB eat yam first a. '"The king (over there) eats yam first" b. "The king (in general) eats yam first:'
(Aboh, p.c.)
Definite determiners ln two English-based creoles 145 (4) Xfa bfbelu read bible "Read the Bible~ (5)
Dawe {ai! a m;~Jn to f6to JlJ, man REL 2sG see be.locate photo on mlax6 wt. brother lSG.POSS FOC "The man you saw on the picture is my brother:'
(Aboh, p.c.)
(Aboh, p.c.)
ce
As examples (1-5) demonstrate, only directly anaphoric definite NPs in Gungbe are accompanied by a determiner, all other types of definite NPs (associative anaphora, situational definite NPs, general knowledge definite NPs, and NPs containing referentestablishing complements and modifiers) as well as generic NPs are bare. The way l;~J is used in Gungbe is thus very different from the way the definite determiner is used in English. As the translations show, in English, in all these contexts the is required. Thus, while English has a definiteness-based pattern of determiner use, Gungbe follows the specificity-based pattern. In the next section, I will consider whether the distribution of definite determiners and bare definites in JC and Sranan shows parallels to Gbe. 3.2
Definite determiners and bare definites in JC and Sranan
3.2.1 The scope of the investigation The investigation of the distributional and discourse-semantic properties of definite determiners and bare definites in JC and Sranan will focus on the contemporary varieties of the two creoles. Due to unavailability of data in the basilect, the analysis of JC will be limited to the mesolect. Most of the examples provided here have been taken from the following sources: Sistren (1986), abbreviated as Sis, and Thelwell (1980), abbreviated as The~ for JC; and Voorhoeve (1962), abbreviated as Vh, for Sranan. The investigation will be restricted to singular and mass definite NPs, which occur with the number-neutral definite determiner di in JC and with the singular definite determiner a in Sranan. Plural NPs marked with the plural definite determiner den in Sranan and with the combination of the definite determiner di with the postnominal plural marker dem or with the plural inflection -(e)s in JC will not be considered here. The principles governing the distribution of Sranan den are not restricted to specificity/definiteness. Firstly, in addition to specificity/definiteness, it is also used to express plurality. Secondly. similarly to plural markers in many other creoles (e.g.• Sao Tome, Cape Verdean, Guinea-Bissau, etc.), itfavours [+human] NPs. The same holds for the JC postnominal plural marker dem. And since this latter can only be realized together with a definite determiner (which is typically either di or a possessive pronoun), when in combination with dem, di may occur in contexts which are not representative of its own discourse- semantic properties. Although the plural inflection does
146 Ekaterina Bobyleva not have the same repercussions for the distribution of di, to keep the two data sets comparable, JC plural definite NPs marked with di ... -(e)s were also excluded.
Definite determiners with specific and non-specific NPs. In the literature on JC, di is often characterized primarily as a specificity marker (e.g., Patrick 2004). Indeed, NPs whose referent has been previously established in discourse are typically preceded by di:
3.2.2
(6) Di square have a upstairs shop and a big ole
(Sis: 3)
parish church. Me never like di church. "The square had an upstairs shop and a big old parish church. I never liked the church~
However, JC di violates the specificity constraint, frequently occurring with definite NPs that are not discourse-linked: (7) Me stepmadda tek khaki cloth and mek one
(Sis: 63)
lickle dolly gemme. One day, me cut off di neck. "My stepmother took khaki cloth and made a little dolly for me. One day, I cut off [its] neck." (8) Look pon de little man hog... "Look at the little hog.. ~ (9)
(Thel: 20)
When the sun bright [... ] an' the sky blue...
(Thel: 66)
"When the sun is bright [... ]and the sky is blue..:' Examples (7 -9) demonstrate that definiteness marking in JC mesolect can be based on the association with another already known referent (7), identification of a nominal description with a referent in a particular situational context (8) or with a generally known referent (9). Sranan data presents us with a similar picture. (10) ...da j bj -a wan ptjin kritji. then 2sG PST have DET small creek
Da m denk a
(Vh: 62)
kritji no dipi.
then 1SG think DET creek NEG deep KThen there was a small creek. Then I thought the creek was not deep:' (11) Dits noo mj um wakti fu artji oten mi thus now lsG to wait PREP hear when lsG
e
-gwe... Pan, mi tjis a
boskop a
(Vh: 74)
oso.
PROG go.away Pang lsG get DET message PREP house "Thus now I had to wait to hear when I was going away... Pang, I got the message at home..."
Definite determiners ln two English-based creoles 147 (12)
Da
pan a
'oorlog broko.
(Vh: 65)
Then pang DET war break "Then pang the war broke out~ (13) Ma a pa dati na a howru testament. but DET father that PREP DET old testament "But that father is in the Old Testament:'
(De Drie 1985: 31)
Examples above demonstrate, that neither JC di, nor Sranan a are restricted to specific NPs. The following examples will provide further support for this observation. Both JC di and Sranan a systematically occur with NPs that contain referent-establishing complements and modifiers, i.e., prepositional phrases and relative clauses. Below I provide parallel examples from JC (14, 16) and Sranan (15, 17): (14) In di evening me get conscious and feel (Sis: 8) di resrllt [a di beating]. "In the evening I regained my consciousness and telt the result of the beating~ (15) Da u b -e -tek a bakasee (fit a pen]... Then 1PL PST HAB take DET backside of DET pen "Then we took the backside of the pen..."
(Vh: 60)
(16) Di table was di place [weh we eat]. "The table was the place where we ate."
(Sis: 75)
(17) D a tnan [dat e -kar i] kon. then DET man COMP HAB call 2SG come "Then the man who called you came:'
(Vh: 58)
Di in JC also systematically occurs when a NP is modified by an ordinal numeral, an adjective in the superlative form or the adjectives same, only, last, next, which all imply definiteness (cf. Hawkins 1978). Sranan a appears to behave less systematically when modifiers of the aforementioned types are involved. In all the cases discussed above, JC di and Sranan a parallel the behavior of the in English. Following the distributional properties of their English counterpart. definite determiners in JC and Sranan also occasionally parallel its semantically vacuous occurrences. For instance, they may occur with weak definites. (18) Mr Iris have to be on the road by six thirty ... KMr Iris had to be on the road by six thirty.. :' (19)
U g
a
oso,
da
u
pot a
radio.
1PL go PREP house then 1PL put DEF radio "We went home, then we put on the radio:'
The definite determiner in JC also occasionally occurs with generic NPs.
(Sis: 75) (Vh: 81)
148 Ekaterina Bobyleva
(20) Di tnanguss chrikifai.
(Stewart 2007: 397)
a. "The mongoose (over there) is cunning:' b. "The mongoose (in general) is cunning." Such semantically vacuous occurrences of the definite determiner are, however, irregular and scarce in JC and nearly absent in Sranan. Generic NPs (21-22) and cases paralleling English weak de:finites (23-24) generally surface determinerless in the creoles: (21) Member seh tnan a green lizard.
(Sis: 45)
KRemember that men are green lizards." (22) Soso Tatnkatnan mag sjwen dape.
(Vh: 62)
only Tamk.aman may swim there "Only Tamk.amen may swim there." (23) She[ ...] use her to carry di excess loa.d pon her
(Sis: 62) head when she going to market. KShe used her to carry the excess load on her head when she was going to the market:' (24) Taka, u b -i -luku san a strati (Vh: 59) Say 1PL PST PROG look thing PREP street KSay, we were looking at things in the street" 3.2.3 Bare definite NPs
Bare definite NPs in prepositional phrases In Section 3.2.1, we observed that the definite determiners in JC and Sranan seem to have a similar distribution to the definite determiner in English. However, unlike English, JC (25) and Sranan (26) allow semantically definite NPs to occur without a determiner. In the contemporary varieties of ]C and Sranan, bare definites can be predominantly found inside prepositional phrases. But in the earlier varieties of the creoles, they occurred in a broader range of syntactic contexts, including subject and object positions. 3.2.3.1
(25) Di toilet used to ha1'e a part like a tennce
(Sis: 16)
outside weh yuh can play Jacks. Even if dem go inna 1 toilet go lock it [...] me know dem in deh ... "The toilet used to have a part like a terrace outside where you could play Jacks. Even if they went into the toilet and locked it [... ] I knew they were in there..:' (26) ...dan w e -kon lat a skoro [...] Da (Vh: 60) then lPL HAB come late PREP school then Although the form inna might look like a contraction of the preposition "in" with a determiner, it should be interpreted as just "ilf. Consider the following examples: inna di slwp "in the shop" (Sis: 7); bma mi grand.foada name "in my grandfather's name" (Sis: 23).
1.
Definite determiners ln two English-based creoles 149
masra 'Van der Geld' ben -tnap a
mojo doro
mister PN PST stand PREP mouth door "Then we came late to school[ ...] Then mister Vander Geld stood in the door opening." The occurrence of semantically definite nouns without a determiner distinguishes the two creoles from English and seems to align them together with Gbe. However, the omission of the definite determiner in cases like (25) and (26) does not provide evidence in favor of the importance of specificity for the distribution of the definite determiners in the creoles. Example (25) violates not only the definiteness-based, but also the specificity-based rules of determiner use. The italicized NP toilet has a discourse antecedent and is therefore [+Specific]. 3.2.3.2 Bare definite NPs with unique referents. There is a special class of definite NPs that can occur without a determiner in ]C and Sranan, namely NPs that have a unique referent. The uniqueness may be either inherent to the semantics of the NP or bound to a particular context. Good examples of such uniquely referring NPs are nouns used to refer to features of local geography. For instance, in JC, town, in the context of Jamaica, is always used to refer to Kingston. A parallel example can be found in Sranan, where foto "town" is always used to refer to Paramaribo. Other NPs that belong to this type are names of professions and titles that within the context of a particular institution refer uniquely to a particular individual. The examples 1 came across in JC are teacher, preacher, masta. "plantation owner" and busha. "overseer': and in Sranan, masra "plantation owner" and basi "boss" (ct~ Bobyleva 2006, 2009). While the nouns listed above are conventionally used in such a name-like fashion in JC and Sranan, NPs may also acquire unique semantics 'on the spo~ when a referent of the NP becomes firmly established in discourse. NPs with such easily accessible discourse referents may also occur without a determiner. Consider an example from Sranan below: (27) Ma now, a pernasi pe Opoko gebore, pe Atyopi gebore, a granmasra dati ben de
wan wreedaardige granmasra. [...] Ma a pernasi dati tussyuru sroysi e broko. [... ] Pernasi feti fu sungu bika smysi boro ... (De Drie 1985: 40, from Bruyn 1995:87) "But now, the plantation where Opoko was born, where Atyopi was born, that plantation owner was a cruel plantation owner. [...] But that plantation, [its) sluice(s) was/were always breaking down. [... ] The plantation was about to flood because the sluice(s) was/were cracked."
Example (27) illustrates that once a referent is firmly established in discourse, its identity becomes so obvious that there is no need to mark it overtly. A similar tendency is observed in some other creoles, for instance, in Cape Verdean (cf. Baptista 2008) and
150
Ekaterina Bobyleva Berbice Dutch (cf. Kouwenberg 2008), in Gbe, as well as in other languages with non-definiteness-based determiner systems, for instance, in Hausa (cf. Lyons 1999). With regard to Gungbe, Aboh (2010: 12) observes that referents that are "contextually prominent and/or known to the discourse participants" can be designated by bare nouns. For instance, a cat in a household can be referred to either with its proper name or with lise "cat': without a determiner. Examples involving bare NPs that are [+definite, +Specific] appear to violate the specificity-based principle of determiner use. In fact, on the basis of the example partially provided here in (27), Bruyn (1995) argues that zero determiners in Sranan are not restricted to non-specific NPs, but are open to both non-specific and specific interpretation. Consequently. she concludes that zero determiners in Sranan do not "express a distinctive value and should be regarded as merely nothing" (Bruyn 1995: 81). In my view, uniquely referring bare NPs neither support nor violate the specificity-based principle of determiner use, as in contrast to other cases of bare definite NPs they do not have a zero determiner in the underlying structure. As I already mentioned above, uniquely referring bare NPs are akin to proper names. I therefore assume that, similarly to proper names, they have no determiner at all and receive a specific definite interpretation by virtue of raising to Spec,DP (cf. Longobardi 1994). 3-2.4 Summary
Based on the data considered above, a conclusion can be made that overt definite determiners in JC and Sranan generally tend to follow the definiteness-based pattern. While the marking of semantically definite NPs closely parallels the behaviour of the in English, semantically vacuous occurrences of the definite determiner in the creoles are less systematic. The occurrence of bare definite NPs is restricted to NPs inside prepositional phrases and uniquely referring name-like NPs. While the contexts favoring definite determiners and bare definites are the same in the two creoles, bare definites are more common in Sranan than in JC, where one observes a stronger tendency towards the English-like pattern of determiner use.
4·
4.1
On the sources of the discourse-semantic and distributional properties of definite detenniners in JC and Sranan The sources of overt definiteness marking
The observed distributional properties of overt definite determiners in JC and Sranan deviate significantly from the specificity-based pattern found in Gbe. The development of the definiteness-based behavior of the creole determiners cannot be accommodated within the substratist hypothesis. Therefore, it requires an alternative explanation.
Definite determiners ln two English-based creoles The development of definiteness marking in JC and Sranan may be either a contact-induced or a language-internal phenomenon. In other words, either it was English that provided the definiteness-based pattern of definite determiner use in the two ereales, or the picture observed in the contemporary varieties ofJC and Sranan represents the result of the process of grammaticalization. Based on the diachronic investigation of determiner use in Sranan, Bruyn (1995, 2008) argues for the grammaticalization scenario. In what follows, I will argue that the most important role should be attributed to English (and, in the case ofSranan, possibly, also Dutch). According to Bruyn. Sranan definite determiners (d)a and den, which developed from the English demonstratives that and them, respectively, and which were ambiguous between demonstratives and definite determiners in early Sranan, gradually underwent semantic bleaching. Having lost their deictic force, they became spedalized as more general markers of definiteness. Bruyn's analysis suggests that Sranan (d)a and den have undergone (and, perhaps, are still undergoing) the same type of grammaticalization from demonstrative to definite determiner as the definite determiners in many Germanic and Romance languages, including English the, which, similarly to Sranan (d)a, is etymologically related to the demonstrative that. The grammaticalization scenario, however, fails to account for the fact that in early Sranan, (d)a and den were already used as definite determiners. Furthermore, they are found in contexts atypical for demonstratives. Example (28) is taken from one of the earliest available sources in Sranan, Van Dyk, dating back to c1765. (28) Hoe tem wi za jam? Q time lPL FUT eat Da tem fa jam. Det time PREP eat "When are we going to eat?" "The time to eat"
(Arends and Perl1995: 133)
Given that cross-linguistically definiteness marking is a late phenomenon, it appears unlikely that the extension of the use of (d)a and den to contexts illustrated in (28), which took place in Sranan at least as early as mid-18th century, is the result of language-internal development, unless we assume that grammaticalization in creoles proceeded at a much faster rate than in other languages. lt appears more plausible to attribute the definiteness-based behavior of (d)a and den to the influence of English. The presence of a similar definiteness-based pattern of determiner use in Dutch may be also relevant considering that Dutch is spoken as a second language by many Sranan speakers. What, on the other hand, can be attributed to the language-internal dynamics is the regularization in the use of (d)a and den as general markers of definiteness observed by Bruyn in her diachronic study.
151
152
Ekaterina Bobyleva
JC provides even stronger arguments in favor of the assumption that English was the most important source of definiteness marking in the creoles. To begin with, unlike the definite determiners in Sranan, JC di derives from the definite determiner the. This makes the grammaticalization scenario proposed by Bruyn inapplicable to the JC case. Early JC sources show that the definite determiner in JC has inherited many of the distributional properties of its English etymon. The following example is taken from early 19th century ]C. It shows that the definite determiner in JC was already then used not only with semantically definite NPs but also with generics. (29) ...de bible bin promise you dat
(D'Costa & Lalla 1989: 40)
Garamighty take care of de good man... "The Bible told you that God Almighty would take care of the good man/good people:' The fact that one occasionally finds semantically vacuous occurrences of the definite determiner (with weak definites and generics) that have direct parallels in English, and the fact that they are more frequentin JC, which has had a much longer and extensive contact with English, than in Sranan presents another argument in favor of English as the main source of the distributional properties of the definite determiners in the two creoles.2
4.2
The sources of bare definite NPs
In the contemporary varieties of JC and Sranan, the omission of the definite determiner is restricted to definite NPs inside prepositional phrases and uniquely referring name-like NPs. Although the occurrence of bare definites represents a common teature of JC and Sranan and Gbe, their distributions in the creoles and in the substrate language are rather different. Firstly, in Gbe, bare definites are allowed in a much broader range of contexts than in JC and Sranan. Secondly, the occurrence of bare definites in the creoles cannot be straightforwardly derived from the specificity constraint, which is crucial in Gbe. The only similarity between Gbe and the two creoles was found in the denotation of unique and discourse prominent referents. However, as already mentioned in Section 3.2.3.2, not only JC and Sranan, but also Cape Verdean Creole and Berbice Dutch use bare NPs in such cases. The latter two creoles do not have Gbe among their substrate languages. This weakens the connection between the presence of this phenomenon in Gbe and in] C and Sranan. Given the lack of strong evidence in favor of substrate influence, in order to account for the instances of bare definites in JC and Sranan, one needs to look for analternative explanation. A distinction needs to be made between the transfer of the discourse-semantic principles of determiner use and the transfer of the distributional properties of a determiner. The implications of this distinction for the analysis of contact-induced change in the discourse-semantic and distributional properties of determiners needs to be further investigated
2.
Definite determiners ln two English-based creoles 153 I assume that the omission of the definite determiner in the contexts illustrated in Section 3.2.3 can be accounted for in terms of what Giv6n (1984) refers to as pragmatic referentiality. As was briefly mentioned in the introduction, Giv6n's account of determiner use in creoles appeals to universal properties of discourse organization. Crucial in Giv6n's hypothesis is the distinction between semantic referentiality (i.e., the supposed existence in the universe of discourse) and pragmatic referentiality (i.e., the importance in the universe of discourse according to the speaker). He observes that while pragmatic reterentiality usually corresponds to semantic reterentiality, it is ultimately pragmatic referentiality that determines whether a NP will have an overt determinerornot It appears to me that Givan's proposal can successfully account for the omission of the definite determiner with semantically definite NPs contained inside prepositional phrases. NPs inside prepositional phrases generally have less discourse prominence than subjects and objects, and can therefore often be characterized as pragmatically non-referential. Despite the differences in the underlying structure briefly discussed in Section 3.2. 3.2, the reasons behind the omission of the determiner in the case of pragmatically non-referential definites and in the case of uniquely referring definites appear to be similar. As Kouwenberg (2008: 449) points out, in both cases the identity of the referent is "not at issue': either because it is considered unimportant or because it is selfevident. Based on this generalization, I assume that the occurrence of determinerless definite NPs in JC and Sranan is determined by pragmatic principles. Although Bickerton, unlike Giv6n, uses the term 'specificity, his understanding of the term is different from the one adopted here, and overlaps with Giv6n's referentiality. While Bickerton (1981) argues that the behavior of determiners in creoles is governed by UG-based principles, he also observes that the grammatical rules related to specificity are sensitive to semantic as well as to pragmatic aspects of the notion. To be more precise, he states that N Ps that do have a specific referent, whose exact identity is either unknown to the speaker or irrelevant to the point at issue are also treated as nonspecific and occur with a zero-determiner. This understanding of specificity comes very close to Giv6n's pragmatic referentiality. Although both Giv6n's and Bickerton's proposals can in principle account for the instances of bare definites in JC and Sranan, Bickerton's proposal appears more problematic as it is based on the questionable assumption that U G- based rules are sensitive to pragmatics. Regardless of the differences in the theoretical backgrounds underlying their ideas, Giv6n's and Bickerton's proposals supported by the data give rise to an important observation, namely. that the pragmatic dimension of referentiality is universally prominent. Given this observation, it can be argued that during the process of creole formation, when the grammar settings of the existing languages become 'loose: the pragmatics-based pattern may 'creep' into the system of the emerging creole.
154 Ekaterina Bobyleva
5·
Conclusion
The present study was concerned with the distribution of definite determiners in two Atlantic English-based creoles, JC and Sranan. Determiner use in the creoles appears to be sensitive to definiteness as well as to pragmatic factors. While overt determiners generally seem to follow the definiteness-based pattern, wherever bare definite NPs occur they express pragmatic non-referentiality. The discussion of the possible sources of the hybrid discourse-semantic properties of the definite determiners in JC and Sranan. has led to the conclusion that the behaviour of the definite determiners in these creoles is a result of the tension between the English-like definiteness-based pattern and the universally prominent pragmatics-based pattern.
References Aboh, Enoch 0. 2004. The Morphosyntax of Compleme11t-Head Seque11ces: Clause Structure a11d Word Order Patterns in Kwa. Oxford: OUP. Aboh, Enoch 0. 2006. The role of the syntax-semantics interface in language transfe.t: InL2 AcquisiNon and Oeole Genesis: Dialogues [Language Acquisition and Language Disorders 42], Claire Lefebvre, Lydia White & Christine Jowdan (eds), 221-52. Amsterdam: John Benjamins. Aboh, Enoch 0. 201 0. The morphosyntax of the noun phrase. In 1bpics in Kwa Syntax [Studies in Natural Language and Linguistic Theory 78], Enoch 0. Aboh & James Essegbey (eds), 11-37. Dordrecht: Springer. Arends, Jacques & Perl, Matthias (eds). 1995. Early Suriname Creole Texts. A CollecNon of 18th-Century Sranan and Saramaccan Documents. Frankfurt & Madrid: Verveurt & Iberoamericana Baptista, Marlyse. 2008. On the syntax and semantics of DP in Cape Verdean creole. In Noun Phrases In Creole Languages. A Multi-faceted Approach [Creole Language Library 31], Marlyse Baptista & Jacqueline Gueron (eds), 61-105. Amsterdam: John Benjamtns. Bickerton, Derek 1981. Roots ofLanguage. AnnArbo MI: Karoma. Bobyleva, Ekaterina. 2006. Nominal Functional Categories in Atlantic Creoles. A comparative approach to Creole genesis. MA thesis, University of Amsterdam. Bobyleva, Ekaerina. 2009. Specificity and definiteness in Creole languages. Unfinished grammattcalization or contact-induced variation? Paper presented at the conference Variation and change in the structure of the noun phrase in Germanic and Romance: Autonomous developments or result oflanguage contact? Amsterdam, January, 2009. Bruyn, Adrienne.1995. GrammaNcalizaNon in Creoles: The Development of Determiners and Relative Clauses In Sranan. Amsterdam: IFOTT. Bruyn, Adrienne. 2008. Bare nouns and articles in Sranan. In Nou11 Phrases i11 Creole Languages. A Multi-faceted Approach [Creole Language Library 31], Marlyse Baptista & Jacqueline Gueron (eels), 339-81. Amsterdam: John Benjamlns. D'Costa, Jean & Lalla, Barbara (eds). 1989. Voice.s in Exile. Jamaican Texts of the 18th a11d 19th Centuries. Thscaloosa AL: The University of Alabama Press. De Drie, Alex. 1985. Sye! Arki tori!, compiled by Trudi Guda. Paramaribo: Afdellng Cultuur Studies van het Ministerie van Onderwijs, Wetenschappen en Cultuur.
Definite determiners ln two English-based creoles 155 E~. Miirvet. 1991. The semantics of specificity. Linguistic Inquiry 22: 1-26. Giv6n. Talmy. 1984. The pragmatics of referentlality. In Georgetow11 lT11iversity Round Table 011 Language a11d Linguistics 1984. Meaning, Form, a11d Use in Context: Linguistic Applications, Deborah Schriffin (ed.), 120-38. Washington DC: Georgetown University Press. Hawkins, John. 1978. De.finitene.ss and Inde.finitene.ss: A Study in Reference and Grammaticality Prediction. London: Croom Helm. Ihsane, Tabea & Puskas, Genoveva 2001. Specific is not definite. Generative Grammar 111 Ge11eva 2: 39- 54. Kouwenberg, SUvia. 2008. Bare nouns in Berbice Dutch Creole. In Noun Phrases 111 Creole La11guages. A Multi-faceted Approach [Creole Language Library 31 ], Marlyse Baptista & Jacqueline Gueron (eds), 437-58. Amsterdam: John Benjamins. Lefebvre, Claire. 1998. Creole Genesis and the Acquisition of Grammar. The Case of Haitia11 Creole. Cambridge: CUP. Longobardi, Giuseppe. 1994. Reference and proper names. Li11gulstic Inquiry 25: 609-665. Lyons, Christopher. 1999. Definiteness. Cambridge: CUP. Patrick, Peter L. 2004. Jamaican Creole: Morphology and syntax. In A Handbook of Varieties of English, Vol2: Morphology and Syntax, Bernd Kortmann. Edgar W. Schneider, Clive Upton, Raj end Mesthrie & Kate Burridge (eds ), 407-438. Berlin: Mouton de Gruyter. Pesetsky, David. 1989. Wh-in-sit:u: Movement and unselective binding. In The Representation of (ln)De.finiteness, Eric Reuland & Alice ter Meulen (eds), 98-129. Cambridge MA: The MIT Press. Sistren. with Ford-Smith, Honor 1986. Dl Lionheart Gal: Life Stories of Jamaican Women. Toronto: Sister Version. Stewart, Michele M. 2007. Aspects of the syntax and semantics ofbare nouns in Jamaican creole. In Noun Phrases in Oeole Languages. A MulN-faceted Approach [Creole Language Library 31], Marlyse Baptista & Jacqueline Gueron (eds), 383-399. Amsterdam: John Benjamtns. Thelwell, Michael. 1980. The Harder They Come. New York NY: Grove Press. von Heuslnger, Klaus. 2002. Specificity and definiteness in sentence and discourse structure. Journal of Semantics 19: 245-274. Voorhoeve, Jan. 1962. Sranan Syntax. Amsterdam: North-Holland.
PART II
Change
Form-function mismatches in (formally) definite English noun phrases Towards a diachronic account Christopher Lucas School of Oriental and African Studies
This article discusses two classes of so-called 'weak definiteS, arguing that their (definite) form is misleading as to their (non-definite) semantics, and outlining a diachronic explanation for why each of these classes (observable in sentences such as Lets go to the pub and He came to the bank of a ri~·er) should exhibit this particular form-function mismatch. For examples such as the pub the loss of an obligatorily definite interpretation is argued to be the result of a semantic reanalysis such that reference is no longer to a specific entity but to the activity conventionally associated with that entity. For examples such as the bank of a river the mismatch is argued to be a consequence of an incompatibility between the semantics of indefiniteness marking and the semantics of relational nouns, which arises when definiteness marking becomes obligatory in a language.
1.
Introduction
This article discusses the meaning and diachronic development of two classes of morphosyntactically definite noun phrases (NPs) with rather puzzling semantics. The italicized phrases in (1) and (2) exemplify the kinds of expressions, labelled 'weak definites' by Poesio (1994), that are the focus of the presentinvestigation. These two classes ofNPs are discussed in Sections 2 and 3 respectively.
Let's go to the pub. (2) He came to the ba.nk of a river.
(1)
'Morphosyntactically definite' is a cumbersome but necessary term, since the major claim to be defended here is that the phrases under discussion do not involve definite reference, despite their morphosyntax. A key assumption underlying this claim, which is probably uncontroversial but which should nevertheless be made explicit, is that there is an abstract semantic
160
Christopher Lucas
pragmatic category of definiteness which is (a) universal, and (b) language-independent. That is, this category is not an emergent property of a given grammatical system (on a par, say, with number marking) but a stable component of human cognition (on a par with the concept of number itself) which is expected to have a range of different manifestations in different grammatical systems, including, in theory, none whatsoever.1 What the correct characterization of this category is remains a vexed issue, and we will not dwell on it here, since the types of expressions we will chiefly be concerned with resist assimilation under any of the major theories of definiteness, whether these are based on uniqueness (e.g. Russell1905), familiarity/identifiability (e.g. Christophersen 1939, Heim 1988), or some combination of the two (e.g. Hawkins 1991). Hence my claim that the relevant expressions, while appearing to instantiate the abstract category of definiteness, are in fact misleading: they exhibit a form - function mismatch. Such a move ought to be controversial. It appears to be a direct flouting of (Modified) Occam's Razor (Grice 1978). Strict application of this principle would lead us to assume, following LObner (1985: 280) in "the spirit of rigid compositionality': that "the definite article has one and the same meaning in all its uses." Certainly this seems like a sensible heuristic: a too-eager willingness to discount apparent instances of the phenomenon under study could result in an overly restrictive theory and the multiplication of entities beyond necessity. Why then would one want to deviate from what seems like simple good scientific practice? In fact I do not. Rather, I want to suggest that the standard conception of what constitutes the definite article (where this is understood as a genuine instantiation of the abstract category of definiteness) is too broad if it includes the examples in (1) and (2). The point may be made clearer by considering the English comparative correlative construction in (3). (3)
The more you eat, the fatter you get.
It should be uncontroversial that the word the which appears at the left edge of each clause in this construction is not the definite article, despite the homophony. Indeed, two recent analyses of this construction which agree in very few respects (Culicover & Jackendoff 1999; Den Dikken 2005), do agree that this the is a degree word and not a determiner. In fact, there is nothing about the syntax of the comparative correlative the that merits labeling it an article, and the homophony with the actual definite article is an accident of their divergent evolution from a common source: both are derived from the Old English distal demonstrative in p-, the comparative correlative degree word from the instrumental case-form py specifically (Jespersen 1961: 7.591). Clearly, to claim that the the in (1) and (2) does not instantiate the abstract category of definiteness is somewhat more controversial than to claim the same of the the in It has been argued, however, (e.g. Leiss 2007) that no language lacks some means (which need not necessarily be nominal determiners) of marking the definiteness status of referents.
1.
Form-function mlsmatches in (formally) definite Engllsh noun phrases (3). The principal reason for this is that there is no doubt that the items in (1) and (2) are derived from the actual definite article. And, of course, they too are (remain) phonologically identical to the actual definite article. Nevertheless, it is undoubtedly the case that certain morphemes which used to mark definiteness in certain languages now do so no longer, at least in a subset of their uses. A particularly well-attested example of this development comes from the history of Aramaic, as pointed out by Greenberg (1978: 59). Old Aramaic had a suffixed definite article, as shown in Table 1 (after Lambdin 1971 ). The ability of this suffix to mark the distinction between definite and indefinite reterence began to be eroded in the Middle Aramaic period, and the present-day Neo-Aramaic varieties have lost the bare indefinite form altogether, such that nouns are essentially always marked with a vestige of the old definite suffix, for example nasa "man, person" (< 'enas-a "the person"). Moreover, certain varieties have innovated a new, preposed indefinite article, which attaches to a noun together with the historical definite suffix, as in (4). (4) Xa. nasa. zille
one man go.3MS.PAST "A man went.:'
(Jewish Arbel Neo-Aramaic; Khan 1999: 196)
It has also been claimed (e.g. Harris 1980) that what was the French definite article now simply marks the gender and number of the noun it precedes. Clearly if this is a monolithic claim about uses of le!la!les as a whole then it is wrong. There remains a strong correlation between the presence of the article in French and definite reference, even if the correlation is less than perfect. However, given that we know articles can progress towards becoming mere noun markers, it seems reasonable to expect that certain contexts will be particularly innovative in this regard, while others will be more conservative. Thus the claim made here for English is similar to that made by Vergnaud & Zubizaretta (1992: 614) for French, whereby, in the relevant contexts, "le functions as an expletive from the point of view of denotation:' My argument for English will be that the contexts in which the definite article can be 'expletive' in this sense are defined semantically, and that this phenomenon arises as a result of diachronic semantic changes (with concomitant syntactic repercussions).
Table 1. Old Aramaic determination system (Lambdin 1971) indef.
sg. m. pl. m. sg. f pl. f
malk malkin malkii. malkii.n
def. malk.-ii. malk.-ayyii. malk.-tii. malk.-ii.tii.
161
162
Christopher Lucas
2.
'Non-specific' weak defi.nites
The first class of weak definites to be considered against this background is that exemplified in (5)-(7).2 (5) a. Let's go to the pu.b. b. #Let's go to the bar!cafe!restau.rant. c. Let's go to a pubi and smash iti up. #. d. #Let's go to the pubi and smash iti up. (6) a. John's always using/on the (tele)phone. b. #John's always using/on the email. c. John uses a (tele)phone as a doorstop. #. d. #John uses the (tele)phone as a doorstop. (7) a. I need to look up a word in the dictionary. b. #I need to look up a rule in the style guide. c. Take a dictionary and remove the first page. #. d. #Take the dictionary and remove the first page.
Take (5) (=1) to begin with. There are various contexts in which this sentence could be uttered such that reference is to a pub which is unique or identifiable or both. However, on the most salient reading of this sentence, at least in British English, there does not have to be any unique or identifiable pub in the discourse context for an utterance of (5) to be felicitous. Notice that the same is not true for any of the semantically similar items in ( 5). For an utterance of (5) to be felicitous there does have to be a uniquely-identifiable bar/ cafe/restaurant in the discourse context This suggests that (5) does not involve some sort of covert satisfaction of uniqueness/identifiability arising pragmatically, since if this were the case one would expect this covert satisfaction to be freely extendable to semantically similar items in similar contexts, which is not the case. Notice also. however, that the weak reading in (5) is not simply equivalent to an indefinite reading either. A weak definite cannot bind a pronoun, as a true indefinite can, (Se-d), and, crucially, 'going to the pub' on the weak reading of the pub involves a specific set of activities where drinking must be high on the agenda. The same is demonstrated in (6) and (7) for the phrases the (tele)phone and the dictionary: these phrases have weak readings not available for semantically similar items, and, where there is a weak reading, there is obligatorily a specific activity interpretation - this type of weak definites does not have the same semantics as true indefinites. To summarize, then, there is a class of morphosyntactically definite NPs which: (a) do not have uniquely-identifiable referents (are 'weak'); (b) form a closed set of apparently arbitrary lexical items; (c) are not interpreted the same way as indefinites;
2..
Sentences marked # are judged to be infelkitous on the weak definite interpretation.
Form-function mlsmatches in (formally) definite Engllsh noun phrases (d) necessarily involve a more-than-compositional 'activity' interpretation; (e) typically occur as the complement of a restricted set of prepositions. Carlson et al. (2006) point out that most or all of these features are shared with typical cases of English bare singular count nouns (cf. Stvan 2007). Thus, 'going to church' in (8) is not the same as visiting a church in order to admire its architectureone has to attend a church service to qualifY as 'going to church: (8) a. Let's go to church. b. #Let's go to cemetery. c. Let's go to a churchi and smash iti up. d. #Let's go to churchi and smash iti up.
-:1-
Moreover, there does not have to be any specific church that is referred to here, the bare singular count noun construction cannot be freely extended to apparently similar lexical items (but see below), and nouns in this construction do not have the same referential properties as ordinary indefinites. The same points can be observed in (9). John is in prison. (9) a. b. #John is in detention centre. c. The warder was sent to a prison. -:1d. #The warder was sent to prison to start his new job. Thus the parallel that Carlson et al. (2006) point out appears to be fairly robust They do not, however, offer any explanation for why there should be this parallel, or how it could have arisen. My suggestion is that certain N Ps, particularly those that refer to purposebuilt locations or tools, are always liable to undergo a diachronic semantic reanalysis such that the activities conventionally associated with those locations or tools become a part of those NPs' semantics, rather than being merely defeasibly implicated by the utterances in which the NPs feature. Moreover, I claim that the crucial factor determining that one of these very similar constructions has no article and the other does is the timing of the reanalysis with respect to the grammaticalization of the definite article. There are certainly clear examples of activity-denoting uses of certain lexical items from the Old English or early Middle English periods - long before the English definite article becomes obligatory. Thus in (10) it is clear that it is an activity of churchservice-attending that is being referred to, rather than an event of moving towards a building, while in (11) and (12) it is clearly the activity oflearning or being under instruction that is at stake: scolasticus does not refer to just anyone who happens to occupy a school building, and it is presumably only useful to send children to a school building if they are going to be taught there. (10) [Context: you (generic readership) do X, Y,Z but you still sin]
Jm gast to chirche you go to church "You go to church:'
(C12th, Morris 1868: 23)
163
164 Christopher Lucas
(11) scolasticus se pe on scole ys 'scolasticus' that who in school is "Scolasticus [means] 'one who is at school'." (12) childru on scole betrecen nytlic ys children to school send useful is "Sending children to school is useful."
(ClOth, Zupzita 1880: 11)
(pre C12th, Chardonnens 2006: 385)
Thus it is clear that this activity reanalysis had already occurred for certain lexical items in certain contexts in this early period; and it is presumably in virtue of this reanalysis that the NPs in question are shielded from the adoption of an article once this becomes obligatory in ordinarily referential NPs in the Late Middle English period. Moving ahead to this period, it should not be surprising to find that the potential for this reanalysis is still there once the definite article has fully grammaticalized. Thus the tavem in (13) and (14) is clearly to be interpreted weakly, in just the same way as its present-day equivalent the pub in (Sa). (13) [Context: talking in generalities about the habits of'gossips']
Sum be at pe tavern IIIse in pe weke "Some [gossips] go to/are in the pub three times a week:' (Late ME, Hales & Furnival1867: 108) (14) [Context: addressing the audience, "Let everyone go their separate ways': no previous mention of any tavern in 4000 lines] For I will rin [... ]to the Tavern. (Early ModE, Chalmers 1806: 155) The only apparent difference with respect to the bare singular count noun construction is that presumably the input to the reanalysis in the case of activity-denoting weak definites must have been NPs that genuinely did satisfy semantic definiteness and were therefore marked as such with the definite article. Notice in this connection that most, perhaps all, of the items that allow for an activity-denoting weak definite interpretation in present-day English (e.g., the pub, the bank, the toilet etc.) can plausibly be supposed to have had referents that were uniquely identifiable in context around the hypothesized time of the reanalysis (cf. Abbott 2001). It should be stressed, however, that drawing these synchronic and diachronic parallels between bare singular count nouns and this class of weak definites does not entail the clearly false prediction that the use of any given present-day English noun in a bare activity-denoting NP dates back to before the grammaticalization of the definite article in the Late Middle English period. Some counterexamples to such a prediction which are clearly grammatical for many native speakers of English (as evidenced by many thousands of Google hits) include: go to mosque, go to synagogue, go to yeshiva, and go to summer camp. Clearly it would be absurd to maintain that summer camp or yeshiva were reanalysed as activities rather than locations some time before the Late Middle English period. My claim instead is that it is the innovation of the construction type 'activity-denoting bare NP' which can be dated to before the grammaticalization
Form-function mlsmatches in (formally) definite Engllsh noun phrases of the definite article. It is not that every token of this construction dates back to this period, though some clearly do, as we have seen in (10)-(12). Thus, although it remains true that new tokens of this construction cannot synchronically be freely generated using any noun you please, it is evident that new tokens of both the bare singular count noun and the activity-denoting weak definite constructions are being gradually innovated on the analogy of existing tokens whose associated activity is very close to that of the new token. Some examples of the latter type include: (15) I'm on I-55 headed north to Memphis. I'm on the cell with a friend. (http:/I axegrinder.blogspot.com/2006 _11_0 1_archive.html) (16) I'm pretty popular but I'm also always on the computer. (http:/I forums.yoyogames.com/forums/3/topics/20662 ?page= 1) (17) Is it ok to go to the ma.ll without a parent during a 'supposedly' school day? (http://answers.yahoo.com/question/index?qid=20081216084912AAjGyrM) The apparent naturalness of the reanalysis that leads to these innovations, combined with its highly contingent item-specific application, suggests that it should be observable in other languages with definite articles, but should not necessarily be expected to involve all and only the translation equivalents of the relevant English items. This does appear to be the case. Let us just note as a representative example that German i m Restaurant "in the restaurant" has a weak interpretation for many speakers, very similar to that of English the pub (Sa), but unlike its translation equivalent the restaurant (Sb). To summarize what we have seen thus far, then: activity-denoting weak definites have the morpho syntax of a definite NP, but a semantics which involves a specific activity performed in a non-specific location or using a non-specific instrument. They are tokens of a construction type that has arisen diachronically through reanalysis of genuinely definite NPs. This reanalysis is similar to that which brought about the activity-denoting bare NP construction type several centuries earlier, before the grammaticalization of the definite article. Although neither construction type allows for the free generation of new tokens in synchrony (thus resembling item-specific idioms), both are capable of gradually recruiting new tokens on the analogy of old ones (thus showing themselves not to be wholly idiomatic). Finally, it is expected that activity-denoting weak definites should be found in other languages with definite articles, but that the lexical items which allow for a weak interpretation cannot be precisely predicted.
3· Relational weak de:finites We now turn to a class ofNPs which share very little with the activity-denoting weak definites other than their weakness. These are the relational weak definites shown in (18) (=2)-(21):
165
166
Christopher Lucas (18) (19) (20) (21)
He came to the bank of a river. I stubbed my toe on the comer of a footstool. I squeezed my neighbor's thigh. The son of a known felon is coming to town.
Itis clear that the bank of a river in (18) does not refer to a unique or identifiable bank. There is no way that it could, since every river necessarily has two banks and, given that the indefinite-marked possessor a river is itself not identifiable, neither can one of its (inalienably possessed) banks be. The same is true for the other examples, all of which involve relational nouns and therefore inalienable possession. On the most salient interpretation of (19) there is no suggestion that I have only one toe, or that the one that I stubbed can be identified or is in any way particularly familiar to the hearer from the previous discourse; and footstools tend to have more than one corner if they have any at all, but the one that I stubbed my toe on is not identifiable. The same holds, mutatis mutandis, for neighbor and thigh in (20) and son in (21). The same does not hold, however, for non-relational heads (possessa) in alienable possessive constructions.3 1his is illustrated in (22)-(23), where the referents of John's plate and my golf club have to be unique or identifiable (in the sense of familiar from previous discourse or general knowledge etc.). (22) I stole John's plate. (23) My golf club is broken. These examples also illustrate the well-known fact that the head (possessum) in an (alienable) English synthetic genitive construction must be interpreted as definite, and it also cannot co-occur with an article (the so-called 'determiner genitive' (DG) phenomenon; Lyons 1999). By contrast. the correct generalization for (18)-(21) seems to be that where a possessive construction is headed by a relational noun, this noun never has to be interpreted as having a unique or identifiable referent. whether it occurs in a synthetic genitive construction as in (20) or a periphrastic ofconstruction with cooccurrence of the definite article as in (18), (19) and (21). Morphosyntactic definiteness, then, seems to have quite different semantic properties when applied to relational nouns as compared to ordinary nouns: ordinary nouns marked definite, whether by the or by a prenominal genitive as in (22)-(23), inevitably force an interpretation in terms of either uniqueness or identifiability, whereas relational nouns marked definite in the same ways do not. Hence, pace StrohWollin (this volume), it will not do to say that merely pragmatics, rather than semantics or grammar, is responsible for determining when morphosyntactic definiteness is acceptable in the absence of a unique or identifiable interpretation.
3· Note in this connection that in the English dialect of Lancashire the distinction between alienable and inalienable possession has not only semantic, but also phonological repercussions, specifically in the realization of the 1sg possessive pronoun my (Hollmann & Siewierska 2007).
Form-function mlsmatches in (formally) definite Engllsh noun phrases 167 Several linguists have noticed that relational weak definites are thus particularly problematic for a fully unified theory of morphosyntactic and semantic definiteness. However, in the spirit of (Modified) Occam's Razor, they have attempted to reclaim such cases as semantically definite nevertheless. Two of the more determined attempts at this project are Lohner (1985) and Barker (2005). To the extent that either of these really demonstrates how relational weak definites are of one semantic kind with uncontroversial cases of definiteness, this is achieved with a suspicious amount of hand waving and, crucially, the introduction of some formal- technical device that essentially sidesteps the problem. There is not space here to review their proposals in detail, but some weaknesses in each can be briefly pointed out. In the case of Barker (2005) it is instructive to consider two quotations from the second-last page of his article. First:
"A successful use of a possessive definite description, then, is one that provides enough information for the listener to reliably pick out the intended kind of object: it's the side of the box I'm talking about, not the bottom, not the top. In other words, what the speaker has in mind is a unique, specific relation, and that specificity is what the definite determiner is marking:' (Barker 2005: 110) This does not seem to capture what is going on here. Surely it is just the semantic content of the noun itself that allows the listener to pick out the intended kind of object, not the definiteness marking: all else being equal, if you say side you mean (uniquely, specifically) "side~ you do not mean "bottom" or "top". Definiteness marking generally then tells you something extra about a particular token of the kind of object denoted by the noun, and the relation of that token to other tokens of the same kind. In fact Barker (2005: 110) tacitly admits this when he goes on to say (in the second relevant quotation): "therefore, with some reluctance, I must assume that the reason possessive weak de.finites don't violate uniqueness is because the uniqueness requirement simply can't apply to relations:' Firstly, this teels uncomfortably stipulative; and secondly, it raises a difficult question, namely: why is it that relational nouns in periphrastic ofgenitive constructions (e.g., (18), (21)) get marked with the definite article if it is both the case that the definite article has one and the same meaning in all its uses (presumably something involving uniqueness), and that the uniqueness requirement cannot apply to relations? LOhner (1985) in a sense takes the opposite approach to Barker (2005), in that he argues that relational nouns, far from being exceptional, are in fact the very essence of definiteness, so it follows naturally that they are marked with the definite article.4 Logically enough, this leads LOhner ( 1985: 311) to the tollowing strongly falsifiable prediction, phrased as a statement of fact: "we have seen that, because of their inherent nonambiguity, functional nouns allow only for the definite article, except of existential
4· More precisely, he argues thls concerning 'functional' nouns, which are relational nouns that logically can only have unique referents, such as father, mother, top, bottom, centre etc.
168
Christopher Lucas
contexts:' The example in (24) (LOhner's (4)) is given as an instance of an existential context where the indefinite article modifying a relational noun is possible. (24) Does a macaque have a tail? We may reasonably assume that a similar sentence that Lohner's theory would have to declare impossible is, for example, (25). (25) ?J was given a tail of a macaque for Christmas. There is little doubt that, for many speakers, (26) is preferable to (25). (26) I was given the tail of a macaque for Christmas. However, native speaker intuitions are far from being unequivocally against examples such as (25 ). This is reflected in online attestations, where we find that there are always a sizeable minority of hits for webpages containing inalienably possessed nouns marked indefinite, as compared to pages where the same structure is marked definite. Thus, on a search of google.com on 13/06/2010 "a tail of a" generates approximately 200,000 results, e.g. (27), while "the tail of a" generates c. 2.5m.5 (27)
[Context: caption for a picture of a whale's tail] A tail of a whale in Long Beach (http:l/latimesblogs.latimes.com/unleashed/2008/08/a-tail-of-a-wha.html)
"A top of a" generates c. 60m hits, versus c. 400m for "the top of a": (28)
[Context: dialogue in a short story by a University of Maryland English teacher] But there are two sides. There's the side that wants to watch live video of a top of a. well, and there's the side that says maybe something better's on. (http://www.guardian.co. uk/books/2007/j an/16/ originalwriting.fiction3)
"A father of a" generates c. 30m hits, versus c. 70m for "the father of a": (29) Once, afather ofa dead soldier called, upset that his son's name had been misspelled on a memorial. (http://articles.latimes.com/2008/ apr/26/nation/na- soldierdad26) This is in no way only a recent phenomenon. The earliest example of a relational noun marked indefinite that I was able to find comes from 1586:
5· As is well known, Google's calculations of the total number of webpages containing a given expression are estimates, which can often be rather inaccurate and are highly liable to change from one search to another. The important point for present pwposes, however, is that, contrary to Lohner's theory, expressions of the form an X ofa Y, where X is a relational noun, are dearly grammatical for many speakers (including newspaper journalists and their editors as in (27)-(29)), although they appear to be less common than equivalent expressions of the form the X ofa Y.
Form-function mlsmatches in (formally) definite Engllsh noun phrases 169 ( 30) In the foregoing, and in the following estimates, I have not pointed out what a father of a family ought to lay by annually, as a provision for his wife and children. (Trusler 1586: 172) These data would appear to falsify Lobner's above prediction, or at least put it in serious doubt. Given, then, that it appears not to be possible to accommodate relational weak definites even by redefining definiteness for their benefit, it may be worth exploring the idea that the examples in ( 18)-(21) are not in fact definite, despite the presence in some cases of the word the. If this idea is accepted, then there are a number of ways one could proceed. One approach to explaining the form-function mismatch in these examples might be to appeal to the ban on articles co-occurring with the synthetic genitive construction, and the inherent definiteness of this construction when it is used to express alienable possession, as illustrated in (22) and (23). Perhaps, by analogy, all possessa should at least be marked definite, whether by means of the prenominal synthetic genitive or the in a periphrastic of genitive construction, even if they are allowed to be interpreted otherwise in the case of relational nouns. A good reason to think that this is not the best approach is that we also find relational weak definites in so-called AG languages ('adjectival genitive, i.e. non-DG ('determiner genitive') languages), where there is nothing inherently definite about ordinary alienable possessive constructions, and co-occurrence of possessives and articles (both definite and indefinite) is perfectly grammatical. (31) and (32) are examples of relational weak definites from two AG languages, Italian and Greek: (31) la
ri1'a di un
flume
DEF.ART bank of INDEF.ART river "the bank of a river"
oxOi enos potamu (32) s-tin to- DEF.ART bank INDERART river "to/at the bank of a river"
(Italian)
(Greek)
How, then, should we approach these cases? One useful way of thinking about them is to consider what functions an article has in ordinary cases beyond marking the definiteness status of a referent Following Hawkins (2004, 2008) we could see the grammaticalization of articles as driven at least partly by processing considerations. Thus one function of articles is to unambiguously construct a (grand)mother NP node as early as possible during online processing. From this perspective the referential properties of articles are merely a by-product of the material recruited for this purpose: a distal demonstrative and the numeral "one" in the case of English (as well as in many other languages). Thus it would not, in fact, be surprising to find clashes arising between the referential properties of certain nouns on the one hand, and the need to unambiguously mark an NP as such by means of one or other of the two articles on the other. My proposal, then, is actually based on the same intuition that LOhner has: that
170
Christopher Lucas
there is some sort of semantic incompatibility between indefinite reference and relational nouns. However, the examples in (27)-(30) militate against claiming that this incompatibility is the result of relational nouns being inherently definite. It may help to understand the true nature of this incompatibility in an intuitive sense from the example in (33). (33) ?J then encountered a side, which turned out to be the side ofa mountain. This sentence is clearly odd, arguably to the point of uninterpretability, and the source of the oddity has to do with the properties of indefinite reference. Whether indefinite reference is conceived of in terms of existential quantification or discourse referents, the route from an indefinite NP to its referent seems to be via the invocation of the category described by the determined noun and the contribution of some token of that category to the proposition expressed. More linguistic material may then express any relations obtaining between this token and other entities, but the semantics of the indefinite article is neutral to any such specific relations. In fact. it seems to presuppose that the category invoked by the noun it determines has an independent reality. However, as ( 33) clearly illustrates, the nature of relational nouns is such that the categories they invoke cannot exist independently of some other category to which they are related. Thus individual tokens of sides cannot exist independently of the entities that they are sides of. A difference between definite and indefinite reference that is crucial here is that the referent of an NP that is (semantically) definite is necessarily evoked in the context of some contextually available set in which it is unique (cf. Hawkins 1991): a relation to other entities is an inherent part of the semantics - pragmatics of definiteness. What this means is not that relational nouns are the essence of definiteness, or that it is impossible to conceive of the entity referred to by a relational noun as being non-unique or non-identifiable, but simply that there is greater common ground shared between the semantics of relational nouns and definite reference than between the semantics of relational nouns and indefinite reference. If this point is accepted along with the idea that the grammaticalization of the articles might have been driven at least partly by purely syntactic processing considerations, then we have the beginnings of an explanation for why the is the preferred NP-constructor in NPs headed by a relational noun with a postposed of-genitive. Before we develop this explanation, however, let us first consider relational nouns heading the synthetic genitive construction. As we have seen, determiners in English and other DG languages never co-occur with preposed genitives. This means that the situation in synthetic genitive constructions headed by relational nouns (as in example (20)) has remained unchanged since before the grammaticalization of the articles. That is, the uniqueness/identifiability of the head was left unspecified then, and remains unspecified now, despite the fact that the head in alienable possessive constructions is now obligatorily interpreted as definite. 6 6. What the correct explanation for this last fact is, is beyond the scope of this article, but see Lyons (1999) and Haspelmath (1999) for some competing suggestions.
Form-function mlsmatches in (formally) definite Engllsh noun phrases Returning now to the case of relational noun heads followed by the younger postposed of-genitive construction, we see that these, in contrast to relational nouns heading a synthetic genitive construction, are not immune to the general requirement since the Late Middle English period that referential NPs be marked with an NP-constructor in the shape of an article. Of course, purely from the point of view of this processingbased requirement, either article would do; but because of the semantic incompatibility suggested above we would expect the co-occurrence of a relational noun and the indefinite article to be dispreterred. By contrast, if speakers are forced to choose an NPconstructor in these inalienable possessive constructions that takes the form of an article, the looks like a better option, thanks to its inherently greater compatibility with relational nouns (the result of a semantics involving a relation to other entities in both cases, as noted above). However, once the definite article has been selected as the NPconstructor of choice with relational nouns, the requirement that anything marked with the be unique or identifiable or both (i.e., semantically definite) will have to be relaxed, since many instances of reference via relational nouns will be to entities which could never be unique and are not identifiable in context, as in (18) and analogous examples. It is worth pointing out here (following Hawkins 2008 once again) that it can hardly be seen as a serious deficit in a language's expressive power that there should be certain NPs for which it is not always possible to signal grammatically whether the head noun is definite or indefinite: for many languages this is the case across the board (but see fn.l). Thus, if we accept the idea that articles might have functions that go beyond marking the definiteness status of referents, it should also not be surprising to find articles (or pseudo-articles) that fail to mark definiteness altogether. That said, the contexts we have identified here in which morphosyntactic definiteness does not signal semantic definiteness are rather narrow, and presumably weak definites have a low token frequency as a proportion of all definite NPs. Thus it remains an overwhelmingly accurate generalization that English NPs must be marked, and interpreted, as either definite or indefinite. It may be the force of this generalization which accounts for the existence of examples as in (27)-(30), despite the oddity for many native speakers of marking the head of an inalienable possessive construction with the indefinite article. It appears that for certain speakers in certain contexts this oddity is preferable to the underspecification of definiteness that comes with marking relational nouns with the. For whatever reason, it is sometimes felt necessary to specifY that the entity in question is neither unique nor identifiable, and the only way of doing this is to mark it with the indefinite article, despite the resulting clash between the semantics of indefiniteness and the semantics of relational nouns.
4·
Conclusion
This article has discussed two classes ofNPs which differ in many respects but share a specific form-function mismatch: they are marked, but not interpreted, as definite. It
171
172
Christopher Lucas was argued that such mismatches, far from contravening the spirit of Occam's Razor, are in fact entirely to be expected, given the fact that definite articles in diachrony have a tendency to spread into contexts where they no longer signal semantic definiteness. Activity-denoting weak definites (as well as tokens of the semantically similar bare singular count noun construction) were argued to arise through a semantic reanalysis of genuinely definite NPs as referring to the activities associated with the locations or instruments to which those NPs originally referred. As such, the weakness of this class is a property of specific lexical items, giving it a semi-idiomatic character. Relational weak definites, on the other hand, were argued to arise from an incompatibility between the semantics of relational nouns and indefiniteness marking. Since this incompatibility concerns some fundamental aspects of reference, unlike the case with activity-denoting weak definites, relational weak definites should presumably be a feature of all languages with obligatory definite and indefinite articles (and of all relational nouns in those languages). Whether or not this is a valid generalization must remain a question for future work.
References Abbott, Barbara. 2001. Definiteness and identification in English. In Pragmatics in 2000: Selected papers from the 7thlntenlational Pragmatics Conference, Vol 2, Nemeth Eniko (ed), 1-15. Antwerp: International Pragmatics Association. Barker, Chris. 2005. Possessive weak definites. In Possessives and Beyond: Semantics a11d Syntax, Ji-Yung Kim, Yury Lander & Barbara H. Partee (eds ), 89-113. Amherst MA: GLSA. Carlson, Greg, Sussman, Rachel, Klein, Natalie & Tanenhaus, Michael. 2006. Weak definite NP's. In Proceedings ofNELS 36, Vol 1, Christopher Davis, Amy Rose Deal & Youri Zabbal (eds ), 179-198. Amherst MA: GLSA. Chardonnens, Laszlo. 2006. Anglo-Saxon Prognostics: A Study ofthe Genre with a Text Edition. PhD dissertation, Leiden University. Chalmers, George. 1806. The Poetical Works of Sir David Ly11dsay, Vol2. London: Longman, Hurst, Rees and Orme. Christophersen, Paul 1939. The Artides: A Study oftheir Theory a11d Use 111 English. Copenhagen: Munksgaard Culicover, Peter & Jackendoff, Ray. 1999. The view from the periphery: The English comparative correlative. Linguistic Inquiry 30(4): 543-571. den Dlkken, Marcel. 2005. Comparative correlatives comparatively. Linguistic Inquiry 36(4): 497-532.
Greenberg, Joseph. 1978. How does a language acquire gender markers? In Universals ofHuman Language, Vol 3: Word Structure, Joseph Greenberg, Charles Ferguson & Edith Moravcslk (eds), 47-82. Palo Alto CA: Stanford University Press. Grice, H. Paul 1978. Further notes on logic and conversation. In Syntax and Semantics 9: Pragmatics, Peter Cole (ed), 113-128. New York NY: Academic Press. Hales, John & Furnivall, Frederick. 1867. Bishop Percy~ Folio Manuscript: Ballads and Romances. London: N. Trubne.t:
Form-function mlsmatches in (formally) definite Engllsh noun phrases 173 Harrls, Martin. 1980. The marking of definiteness: A diachronic perspective. In Papers from the 4th Inten1atlonal Conforence on Historical Linguistics, Elizabeth Closs Traugott, Rebecca Labrum & Susan Shepherd (eds), 75-86. Amsterdam: John Benjamins. Haspelmath, Martin. 1999. Explaining article-possessor complementarity: Economic motivation in noun phrase syntax. La11guage 75(2): 227-243. Hawkins, John. 1991. On (in)definite articles: Implicatures and (un)grammaticality prediction. Journal of Linguistics 27(2): 405-442. Hawkins, John. 2004. Efficiency and Complexity in Grammars. Oxford: OUP. Hawkins, John. 2008. The typology of noun phrase structure from a processing perspective. Paper presented the RCEAL seminar, University of Cambridge, October 2008. Heim, Irene. 1988. The Semantics ofDefi11ite and Indefinite Noun Phrases. New York NY: Garland. Hollmann, Willem & Siewierska, Anna. 2007. A construction grammar account of possessive constructions in Lancashire dialect: Some advantages and challenges. English Language a11d Linguistics 11: 407-424. Jespersen, Otto. 1961. A Moden1 E11glish Grammar 011 Historical Principle.s, 7 Vols. London: George Allen and Unwin. Khan, Geoffrey. 1999. A Grammar ofNeo-Aramaic: The Dialect of the Jews of Arbel. Leiden: Brill. Lambdin, Thomas. 1971. The junctura! origin of the West Semitic definite article. In Near Eastern Studies in Honor ofW.F. Albright, Hans Goedtcke (ed), 315-333. Baltimore MD: John Hopkins University Press. Lelss, Elisabeth. 2007. Covert patterns of definiteness/indefiniteness and aspectuality in Old Icelandic, Gothic and Old High German. In Nominal Determination: Tjpology, Context Constraints, and Historical Emergence, Elisabeth Stark, Ellsabeth Leiss & Werner Abraham (eds), 73-102. Amsterdam: JohnBenjamins. Lohner, Sebastian. 1985. Definites. Journal of Semantics 4: 279-326. Lyons, Christopher. 1999. Definiteness. Cambridge: CUP. Morris, Richard. 1868. Old English Homilies and Homiletic Treatises of the Twelfth and Thirteenth Centuries. London: N. Triibner. Poesio, Massimo. 1994. Weak definites. In Proceedings of SALT IV, Mandy Harvey & Lynn Santelmann (eds), 282-299. Ithaca NY: Cornell University. Russell, Bertrand. 1905. 0 n denoting. Mind 14: 479-493. Stvan, Laurel. 2007. The functional range of bare singular count nouns in English. In Nominal Determination: Typology, Conte.r:t Constraints, and Historical Emergence, Elisabeth Stark, Elisabeth Leiss & Werner Abraham (eds), 171-187. Amsterdam: John Benjamins. Trusler, John. 1586. The London Adviser and Guide: Containing Every Instruction and Information Useful to Persons Living In London. London. Vergnaud, Jean-Roger & Zubizaretta, Marla Luisa 1992. The definite determiner and the inalienable constructions in French and Engllsh. Linguistic Inquiry 23(4): 595-652. Zupitza, Julius. 1880. JFlfrics Grammatik und Glossar. Berlin: Weidmannsche Buchhandlung.
The emergence of the definite article in English* A contact-induced change? Paola Crisma Universita di Trieste
In this paper, based on quantitative evidence collected on the YCOE, I argue that the establishment of the definite article in English is no later than the last quarter of the 9th century, being regularly used in Old English prose of the ~fredian' period. This dating enables one to exclude the possibility that the levelling of inflections, the Norman invasion or the contact with the Danes might have played a role in introducing the definite article as an innovation, and leaves as the sole possible 'external' influence the Celtic substratum/adstratum.
1.
The debate on the definite article in the history of English
In historic times, many Indo-European (IE) languages developed articles, independently of one another but along similar paths: the definite article before the indefinite one, the former in many cases a weakening of the distal demonstrative, the latter of the numeral for "one': The establishment of the definite article in English has received its share of attention, and has been alternatively attributed to the decay of the case system, to Scandinavian influence, to French influence after the Conquest. 1 1his work, based on evidence from the YCOE, 2 addresses a crucial point in the debate, namely the date of this change. There is in tact no general consensus as to when English can be said to have a definite article, the controversy deriving from the lack of agreement as to what having a definite
*
Most of the materlal of this paper was elaborated while I was spending a six-month period as a Visiting Scholar at UCLA with a Fulbright Research Grant I am particularly indebted to Donka Minkova for fruitful discussion.
1.
For an outline of the discussion, see McColl Millar (2000: 306-325).
Taylor, Warner, Pintzuk and Beths (2003). The YCOE is a tagged corpus of about 1,500,000 words comprising various prose texts dating from the late 9th to the 12th century.
2.
176
Paola Crisma
article really means. The situation of early Middle English (ME) is straightforward: it is already like Present-Day English (PDE), having two distinct morphemes, the article pel the and the demonstrative pcet!that. On the other hand, the situation in Old English (OE) is less clear-cut, for there is one morpheme- se- that sometimes corresponds to the demonstrative tha.t,3 others to the definite article the, while often it is impossible to decide between the two. 4 Since there is no definite article phonologically5 distinct from the demonstrative, opinions vary as to whether OE se can be called a definite article or not, and labels such as 'near-article' (McColl Millar 2000) or 'demonstratifarticle' (Masse 1945) have been assigned to it According to some authorities in the field (Quirk & Wrenn 1957, Mitchell1985 ), the debate over the definite article in 0 E is an "unreal problem~ a mere matter of terminology: given the impossibility of applying the categorizations used for PDE to OE, one is simply bound to conclude that "OE did not have a definite article" (Mitchell1985: 133, fn. 87).1his statement has received the strictest interpretation in some recent literature on the topic dating 'the emergence of the definite article' in the transition from the 0 E to the ME period. 6 However, this view, which maintains that a morpheme can be called a 'definite article' only if it is formally distinct from any other morpheme one can call a 'demonstrative, is certainly too extreme: in tact, it would force one to conclude that modem German, analogous to OE in this respect, 7 does not have a definite article either. It is clear that the first problem in discussing OE se is that of delimiting the properties that define definite articles, in order to check whether OE se had those properties. In this paper, I will draw the line between demonstrative and definite article along the lines of Greenberg (1978:§3.3): "the point at which a discourse deictic becomes a definite article is where it becomes compulsory and has spread to the point at which it means 'identified' in general': In this, I agree with Christophersen (1939: 83). I will also reach Christophersen's conclusion that OE at Alfred's time "possessed a real definite article" (p. 92); however, while he supports this idea with impressionistic observations,
3· Given the nature of the surviving source material, which comprises only very few passages of direct speech, the demonstrative use of se is normally anaphoric, that of a discourse deictic. The distinction between deictic use and anaphoric use may be not crucial as far as definiteness is concerned, see Heim (1982: 309, 314). 4·
In some cases it even corresponds to this, see Mitchell (1985: 132) and example (5) below.
5· The two categories mlght be just homographs, as Sweet (1905) seems to suggest: "Se is both a demonstrative and a definite article. It is also used as a personal pronoun. [... ] seas a demonstrative and personal pronoun has its vowel long." (p. 21). 6. Ackles (1997), Philippi (1997), Wood (2003), Watanabe (2009). 7· The analogy goes further: both in OE and in modern German the morpheme used as a definite article is also used as a relative or as an independent pronoun. If one accepts that modern German has a definite article, taking independent use as the demarcation criterion between article and demonstrative, as in Wood (2003:§§3.2.2-3), turns out to be untenable.
The emergence of the definite article in Engllsh 177
justly rejected by Mitchell ( 198 5: 132 ), I will provide quantitative evidence that this is indeed the case. Of course, the notion that definite articles are "compulsory" needs qualification, since it is a trivial observation that in languages that do have a definite article, the latter is not found with every noun phrase.8 I will individuate two distinct requirements that make the definite article a Kcompulsory" morpheme. The first requirement is that in languages that have the definite article (e.g. modern Romance and Germanic languages, Greek, Bulgarian, etc.) a nominal argument must not be ambiguous between being interpreted as definite or as indefinite.9 Thus, what is "compulsory" in this type of languages is the unambiguous evaluation ofthe feature [±definite] associated with noun phrases in argument function: an argument noun phrase is interpreted as definite only if recognizably marked as such (lexically, morphologically, or syntactically), and is indefinite otherwise. Strategies for marking a noun phrase as definite may vary cross-linguistically,10 but there is a generalization that is solid enough: languages of this type always have a minimal morpheme whose interpretive content is only [+definite] (plus phi-features, where applicable), i.e. some form of definite article. The requirement imposing the overt marking of [+definite] can be expressed as follows: (1) REQUIREMENT 1: [±definite] is a grammaticalized feature. A noun phrase can be interpreted as [+definite] only if overtly marked as such (lexically, morphologically, or syntactically). In languages where Requirement 1 does not apply (typically older IE languages, e.g. Latin), a nominal argument can be interpreted as definite even in the absence of overt marking distinguishing it from indefinites; these languages may have demonstratives but lack a definite article. The second requirement is that of obligatorily spelling out the category D when a noun phrase is usedinargumentfunction (subject, object, etc.), but not when the same appears as a non-argument (predicate, vocative, etc.). This argument/non-argument
Here and henceforth, 'noun phrase' ls used informally to refer to any nominal constituent, and it is neutral as to the presence/absence of functional projections (notably, a DP layer).
8.
9· I assume as a background the model of Pile Change Semantics developed in Heim (1982), which rejuvenates the traditional 'familiarity theory' of definiteness presented e.g. in Chrlstophersen (1932). Heim's model incorporates in a formal framework the pragmatic notion of conversational domain. In her model, during a conversation each listener is constantly updating a 'file' containing the information conveyed by the conversation. When an indefinite ls used, the llstener adds a 'new card' to the file, while a definite brings about the updating of an 'old card: This model makes no dlstinction between specific and non-specific indefinites, a question which I will not address. 10. The definite article can be a phrase-initial morpheme, free (as in Engllsh) or enclitic (as in Rumanian and Bulgarian), or else a lower affix (as in Scandinavian languages). Also, languages may vary as to whether they allow for alternative strategies to mark [+definite], hence for the omission of the definiteness marker.
178
Paola Crisma
asymmetry was originally noted by Szabolcsi (1987) and Stowell (1989), who argued for the existence of a systematic difference in the structural representation of nominal constituents: arguments would always require aD (an article or other determiner) to be present in the representation, while non-arguments could be simple NPs. The second requirement can be expressed as follows: (2) REQUIREMENT 2: A DP can be an argument, an NP cannot Szabolcsi and Stowell's proposal is not without problems, especially because the possibility for non-arguments to appear determinerless is not generalized. Also, its universal validity has been challenged. e.g. by Chierchia (1998), who suggested that it may be subject to parametric variation. 11 For the purposes of the present discussion, I will limit myself to noting that this requirement is certainly responsible for the "compulsory" use of the definite article in cases where definiteness is expressed by other means, and its presence can only be due to the necessity of spelling out the category D. These two requirements together characterize IE languages that have the definite article. Since PDE falls into this category and older stages ofiE languages- including Germanic- do not seem to have these properties, it is clear that some change took place at some point along the path that led to the present situation. I will argue that this happened quite early in the history of English, fur there is clear evidence that already in Alfred's times OE was subject to the requirements in (1) and (2), therefore sewas a definite article in the same sense as German der is.
2.
Requirement 2 in OE: Se and proper names
In languages that obligatorily value the [±definite] feature for each argument nominal, it is often the case that proper names (henceforth ProperNs), being rigid designators and therefore intrinsically definite, appear without an article (Longobardi 1994: 63 5).12 However, when an adjective precedes the Proper N, the insertion of an article becomes obligatory in some languages (e.g. Italian and, to some extent, German) but not in others (e.g. PDE): (3) a.
Ho incontrato (*il) Gianni13 Have-1sG met the Gianni
u. In his view, Germanic languages have an intermediate status, allowing for some arguments to be NPs. I argued (Crisma 1999) that Germanic languages are better treated assuming that they fulfill this requirement but allow for a wider use of a null determiner. 12. Some varieties may use an article before proper names, the definite article or a special morpheme (as ln Catalan). 1his phenomenon is irrelevant to the present discussion. 13. The grammaticality judgments are those of standard Italian. The definite article with unmodified ProperNs is acceptable ln many regional varieties, particularly with feminine nouns.
The emergence of the definite article in Engllsh 179
Table 1. Determiners and adjectives with ProperNs in OE prose Unmodified ProperN with se!]1es w/o se/}1e.s Argument Non-Argument
Adj + ProperN with se!]1es w/o se!]1es
1,097
43,302
1,789
(2.5%)
(97.5%)
(94.3%)
109 (5.7%)
33 (0.4%)
7,818 (99.6%)
60 (23.8%)
192 (76.2%)
b. Ho incontrato *(il) vecchio Gianni Have-1samet the old Gianni "I met (old) Gianni"
Typically, this phenomenon displays argument'non-argument asymmetries, thus the article may be absent with predicates or vocatives: (4) Si
e mascherato da vecchio Cameresi
is disguised as old Cameresi "He disguised himself as old Cameresi" REFL
(Longobardi1994:627)
Independently of the formal account that one can give of this phenomenon, the article in sentences such as (3b) has the crucial property ofbeing redundant, for it does not add to the interpretation: 14 itis an 'expletive' article in the sense ofVergnaud & Zubizarreta (1992) and Longobardi (1994), and its presence is plausibly due to Requirement2. A search on the YCOE, presented in Table 1, shows that an analogous effect of Requirement 2 is clearly observable in OE: one sees in fact that in OE ProperNs 15 are normally used without a determiner, though there are a few cases (2.5%) in which a determiner does appear. Typically, these are cases when a new character is introduced in the narrative, and then he/ she is described at some length, as in (5 ); in such cases the determiner (most often se) is a discourse demonstrative with a deictic/anaphoric meaning, and is better rendered in PDE as "this": (5) On Ualeria j:N:e.are mreglJe wres sum swilJe halig wer
In Valeria SB province was some very holy man pam wres nama JEquitius [... ] SolJlice se JEquitius SB-DAT was name Equitius [...] Truly SB Equitius
14· M witnessed by the fact that it can be altogether missing ifthe D position is filled via N-toD movement (see Longobardi 1994). 15. The name for God is consistently treated as a ProperN. It is so labeled in YCOE, and is included in the data discussed here. On the other hand. nouns labeled as ProperNs in YCOE that are used as indefinites (e.g. rej:Jiende Langbearda becomon "raiding Longbards arrived~ GDPref_ and_4_[C]:22.292.10.4323) have been excluded from Table 1.
180
Paola Crisma
for preare micelnysse his halinysse for
weariJ manigra
his holiness-GEN became many-GEN.PL mynstra feeder & la.reow on preare ylcan m~giJe monastery-GEN .PL father and master in SB same province "In the province of Valeria there was a very holy man called Equitius. [...] Indeed this Equitius, because of the greatness ofhis holiness, became the father and the governor of many monasteries in that province:' (GD_1_[H] :4.26.10.250) SE
greatness
The striking piece of evidence in Table 1 is that, when an adjective is present, the rate of determiners with Prop erN s soars to 94.3%. Some of these determiners are not redundant, as in (6), where the ProperN is not a rigid designator, for there are two kings with the same name: (6) ForiJon se cyning prere ilcan peode Sigeberht, se because SE king SB-GEN same people Sigberht, SB
cifter pam oiJrum Sigeberhte riice onfeng,
w~s freond
after
was friend
SE
Oswires
other Sigberht reign took
pres
cyninges.
Oswia-GEN SE-GEN king-GEN (Bede_3:16.224.9.2294) "Because Sigber ht, king of this same people, who ascended to throne after the other Sigberht, was a friend of king Oswia's" The vast majority of cases, however, are like the two occurrences in (7), where the determiner is only a filler of the D position, for the ProperN is intrinsically unique: 16 (7) se arwuriJa JEquitius ongann ormretlice to pancienne SE venerable .IEquitius began immensely to thank
pam relmihtigum Gode SE almighty God (GD_1_[H]:4.38.3.390) "the venerable Equitius began to greatly thank Almighty God"
Also noteworthy is the asymmetry between arguments and non-arguments: the latter, whether vocatives (cf. (Sa)), predicates (cf (8b)) or parentheticals (cf. (8c)), lack an article in 3 cases out of 4. (8) a.
Da. cwred se kyning to pan abbode. La leof Sa:xulf then said SE king to SE abbot: Oh dear Seaxulf "then the king said to the abbot: 'Oh dear Seaxwulf"' (INTERPOLATION, ChronE_[Plummer]: 656.12.39)
16. The cases in (7) cannot be assimilated to the d.eictk/anaphoric use illustrated in (5): this whole chapter of Gregory~ Dialogues is about the life ofEquitius; after he has been introduced as in (5), he is simply referred to as lEljuitius, without se, though se is found in appositional constructions, e.g.lEljuitlo pam abbode "Equitius the abbot': pam Godes peowe .A!:quitie "God's servant Equitius': AB for God, its name appears in a quasi-formulaic "thank God" construction, but he does not play any discernable active role in the episode.
The emergence of the definite article in Engllsh
b. Pret bilewite seep lJa gefullode lJone arleasan SB meek sheep then baptized SB impious/cruel Saulum, & worhte hine arfa:stne Par4lum. Saul and made him pious/merciful Paul "then, the meek sheep baptized the cruel Saul and turned him into the merciful Paul" (.tECHom_I,_27:403.90.5316) c.
heo nrefre blon
hire sceppende a:lmehtegum Gode
they never ceased their creator
almighty
God
poncunge don, thanking give (Bede_4:24.338.15.3394) "they never ceased to give thanks to their creator Almighty God" The two sharp contrasts in Table 1 (presence/absence of an adjective and argument/nonargument status) point to the conclusion that in the construction under discussion se is functioning as an 'expletive' article, whose presence is demanded by Requirement 2. The data in Table 1 cover a period spanning from the late 9th to the 12th century. A text-by-textcomparison, anyway, does not reveal any detectable diachronic development with respect to this phenomenon: earlier and later prose documents pattern alike. Some development may emerge by examining some of the purported earliest specimens of the language, i.e. Beowulfand the poetic Genesis and Exodus. Table 2 presents the pattern observed with ProperNs preceded by an adjective in these three poemsP Comparing the data in Table 1 and Table 2 has some obvious limits. First, the figures presented in the two tables belong to different orders of magnitude. Second, the language of poetry is certainly more artificial than that of prose, no matter how elaborate the latter. Third is the fact that the dating of the earliest 0 E poetry is anything but agreed upon; therefore it is not the firmest basis for the reconstruction of the diachronic development of English. Still, the results are quite clear: the grammars represented in Table 1 and Table 2 are two different grammars, the latter showing no hint of obligatory use of expletive articles with ProperNs. To the extent that these data can be taken to witness a diachronic change, one must conclude that the change was complete Table 2. ProperNs preceded by adjectives in early OE poetry Beowulf with se!]1es w/o se/}1e.s Argument Non-Argument
17.
0
20
0
14
Genesis & Exodus with se!]1es w/o se!]1es 14
0
0
Data collected on the York Poetry Corpus, www-users.york.ac.uk/-lang18/pcorpus.html
181
182
Paola Crisma
by the last quarter of the 9th century: the evidence shows that Requirement 2 became effective already at this stage. IS
3· Requirement 1 in OE: Noun phrases without an overt determiner Having shown that OE prose was like PDE in having clear effects of Requirement 2, 19 it is crucial to establish whether in OE Requirement 1 held as well namely, whether a noun phrase could be interpreted as definite only if overtly marked as such. In PD E this marking can be of various kinds: it can be an intrinsic semantic property, as with ProperNs (cf. 9a); it can be obtained by means of an overt morpheme, e.g. the definite article (cf 9b ); it can be the result of a syntactic operation of definiteness agreement with a genitive contained in the noun phrase (cf. 9c): (9) a. John left b. 'The book is on the table John's book is on the table
It is quite straightforward to show that in a given language Requirement 1 does not hold (i.e. [±definite] is not a grammaticalized feature): in order to prove it, it is sufficient to find a few instances of nominal arguments dearly interpreted as definite even in the absence of any overt marking. This proves to be the case in OE poetry. Consider the following passage, narrating the killing of a dragon by the warrior Sigmund: (10)
[... ] Sigemunde gesprong Sigemund-DAT rose
(Beow. 884)
cifter dealJdrege dom unlytel after death-day glory great
syplJan wiges since
heard
wyrm
acwealde,
war-GEN hard-NOM serpent-Ace killed
hordes
hyrde
[... ]
treasure-GEN guardian
hwrepre him yet
gesrelde
lJret pret swurd purhwod
him-DAT happened that
wrretlicne
wyrm
wondrous-Ace serpent-Ace
dryhtlic iren; draca morlJre lordly
SB
sword went-through
pret hit on wealle retstod SB
it on wall
swealt.
stayed (Beow. 892)
iron; dragon murder-DAT died.
In this 0 E turns out to be more similar to modern German than to PDE, (cf. the translation of the examples in (3)). This use of the 'expletive' article was therefore lost at some point. I will not investigate this further change.
18.
19. But see fn. 18.
The emergence of the definite article in Engllsh 183
Hcifde aglreca Had
elne
gegongen
fierce-warrior strife/courage-nAT gained
pret he bea.hhordes
brucan moste
that he ring-hoard-GEN enjoy was-able
selfos
dome; [... ]
self-GEN choice
[... ] Wynn hat gemealt
(Beow. 897) Serpent hot melted "great glory came to Sigemund after his death, since hardened by war, he killed a dragon, guardian of treasures. [...] So it befell that his sword went through the wondrous serpent and stuck on the wall, lordly iron; the dragon died, killed. The fierce warrior had obtained through his courage that he could enjoy the treasure at will. [...] The dragon melted in its heat" The dragon is first mentioned in line 886 (underlined wyrm): here no determiner is present, but the interpretation can very naturally be [-definite] "a dragon". In the following lines (891, 892, 897) the same dragon is mentioned again (boldface wynn or draca ), and the interpretation is here necessarily [+definite] "the dragon': but no determiner precedes it. Note also aglreca. in line 893, which refers to Sigmund and is thus [+definite] "the fierce warrior" and not (-definite] Ka fierce warrior". 20 This passage unambiguously shows that in the language of OE poetry, which we have seen is not subject to Requirement 2, is not subject to Requirement 1 either. Given that OE poetry and OE prose instantiate two different grammars with respect to Requirement 2 (cf. §2), one might wonder whether there is an analogous difference with respect to the effects of Requirement 1, namely whether [±definite] became grammaticalized in the language of OE prose. Proving this turns out to be a non-trivial task: basically it amounts to showing that no noun phrase is ever interpreted as definite unless it is headed by a proper name, or introduced by se, or by a genitive or a possessive pronoun (cf. 9); in other words, that cases such as those exemplified in (10) do not exist in OE prose. Taken literally, this would mean checking the interpretation of all the 81,7 30 noun phrases that satisfy the required conditions, which is obviously unrealistic. The only feasible strategy is that of selecting a random sample of manageable size, and showing that no item in the sample violates Requirement 1. The criterion chosen to select the sample was the initial letter of the head noun: 21 this criterion guarantees that all texts and periods are represented, and is at the same time blind to the semantics and additional syntactic specifications. Table 3 shows that the set of noun phrases having a head noun beginning with h-is a representative sample of the entire collection of noun phrases in the Corpus. The
20. The use of pat before swurd in line 890 ls unexpected. for the sword ls mentioned for the first time and a demonstrative would be inappropriate. Here it may be a case of dittography.
21.
ProperNs are excluded.
184 Paola Crisma Table 3. Distribution of various types of noun phrases not headed by a ProperN 'D-less' Nouns
D+N
Q/NUM+N
Gen/Poss+ N
XXX+N
TOT
all
81,730
26,390 (10.7%) 1,623 (8%)
60,005 (24.2%) 4,944 (24.3%)
678 (0.3%)
#h-
78,668 (31.8%) 6,752 (33.2%)
247,471
33% 6,944
72
20,335
34.1%
(0.4%)
distribution of the various types of noun phrases 22 is in fact virtually identical in the two groups. A sample of 6944 tokens can still be discouraging, but in fact it contains a lot of items that can be excluded a priori. First, all non-arguments (cf. Requirement 2 above). Second, noun phrases appearing as adverbials and as complements of prepositions: in fact, even in languages with a definite article, these two categories are often characterized by a great deal of idiosyncrasies as to its use, possibly due to their uncertain reterential status (She goes to school vs. She goes to the restaurant, where it is not obvious that school or restaurant denote a particular building). For this reason, even when investigating Requirement 1 in modern languages, the safest information comes from noun phrases occurring as subjects or objects: it was then only natural to restrict the analysis of OE to these two kinds of arguments. This way the sample was reduced to 1300 tokens, 23 whose interpretation was then checked against Requirement 1.
3.1
Bare nouns (Carlson 1977)
About two thirds of the tokens in the sample turned out to be 'bare' nouns in the sense of Carlson (1977), namely D-less plural or mass nouns with an existential (as in lla) or a generic (as in 11 b) interpretation: (11)
a.
[...] pret hyrdas wrenm on pam ilcan Iande wreccende that shepherds were on SE same land awake "[we heard] that shepherds were waking in the same land" (ScraggVerc_5:151.931)
11. The types of noun phrases are (from left to right, skipping the first column): noun phrases with se or pes as the first element, noun phrases with a quantifier or a numeral as the first element, noun phrases with a genitive or a possessive pronoun as the first element, noun phrases with miscellaneous material as the first element. In the first column in the table, the relevant group for the present discussion, namely noun phrases that do not have any of the preceding categories as the first element. They will be referred to as 'D-less' noun phrases. 13.
Representing 6.3% of the total of noun phrases satisfying the required conditions.
The emergence of the definite article in Engllsh 185 b. Dumbe hundas ne magon beorcan dumb dogs NEG be-able bark "Dumb dogs cannot barl(
(CP:15.89.16.578)
Since bare nouns are [-definite], they do not constitute a violation of Requirement 1, and are common in languages that have a definite article. Alongside bare nouns virtually identical to PDE ones, one finds bare singular count nouns, as in the examples in (12): (12)
a.
& wreg mid hine twiecge handseax gerettred and carried with him two-edged dagger poisoned "and [he] carried on him a poisoned two-edged dagger" (Bede_2:8.122.11.1155)
b. Sua sua mid lilJre wislJlunga mon hors gestillelJ, sua eac so so with soft whistling MAN horse appeases so also
mid lJre,.e illcan wistlunga mon mreg hund astyrigean with S1l same whistling MAN may dog stir-up "just as with soft whistling one appeases a horse/horses, so with the same (CP:23.173.21.1178) whistling one may stir up a dog" Bare singular count nouns do not constitute a violation of Requirement 1 either: they are interpreted as existential or generic indefinites, as shown by the fact that in these sentences PDE uses the 'indefinite articli. 24 Table 4 presents the frequencies 25 of the various types of bare nouns.
14. The latter ls probably better analyzed as a morpheme marking a +value for a [±count] feature rather than as a marker of [-definite]: in fact, in languages where Requirement 1 applies, the indefinite reading ls the default one, which obtains every time a noun phrase is not overtly marked for [+definite] (cf. Ackles 1997 and Crlsma 1997, 1999). Languages that have no 'indefinite article: on the other hand, are characterized by a potential ambiguity between a mass and a count reading for singular nouns: i. Eg borlJalJi hdkarl a ]Jorrabl6tinu (Icelandic) I eat-PAST-1sG shark at winter-festival-DEF "I ate a shark/shark meat at the Wlnter Festival" 15. There are a few cases of ambiguity between a singular and a plural reading (e.g. hors in example (12b), between amass or a count reading (e.g. hlaf"bread!a loaf"), or between an existential or a generic reading, as in (i): [... ] & of horses jlasce cumatJ eac hyrnetta [... ] and from horse-GEN flesh come also gadflies (lEHom_1:267.139) "[and from ass flesh wasps come] and from horse flesh also gadflies come" In such cases, the ambiguous token has been counted in the more numerous category. i.
186
Paola Crisma Table 4. 'Bare nouns' with non -definite interpretation
Plural Singular Mass Sl..ngular Count
Existential
Generic
236
17
569
11
194
14
Their total amounts to 1041, about 80% of the sample. Thus, out of the original1300 D-less noun phrases, we are left with 258 which might potentially constitute a violation of Requirement 1.
3.2
Special lexical items
Among the remaining potential violators of Requirement 1, the most numerous group is made up of D-less nouns that seem to behave as Prop erN s, though they are not labeled as such in the YCOE. Among these are heafon "heaven': hell "hell" and compounds thereof. 26 Also two metonyms for sacraments, husel "housel" (for eucharist) and haligreft "veil" (for ordination), are normally found without se (35 tokens), unless preceded by an adjective or followed by a relative clause. A peculiar case is that of Hrelend "Savior": normally it is preceded by se, 27 but there are 41 instances, mostly in the BUckling Homilies, without se. Here Hrelend is probably assimilated to Dryhten "Lord': which is normally used as a ProperN (and is so labeled in the YCOE). In all, these ProperN -like elements, all denoting abstract or concrete entities related to religion, amount to 219 tokens. Among 'special lexical items: one is somehow problematic: higon!hiwan, a plural noun meaning "community': In 11 occurrences without se it seems to receive a definite interpretation. It is certainly not a coincidence that all these occurrences are found in legal documents like wills and laws, which suggests that this use was a peculiarity of the genre. 3·3
Other D-less noun phrases
Excluding bare nouns and special lexical items, one is left with 29 D-less noun phrases, all interpreted as definite. Of these 29 cases, all but 3 fall into precisely delimited categories, which can occur without a definite article in many languages that otherwise comply with Requirement 1. A detailed formal treatment has been provided in Crisma (2008). To sum it up:
26. Note that the PDE developments are also used without the article.
27. Unless it is used together with the ProperN Crist, as in Hrele11d Crist.
The emergence of the definite article in Engllsh 187 A. 7 tokens are coordinations of two D-less noun phrases: (13)
[... ] pa. w~pna to pam ga.stlicum weorce [...] pret the weapons for SB spiritual work that
synd pa halgan bee:
saltere and pistolboc,
are the holy
books: psalter and book-of-the-epistles
godspellboc and
m~.sseboc,
sangboc
and handboc,
gospel and missal singing-book and handbook "[he shall have also] the weapons for the spiritual work [before he is ordained], namely the holy books: psalter and book of the epistles, gospel and missal, (£Letl_[Wulfsige_Xa] :52.64) singing book and manual .. ~ In such cases, a definite interpretation is admitted also in PDE and many other languages that have a definite article (Heycock & Zamparelli 2003, Roodenburg 2004).
B. 1 is a singular (mass) noun phrase introduced by swelce Ksuch': again a possibility admitted in PDE: "such books~ "such discourse-anaphoric, interpretation:
wine~
"such a boo!.(, all receive a definite,
(14) ne a.n his mode ne gebrohte swelce hreowsrlnga NEG on his mind NEG brought such repentance Knor brought [he] such repentance on his mind"
(CP:53.415.1.2861)
C. 4 tokens are marked as definite via long-distance agreement of a null D with a postnominal definite genitive (as in Icelandic, in Celtic and Semitic) rather than locally with a prenominal one (as in PDE): (15)
cycenan, [... ] pret prer stod gesund hus prere that there stood sound house SE-GEN.SG kitchen "[and they saw] that the undamaged building of the kitchen stood there" (GD_2_[C]:10.124.7.1498)
In presenting a formal account thereof, Crisma (1997, 1999) has argued that this possibility is to be typologically related to the absence of the 'indefinite article' and of the obligatory marking of [+count]. D. 6 tokens are kind-referring singular count nouns: (16) Hwrel is ealra fixa mrest whale is all-GEN.PL fish-GEN.PL greatest "The whale is the greatest of all fish"
(£LS_[Maccabees] :572.5204)
This possibility is also likely related to the absence of obligatory [+count) marking, as these noun phrases are the expected counterpart ofkind-reterring bare mass/plurals of PDE, modulo such a minimal difference.
188
Paola Crisma
E. 4 tokens of inalienably possessed nouns might fall into the same pattern: (17) to prere stowe lreded wres, prer him mon to SE place led was where him-DAT MAN
sceolde
heafud oftlean
must-PAST head cut (Bede_5:17.456.5.4579) "and was led to the place where he was to be beheaded" In these cases, if the identification of the referent does not necessarily hinge on the definite article, and the latter occurs as an expletive in this construction, as suggested by Vergnaud & Zubizarreta (1992), its function would reduce to neutralizing the [-count] reading of a singular noun. Hence, it could be omitted in languages without obligatory marking of [+count].
F.
2 nouns with superlative adjectives, which extend to include semantically analogous adjectives (denoting uniqueness of position in an ordered set), like ordinal numerals and the word for "other" (2 tokens). (18)
[ ... ] &
gecunnian, hwelc heora
swiftost hors haifde
and test which them-GEN swiftest hors had "[that he allowed them to run a race] and test which of them had the swiftest horse" (Bede_5:6.398.30.3999) These adjectives too might possibly induce semantic definiteness via long-distance agreement with a null D, but again only in languages where the latter need not be obligatorily marked for [+count]. The 3 cases that do not receive a formal account are the following: (19) a.
Witodlice eahta synt heafodleahtras, indeed eight are capital sins "indeed, the capital sins are eight"
(ScraggVerc_3:19.402)
wu.lj; on rermergen he itelJ hlolk b. Beniamin is risende Benjamin is rapacious wolf on day-break he eats booty &
on aifenne herereaf drelelJ
and on evening spoil shares "Benjamin is a greedy wolf: in the morning he eats (the) booty and in the evening he shares (the) spoil" (Bede_1:18.92.10.846) (19a) is problematic only in the proposed translation, but alternative translations with a PDE bare plural, such as "Capitals sins are eight" or "There are eight capital sins~ are conceivable. (19b ), a quote from Genesis (xlix 27), is part of a sort of oracle known as the "Blessing of Jacob". Its interpretation is obscure, as it fits oracles. Moreover, as is often the case with poetic passages of the Old Testament, it is filled with unfiuniliar
The emergence of the definite article in Engllsh 189
allusions, and the Hebrew original uses rare lexemes of unclear meaning. 28 It is obvious that this passage instantiates a special register, and the omission of articles might just be a peculiarity of this register. In sum, almost all the 1300 D-less nominals examined are either bare nouns or are headed by nouns assimilated to ProperNs, and there are 8 D-less noun phrases interpreted as definite in constructions admitted also in PDE (see points A and B above). Potential problems for Requirement 1 are 21, 1.6% of the total, a rate that in itself is sufficient to show that 0 E prose was not like Latin or like the language of Beowu.lf (see (10)). Furthermore, 18 out of these 21 cases (points C-F above) may receive a formal account, and, most interesting, a unitary account: I proposed that they can be traced back to the absence of 'indefinite article' in OE, i.e. of obligatory marking of [+count]. This survey has therefore shown that OE prose, different from the earliest OE poetry, was subject to Requirement 1.
4· The possible causes for the emergence of the definite article
The evidence presented has shown that, by the late 9th century, OE prose, unlike earlier poetry, was subject to both Requirements 1 and 2, i.e. it had a definite article (but not an 'indefinite' one, like e.g. Icelandic, Celtic and Semitic). This dating allows for some speculation on its 'emergence'. I will take as a starting point Keenan's (2002: 327) Inertia, the idea that grammatical change does not happen without some external cause (where decay is also reckoned among external causes). On the basis of the proposed date one can exclude some potential causes. First, the leveling of inflections: though there are minor differences between the inflectional system in Beowulfand that of prose, the two systems are functionally equivalent, it is only at a much later stage that massive syncretism and simplifications are observed. Second, semantic bleaching: an account along these lines is also untenable, for in OE prose se was still a full-fledged demonstrative in many contexts, and was still both a relative and an independent pronoun. Hence, neither phonological reduction nor semantic bleaching plausibly played a role in the establishment of the definite article. Another possible cause of change compatible with Inertia is language contact Timing enables one to exclude any influence of the Normans, but also of the Danes: it is only with .iEthelred's reign (865-871) that the Danes began to settle in Britain, after a period of occasional raids and plunders (hardly a tavorable situation tor language contact). Since Alfredian prose dates from 890, it is probably too close for Scandinavian to have had any real effect. Inddentally, note that the records of the language at the relevant period are insufficient to establish whether Scandinavian had the definite 28. In this particular case, hloOe translates Latin praedam, but the meaning of the original Hebrew 'ad is not dearly established; the Septuagint, in fuct, simply omits it.
190
Paola Crisma
article at the time. The only remaining possibility is that of Celtic influence. which has been advocated for other grammatical properties by the proponents of the 'Celtic hypothesis' (Tristram 2004, Filppula, Klemola and Paulasto 2008 a.o.). What makes this option extremely attractive is the surface resemblance of the determiner systems in 0 E and Old Welsh and other old Celtic languages: a phrase-initial definite article; no 'indefinite article'; definiteness inheritance from genitives and possessives; the possibility of having such definiteness inheritance from a postnominal position (across the board in Celtic, sporadically in OE- point C above); optional use of the definite article before ordinal numbers (point F). Of course, it is only with a systematic comparison of OE and Celtic languages, analyzing all the structures in C- F. that this attractive hypothesis could be turned into a sound and solid one. This is left for future research.
Sources JEHom: Pope, John C. 1968. Homilies oflElfric, A supplementary Collection. EETS 260. London: Oxford University Press. lECHom_l: Clemoes, Peter. 1997. lElfric's Catholic Homilies: The First Series. EETS s.s. 17. Oxford: Oxford University Press. lELetl: Fehr, Bernhard. 1914. Die Hirtenbriefe Aelfrics in Altenglischer und Lateinischer Fassung. Bibliothek der Angelsaechsischen Prosa. IX: 1-34. Hamburg: Verlag von Henri Grand lELS: Skeat, Walter W. 1966 (1881-1900).1Elfric's Lives ofSaints. EETS 76, 82, 94, 114. London: Oxford University Press. Bede: Miller. Thomas. 1959-1963 (1890-1898). The Old English Version of"Bede's Ecclesiastical History of the English People". EETS 95, 96, 110, 111. London: Oxford University Press. Beow: Dobbie, Elliott Van Kirk 1953. Beowulf and Judith. The Anglo-Saxon poetic records, IV. New York: Columbia University Press. ChronE: Plummer, Charles. 1965 (1892-1899). TWo of the Saxon Chronicles Parallel. Oxford: Clarendon Press. CP: Sweet, Henry. 1958 (1871). King Alfred's West-Saxon Version of Gregory's Pastoral Care. EETS 45, 50. London: Oxford University Press. GD, GDPref. Hecht, Hans. 1965 (1900-1907). BischofWa:rferth von Worcester Obersetzungder Dialoge Gregors des Grossen. Bibliothek der Angelsaechsischen Prosa. V. Darmstadt: Wissenschaftliche Buchgesellschaft:. ScraggVerc: Scragg, Donald G. 1992. The Vercelli Homilies and Related Texts. EETS 300. Oxford: Oxford University Press.
References Ackles, Nancy M. 1997. Historical Syntax of the English Article in Relation to the Count/Noncount Distinction. PhD dissertation, University of Washington.
The emergence of the definite article in Engllsh Carlson, Greg N. 1977. A unified analysis of the English bare plural. Linguistics a11d Philosophy 1: 413-456. Chierch!a, Gennaro. 1998. Reference to kinds across languages. Natural La11guage Semantics 6: 339-405. Christophersen, Paul. 1939. The Artldes: A Study of their Theory and Use in E11glish. Copenhagen: Munksgard Cr!sma, Paola. 1997. L'artkolo nella Prosa Inglese Antica e la Teoria degli Articoli Nulli. PhD dissertation, Universita di Padova. Cr!sma, Paola. 1999. Nominals without the article in the Germanic languages. Rivista di Grammatica Ge11erativa 24: 105-125. Cr!sma, Paola. 2008. Pursuing an old will-d-the- wisp: The nature of Old English se. Paper presented at DIGS X, Cornell University,. Ithaca NY, August 2008. Filppula Markku, Klemola, Juhani & Paulasto, Hell (eds). 2008. English and Celtic in Contact. London: Routledge. Greenberg, Joseph H. 1978. How does a language acquire gender markers. In Universals of Human Language, VoL 3, Joseph H. Greenberg, Charles A Ferguson & Edith A. Moravcslk (eds), 47-82. Stanford CA: Stanford University Press. Heim, Irene. 1982. The Semantics of Definite and Indefinite Noun Phrases. PhD dissertation, Univers tty of Massachusetts, Amherst. (Also published by Garland, 1988, New York NY). Heycock, Caroline & Roberto Zamparelli. 2003. Coordinated Bare Definites. Linguistic Inquiry 34: 3.443-469. Keenan, Edward 2002. Explaining the creation of refiexlve pronouns in English. In Studies In the History of the English Language, Donka Minkova & Robert Stockwell (eds), 325-354. Berlin: Mouton de Gruyter. Longobardi. Giuseppe. 1994. Reference and Proper Names. Linguistic Inquiry 25(4): 609-665. McColl Millar, Robert 2000. System Collapse System Rebirth. Bern: Peter Lang. Mitchell, Bruce. 1985. Old English Syntax, Vol. I. Oxford: Qarendon Press. Mosse, Fernand 1945. Manuel de lbnglais du Moyen Age: Des origines au XIVe siecle. Tome I: Vleil-anglais. Paris: Aubier Montatgne. Philippi, Julia. 1997. The rise ofthe article in the Germanic languages. In Parameters of Morphosyntactic Change, Ans van Kemenade & Nigel Vincent (eds), 62-93. Cambridge: CUP. Quirk, Randolph & Wrenn, Charles Leslie. 1957. An Old English Grammar. London: Methuen&Co. Roodenburg, Jasper. 2004. French bare arguments are not extinct: The case of coordinated bare nouns. Linguistic Inquiry 35(2): 301-313. Stowell, Timothy. 1989. Subjects, specifiers, and X-bar theory. In Alternative Conceptions of Phrase Structure, Mark Baltin & Anthony Kroch (eds), 232-262. Chicago IL: University of Chicago Press. Sweet, Henry. 1905. An Anglo-Saxon Primer with Grammar, Notes and Glossary. Oxford: Clarendon Press. Szabolcsi, Anna. 1987. Functional categories in the noun phrase. In Approaches to Hungarian 2, IstvanKenesei (ed), 167-190. Szeged: JATE. Taylor, Ann, Warner, Anthony,. Pintzuk. Susan & Beths, Frank. 2003. The York-Toronto- Helsinki Parsed Corpus of Old English Prose. <www -users.yorkac.uk/-lang22/YCOE>
191
192
Paola Crisma
Tristram, Hildegard L.C. 2004. Diglossia in Anglo-Saxon England, or what was spoken Old English like? Studia A11glica Posnanie,ISia 40: 87-110. Vergnaud, Jean-Roger & Zub!zarreta, Maria-Luisa. 1992. The definite determiner and the inalienable constructions in French and English. Linguistic Inquiry 23: 595-652. Watanabe, Akira. 2009. A parametric shift in the D-system in Early Middle English: Relativization, articles, adjectival inflection, and indeterminates. In Historical Syntax and Linguistic Theory, Paola Crisma & Giuseppe Longobardi (eds), 358-374. Oxford: OUP. Wood, Johanna L. 2003. Definiteness and Number: Determiner Phrase and Number Phrase in the History of English. PhD dissertation, Arizona State University.
On the syntax of Romanian definite phrases Changes in the patterns of definiteness checking Alexandra Cornilescu & Alexandru Nicolae University of Bucharest & "Iorgu Iordan Institute of Linguistics, Bucharest
Al. Rosetti"
We discuss a type of variation in the pattern of definiteness valuation in Old Romanian (XVIth to XVIIIth century), which has never been noticed before, and examine its significance for the evolution of the DP. In Old Romanian, the suffixed definite article variably occurs either on the first or ott a lower [+N] constituent (noun or adjective), so that an indefinite adjective may precede the definite noun. In contrast, in Modern Romanian, it is always the first noun which bears the definite article, while, in case an adjective precedes the noun, the definite article occurs on the adjective. The existence of this lower (definite) article raises several questions (the contexts where it occurs, its significance for the emergence of the enclitic definite article, etc.), to which this paper provides tentative answers. We propose that the existence of a lower definite article combined with a tendency for economy made possible the extension of the use of the article to (pre-nominal) adjectives, gradually leading to stricter conditions in the valuation of definiteness in Modern Romanian (Local Agree). At the same time, the lower article is evidence that the Romanian enclitic definite article originates in a post-posed demonstrative.
1.
1.1
Aim of the paper. Theoretical assumptions The problem
It is well known that in Modern Romanian (=MR) the definite article -(u)l is a suffix whose position is fixed: it always occurs on the first N (oun) or A( djective) in the group. This distribution is illustrated in (1): (1) a. fata frumoasd. girlthe beautiful "the beautiful girl"
b. frumoasa fata beautiful. the girl "the beautiful girl"
194
Alexandra Cornilescu & Alexandru Nicolae
*fata frumoasa beautiful.the girl
d. *frumoastl beautiful
fata girl. the
In Old Romanian (=OR), however, the definite article variably occurs either on the first Nor A, or on a.lawer N, so that another nominal constituent. tor instance, an (indefinite) A may precede the definite N: (2) OR: spre tictlloase crwintek_ mele a.udzul t"fi pleaal... (Cantemir) to VICIOus words. the my hearing your turn "Lend your ear to my vicious words." (MR: spre tici:tloasel~ cuvinte ale mele audzul tti pleactl ... ) This type of variation has never been discussed before (cf. Starn 2008). It is the a.im of this paper to explore it and to examine its significance for the evolution of the Romanian DP. The article which appears on a constituent different from the first Nor A will be referred to as 'the lower article' for convenience. Before describing the lower definite article, it is appropriate to present our assumptions regarding definiteness checking in MR, so as to later identify the source of the contrast between the two stages in the evolution of the D P. 1.2.
Theoretical assumptions
Definiteness as an (un)interpretable feature It is likely that in UG, the D head is uninterpretable [~] and interpretable definite (i.e., [u~. idef] ), since, in as much as it is interpretable, definiteness is tied to "referentiality" Thus, definite DPs (proper names, pronouns, definite and demonstrative descriptions) have determined reference (Farkas & von Heusinger 2003, Farkas & de Swart 2007), requiring unique discourse referents. From a syntactic perspective it is the D-layer which secures argumenthood (Stowell1989; Longobardi 1994; Giusti 1996, 2005; Borer 2005). From a semantic perspective, in theories like DRT (Kamp & Reyle 1993, and references above), in aD+ NP structure, it is the D which introduces the (unique) discourse referent. while the NP supplies a descriptive predicative condition. Following the theoretical suggestions ofPesetsky & Torrego (2007) on the syntax of valuation. definiteness will be considered a nominal property, uninterpretable on the noun ([udef] ) and interpretable (though unvalued) on the determiner ([idef]). Yet. definiteness may be valued on certain types of nouns, i.e., certain categories of nouns may be marked as [u + def] from the lexicon. In UG, proper names are inherently [+definite] and value the [idef] feature of D, as proposed in Longobardi (1994). Similarly, we propose that in languages where nouns morphologically vary for definiteness, like Romanian, nominal morphology may supply the value of the definiteness feature in D. Concluding, definiteness in Dis interpretable and unvalued [idef], and it will be valued either by external merge of a lexical determiner or by internal merge of an NP/ DP which is morphologically definite, such as a noun suffixed by the definite article. 1.2.1
On the syntax of Romanian definite phrases 195 Assuming that feature valuation is consequent upon external merge of the article, definiteness valuation for a language that has free-standing definite determiners like English might look like the following: (3)
DP
~
D [utjl]
NP [icjl]
[i + def] the
The D head agrees with theN head valuing its cj>-features. At the same time. the definite article values the [idef] feature of the D head. In agreement with other analysts, we assume that in MR, the enclitic article -(u)l is a suffix (cf. Ortmann & Popescu 2000 among many.), subcategorized for an N-1 Acomplement, with which it merges in the lexicon: (4) a. trandafirul (jrumos) rose.the (beautiful) "the beautiful rose"
b. frumosul trandafir beautiful.the rose "the beautiful rose"
Let us detail the mechanism of definiteness valuation. When the N is suffixed with the definite article. the result is a definite noun, therefore an NP which is valued for definiteness, marked [u + def, icj> ]. These features of theN are used to value the corresponding interpretable (but unvalued) features of the D head, as shown below: (5)
DP
~
D [ucjl] [idef]
Agree
NP [icjl] [u + def] trandafirui
As mentioned, the definite article may also be suffixed to an A-head. Importantly, MR As may be suffixed with the definite article only when the AP is attributi11e and pre-nominal; in such cases, the A merges as a specifier of the NP, so that the adjectival head is in a configuration oflocal Agree with the NP: (6) frumosul trandafir beautiful.the rose "the beautiful rose"
When the A is predicative (e.g., a post-copular predicative (7a) or a post-nominal adjunct (7b )), the definite article is impossible: (7) a. *Trandafirul este frumosul. rose. the is beautiful.the
b. *trandafir(ul) frumosul rose(.the) beautiful
196
Alexandra Cornilescu & Alexandru Nicolae
Locality of agree and move in modern Roma.nian DPs Definiteness is therefore an agreement feature for As. We assume that MR As, by virtue of being -complete, may also bear an [udef] feature which is never valued by As themselves, but may be valued by a definite N. The A probes the nominal that it c-commands and will agree with the N in definiteness and <1>- features, so that the A, which enters the derivation [ucj>, udef], may end up being [u<j>, u + def], its features being thus identical with theN's. When this happens, phonology always realizes [u + def] on the highest copy below D, i.e., the highest N or A below D, which will bear the definite article at PF. This highest copy is the one that values the [idef] feature of the D head. The process of definiteness feature transmission assumes the form of a series of Agree relations ((8b), (8c)); finally, the definite A immediately below D values the [idef] feature ofD (8c).
1.2.2
(8) a. frumosul
trandafir beautiful.the rose "the beautiful rose"
b.
FP
~
AP
F'
I~
A
F
NP
I
N
[u+J
[i+l
[u + d.ef] jrllmosul
[u+def] trandafir Agree
c.
DP
~
FP
D
~
I
[u+J [i + def]
AP
I
A
F'
~
F
[u+J [u + def]
NP
I N
[i+l jrllmosul
[u+ def] trantkjir
Agree
The description given above shows that in MR Agree is strictly local, as in (9):
On the syntax of Romanian definite phrases 197
(9) Definiteness valuation in Modern Rotnatdan -Local Agree (LA)): The [+def] GoalP which values [idef] in D must be a [+N] phrase immediately below D. A definite AP/NP must end up immediately below the D, in a position where the [u + detl feature of the definite N/A is accessible to D head tor strictly Loca.l Agree. This is why (lc) and (ld) are impossible. The article is too remote to function as a Goal for the Probe in D. There is an opaque intervener (the NP in (lc), the AP in (ld)) between the Probe and the Goal Pairs like (la) and (lb) (=(10)) represent different derivations ((lOa) has the derivation in (8c) above; (lOb) has the derivation in (11)- irrelevant details aside), both observing the same generalization in (9): ( 10) a. frumosul
trandafir beautiful the rose "the beautiful rose"
(11)
b. trandafiml frumos rose.the beautiful "the beautiful rose"
DP
~
D
NP
[u4J] [i + def]
~ NP
I
AP
I
N
A
[i4JJ [u+ def] trandajind
[u4J] fru11Ws
Agree
We conclude that in MR, it is the first [+N] constituent (N or A) of the DP which values the feature in D, by Local Agree. Romanian and French contrast with English in the following modifier + proper name structure: (12) a. le vieux Paris the old Paris "old Paris"
b. vechiul Paris old.the Paris "old Paris"
old Paris Unlike the two Romance languages, English allows Long Distance Agree (LDA), in the sense that the N which values the feature in D may be separated from D by an intervening adjective. French and Romanian disallow LDA, but use alternative strategies to value definiteness: a free standing definite article (French) or a definite article suffixed on an A immediately below D (Romanian). Locality of Move For the analysis that follows, it is also important to mention one more property of the MR DP related to the post-nominal Dem(onstrative) and, more generally, to the application of DP-internal Move. We claim that Move, like Agree, is local; it cannot cross a higher specifier. As to demonstratives, like Spanish, MR disposes ofboth pre-nominal (13a) and post-nominal Dems (13b ). The former are similar
198 Alexandra Cornilescu & Alexandru Nicolae
to their Romance counterparts and need no discussion. The latter have special locality constraints, namely, the post-nominal Dem can only be preceded by a definite N, strictly adjacent to it (13b) (see the ungrammaticality of (13c) and (13d)). (13) a.
acest (joarte) frumos
portret
this (very) beautiful portrait "this (very) beautiful portrait" b. portretul acesta (foarte) frumos portrait this (very) beautiful "this (very) beautiful portrait" c. *portretul (joarte) frumos acesta portrait (very) beautiful this
d. *(foarte) frumosul acesta portret (very) beautiful this
portrait
Strict adjacency of the post-nominal Dem to the definite head has standardly been interpreted as an instance of definite N°-Movement to D across a phrasal Dem in the specifier below D, as shown in the intermediate configuration (14) (cf. Cornilescu 1992, Dimitrova-Vulchanova & Giusti 1998 among many). (14)
DP
~ D' ~
D [idef]
[u-+J
[+EPP]
FP
~ DemP
F'
I~
Dem [u+def]
[u+J
F
NumP
~
Num
NP
I
N [u+def]
[i+l porlreftd acesta
The examples in (13) show a sharp contrast between APs and NPs in the demonstrative construction. In contrast to definite Ns, definite As cannot move across the DemP, as known since Grosu (1988). In MR, As always move as phrases (13d) and, if Move is local, APs cannot move past the demonstrative specifier. This raises the issue of why N-Raising is possible, while AP-Raising is not. Notice that the DemP and the definite
On the syntax of Romanian definite phrases 199 NP/AP share their[~) and [+def]-features, which makes them equally good goals for the probe in D. Consequently, it is the DemP, which is closer to D, that should value definiteness. This is what blocks the raising of the AP, as expected. Once again, the definite NP and the DemP are equally good goals for the probe in D, but theN-head is lighter and as such is preferred as a mover. The preference for N°-Movement thus follows from a principle of economy which requires piping only as much material as is necessary for convergence (Pied Pipe Less Weight - Stateva 2002, or Attract/Move Smallest- Akiyama 2004). The N successive-cyclically moves crossing the DemP and ends up in D. Conclusion What has been said so far proves that the syntax of the MR DP shows locality constraints on both Agree and Move.
Long distance agree in Old Romanian
2.
2.1
The lower definite article
Against the background sketched in Section 1, we may now turn to the variation in the pattern of definiteness valuation in OR (XVIth to XVIIIth century), which we mentioned above and which has gone unnoticed so far; no mention of it is made of it in important histories of Romanian (Ghepe 1975; Rosetti 1968; Densusianu 1961; Dimitrescu 1978; Brancu~ 2004; Niculescu 1990 ). In 0 R, the definite article variably occurs either on the first [+N] constituent or on the head N itt a lower position; another nominal constituent, for instance, an A, will precede the definite N in such cases: (15) ca mare scdf}netul
roatelor like great grinding.the of the wheels "like the strong grinding of the wheels"
(Cantemir)
Examples of this type show that locality conditions are not so strict in OR, allowing a different pattern of definiteness checking: (16) Definiteness valuation in Old Romania, -Long Distance Agree (LDA): The goal that values the probe in Dis a c-commanded nominal phrase (NP, AP) which need not be the first (nominal) phrase c-commanded by D. This amounts to saying that the [+def] feature is realized either on the first or on a lower nominal constituent of the DP. The definite article which shows up on aN which is not the first constituent of the DP is what we have called the "lower definite article". 2.2
The extension and range of the lower definite article
The lower definite article is present from the oldest Romanian texts of the XVIth century up until the first half of the XVIIIth century. This phenomenon thus belongs
200
Alexandra Cornilescu & Alexandru Nicolae to OR, as opposed to MR, assumed to start at the end of the XVIIIth century (cf. Ghetie 1975): (17) a.
...au venit egumenul de Bistrifd has come abbot.the of Bistrita
cu cinstitd cartea mdrii tale (XVIth- DlR) with honored letter.the highness.theGen your "...the Abbot of Bistrita came with your highness' honored letter" b. au ajlat cap ~i fncepatura. mo§ilor [... ] they.had found head and beginning ancestors.the.Gen
ca sa
nu se tnece
so that not be drowned a
toate ftlrile
ani!
trecufi
(XVIIth, 1641- Ureche)
of all countries.the years.the passed "They found the origin and the beginning of their ancestors so that the passed years of all countries may not be drowned in oblivion" c. plecat robul Mdriei Tale, Radu logofat. .. humble servantthe highness.the.Gen Your, Radu Chancellor (XVIIth, 1688- Biblia) "your highness's humble servant, Radu Chancellor"
d. Umblam dupa go.we
after
a
lumii
i~eliitoare fa~
(XVIIth, 1671 - Costin)
world's deceitful face. the "We are after the world's deceitful face" ALgenitival art
e. A§a, ford veste el thus suddenly he
In
vriijm~i colffl.
crocodilului
ajM.ndu-se
in inimical teeth.the crocodile.the.Gen being (XVIIIth- Cantemir) "Thus, suddenly, he was in the crocodile's inimical teeth"
As to the range of the constituents that may precede the lower definite article suffixed on the N, they may be of two types: As and pre-nominal genitives (Gens). A. an (indefinite) adjective
crocodilului ajM.ndu-se (18) A§a, fdra veste el tn vriijm~i colfi! thus suddenly he in inimical teeth. the crocodile.the.Gen being (Cantemir) "Thus, suddenly. he was in the crocodile's inimical teeth"
On the syntax of Romanian definite phrases B.
Gen Phrase (Gen D P) (19) ...ca. sa nu. se fnece so that not be drowned
a toate ftlrile ani! trecufi (Ureche) of all countries. the years. the passed "... so that the passed years of all countries may not be drowned in oblivion" One should know that in MR a pre-nominal Genis followed by an indefinite N. Just as in English, a DP containing a pre-nominal Genis interpreted as [+detl, and it is the pre-nominal Gen DP which checks the [idef] feature of the DP. The pre-nominal Gen DP in MR functions as a definite Determiner Gen (Huddlestone & Pullum 2002). It also occurs in the first position of the DP, presumably in [Spec, DP]. (20) a.
al
regelui
flu
king.the.Gen son "the king's son" ALgenitival artide
b. fiul regelui son.the king.the.Gen "the king's son" Examples like (19) contrast with MR (20a), and are no longer found in MR. The lower definite article raises several questions: 1. What is the interpretation of this phenomenon in the framework sketched above? 2. What made the lower article likely or possible in OR? 3. What are the contexts that favored the occurrence of the lower article? 4. What are the causes that led to the elimination of this pattern? 5. Are there other OR DP structures which relate to the existence of LDA? In the following sections of the paper we supply tentative answers to these questions. 2.3.
Interpreting the facts of Old Romanian in the framework sketched in Section 1
Two morpho-syntactic properties of Romanian N/A have combined to produce the strict locality conditions on definiteness valuation in MR. The first is the suffixal nature of the article which allows N s to be valued tor definiteness, even if their definiteness feature is uninterpretable (i.e., [u + def] ). The second significant property is that, at some point in the evolution of Romanian, As must have acquired the possibility of optionally incorporating an uninterpretable unvalued definiteness feature [udetl. This feature was valued through Agree with theN, as shown above. Since Agree was long distance, and the nominal valued for definiteness did not need to be the first NP/ AP below D, we expect the following alternations in 0 R, all of which are attested: i.
The definite article shows up on the first NP or AP of the DP (the MR pattern, available in all attested stages of Romanian):
201
202
Alexandra Cornilescu & Alexandru Nicolae (21)
au purees
fortl
numai din
it.happened only outof veche~
p
rtinced~
pizmaluire
(Cantemir)
old. the and rancid. the envy "It all happened out the old and rancid envy" ii.
The definite article shows up on an NP which is not the first phrase of the DP (this is the lower article): (22)
spre ticaloase cuvi,tek_ mele audzul lfi
pleactl ...
(Cantemir)
to vicious words.the my hearing your tum "Lend your ear to my vicious words" iii. Sporadically, the definite article could also be present on more than one constituent, i.e., the DP shows multiple definites (cf. Croitor 2008). This is consistent with the view that definiteness had become a concord feature in OR, and Romanian morphology allows it to be uninterpretable but valued and realized on both Ns and As. Multiple definites are present both in the order A+ N (examples (23), from Croitor 2008), and in the order N +A (examples (24), likewise from Croitor 2008); apparently, multiple definites were lost (in these patterns) at the end of the XVIIIth century: A. Multiple definites: A+ N (23) a.
Ce i-au
ttliat atuncea curdnd puterni~
that to-him.have cut then
soon
mtin~
lui
mighty.the hand.the of
Dumnedzetl zilele
(Costin)
God days.the "God's almighty hand took his/her days" Paraschev~ (Costin) b. ... moa§tele a sfin~ prepodobn~ ...relics.the of saint the beautifully-adorned. the Parascheva, "... the relics of the holy, beautifully-adorned (Saint) Parascheva"
B. Multiple definites: N + A (24) zidul cetapi marei ¥i frumoasei wallthe city.the.Gen big.the.Gen and beautiful.the.GEN
tn Spania
( Cantacuzino)
in Spain
"the wall of the big and beautiful city in Spain" It is reasonable to assume that it was precisely the possibility of 1'a.luing definiteness on pre-nominal As that led to stricter locality conditions on Agree, that is, to the MR requirement that the constituent that values the [idef] feature of D should be the first AP/NP below D.
On the syntax of Romanian definite phrases 203 Gradually, as a result of a general tendency towards economy, the constituent that values [idef] in D (i.e, which has an LF effect) got to be the only one which phonologically realizes definiteness. This was either a definite pre-nominal A or a definite N. Multiple occurrences of the suffixal article are now ruled out. Moreover, only pre-nominal As are ever suffixed by the definite article, since only pre-nominal As can be closer to D than the N. Definiteness thus turns into an exclusive property of attributive As as opposed to predicative ones (see above). This means that As that merge as attributes (specifiers or pre-nominal adjuncts) are obligatorily specified as [udetl This is what guarantees that if the head N is [u + detl, and there is a pre-nominal A in the DP, definiteness will be realized on the pre-nominal adjective immediately below D.
3· 3.1
What the lower article suggests about the emergence of the enclitic article An open question
The account proposed above relies on the contrast between languages which value definiteness in D by merging a determiner and languages which value D by means of a lexical category morphologically marked as valued for definiteness. From a Romance diachronic perspective, one of the still hot puzzles is the manner in which the same Latin demonstrative ille (cf. Iordan & Manoliu 1965) led to proclitic free-standing articles in other Romance languages, but to an enclitic article in Romanian. We claim that the lower article tilts the balance for the hypothesis that the enclitic article developed out of the post-nominal Latin Dem ille rather than out of a pre-nominal demonstrative, as in Giusti's (1998) analysis of the history of the definite article in Romance. Both proposals have been advanced for Romanian as well (see Renzi 1993 for the pre-nominal demonstrative hypothesis, and Coteanu 1956 and Graur 1967 for the post-nominal demonstrative hypothesis). The lower article matters in this on-going debate, since it is so far the only attested construction that does not show the Romanian definite article on the head of the first NP/AP of the DP. The standard Khigh article" is consistent with analyses where the article merges in D and is lowered on the first [ +N] constituent (as recently proposed by Dobrovie & Giurgea 2006), or where the article merges in D and there is movement into the D-area (N-to-D, AP-to-[Spec, DP]) as assumed in older studies (DobrovieSorin 1987; Grosu 1988). The lower article should be construed as evidence that the article merges low, and 'moves' higher through Agree, reaching the position below D. The aim of this section is two-fold: we propose a possible path from the Latin post-nominal demonstrative to the enclitic definite article; secondly we argue that, if correct, the diachronic proposal is consistent with an analysis of the enclitic article as a suffix rather than a (second position) clitic.
204
Alexandra Cornilescu & Alexandru Nicolae Before detailing the change from post-nominal Dem to enclitic definite article, we spell out our (minimal) hypotheses regarding the architecture of the DP. Following a number of researchers (Borer 2005; Giusti 1993; Julien 2005; Roehrs 2006), we assume that articles, or rather, determiners, merge lower, say in an Art(icle) P (valuing the [<1>] and [idetl features of the Art head), and then move to D or [Spec, DP] to value the features there (deixis, specificity, definiteness, [u<j>] ), if we assume, as mentioned above (cf. also Giusti 1993; Longobardi 2001) that an argumentDP is interpretable only if its D projection is visible. The space between DP and ArtP may (but need not) host periphery or quantificational projections, as in the proposal put torth by Roehrs (2006): (25) DP > CardP > ArtP > NumP > NP It follows that the Dems merge in [Spec, ArtP], under the assumption that they are phrasal. We also accept that pre-nominal As in Romance merge as specifiers of functional projections (Bernstein, 1991, 1993), while post-nominal As merge as adjuncts (Giurgea 2008). As shown, MR relies on LA and Local Movement. OR appears to have employed LDA (checking of a feature across an intervening specifier), and Long Distance Movement (movement over a relevant specifier). In agreement with proposals for Scandinavian (see especially Roehrs 2006: 49-64) and Romance (Renzi 1993, 1997; Lyons 1999; Giusti 1998), we might envisage the following scenario tor the rise of the enclitic article. The enclitic article emerged through the reanalysis of the Dem ille in the context of the post-nominal Dem construction; reanalysis amounted to a change in its c-selection properties. This hypothesis is supported by the existence of post-nominal Dems in all written phases of Romanian (26):
acela (26) a. Au trimis Pa§a pre tlilmaciul have sent Pasha PE translator.the that "Pasha sent that translator" b.
~i au ctlzut In ri'izboiul acela and have fallen.they in war.the that "And they fell dead in that war"
(Costin)
(Ureche)
The post-nominal position of As, possible in Latin, was reinforced by the contact with the local Dacian idioms, which strongly preferred to post-pose the A, including the (adjectival) Dem (Brfulcu~ 2004; Graur 1967). In all attested stages of Romanian, the pre-nominal or post-nominal position of the Dem depended on its textual, prosodic role (deictic or anaphoric). Consequently, it may be believed that either the NP or the DemP alternatively moved to [Spec, DP] to check the unvalued features there, thus deriving the alternative orders ((27a) and (27b )).
On the syntax of Romanian definite phrases (27) a.
DP
~
NP
D'
~
D [+def]
ArtP
~ DemP
Art'
~
Art
b.
~
ille [+foe]
homo [-foe]
NumP Num'
DP
~
DemP
D'
~
D
ArtP
~
Art'
toemP
~
Art
NumP
~
ille
NP hamo
Num'
As known, in two-member structures like omul acesta (man. the this) the post-posed Dem is normally focused (Bernstein 2001 a.o.), and may have a [locative-deictic] feature, while the Romanian pre-nominal Dem (acest om "this man") is mostly anaphoric (Manoliu-Manea 1993). It is, then, more plausible to assume that the re-analysis of the Dem as an article occurred in a three-term construction, which would shift the nuclear stress on the last (most deeply embedded) third term (Cinque 1993). Graur (1967) thus suggested as a basis for re-analysis the three-term construction: homo ille bonus, N + Dem + Adj. Since the Dem is not in focus, and thus probably not stressed, it is likely that it was "weakened" and re-analyzed as a head moving to D 0 rather than to [Spec, DP], following the general evolution of Romance or Germanic (Roehrs 2006 for Scandinavian). lt is true that, as remarked by one of the reviewers, since in the earliest Romanian texts the article is already fully grammaticalized, there is no decisive piece of evidence that re-analysis occurred in the three term structure. The demonstrative might perhaps have been distressed as a result of its anaphoric function in the simpler two-term construction as well. The change from a phrase to a head moving to the D-head position would have been the same. The often made suggestion that
205
206
Alexandra Cornilescu & Alexandru Nicolae re-analysis is based on structures including post-nominal modifiers probably takes into account the fact that the definite article is still required to license a post-nominal modifier/argument in prepositional constructions, where the nominal head is otherwise determinerless (Isac 2006): (28)
a.
Cartea
este pe masa.
book.the is on table "The book is on the table:'
b. Cartea.
este pe masa rotunda. book.the is on table.the round "The book is on the round table."
Head movement of Dem to D frees [Spec, DP] for NP-movement; suffixation takes place in this Spec-Head configuration. Re-analysis of the Demas a suffix represents a change of its c-selection feature, which becomes [+N-], satisfied by combining it with a nominal head (an N°, A0 constituent). Head-adjunction of the Dem-article to theN causes the former to undergo phonological reduction, dropping its first syllable ILLE ~ -(U)L. Significantly, the pre-nominal definite article of French, which has developed out of a pre-nominal demonstrative (lordan & Manoli u 1965; Giusti 1993, 1998 ), has also further developed from an independent head into a prefix in some of the ereales based on French (Mauritian Creole), being reanalyzed as part of the noun stem (Lyons 1999: 331; examples from Lyons 1999): (29) a.
le lit
(Standard French)
the bed "the bed"
b. lili
la
(Mauritian Creole)
the.bed there "that bed" The change from Dem to article also meant a loss of semantic features (bleaching), from the richer matrix of the Dem [(locative)-deictic, definite, anaphoric, 3rd/6th person, adjective/pronoun] to the more reduced feature matrix of the article [definite, anaphoric, adjective] (cf. Giusti 1998; Lyons 1999).
On the syntax of Romanian definite phrases 207 (30)
DP
~
D'
NP
I
N
~
D
ArtP
I~
Dem+D DemP
hamo
I
(ll)le
""-/ am+ (u)l
Art'
~
Art
NumP
~
Nwn'
NP
~
AP bonus
/
Num ....
..... bun
The lower article preceded by an A is a precious missing link in the chain leading from a post-nominal Dem to an article placed on the first nominal constituent of the DP. Thus Renzi (1993: 308), expressing skepticism about the post-nominal Dem hypothesis, was wondering: "Why starting from omul btltran [man. the old "the old man"], we have btltrdnul om [old. the man "the old man"], and not batran omul [old man. the]?': in other words why is there no evidence of an article on a lower term than the first? The examples that we have surveyed are exactly of the type expected by Renzi (1993) to occur under the hypothesis of deriving the article from the post-nominal Dem.
3.2
Consequences for the analysis of the article: The Romanian definite article is a suffix
The historical path suggested above is consistent with the claim that the combination of the article with the N or A is the outcome of a morphological rule, rather than the effect of syntactic movement; in particular, it is not the output of N-to-D or AP-to[Spec, D], as previously believed, since the important work ofDobrovie-Sorin (1987) and Grosu (1988). Strong evidence shows that the definite article is a suffix (Ortmann & Popescu 2000; Dobrovie-Sorin & Giurgea 2006; Tasmowski 2009), rather than a second position clitic/a Wackemagel clitic (Renzi 1993). The morphologic status of the article as a suffix rather than a eli tic follows from the following facts: (i) the article is not always in second position, since, inside the AP, degree words (and prepositions) may precede the A to which the article attaches, as in atdt delfoarte lungi + le drumuri ("such of/very long. the roads"); (ii) the article attaches to both conjuncts in a coordination, as infrumos + rd ~i mare*(+le) or~ (beautiful+ the ~i big*(+the) city), an unexpected repetition fora clitic (Zwicky & Pullum 1983); (iii) the article shows allomorphy, changing its form as a function of the last phoneme of the nominal base and of the
2o8 Alexandra Cornilescu & Alexandru Nicolae
inflectional class of the stem (e.g. masculine nouns ending in a consonant take the basic allomorph -ul, as in copil, child~ copilul, child. the, while masculine nouns ending in -e take the allomorph -le, as in perete, wall~ peretele, wall. the); (iv) the article occurs in a constant position, i.e., on the first N or A in the DP. In contrast, Romanian clitics are inconsistent with respect to their position (since they may occur both before and after their syntactic host: l-am luat/am luat-o hima_-haveAUX. lst person taken/ haveAUX.lst person taken-hera_), as well as to the type of constituents they cliticize on (verbs, auxiliaries, Ns, complementizers).
4· Contexts of occurrence of the lower definite article A relevant question, already formulated above, is what contexts require or allow the use of a lower article. To answer this question, a body of texts ranging from the earliest Romanian writings of the XVIth century to the first half of the XVIIIth century was examined. The texts show variation between the 'high' article, used in most cases, and the 'lower' article. The lower article predominantly occurs when the definite head N is followed by another constituent (especially by a Gen(itive)). It is for modified and complemented (by a Gen) DPs that we have checked the relative frequency of the high vs.lower article (see (31) below). (31)
Text Miron Costin
Letopisetul Cantacuzinesc R. Popescu R. Greceanu Constantin Cantacuzino Cantemir TOTAL
AnEF+ N + GenDP (high article) 0 3 0 2 0 1 6 High article: 6 (12,5%)
A+NnEF+ GenDP (lower article) 15 1 1 3 5 17 42 Lower article: 47 (87,5%)
The examination of the data shows that a lower definite article on N appears overwhelmingly (87,5%) in contexts where the head N is followed by a Gen DP complement. Other post-nominal modifiers may also sometimes trigger the presence of the lower article as in ( 34 ). If there is no camp lement or modifier, the article emerges on the pre-nominal A, as in MR.
On the syntax of Romanian definite phrases A. A[+def] + N[-def] (no complement/modifier) (32)
au purees
fo.ra numai din
it.happened only out.of
vechef!
~
rt2ncedf! piztniJluire
(Cantemir)
old. the and rancid. the envy "it all happened out the old and rancid envy" B. A[ -def] + N [def] + GenD P (33) a.
...ca mare sdr~etul roatelor like great grinding.the of the wheels "... like the strong grinding of the wheels .. :'
(Cantemir)
uscate vinele b. ... Corbul de the Raven against dried-up veins.the ciolane clonful ~i-ar ciocni (id.) empty.theGen bones bill his would knock "The Raven might knock his bill against the dried-up veins of his bones."
goalelor
c.
... de dulce otrav!! Hulpii ...with sweet poison. the of the Vixen
tare
se
amefirit
(Cantemir)
much they.got drunk "...they got quite drunk from the sweet poison of the Vixen" C. A[ -def] + N [def] +Modifier (PP or AP modifier) (34) a.
Nqtiutor gt2ndul
omenesc [... ]
Ignorant thought.the human
Ia
ce merge...
(Costin)
what is heading for "What is the ignorant thought of man aiming at?"
intunecoastl ceaf!! aceia•.• b. ...§i prin and through dark fog.the that "and through that dark fog"
(Cantemir)
This distribution signals a tight relation between the inflectional Gen (and other modifiers) and the lower article, a fact which should be accounted for. In MR, i.e., after 1780, the lower article disappears. There are, however, two types of motivated exceptions. The lower article is still part of religious and other obsolete texts ( 35), and in such cases it is again mostly followed by the Gen. Secondly, there is a small group of quantificational or evaluative As that may function as definite quantifiers, and may or must be followed by definite Ns (see also GALR 2005; Barbu 2004): tntreg ("whole"), singur ("unique") ditamai ("big"), gogeamite ("big"), as in (36). No Gen modifier is required.
209
210
Alexandra Cornilescu & Alexandru Nicolae (35)
Miluie~te-mtl,
Doamne,
Have-mercy-on-me God,
dupa
mare mila
Ta!
according to great mercy. the Your "God, have mercy on me, according to your great mercy." (36)
a.
intreg
or~l
/intregul
OTfl1
whole city.the whole.the city b. ditamai prostul/*ditamaiul prost big foolthe We may conclude that one significant change in the syntax of the Romanian D P relates to the locality conditions of definiteness valuation. The definite feature strengthens, requiring valuation by a strictly local nominal constituent (N or A). Long Distance Agree is ruled out.
5· 5.1
The (Lower) article and the reorganization of the genitive system The inflectional and the prepositional genitive
Romanian differs from other Romance languages in that it disposes of an inflectional Gen. while in other Romance languages the Gen is prepositional, marked by de "of (Grosu 1988 a.o.).The creation of a nominalinflectional Gen in Romanian was often viewed as an 'effecf of the suffixal definite article. Also, an influential traditional opinion was that "in Romanian, the maintenance of oblique cases is the first and most important cause of the post-position of the definite article" (Coteanu 1956: 67). As also underlined by one of the reviewers, there is no demonstrable correlation between enclisis and the development of an inflectional Gen, or the other way round. It is true to say, however, that the development of the article system favored the reorganization of the Gen system: the Gen case has developed parallel. inflectional and prepositional forms (see (37)), and it is the suffixal article which varies for Case in Romanian. Thus not only did Romanian develop an inflectional Gen, as is well known, but the prepositional Gen, based on the same preposition DE as in all Romance, did not disappear, either. Rather it became very limited and specialized (see Cornilescu 2004 for details). In OR, the inflectional and the DEGen are in free distribution (at least in post-nominal position where both occur) as shown by Pana Dindelegan (2008). Romanian has developed a morphological distinction between 'anchoring Gens: always DPs, and 'nonanchoring (Prepositional) Gens: always syntactic NPs (in the sense of KoptjevskayaTamm 2005), thus verifying the typological generalization that only languages that have articles may develop specialized forms for anchoring vs. non-anchoring Gen. The two forms show a very different cluster of morpho-syntactic and semantic properties, summed up below:
On the syntax of Romanian definite phrases (37) a.
citirea
cat
mai des
a autorilor clasici
reading.the more frequently of authors
classical
frecventa de romane polifiste b. citirea reading.the frequent of fiction crime (38) Anchoring Gens a. inflectional b. DP referential, < e>-type denotation
Non-anchoring Gens (limited to nominalizations, see below) a. prepositional b. NP c. <e, t> denotation
In sum, Romanian has developed a reliable syntax-semantics correlation in the domain of the Gen, by developing an inflectional Gen system alongside of the prepositional DE one. 5.2.
Significance of the statistical correlation between the lower definite article and the inflectional genitive
As noticed in Table (31), there is a strong statistical correlation between the lower article and the inflectional Gen. This correlation may be interpreted from a double perspective: i.
ii.
One may adopt a functional-pragmatic perspective, considering as significant the referential interpretation of the inflectional Gen, as opposed to the property, generic interpretation of the prepositional Gen. One might say that the lower article always accompanied that Gen form which served as a contextual anchor tor the head. The article, initially a post-posed Dem, had a similar anchoring role, probably being interpreted as [locative-deictic] (Lyons 1999; Bruge 2000; Coteanu 1956; Manoliu-Manea 1993). The functional role of the post-posed article may have been that of a weak Dem, i.e., a form intermediate between a Dem and a definite article (Lyons 1999; Giusti 1998 ). It is the correlation between the anchoring role of the Gen and the presence of the post-posed Dem which is still visible in the lower N [+def] examples noticed above. The correlation between the lower article and the post-posed inflectional Gen can also be viewed from a mor-e na,.rowly dist,.ibu.tional perspective, more likely to be the correct one. lt is known that the inflectional Gen in Romanian is realized either by a bare inflected DP (= the bare Genitive (39b)), or as an inflected DP preceded by the genitival article AL (39a) (for a description of the Gen article see Cornilescu 1995, 2004). The two forms are in complementary distribution.
Roughly, the AL Gen occurs whenever the head N is indefinite (39a), while the bare Gen occurs when the head N is definite and the Gen is adja.cen t to the head N ( 39b ):
211
212
Alexandra Cornilescu & Alexandru Nicolae (39) a.
doi prieteni ai
copilului
two friends ALJFitival artide child. theGen "two friends of the child" b. prietenul copilului friend. the child. theGen "the child's friend" It is the syntax of definite DPs containing Gens that is of interest. With definite heads, the AL Gen occurs in two situations: (i) when there is an intervening modifier between the definite head and the Gen (40a); (ii) when the Gen is adjacent to the head, but the article is on a pre-nominal A (40b). (40) a. prietenul bun al copilului friend. the good ALgenitival artide child theGen "the child's good friend"
prieten al copilului b. bunul good.the friend ALgenitival artide child. theGen "the child's good friend" c. *bunul prieten copilului good.the friend child.theGen Generalizing, the bare Gen occurs only when it immediately follows a definite N, while the AL Gen occurs elsewhere. The bare Gen is the preferred form since it is more economical (economy of representation). The preference for the bare Gen clearly must ha1'e
been felt itt OR as well. This is what explains the use of the lower definite article with inflectional Gen. If the article is placed on theN, instead of being placed on the higher A. the Gen is a.djacent to the definite article and it is possible to employ the more economical ba.re Gen (41a) (= 17a), replacing theAL Gen (41b): (41) a. OR: cu cinstita with honored
cartea
mllrii
tale
(Bare Gen)
letter. the highness.theGen your
b. MR: cu cinstita carte with honored.the letter
a ALgen.art
mllrii tale (AL Gen) highness.theGen your
Thus both formal economy considerations and functional semantic considerations may be invoked to account for the preference for the lower article in the context of an inflectional Gen.
6. Other or DP patterns where agree and move operate long distance In this section we show that the possibility of valuing the [idef] feature in D across a specifier, that is, LDA, was very general in OR and it combined with long distance
On the syntax of Romanian definite phrases move, therefore with movement across a specifier, a fact which is no longer allowed in MR. These hypotheses allow us to understand a number of other structures possible in OR, but systematically excluded in MR. So far we have only examined cases where LDA is expressed by a lower article. There are, however, other constituents that may value D-definiteness, such as Dems or inflectional Gen DPs. In OR, these types of constituents could, like the definite article, valueD-definiteness either by LA, or by LDA, i.e., at a distance from the D position. Vatiation in the pattern of definiteness valuation thus also had consequences for DPs that did not include the definite article, but which show Dems and Gen DPs in constructions no longer available in MR. In this section, we review some of these constructions which involve LDA or LD move.
6.1
Adjectives before demonstratives
As shown in Section 1, the post-nominal Dem cannot be preceded in MR by As. In contrast, in OR, both indefinite and definite As could precede the Dem. Consider indefinite As first: (42) a.
0 R: cumplite aceste vremi de acu.m terrible these times of now "these terrible times of now"
b. MR: aceste cumplite vremi de acum these terrible times of now The A in (42) is surely in an emphatic periphery position. As proposed by Laenzlinger (2005), periphery As merge or move to the left periphery of the DP defined by him as the space between a lower DAgreementand a higher DDeixis" This proposal is rather similar to Roehrs' (2006) in (25) above, who also proposes that Agreement features are checked in ArtR while 'referential'/deictic features are checked in the higher D position. Under these assumptions, an example like (42) would have the structure in (43): the Dem merges in [Spec, ArtP], the A is above it in a periphery projection, and the Dem checks its deictic feature across the A by LDA.
213
214
Alexandra Cornilescu & Alexandru Nicolae (43)
DP
~
D [idef]
EmphP
~ Emph'
AP
~
Emph
umplite
ArtP
~
DemP
aceste [u+def]
Art'
~ NumP I
Art
NP.... vremi
The absence of this pattern in MR is the result of the disappearance of LDA. 6.2
Definite NP + Dem
In OR, it is also possible for a complex definite NP to precede the Dem, contrary to MR, where the N° alone crosses the Dem (see Section 1).
piimt2nttl acesta sa vie (44) a. pii ticdlosul on wretched.the earth this come.Subj "... that he should come on this wretched earth .." ta aceasta b. itdma -heartthe your this "this heart of yours"
(Greceanu)
(Greceanu)
The derivation of these examples involves movement of a phrase across a specifier, i.e., LDM. A relevant intermediate structure is the one in (45). Assuming that there is NP movement (at least) to NumP in Romance (cf. Cinque 2004 among many authors) the definite NP is in [Spec, NumP] functioning as a Goal for the unvalued [cj>] and [def] features of the Art head The Dem merges in [Spec, ArtP], valuing the features of the higher D, through LDA.The definite NP moves to [Spec, DP] presumably to avoid the focus interpretation. ltis apparent that in moving to [Spec, DP], the definite NP crosses a phrasal constituent in [Spec, ArtP], this being an instance ofLDM.
On the syntax of Romanian definite phrases 215 (45)
DP
~
D [i+def]
EmphP
~ EmphP
ArtP
~
DemP
Art'
I~
acesta
Art [+"'] [+def]
NumP
~
FP
~
AP
F'
I~
tic!los+ ul
F
NP
p!m!nti:l
Expectedly, the structure in (45) is impossible in MR, where Move is strictly local. Notice also the examples below in (46), where a definite A alone has moved across the Dem, after checking definiteness against the ArtP head, as in (45) above. ( 46) a.
0 R:
nenorocitele
acestea vremi
(Greceanu)
unfortunate. the these times "these unfortunate times" (MR: vremile acestea nenorocite)
6.3
Genitive DPs
A second example ofLDA in DPs that do not contain the definite article is offered by pre-nominal Gen DPs. With pre-nominal Gen DPs, as with Dems, definiteness is checked either at a distance or in a local configuration. First. there are DPs where the Genis pre-nominal but is not involved in valuing definiteness. A definite determiner (the definite article, a Dem) or an indefinite one values the feature in D. In such cases, the pre-nominal Gen has nothing to do with the checking of definiteness (the valuer of the [idetl feature in Dis the definite article (47a), the Dem (47b) or an indefinite determiner (48)): (47) a.
ca
sa
nu. se tnece
a toate f4rile
so that not be drowned of all
anij_
trecufi
countries.the (Ureche)
years. the passed " [... ] so that the passed years of all countries may not be drowned in oblivion"
216
Alexandra Cornilescu & Alexandru Nicolae b. aceste ale Ciacalului cuvinte these AL Jackal.the.Gen words "these words of the Jackal"
(Cantemir)
(48) acele jigtlnii, carele [... ] tntr-alttl a trupului parte those beasts, which[ ... ] in other AL body.the.Gen part arme [...] poarta (Cantemir) arms [] carry "those beasts which carry [] arms in some other part of their body"
Interestingly. however, a Gen DP may incorporate a [+def] feature and represent the only definite constituent of the containing DP, checking definiteness by LDA, as in (49). ln such cases, what matters is that the Gen DP is not in DP initial position, i.e., [Spec, DP], being preceded by periphery As, a structure impossible in MR. (49) a.
din cumplita.
vra~md§ie
frumos mirositoare
of cruel.the hostility
!!
dragostei flori [... ] a
sweet
smelling
ritstlri
AL love flower will spring "... the sweet smelling flower oflove will arise out of that cruel hostility.. ~ b. tnsa ~i aceasta pre mai mare yet and this PE bigger
!!
viclqugului captu1ala o
fiJcea
AL cheating hiding itcL made ...but this she was doing to hide her cheating all the more"
(Cantemir)
In the same texts, however, a pre-nominal Gen DP may be in first position and is sufficient to trigger a definite interpretation of the containing D P. just as in English or in MR. This may be interpreted as an instance of LA. As already mentioned above in Section 1, LA with the pre-nominal Genis the only option ofMR, examples (SO) being syntactically perfect in MR.
(SO) a.
...lzul.ndu-se dupa !! .fEjJ. sale simfire ...taking-REFL after of his nature feelings ... "following the feelings of his own nature .. :'
ale mele impleticite protase b. nici va miera.fi de nor wonder about AL my clumsy sentences ... "nor do you wonder about my clumsy sentences ... ~
(Cantemir)
(Cantemir)
Thus definiteness valuation across a specifier is a general option of OR for all constituents that could, in principle, value the definiteness feature in D.
On the syntax of Romanian definite phrases Syntactic ambiguity and the loss of the lower definite article
7·
The co-existence of LA and LDA already mentioned above resulted in the occurrence of several systematically ambiguous structures. We briefly discuss two such cases. 7.1
The pre-nominallexical genitives
As shown in the previous section, the pre-nominal Gen could verify definiteness locally or at a distance, a phenomenon which could lead to interpretative ambiguities. Here is an example: (51)
raiul
spandzurat eel
paradise. the suspended that
ce
din
[fapte ale
lutnii
minu"~
unul ieste
(Cantemir)
which out of seven of.the world wonders one is Assuming the DP structure in (25) above and also that the pre-nominal Gen sits in a PossP analogous to the English determiner Gen, example (51) has the following two representations: (52) a. DP [ -def] QP
Po ssP ale lumii ~ b. DP [ +def] QP Po ssP ~apte ale lumii[+detl ~apte
ArtP/NumP minuni
LA
ArtP/NumP minuni
LDA
Representation (52a) is indefinite, "seven wonders of the world"; representation (52b) is definite, meaning "the seven wonders of the world". (52b) involves long distance valuation of definiteness across the QP, while in (52a) the cardinal values the feature on D by means of LA. The ambiguity in the data made it difficult for the interpreter to establish the type of (in)definiteness checking involved. 7.2
DP-internal inverted predicative adjectives
A second type of ambiguity regards the interpretation of pre-nominal As followed by the lower article.ln the first place, these are attributive As in a LDA configuration. This is the only interpretation, for instance, after prepositions (53a). On the other hand, the indefinite A may be an emphatic inverted predicative A, originating in a Kaynean small clause, and appearing at the periphery of the DP in a construction devoid of the copula, possibly as in (53b ), or across the copula otherwise (53c). (53) a.
...au venit egumenul de Bistrifit has come abbot.the of Bistrita
cu
cinstittl
cartea
mtlrii
tale
with honoured letter.the highness.theGen your "...the Abbot of Bistrita came with your highness' honored letter"
(DlR)
217
218
Alexandra Cornilescu & Alexandru Nicolae b. iar ascunsfl giude(ul lui Dum,edzau and hidden judgmentthe of God
toate gtindurile oamen~ti le stramutd. (Costin) them 0 moves all thoughts.the human "and the hidden judgment of God moves all human thoughts" "and the judgment of God, being hidden, moves all human thoughts" c. Ia.rt:t lu~coasa szmt lucrtlrile razboaielor and slippery are things.the of.wars
si tn puterea lui Dumnezeu stau. (Costin) and in power of.God stand.they "And the matters of wars are slippery and lie in the power of God:' Ambiguities of this type led to difficulties in establishing the value of the Agree Parameter: Local Agree or Long Distance Agree? Such situations eventually triggered the re-setting of this parameter, in favor of the more constrained grammar, the one in which definiteness is valued by means of LA, a system which generates fewer ambiguities. The re-setting of definiteness valuation illustrates an instance where the Subset PJinciple is diachronically relevant. Roberts (2007: 260) shows that "the Subset Principle might[ ...] lie behind the phenomenon of'restriction offunction: whereby in one system a given operation applies more freely than in another': This work was supported by CNCSIS-UEFISCSU, project PN II- IDEI 1979/2008.
8.
Conclusions
1.
In Old Romanian, the definite article suffixed to the noun/adjective may occupy the first position of the DP, but also a lower position. In particular, indefinite constituents such as indefinite quantifiers and adjectives or indefinite Gens may precede the definite noun. 0 ld Romanian thus disposes of Long Dista.nce Ag,.ee in the valuation of definiteness. Distributionally, the lower article is conditioned mostly by a post-nominal Gen or modifier. The lower article is evidence that the Romanian enclitic definite article originates in a post-posed demonstrative, following the same steps as suggested for Scandinavian by Roehrs (2006). It also confirms that the article should be viewed as a suffix combining with the N in the lexicon. The presence of the suffixed definite article leads to a different system of valuing the [idef] feature in D by the [u + def] feature carried by the definite noun. At some point in the evolution of Romanian, definiteness became a concord feature for adjectives; the latter optionally entered the derivation with an uninterpretable unvalued definite feature, valued by Agree with the noun. It is reasonable to assume that it was precisely the possibility of valuing definiteness on pre-nominal adjectives that ultimately led to the Modem Romanian requirement that the constituent that values the [idef] feature of D should be the first AP/NP below D.
2.
3. 4.
On the syntax of Romanian definite phrases
5.
6.
In Old Romanian, Long Distance Agree apparently co-exists with Local Agree. a factor that may cause ambiguity. Consequently, Romanian settles for Local Agree, selecting the more restrictive grammar. While in Old Romanian there is Long Distance Agree and Long Distance Move, Modern Romanian loses both options. Changes in these parameters lead to the disappearance of a number ofDP structures involving Long Distance Move/ Agree. Thus, all patterns involving movement of an XP across the demonstrative or checking of the definiteness feature across a demonstrative are lost.
References Akiyama, Masahiro. 2004. Multiple nominative constructions in Japanese and economy. Linguistic Inquiry 34(4): 671-683. Barbu, Ana Marla. 2004. Sinta.t:a Determinatorllor. Anallzd lingvistlcil ~~ computaflonalil. Bucur~ti: All Bernstein, Judy. 1991. DPs in French and Walloon: Evidence for parametric variation in nominal head movement. Probus 3(2): 101-126. Bernstein, Judy. 1993. The syntactic role of word markers in null nominal constructions. Probus 5:5-38.
Bernstein, Judy. 2001. Focusing the 'right' way in Romance determiner phrases. Probus 13: 1-29. Borer, Hagit. 2005. In Name Only. Structuring Sense. Oxford: OUP. Bdncu~. Grlgore 2004. Introducere in Istorla Limbli Rombne. Bucure~ti: Editura FundatJ.ei Romania de Maine. Bruge, Laura. 2000. Categorie Funzionali del Nome nelle Lingue Romanze. Milan: Cisalpino. Cinque, Guglielmo. 1993. A null theory of phrase and compound stress.lJnguistic Inquiry 24(3). Cinque, Gugl.lelmo, 2004. A phrasal movement analysis of the Romanian DP. In Studia Lingulstica et Phllologlca In honorem D. Irlmia, Ana-Maria Minut & Eugen Munteanu, 129-142. l~l: Ed Editura Universitil.tJ.i A. I. Cuza. Cornllescu, Alexandra. 1992. Remarks on the determiner system of Rumanian: The demonstratives ALand CEL. Probus 4: 189-260. Cornllescu, Alexandra. 1995. Romanian genitive constructions. In Advance.s 111 Romanian Ll11guistlcs, Guglielmo Cinque & Giullana Giusti (eds), 1-52. Oxford: OUP. Cornllescu, Alexandra. 2004. Romanian genitives revisited Bucharest Working Papers i11 IJnguistics. Coteanu, Ion. 1956. Despre pozttia articolului in limba RomAni.!.. Studii ~~ Cercetilri Lingvistice VII: 1-2.
Croitor, Blanca. 2008. Aspecte privind acordul in determinare in limba romfuJA veche. In Limba romdnil - Dinamica limbi~ db1amica i11terpretilri~ Gabriela Panil. Dindelegan, 213-218. Bucure~ti: EUB. Densusianu, Ovid 1961. lstoria Limbii Romdne. Bucure~tl: Editura ~tllntificil.. Dimitrescu, Florica (ed). 1978. Istoria Llmbii Rombne. Buc~ti: Editura Didactic!\ ~i Pedagogidl.. Dimitrova-Vulchanova, Mila & Giullana Giusti. 1998. Fragments of Balkan nominal structure. lT11iversity of Venice Working Papers i11 Linguistics 8( 1): 141-172.
219
220
Alexandra Cornilescu & Alexandru Nicolae Dobrovie-Sorin, Carmen. 1987. A propos de la structure du groupe nominal en Roumain. Rivista di Grammatica Generativa 12: 123-152. Dobrovie-Sorin, Carmen & Giurgea, Ion. 2006. The suffixation of definite articles in Balkan languages. Revue Roumab1e de Li11gulstlque Ll(1): 73-104. Farkas, Donka & de Swart, Hemii!tte. 2007. Article choice in plural generics. Li11gua 117: 1657-1676.
Farkas, Donka & von Heusinger, Klaus. 2003. Stability of reference and object marking in Romanian. Paper presented at the workshop on Direct Reference and Speclfictty. Vienna, August 2003. Ghetie, Ion. 1975. Baza Dialectaltl a Romd11ei Literare. Bucur~ti: Editura Academiei. Giusti, Giuliana. 1993. La Sintassi del Determina11ti. Padua: Unlpress. Giusti, Giuliana. 1996. Is there a focus and a topic phrase in the noun phrase? Ms, University of Venice. Giusti, Giuliana 1998. The rise of a functional category. From Latin ILLE to the Romance article and personal pronoun. University of Venice Working Papers in Li11guistlcs 8(2): 53-71. Giusti, Giuliana. 2005. At the feft periphery of the Romanian noun phrase. In 011 Space and Time in LanguagE, Martine Coene & Liliane Tasmowski (eds), 23-49. Cluj Napoca: Clusium. GALR 2005 - Gramatica Limbii Romdne. 2005. Bucure~ti: Editura Academiei Romil.ne. Graur, Alexandru. 1967. De nouveau sur !'article postpose en Roumain. Revue Roumaine de Linguistlque 1: 3-18. Gturgea, Ion Tudor. 2008. Recherches sur la structure interne des pronoms et des expressions nominales sans nom exprime. PhD dissertation. Universite Paris 7, Denis Diderot. Grosu, Alexander. 1988. 0 n the distribution of genitive phrases in Romanian. Linguistics 26: 931-949.
Hudd.lestone, Rodney & Pullum, Geoffrey. 2002. The Cambridge Grammar of the English Language. Cambridge: CUP. Iordan, Iorgu & ManoH.u, Maria. 1965. Introducere fn lingvlstica romanictl. Bucure~ti: Editura Didactici:l ~t Pedagogic!\ (Spanish edn- 1972, Manual de lingi.Ustica roman lea, rev., reelaboract6n parcial y notas pro Manuel Alvar, Madrid: Editorial Gredos). Isac, Daniela. 2006. In defense of a quantificational account of definite DPs. Linguistic Inquiry 37(2): 275-288.
Julien, Marit. 2005. Nominal Phrases from a Scandinavian Perspective [Ltnguistik Ald:uell!Linguistics Today 87]. Amsterdam: John Benjamtns. Kamp, Hans & Reyle, Uwe. 1993. From Discourse to Logic. Dordrecht: Kluwer. Koptjevskaya-Tamm, Maria. 2005. Marla~ Ring of Gold: Adnominal possession and non-anchoring relations in European languages, Possessives and Beyond: Semantics and Syntax, Ji-yung Kim, Yu Lander & Barbara H. Partee (eds), 155-181. Amherst MA: GLSA. Laenzltnger, Christophe.t: 2005. Some notes on DP- internal movement. GG@G 4: 227-260. Longobardi, Giuseppe. 1994. Reference and proper names: A theory ofN-movement in syntax and logical form. Linguistic Inquiry 25(4): 609-665. Longobardi, Giuseppe. 2001. The structure ofDPs: Some principles, parameters and problems. In The Hi.mdbook of Co11temporary Syntactic Theory, Mark Baltin & Chris Collins (ed), 562-603. Oxford: Blackwell.
Lyons, Christopher. 1999. Definiteness, Cambridge: CUP. ManoH.u-Manea, Maria. 1993. Gramatictl, Pragmasemantictl ji Discurs. Bucu~ti: Litera. Niculescu, Alexandru, 1990. Outline History of the Romanian Language. Padua: Unipress.
On the syntax of Romanian definite phrases
Ortmann, Albert & Popescu, Alexandra. 2000. Romanian definite articles are not clitics. In Clitlcs 111 Phonology, Morphology and Sy11tax, Birgit Gerlach & Janet Grijzenhout (eds). Amsterdam: John Benjamins. PanA Dindelegan, Gabriela. 2008. Tipuri de Gramatical!zare. Pe marginea utilizil.rllor gramaticalizate ale prepozipllor de ~i la. InLimba Rom4ml- Dina mica limbi~ di11amica interpretdrii, Gabriela PanA Dindelegan (ed), 227-239. Bucharest: EUB. Pesetsky, David & Torrego Esther. 2007. The syntax of valuation and the interpretability of features. In Phrasal a11d Oausal Architecture: Syntactic Derivation and b1terpretation [Linguistlk AktuellJLinguistics Today 101], Simin Karimi, Vida Samiian & Wendy Wilkins (eds), 262-294. Amsterdam: John Benjamins. Renzi, Lorenzo. 1993. I:articolo posposto Rumeno in dlacronia e in sincronia. Revue Roumaine de Linguistique XXXVIII(4): 307-323. Renzi, Lorenzo. 1997. Fissione di Lat ILLE nelle Lingue Romanze. Italica et Roma11ica. Festschrift for Max Pfister zum 75, Gunter Holtus, Johannes Kramer & Wolfgang Sweckb.ard (eds). 'llibingen: Niemeyer. Roberts, Ian 2007. Diachronic Syntax, Cambridge: CUP. Roehrs, Dorian 2006. The Morpho-Syntax of the Germanic Noun Phrase: Determiners move into the Determiner Phrase. PhD dissertation, Indiana University. Rosetti, Alexandru. 1968. Istorla Limbll Rom4ne de la Origini pttnit fn Secolul al XVII-lea. Bucure~ti: Editura pentru Literaturi\ ~i Art!\. Stan, Camelia. 2008. Grupul nominal romi!nesc (aspecte diacronice). In Llmba Romttnit Dinamlca limbii, dinamica lnterpretdrll, Gabriela Pan!\ Dtndelegan (ed), 239-243. Bucharest: EUB. Stateva, Penka, 2002. Possessive ditics and the structure of nominal expressions. Lingua 112: 647-690. Stowell, Tim. 1989. Subjects, specifiers, and X-bar theory. In Alternative Conceptions of Phrase Structure, Mark Baltin & Anthony Kroch (eds). Chicago I L: The University of Chicago Press. Tasmowski, Liliane. 2009. Le Roumain, entre langues Romanes et langues Balkaniques. Paper presented at the University of Bucharest, May 2009. Zwicky, Arnold & Pullum, Geoffrey. 1983. Cliticization vs. inflection: English n't. Language 59(3): 502-513.
Corpus DIR- Documenre §I tnsemnitri rom4ne§ti din secolul al XVI-lea, text stabllit ~i indice de Gheorghe Chivu, Magdalena Georgescu, Magdalena IonJ:ti\, Alexandru M~ ~i Alexandra Roman-Moraru, Introducere de Alexandru Mare~. Bucure~ti: EdituraAcademiei, 1979. Costin- Miron Costin, Opere alese, P.P. Panaitescu (ed), Bucu~ti: Editura Tineretului, 1966 Cro11icarii .Mu11teni, Al Piru (ed), Bucur~ti, Editura Tineretului, 1964. Letopisetul Cantacuzinesc, in Cronicarii MuntenL R Popescu, LetopisetuJ. Bi\lenllor, in Cronicarii Munreni. R Greceanu, Cronica, in Cro11icarii .Mu11teni. Constantin Cantacuzino, Istoria 'fi\rii RumAn~ti, in Cro11icarii .Munreni. Cantemir - Dimitrie Cantemir, lstoria ieroglifica, Coleqia Biblioteca pentru totJ.. 1983. Editura Minerva, Bucure~t!. Ureche- Grigore Ureche, Letopiseful 'fhr4i Moldovei, P.P. Panaitescu (ed). 1955. Bucur~ti: ESPLA. Biblia- Biblia sau dumnezeiasca scripturit, Bucu~ti, 1688.
221
Coexisting structures and competing functions in genitive word order* Elisabetta Magni University of Bologna
This paper examines the correlations between word order at the clause level and word order at the noun phrase level in languages with no dominant genitive order. In this perspective, the comparison between Latin and English is a revealing domain for investigating both synchronic variation and diachronic evolution. The analysis uses a functional-typological approach; the assumption is that coexisting and competing patterns. as well as regularities and irregularities in word order, can be analyzed through diachrony. Results show that (1) in both languages, structures developing from different sources are assigned different functions according to multiple processing factors; (2) the mechanisms of word order flexibility observed in Latin, mutatis mutandis, can provide some useful insights for English as well.
1.
Possessive constructions and word order types
The relationship between OV/VO orders in clausal syntax and the ordering of genitive modifiers in the noun phrase was one of the stronger Greenbergian correlations. According to Dryer (1992, 2008), however, it differs from other correlating pairs because SVO&NG languages are as common as SVO&GN languages. In tact, while the VO order predominates over much of Europe, this is much less true for the NG order: Scandinavian and Baltic languages, for instance, are GN, and there are also languages in which neither order predominates. Languages in which there is no dominant order in possessive (or genitive) constructions are of particular interest because they allow investigation of factors motivating synchronic variation and mechanisms of change. In principle, synchronic variation can be motivated by the complexity of the genitive relation, which can be described
*
I am grateful to the editors of this volume for various valuable suggestions. I also thank the two anonymous reviewers for their comments on an earlier version of this article. Any remaining errors are of course my own. This work was supported by the Italian Ministry of Education, University and Research (PRIN grant 2008EHLWYE-002).
224 Elisabetta Magni
Table 1. Word order types in genitive constructions (from Dryer 2008) Word order (in a sample of 11 OS languages)
No dominant order
GN NG The same construction: GN/NG Different constructions: GN & NG
Languages Finnish Italian Macedonian English
Number 606 416 83
"as simply involving some semantic relation between a noun phrase and a noun, the particular relation being determined by inference from the semantics of the words involved and from the context" (Dryer 2008: 350). Cross-linguistically, however, word order in the noun phrase tends to be less flexible than word order in the clause, and genitive constructions are indeed so strictly grammaticalized that only 7.5% of the languages investigated by Dryer (2008) admit both theGN order and the NG order. This is shown in Table 1. Languages with no dominant order in genitive constructions fall into two subtypes: languages in which there is essentially one construction but both orders are allowed, and languages with two distinct structures: a GN construction and an NG construction. Macedonian, for instance, has free word order, and the very same genitive noun may precede or follow the head noun. English is coded as a language of the second type, where the s-form and the of form provide two complementary positions for the possessor. This study deals with coexisting genitive patterns and compares their behavior in two diverse language types. Through a functional-typological approach, this analysis explores the emergence and functioning of competing structures in English and Latin. Section 2 will focus on English, a SVO language; Section 3 will focus on Latin, a flexible order language. The paper will conclude with reflections on similarities and differences in genitive variation, with a focus on mechanisms of word order flexibility and the role these have played in the evolution of English and Latin.
2.
2.1
GN and NG in English The diachrony of the genitive variation
Old English is an inflectional language, where the of-genitive is a marginal construction with respect to the s-genitive. In a similar way to Latin, the s-genitive displays word order variation allowing the possessor to be realized both prenominally and postnominally. As postnominal s-genitives decrease their frequency, the ofgenitive gradually gains ground, and in the course of Middle English it becomes the major construction. Moreover, in late Middle and Early Modern English, the loss of
Coexistl..ng structures and competl..ng functions in genitive word order 225
inflectional endings leads to the use of juxtaposed structures like the man name, or even the Duke of Somerset dowther, where the s-less genitive is a competitor to the s-genitive (Rosenbach 2002: 206). Nonetheless, from 1400 to 1630, the s-genitive increased its relative frequency, and in present-day English it challenges the of-genitive. Rosenbach (2002: 212-218) discusses this revival of the s-genitive in connection with the debated development of the s- marker towards a clitic. According to some scholars (e.g. Janda 1980), the evolution of theGN structure implies a reanalysis due to the chance homonymy between one of the genitive markers and the unstressed possessive pronoun hys/ys, which occurs in left dislocations like those in (1), from Wyld (1953: 315):
Seynt Dunstone his lore b. my lord ofExcetre is tenants c. the Busshoppe ofRome his Iawes d. the Bishop of London and Coventre ys wiff
(1) a.
(c. (c. (c. (c.
1420) 1450) 1530) 1550)
Although there is no general consensus on the role of the his-genitive in the development of the possessive morpheme, it seems reasonable to explain coexisting patterns with a process of 'reactivation: through which the old inflectional genitive ending revives through "the apparent structural as well as functional overlaps with another possessive construction[ ... ], which has also been available in the English language" (Rosenbach 2002: 216). 1 If so, the peculiar situation of modem English, where clausal word order is more constrained than word order in genitive constructions, could be explained by the fortuitous survival of a structure that was on its way out of the language. 2.2
The synchrony of the genitive variation
At present, the choice between the two patterns can be viewed as a case of probabilistic variation. Factors that determine it in fact "do not determine categorically which construction is to be used, but rather the likelihood with which the two genitives are used, 1. Janda has been criticized in particular by Allen (1997), who claims that s-genitives cannot be derived causally from his-genitives, and that in Middle English examples like Margere ys dowgther,ys/hiswas onlyanorthographical variant ofinfiectional-s. Janda's (2001) recent modified account, however, postulates a reanalysis of -(e/ily)s to an invariant ditic his due to hypercorrection, also explaining that in similar cases "accidental developments involving items already slmilar in form and meaning to a grammaticalized affix caused upgrading via reanalyses which were unrestrained (and unconstrained) by the previous histories of the relevant morphemes" Oanda 2001: 303). Since "both lnterpretattons are possible, and we cannot decide[...] which is the correct one': the last word on this issue has not yet been spoken (Rosenbach 2004: 76); more references pro and against Janda's proposal are found in Rosenbach 2002: 212-217. For an updated discussion of similar phenomena and constructions in other Germanic (in particular Scandinavian) languages, see also Koptjevskaja-Tamm 2003: 665-670.
226 Elisabetta Magni
i.e., their frequency distribution" (Jager & Rosenbach 2006: 947). Animacy, topicality, and the type of possessive relation are well-known factors affecting genitive variation in English. These tend to correlate and cluster (both topics and prototypical possessors are usually animate), and an empirical analysis needs to tease them apart in order to test their relative strength. According to Rosenbach's data, the relative ranking of factors is: animacy> topicality> possessive relation (Jager & Rosenbach 2006: 949). Animacy in particular is shown to correlate with syntactic weight. since animates tend to be shorter than inanimates (Jager & Rosenbach 2006: 951). Predictably, the less frequent the s-genitive the longer the possessor. For human possessors, however, its frequency also depends on the number of premodifiers. The preference for the s-genitive with possessors premodified by up to two elements (i.e. the other person's nose is more likely than the nose of the other person) indicates that animacy has a relatively stronger role than that of weight (Jager & Rosenbach 2006: 949). Synchronic analyses, which can surely be refined by paying due attention also to cognitive-pragmatic factors, thus show some convincing rationale for the distribution and frequency of the two pattems.2 From a different angle, however, it is perhaps more interesting to consider how present-day variation, in the sense of gradience,3 ties in with processes of change, and how work on grammaticalization can help to understand synchronic states of affairs. In this perspective, for instance, it is arguable that the correlation between different functions and different word orders is also diachronically mediated and to some extent dependent on the sources of the two structures (Giv6n 2001:256-258, Bybee, Perkins & Pagliuca 1994). In fact, the NG and GN patterns seem to retain some characteristics of the constructions from which they originate. In the of-genitive, for instance, we can observe that the possessor has the same semantic teatures as the complement of a preposition, which in this case (cf. the original ablative value of oj), is typically inanimate. Besides, if pronominal modifiers (i.e. his-genitives) play a role in the evolution of the s-genitive, its association with animate and definite possessors is unsurprising. These issues concerning the influence of the past on present variation will be discussed further in the last section, after the analysis of genitive variation in Latin.
3· GN/NG in Latin 3.1
The situation in early Latin
When discussing word order in Latin, cases concerning other old Indo-European languages are often taken as a starting point. As prenominal genitives are commonly 2. For a critical evaluation of the different approaches and a detailed reanalysis of pre nominal genitives In English, see also Keizer (2007).
3·
A stimulating perspective on variation, recently explored in Traugott & Trousdale (2010).
Coexisting structures and competing functions in genitive word order 227 attested in Sanskrit, Ancient Greek, and Hittite, traditional grammars assume that GN was the ancient pattern. 4 This is also confirmed by the data from Oscan and Umbrian (Rosenkranz 1933). These are archaic languages closely related to Latin and therefore may provide useful information on Latin, see for instance the examples in (2), taken from coeval texts (c. II century B.C.E.): (2) a. senate{s tanginUd b. senatuos sententiad senate.GEN .SG judgement.ABL.SG "by the judgment of the senate"
(Oscan, CippusAbellanus B) (Latin, CIL 12 581)
The so-called Cippus Abellanus presents only one example of the reverse order, see (3a); similarly. in Early Latin postposed genitives seem to be limited to specific usages or contexts, as in (3b). For Adams they can be explained by 'emphasis': 5 (3) a.
sakaraklum
Hereklefs
(Oscan, Cippus Abellanus 11)
temple.ACC.SG ofHercules.GEN.SG "the temple of Hercules" b.
apud aedem
Duelonai
(Latin, CIL 12 581)
at temple.ACC.SG ofBellone.GEN.SG "at the temple of Bellone" Even if this view is accepted, in Classical Latin the GN pattern ceased to be the 'unmarked' option, and from Plautus onwards the NG order as well was frequently employed.6 3.2
The complex distribution of the two patterns in Classical Latin
Evidence suggests that, at the level of the clause, Classical Latin was a flexible order language, where multiple factors combined in determining the order of the constituents (Pinkster 1980: ch.9; Bolkestein 2001; Magni 2008: 226-228). Accordingly, on the noun phrase level, variation was conditioned by a number of concurrent factors, such as semantics, information structure, heaviness or structural complexity, prosody.
4 "Le complement determinatif au genitif devait anciennement preceder le substantif dont il dependait" [Formerly, the determinative complement in genitive was likely to precede the noun on which it depended] (Ernout & Thomas 1953: 163). 5· "It would seem that in the official religious language it was a customary act ofptty to throw the name of god into relief" (Adams 1976: 76). 6. "Mais, dans l'usage courant a l'epoque historique, le genitif complement suit le nom qu'il deter mine (orbis terrarum) aussi bien qu'ille precede (terrae motus)" [But in current usage durIng the historical period the genitive complement may follow the noun it determines (orbis terrarum) as well as precede it (terrae motus)] (Ernout & Thomas 1953: 163). See the data in Adams (1976), and the discussion in Bauer (1995).
228 Elisabetta Magni
However, we have no extensive data on the distribution of the two genitive constructions in the diachrony of Latin. Moreover, the comparison of the preceding results, which range from Marouzeau's (1922) monumental work to more recent (e.g. Adams 1976) and detailed synchronic analyses (e.g. Devine & Stephens 2006, Viti 2010) is difficult, because scholars use different corpora, apply different criteria of inclusion and exclusion, and have different ways of classifying data. Generally speaking, except tor authors like Caesar and Tacitus, where instances of NG outnumber instances of GN, preference for prenominal or postnominal genitives is not clear, and until Late Latin their distribution is similar. It also appears that in each category of genitive, different words have different distribution and rules? In my opinion, rough percentages of GN vs. N G fail to account for the distribution of genitives because the frequency of the two patterns is relative to the functions they convey. Here, I will try to define the functional domains where prenominal and postnominal possessors tend to occur. I will focus on the position of adnominal genitives in phrases with a (virtually) fixed word order. This choice is not merely motivated by reasons of space, but also by the belief that expressions with entrenched and routinized recurrent patterns may help to disclose the functional motivations underlying frequent orders. 8 3·3
Functions of the GN structure
Since Early Latin, GN structures occur in the standard onomastic formula: (4) L. Aimilius L.(uci) f(ilius) L. Aemilius:NOM.SG Lucius.GEN.SG son.NOM.SG KLucius Aemilius, son of Lucius"
(CIL 12 614)
fixed expressions like the following ones: (5) a.
reliquiarum
reliquias
(Pl. Cure. 388)
remnanlGEN .PL remnant.ACC.PL "the remnants of the remnants" 7· "It is not the case that objective genitives always follow the noun and subjective genitives always precede it, or that possessive genitives always precede and partitive genitives always follow the noun, or that deverbal nouns always take a genitive to the left and relational nouns always take a genitive to the right" (Devine & Stephens 2006: 314). Moreover, "many complements that apparently do not move but stay to the right of the head are also focused, and some complements that do move to the left of the head are not focused" (Devine & Stephens 2006: 380). 8. On this point see also the discussion in Viti (2010: 78). All the examples have been carefully checked throughout the corpus provided by the LLT-A (2008) database. The analysis restricts attention to phrases with lexical (i.e. non pronominal) genitives; genitives dependent on verbs or adjectives, and genitives that are discontinuous with respect to their head noun are disregarded, but see Bolkestein (2001).
Coexisting structures and competing functions in genitive word order 229 b.
divum
deo
god.GEN .PL god.DAT.SG " (to) the god of gods"
(Carm. Sal. in Varr. LL 7,27)
old formulaic and juridical phrases: (6) a.
deorum
beneficio
god.GEN .PL help.ABL.PL "by the help of the gods"
b. vitae
necis=que
potestas
life.GEN.SG death=and.GEN.SG power.NOM.SG "power of life and death" genitive constructions destined to become compounds because they were more likely to be interpreted as a single concept or structural unity: (7) iuris consultus "iurisconsult" ma.nus missio "manumission" plebis sci tum "plebiscite" aquae ductus "aqueduct" These examples seem to confirm the crucial role of conceptual individuation: "the degree to which the genitive plus head combination expresses a single recognizable concept, as opposed to an ad hoc combination of two independent concepts" (Devine & Stephens 2006: 388). This notion, which correlates with predictability and importance, as well as anchoring and non-anchoring relations, shows a cluster of functions operating in between pragmatics and syntax. Their pertinence will also emerge from the analysis of some typical NG structures. 3·4
Functions of the NG structure
The NG order regularly occurs not only with the names of gods, but also in a set of fixed noun phrases indicating social, religious or public functions, as in (8): (8) paterfamilias!deum "father of the family/of the gods" praefectus urbis/cohortium "prefect of the city/ of the cohorts" magister equitum!navis "chief of the cavalry/captain of the ship" tribunus plebis/militum "tribune of the people/of the army" According to Rosenkranz ( 19 33: 139 ), in these structures "liegt ein latenter Gegensatz" [there is a latent contrast], hence postposition is due to the stress on the element that expresses the 'potential contrast'. Both new and contrastive information entail unpredictability: the former selects an entity from an unlimited set of possibilities, the latter selects an entity as opposed to another entity or a limited set of other entities. Thus, in
230
Elisabetta Magni the expressions in (8), the potential interference of referents belonging to the same semantic field makes information retrieval less immediate. ln some cases, examination of a specific author can refine statistical data, and shows how Latin word order is sensitive to contrast. For instance, Adams (1976: 74) notes that in the ancient senatorial expressions the genitive senatus always precedes words like consultum "decree" and sententia "decision': as in (2b), but Livy prefers the reverse order with auctoritas 'authority'. As a matter of fact, senatus auctoritas occurs only in 5 instances but, on closer inspection, in 19 cases out of 30 sena.tus follows the expression ex auctoritate, in a formula that admits also other (synonymous) genitives: e.g. ex auctoritate patrum "on the authority of the senate': or ex auctoritate pop-
uli Romani. The N G order is also typically found in a set of 'appositive' structures, where the genitive describes, defines, qualifies the content of a generic word (Emout & Thomas 1953: 42-43), like in homo nihili "a fellow of no account" (PL Trin. 321), and in colloquial expressions like jlagitium hominis "a fiend ofa man"(Pl. As. 47 3), scelus viri "a shame of a man" (PL Cure. 613), monstrum mulieJis "a fright of a woman" (PL Poen. 273). 3·5
Tendencies and inconsistencies
Postnominal genitives are associated with new, less predictable and/or accessible information, as well as with possessors that combine with their head noun in structures with a kind-referring function, and a relatively low degree of integration. Conversely, prenominal genitives are preferred for old, predictable, conventionalized possessibility and/or tighter relations. 9 1hese structures are mainly associated with non-disruptive, or unimportant information, with anchoring functions and referent identification. 10 Notions like saliency, predictability and referential unity, however, are not so straightforward. In Latin the ordering of genitive constructions seems to depend on a tricky set of 'competing motivations~ with inconsistencies due to the clash between pragmatic and semantic factors. As for inconsistencies with kinship terms, which typically entail anchoring relations, Devine & Stephens (2006: 389) say that in filial relations we find GN structures because the nucleus of information is someone's identity (i.e. one concept, as in the modern first and last name combination). Conversely, N G structures are used when the nucleus of information is the nature of the relationship (i.e. two concepts, two
Anaphoric accessibility is a prevalent feature that also explains the large amount of pronominal genitives (i.e. with elus "of him~ eorum "of them~ etc.). See Bauer (1995: 55-56) for details and comments. to. Anchoring relations are those in which the possessor can be viewed as a sort of'anchor' that narrows and specifies the referent of the NP (see Taylor 2000, Koptjevskaja-Tamm 2003). 9·
Coexistl..ng structures and competl..ng functions in genitive word order 231
related individuals). 11 In my opinion, such splits in the animacy/inalienability domain may also be due to the fact that in the onomastic formula the father's name is not only predictable- being often the same as the son's name (see 4)- but it also represents information that usually decays in the subsequent discourse. If information does not decay, even with filius we get the NG order, as in ( 10 ): 12 (10) praeerat Arruns Tarquinius,filius command.iMPF.3SG Arruns .NOM.SG son.NOM.SG
regis;
rex
ipse
king.GEN.SG king.NOM.SG himself.NOM.SG
cum legionibus
sequebatur
with legion.ABL.PL follow.IMPR3SG "Arruns Tarquinius, the king's son, was in command: the king himself followed with the legions." (Liv. 2,6) GN structures can convey new, but decaying information also when the dependent is non- or less definite, as in regis filius "son of a king" vs.filius regis "the kings son': 13 As we have seen, low contrast is connected with information unpredictability and tavors postnominal genitives. However, high contrast and emphasis - often signaled by intensifiers and quantifiers - may trigger genitive fronting, whereby the genitive referent is presented as more salient and worthy of attention (Viti 2010: 81-83 ). Claims about pragmatic markedness of prenominal genitives, however, should be limited to 'true' possessive genitives (cf. DeJong 1983), because GN structures can also involve genitives with typical non-anchoring functions, as in the example tridu.i iter "a journey of three days" and argenti minae "silver coins': Interestingly, however, structures expressing duration, quantity, material or content are in no way consistent in word ordering, and in u. "For filius we find that prehead genitives were almost exceptionless, for uxor almost half of the genitives were posthead and for avunculus and patruus all were posthead" (Devine & Stephens 2006: 89). 11. An anonymous reviewer suggests a different interpretation: in (10), he/she says, there is a contrast between the son of the king and the king himself, which explains why regis does not precede filius and the NP exhibits the unmarked pragmatic order. However, filius regis is all but 'unmarked' (see also fn. 11). In fact, looking for the NP fill* (ie.filius,filia,filium, etc.) regis in the LLT-A database (period Antiquity, hundreds of Latin authors and works from 200 BCE to 200 CE), we get 15 occurrences, whereas the reverse regis fill* is found 26 times. The fact that postnominal regis is associated with a relevant (and/or distinct) piece of information is also confirmed by the followl..ng example (where there is no contrast at all): legati Ariarathis puerum
filium regis se cum adducentes Romam uenerunt; quorum oratio fuit regem educendum filium Romam mlsisse "messengers of Ariarathes arrived, bringing with them the king's young son. They explained that the king had sent his son to be brought up in Rome." (Liv. 42,19). 13. Cf. Medicum aiunt, cum regisfiliam curare deberet"There is a story that once a physician had to cure the daughter of a king." (Cic. De ira, 3,39). This usage is still found in later authors; more examples in Magni (2008: 242 fn. 45).
232 Elisabetta Magni
Classical Latin we also find N G patterns, as in aceruom lignorum "a heap of wood" ( Cato Agr. 50,2) and corbitam cibi "a bower full of food" (Plaut Cas. 778). The lack of a strict labor division between the two construction types can be related to the fact that specialized structures for anchoring and non-anchoring genitives may develop only in languages that have articles (see also the observations in Cornilescu & Nicolae, this volume). More crucially, however, the overlap between constructions with genitives, i.e. synchronic gradience, "can be seen as something steering gradual change" (Rosenbach 2010: 150). In fact, as we will see, the increasing frequency of postnominal genitives in 'appositive' structures and non-anchoring relations is particularly relevant to the change observed in Late Latin. Besides, in Latin as in English, the diachronic perspective can help to understand coexisting and competing patterns. 3.6
Sources and evolution of genitive constructions
Concerning the functions of the Latin genitive, two main correlations have been proposed: a correlation with the verbal phrase, and a correlation with the adjective. According to Benveniste (1966), the original function of genitive constructions can be seen in nominalizations that transform a verbal phrase into a noun phrase as in the example in (11): 14 ( 11) Finite clause, referring 0, OV: aquam due ere "to canalize water" Nominalized VP, referring 0, GN: aquae ductus "duct of water" Nominalized VP, non-referring 0, NN: aquaeductus "aqueduct" Since these structures grammaticalized during a period of OV syntax, many (older) genitive constructions and compounds display a GN order. For Benveniste, this basic transposing relation spreads from intransitive verbs (e.g. sol oritur "the sun rises" > so lis ortus "the rise of the sun") to nominalizations. It then extends from nominalized structures to 'true' possessives, and gives origin to all the genitive uses. Some genitive relations, however, suggest a different correlation and a different evolution. As noted by Woodcock (1959: 50), the genitive inflection "turns a noun or a person into a sort of indeclinable adjective, which is sometimes interchangeable with an adjective". This claim is reminiscent ofWackernagefs hypothesis that in Indo-European languages the use of adjectives expressing possession is not only preferred, but even anterior to genitive constructions. 15 In Latin, this usage is still preserved with the 14. "La fonction du genitif se definit comme resultant d'une transposition d'un syntagme ver-
bal en syntagme nominal" [The genitive function is defined as the result of a transposition of a verbal phrase into a noun phrase] (Benveniste 1966: 148). 15. "In den klassischen Sprachen ist hier dW'chweg das Adjektiv das Primitivere; der Genitiv mehr der ji.ingern iiberhaupt analytischen Sprachstufen etgen" [In Classical languages the adjective is everywhere the more primitive, the genitive being more characteristic of the more recent analytical stages of the language] (Wackernagel1908: 145).
Coexisting structures and competing functions in genitive word order 233 nouns of gods and those indicating relevant individuals, or social and familiar relationships and explains many interesting cases where N G structures display corresponding adjectival constructions, as in (12): (12) a. b.
3·7
sacerdos Veneris "priestess ofVenus" sacerdos Veneria
(Pl. Ru.d. 433) (Pl. Rud. 329)
arnica eri "master's lover" arnica erilis
(Pl. Mil. 105) (Pl. Mil. 114)
Genitives and adjectives
ru is well known, Dryer (1988, 1992) did not find evidence tor a correlation between OV/VO order and the order of adjective and noun. ln fact, on the one hand, the NA order is more common than the AN order in both OV and VO languages; on the other hand, in many languages, as in Latin, adjectives do not form a distinct word class. Moreover, the inflectional marking of adjectives is connected with a relative independence, rather than dependence in noun phrases (Bhat 1994: 170-171). As.;cordingly. Latin adjectives form loose paratactic structures where the modifier-modified distinction is left unspecified, and items from the same category are juxtaposed, also shifting in different positions, according to the following rule (adapted from Bauer 1995: 73): (13) [[Qualifying Adj.] Noun] in unmarked order; NA in marked order [Noun [Classifying Adj.]] in unmarked order; AN in marked order
The pairs in (14a) and (14b) illustrate a semantic and functional contrast that relates both to different (and variously labeled) adjective categories and to their usages in different contexts. Latin adjectives can thus code the difference between qualifying and classifying function, or also, as in the Romance languages, the difference between figurative and literal sense: 16
dulces voluptates "sweet-flavored pleasures" a.qu.a du.lcis "fresh water"
(Cic. Fin. 2,30) (Cat Agr. 106,1)
b. herba sanguinaria Kherb that stanches blood" sanguinaria iuventus "blood-thirsty young men"
(Col. RR 7,5,9) (Cic. Att. 2,7,3)
(14) a.
16. Marouzeau (1922: 14) uses the terms adjecNf distinctlf and adjecNf descriptlf, whereas for Adams (1976: 88-89) "objective adjectives regularly follow the noun at all periods. The adjectives which par excellence belong to this category are those derived from proper names [... ] and the possessives meus, tuus, suus, etc. On the other hand subjective adjectives - adjectives expressing a personal judgment, affective or emotive words (e.g. bonus, malus, justus, improbus, etc.)- usually precede the nouns': See Touratier (1990), and Giusti &Oniga (2006: 87-93) for a detailed discussion and more references.
234 Elisabetta Magni
Following Lehmann (1991: 222-223), we may thus observe that, at the general level of the attributive relationship, ordering is crucially related to handedness. More specifically, the prenominal position entails a tighter syntactic bond between attribute and head noun, which for the adjectives corresponds to the expression of inherent and essential properties, and to the function of description and concept formation. This position is incidentally more consistent with OV syntax, but has in fact little to do with typological correlations, being rather used to mirror the higher level of dependency of attributes. Conversely, the postnominal position entails a looser nexus, which is more suited to the expression of accidental properties, to object identification, and to discriminate an entity. An explanation for the increasing usage of postnominal attributes can thus be found in the very nature of classifying adjectives. These are often derived from nouns and, differently from qualifying adjectives, do not have comparative and superlative forms because they do not focus on a single property, but rather on the cluster of properties suggested by their general meaning "related or pertaining to X". ClassifYing adjectives contribute to participant-identification, are closer to nouns and more likely to recategorize as nouns. Moreover, classifying adjectives form appositive structures that entail a sort of 'possessive' relationship because they are more independent than qualifYing adjectives. Hence, it is reasonable to assume that, when the genitive supersedes the adjective in possessive constructions (§3.6), the new structure parallels the older one, and postnominal genitives replace postnominal adjectives. In this way, the NG order conveys a wider range of (non prototypical) possessive relations and encroaches on the functions of the GN pattern. "Cross-linguistically, it is not at all unusual that languages treat Ds and adjectival attributes (as well as other types of adnominal dependents) in the same way" (Koptjevskaja-Tamm 2003: 698). This is mainly due to the fact that nominals in attributive function and classifying adjectives are in between prototypical instances of nouns and adjectives. Besides, in languages in which adjectives are morphologically noun-like (i.e. carry the same gender/number and case marking morphology), the adjectival pattern can be easily extended to genitival modifiers: "[u]nlike the abstract xbar prindple ('harmony'), these analogical extensions are much more concrete and plausible, driven by close morphological and syntactic similarities" (Giv6n 2001: 258). 3.8
Towards new word orders
The replacement of appositive structures with a genitive is part of a more general change that can be easily traced in geographical names, where the expression urbs Roma, common in Old and Classical Latin, gave way to urbs Romae "the dty of Rome': especially frequent in tamiliar and Late Latin. Similarly, in the course of time the appositive construction with arbor "tree" in (lSa) gave way to the genitive structure exemplified in (lSb):
Coexistl..ng structures and competl..ng functions in genitive word order 235 (15) a. arbor olea "olive tree" b. arbor fici "fig tree"
(Varro, RR 1,2,20) (Col5,11,14)
Far from being an isolated phenomenon, this type of construct, which is semantically analogous to the partitive function, is integrated in the overall development of subordination.17 The attributive element, which was appositive in the archaic structure, becomes grammatically marked as the camp lement of the head, and the last stage of this shift is accomplished when the genitive is replaced by a prepositional phrase in the Romance languages. As it is well known, this process begins with the partitive function of the analytical structure de+ ABL, concerning reference to the source of an object, or the starting point of movement. Significantly, in Old French the usage of juxtaposed possessive noun phrases (e.g.la.fille le roi "the king's daughter") for (extended) inalienable possession (body-parts and relatives, but also swords and houses) still coexists with prepositional structures (Vaananen 1982:202, Koptjevskaja-Tamm 2003: 706). 18
44-1
Variation and change in Latin and English Word order flexibility in Latin
The fluidity of the syntactic cohesion within the Latin nominal group reflects the relative autonomy of the dependent in dependent-marking languages (Nichols 1986). Moreover, adnominal modification is known to show peculiar features in some flexible order languages, like Kalkatungu (Queensland, extinct) and Yimas (Papua New Guinea), where "noun modifiers are not fully integrated constituents of the noun phrase" (Rijkhoff2002: 19), and appositional modification is the rule rather than the exception. Similarly. the nominal group in Latin Kis not integrated tightly enough to be called a phrase. There is neither tight syntactic cohesion in such syntagms nor a pronounced hierarchical inequality between the substantival head and the various kinds of modifiers" (Lehmann 1991: 229). On a synchronic level, word-order flexibility allows the use of either tighter or looser syntactic bonds depending on the informational and semantic properties of the attributes. Easily available information is usually placed first in linear order and more prototypical, inherent and predictable relations are preferably encoded with more bounded constructions. Accordingly, Classical Latin employs more compact GN structures when the linkage between G and N is obvious from the context, and looser NG structures when unpredictable and less prototypical relations intertere with the anchoring function. In a diachronic perspective, the increasing usage 17. "Un genitif expltcatif s'est introduit pour etablir un lien de dependance dans plus leurs constructions appositionnelles" [In many appositional constructions, an explicative genitive arose in order to establish a dependency link] (Ernout & Thomas 1953: 43), see also Bauer 1995: 62. 1.8. Standard Italian is "an interesting exception. synchronically distinguishl..ng between the possessive preposition di and the ablative preposition da" (Koptjevskaja-Tamm 2003: 686).
236 Elisabetta Magni
of postnominal patterns paves the way to the rigidification of the new word order, and to a system that fixes word order regardless of any semantic or communicative nuances. The complex transition from the Latin flexible order to the SVO type of Romance languages can thus be viewed as reflecting a reordering that took place at different periods in a set of structures where competing factors governed the choice of constituent order. 4.2
The evolution of genitive patterns in English
Concerning genitive noun phrases, Classical Latin and Old English share many properties, such as the use of inflected genitives; no obvious preference for adnominal genitives, realized either prenominally or postnominally; the possibility for two genitives to precede (or follow) the head noun. 19 Both languages also display adjectival modifiers agreeing in gender, number and case with the head noun. All this is in perfect keeping with word order flexibility, which "is often part and parcel of the diachronic mechanisms via which one rigid word order changes into another" (Giv6n 2001: 271). Given that English changed from SOV to SVO, it is thus possible that some noun modifiers grammaticalized earlier, during the old OV syntax, while others grammaticalized later, and consistently with the new VO syntax. But grammaticalization may also take place independently from clausal word order, for instance through 'condensation' ofloose paratactic structures into tighter syntactic patterns (Giv6n 2001: 256). As we have seen (§2.1), the evolution ofGN patterns in English involves an indisputable stage where prenominal modifiers were both appositional and phrasal (s-less genitives), and a probable stage where these features are entrenched (overlap with hisgenitives). In fact, according to traditional analyses, "linking pronouns form a co-constituent with the possessee nominal and have a co-indexing function with respect to D, which identifies the possessor itself" (Koptjevskaja-Tamm 2003: 671). Possessive constructions with linking pronouns can be found in many linguistic areas, and frequently among creoles, where they are found to constitute one of the sources for head-marked possessive constructions (e.g. Krio, Sierra Leone). Conversely, in English, the linking pronoun seems to grammaticalize into a dependent-marker through conflation with the older inflectional construction, which thus becomes 'reactivated: All this is consistent with the predominant dependent-marking profile of Europe; at the same time, however, it results in typological inconsistency ( GN order) and peculiarity (no dominant order). Such characteristics, as in Classical Latin, may originate from a transition stage where some genitive relations are treated also as appositional structures.
19. On GGNfNGG structures In English (ct pas cynlnges wif"the.GEN klng's wifif, OrHead
1,11,12) and Latin (ct:Jormam Epicuri vitae beatae "Epicurus' notion of a happy lifif, Ctc. Thsc. 3,38), see the discussion In Crisma & Gianollo 2006.
Coexisting structures and competing functions in genitive word order 237 43
The role of the sources and the phenomena of gradience
As a matter of fact, both the different origin and the overlap with other constructions are crucial factors in determining past and present variation of the genitive patterns in English. Following Bybee, Perkins & Pagliuca (1994: 15 ff.), we assume that the different sources of the two patterns can contribute to explaining their different properties, in the sense that some specificities of the source construction may be maintained through the various stages of grammaticalization. Retention of earlier meanings and functions can thus determine the evolution and the subsequent behavior of a structure. As suggested in §2.2, the NG pattern in English evolves from a prepositional construction where of (cf. the cognate Latin ab) was typically associated with inanimate complements, because it originally expressed a spatial movement from a source.20 The s-genitive, on the other hand, is etymologically different, and its possible evolution from a construction with the pronominal modifier his seems to motivate the preference for animate/definite (and topical) possessors. 21 Also, coming back to the relation with appositional structures, we know that the overlap between s-genitives and noun-noun constructions (cf. (the mornings paper/ the morning paper) is a phenomenon of synchronic gradience (Rosenbach 2010). But present variation can be illuminated by diachrony, when s-less genitives mentioned in §2.1 are viewed as the bridging constructions that paved the way for structural analogy (Rosenbach 2010: 170 ff.). This long lasting proximity between s-genitives and appositional structures can also help to motivate the syntactic properties of the G N pattern discussed in §2.2. In tact, in several languages appositional modification reflects some restrictions on complexity in the prefield of the noun (Rijkhoff 2002: 20). In Yimas, for instance, noun phrases may consist of only two constituents, a modifier and a head, in this order, and the scrambled structure (with two noun phrases in apposition) occurs when a noun has more than two modifiers (Foley 1991: 188). This could thus explain why. in English, the s- genitive is still preferred for possessors premodified by up to two elements. The s- form, which was the 'scrambled' order with respect to the emergent NG structures, survives throughout the transition towards the VO order because of the cumulation of semantic and syntactic features that strengthen its usage. 22 10. On the typology ofthe gramm.aticalization path: Ablative> Possessive, see Heine & Kuteva 2002: 34. n. As is well known. pronouns are typically high on the scale of animacy/definiteness. Besides, a construction like the so-called post-genitive (e.g. a friend ofhis) does not constitute a counterexample. In fact, "the independent genitive is not in this case elliptical. Rather, the post-genitive contrasts in terms of indefiniteness or unfamiliarity with the normal determinative genitive': (Quirk et al.1985: 331). In other words, his friend presupposes definiteness, whereas in a friend of his the presupposition is one of indefiniteness, which motivates the choice of the prepositional structure. 22. Evidence for ongoing change of the s-genitive in American English in Rosenbach (2002).
238 Elisabetta Magni
5·
Conclusions
While the existence of two (apparently) synonymous patterns may seem uneconomic and contrary to the principle of isomorphism, this study of genitive variation in Latin and English confirms that word order options can be diachronically-mediated and synchronically suited to various processing needs. As noted by Giv6n, "[i]n principle, each modifier in the NP grammaticalizes on its own, following its own diachronic trajectory, that may or may not be determined by clausal ( OV vs. OV) syntax" (Giv6n 2001: 245). As a matter of fact. when investigating similarities and differences in (possessive) noun phrase variation, not only the OVNO dichotomy is at stake. And in this respect, the mechanisms of word order flexibility, which have proven to be crucially relevant for Latin, have been shown to provide some useful insights also for English.
Abbreviations ABL = ablative; Ace = accusative; AN = adjective-(head) noun; D = dependent; GEN = genitive; GN = genitive-(head) noun; IMPF = imperfect indicative; NA = (head) nounadjective; NG =(head) noun-genitive; NN =noun-noun; NOM= nominative; PL = plural; SG = singular; (s)ov = (subject)-object-verb; (s)vo = (subject)-verb-object; VP = verbal phrase.
References Adams, John N. 1976. A typological approach to Latin word order. b1dogermanische Forschungen 81: 70-99. Allen, Cynthia L 1997. Investigating the origins ofthe 'group genitive' in English. In Papers from the English Historical Syntax Meeting, Manchester, 11-12 May 1996, David Denison & Nigel Vmcent (eds). Transactions of the Philological Society 95(1 ): 111-131. Bauer, Brigltte L.M. 1995. The Emerge11ce and Development of SVO Patterning 111 Latin and French. Oxford: OUP. Benveniste, Emlle. 1966. Probleme.s de Li11guistique Generale. Paris: Editions Gall!mard Bhat. D.N.S. 1994. 1he Adjectival Category. Criteria for differentiation and identijicatio11 [Studies in Language Companion Series 24]. Amsterdam: John Benjamins. Bolkestein, Machtelt. 2001. Random Scrambling? Constraints on discontinuity in Latin noun phrases. In De Lingua LaNna Novae Quae.stio11es, Claude Moussy (ed), 245-258. Louvain: Peeters. Bybee, Joan L, Pagliuca, William & Perkins, Revere D. 1994. The Evolution of Grammar. Chicago IL: The University of Chicago Press. Crisma, Paola & Gianollo, Chiara. 2006. Where did Romance N-raising come from? A parallel study of parameter resetting in Latin and English. In Romance Language.s and Li,~guisNc
Coexisting structures and competing functions in genitive word order 239
Theory. Selected Papers from 'Going Romance' 2004 [Current Issues in Linguistic Theory 278], Jenny Doetjes & Paz Gonzalez (eds), 71-93. Amsterdam: JohnBenjamins. Devine, Andrew M. & Stephens, Laurence D. 2006. Latin Word Order. Structured Meaning and Information. Oxford: OUP. Dryer, MatthewS. 1988. Object-verb order and adjective-noun order: Dispelling a myth. Lingua 74: 185-217. Dryer, Matthew S. 1992. The Greenbergian word order correlations. Language 68: 81-138. Dryer, MatthewS. 2008. Order of genitive and noWL In The World Atlas of Language Structures Online, Martin Haspelmath, MatthewS. Dryer, David Gil & Bernard Comrie (eds), Ch. 86. Munich: Max Planck Digital Library. (7 April2010). Ernout, Alfred & Thomas, Fran~ois. 1953 Syntaxe Latine. Paris: Klincksieck. Foley William A. 1991. The Yimas Language of New Guinea. Stanford CA: Stanford University Press Giusti. Giuliana & 0 niga, Renato. 2006. La struttura del sintagma nominale Latino. In Affl della Giornata di Lingulstica Latina (Venezia, 7 magglo 2004), Renato Oniga & Luigi Zennaro (eds), 71-100. Venezia: Cafoscarina Giv6n, Talmy. 2001. Synta.r:: An Introduction, Vol 1. Amsterdam: John Benjamins. Jager Gerhard & Rosenbach, Anette 2006. The winner takes it all - almost: Cumulativity in grammatical variation. Linguistics 44(5): 937-971. Janda, Richard D. 1980. 0 n the decline of declensional systems: The overall loss of 0 E nominal case infiections and theME reanalysis of -es as his. In Papers from the Fourth International Conference on Historical Linguistics, Elizabeth C. Traugott, Rebecca Labrum & Susan C. Shepherd (eds), 243-252. Amsterdam: John Benjamins. Janda, Richard D. 2001. Beyond 'pathways' and 'unidirectionality': On the discontinuity oflanguage transmission and the counterability of grammaticalization. Language Sciences 23: 265-340. de Jong, Jan R 1983. Word order within Latin noun phrases. In Latin Linguistics and Linguistic Theory [Studies in Language Companion Series 12], Harm Pinkster (ed..), 131-144. Amsterdam: John Benjamins. Heine, Bernd & Kuteva, Tania. 2002. World Lexicon of Grammaticalization. Cambridge: CUP. Keizer, Evelien. 2007. Prenominal possessives in English. Function and use. In Structural-functional studies in English grammar in honour of Lachlan Mackenzie [Studies in Language Companion Series 83], Mike Hannay & Gerard J. Steen (eds), 59-82. Amsterdam: John Benjamins. Koptjevskaja-Tamm, Marla. 2003. Possessive noun phrases in the languages of Europe. In Noun Phrase Structure in the Language.s ofEurope, Frans Plank (ed..), 621-722. Berlin: Mouton de Gruyter. Lehmann, Christian. 1991. The Latin nominal group in typological perspective. In New Studie.s in Latin Linguistics, Robert Coleman (ed), 203-232. Amsterdam: John Benjamins. LLT-Series A. 2008. Library of Latin Texts [Centre 'Traditio Litterarum Occidentalium1, Paul Tombeur (ed). Thrnhout: Brepols. Magnl. Elisabetta. 2008. The evolution of Latin word (dls)order. In Universals of Language Today, Sergio Scallse, Elisabetta Magni & Antonietta Bisetto (eels), 225-251. Berlin: Springer. Marouzeau, Jules. 1922. J:ordre des mots da,IS la phrase Latine, 1bme I: Les groupes 11ominaux. Paris: Champion. Nichols, Joanna. 1986. Head-marking and dependent-marking grammar. Language 62: 56-119. Pinkster, Harm. 1980. Latin Syntax and Semantics. London: Routledge.
240
Elisabetta Magni Quirk, Randolph, Greenbaum, Sidney,. Leech, Geoffrey & Svartvlk, Jan 1985. A Comprehe,ISive
Grammar of the E11glish La11guage. London: Longman Rijkhoff, Jan 2002. The Noun Phrase. Oxford: OUP. Rosenbach, Anette. 2002. Ge111tive Variation in English: Co11ceptual Factors 111 Synchronic and Diachronic Studie.s. Berlin: Mouton de Gruyter. Rosenbach, Anette. 2004. The English s-genitive. A case of degrammaticalization? In Up and Down the Cli11e. The Nature of Grammaticallzatlo11 [Typological Studies in Language 59], Olga Fischer, Muriel Norde & Harry Perrtdon (eds), 73-96. Amsterdam: John Benjamins. Rosenbach, Anette. 2010 How synchronic gradience makes sense in the light oflanguage change (and vice versa). In Gradience, Gradualness and Grammaticalizatlon [Typological Studies in Language 90]. Elizabeth Closs Traugott & Graeme Trousdale (eds), 149-179. Amsterdam: John Benjamins. Rosenkranz, Bernhard 1933. Die Stellung des Attributiven Genettvs tm Italtschen. Indogermanische Forschungen 51: 131-139. Taylor, John R. 2000. Possessives in English: An Exploration in CogniNve Grammar. Oxford: OUP. Touratier, Christian. 1990. Adjectif et syntagme nominal. In New Studies in Latin Linguistics [Studies in Language Companion Series 21], Robert Coleman (ed), 233-249. Amsterdam: John Benjamins. Traugott, Elizabeth Closs & Trousdale, Graeme. 2010. Gradience, Gradualness and Grammaticalization [Studies in Language Companion Series 90]. Amsterdam: John Benjamins. Vaaniinen, Veikko. 1982. Introduzione allatlno volgare (transl. of Introduction au latin vulgaire. Paris: Klincksieck. 1963). Bologna: Patron. Viti. Carlotta. 201 0. Observations on genitive word order in Latin. In Le syntagme nominal en latin. Nouvelles contributions Actes de l~telier du Centre Alfred Ernout, Universltf de Paris-Sorbonne, 11 octobre 2008, Olga Spevak(ed), 73-92 (7 April2010). Wackernagel, Jacob. 1908. Genetiv und Adjektiv. In Mllanges de linguistique offerts aM. Ferdinand de Saussure, 125-152. Paris: Champion. Woodcock, Eric C. 1959. A New Latin Syntax. London: Methuen & Co. Wyld, Henry C. 1953. A History of Modern Colloquial English. Oxford: BasU Blackwell. (Reprint).
Anaphoric adjectives becoming determiners* A corpus-based account Freek Van de Velde Research Foundation FWO/University of Leuven
Standard accounts of determiners typically deal with the few well-known elements that fall under this category: articles, demonstratives, possessives and (some) quantifiers. It can be shown, however, that the determiner slot in Dutch can also be occupied by certain elements that do not regularly feature in reference grammars, namely the anaphoric adjectives like voornoemd ("aforementioned"). Their syntax is subject to variation in Present-day Dutch, and possibly to change as well: a corpus study reveals that they are increasingly used as unequivocal determiners, irrespective of their token frequency.
1.
Determiners
In both formal and functional approaches to syntax, it is commonly assumed that noun phrases in English and Dutch have a determiner slot Opinions diverge on the issue whether this determiner is the head of the whole constellation (the DP approach) or whether the determiner is some sort of modifier of the head noun (the traditional NP approach) (see Abney 1987; Hewson 1991; Van Langendonck 1994; Hudson 2004; Escribano 2006; Van Eynde 2006, amongst others for arguments for either sides). Respective abstract templates are given in (1) and (2). This article takes an NP approach, but the results do not crucially hinge on it (1)
[DP ... D ... [NP ... N ...]]
(2) [NP [... D ... [... N ... )]]
Another point of discussion is the question whether determiners are a universal category, that is, whether all languages have determiners and whether they always have had determiners during their histories. This idea appears to be flouted by languages
*
I would like to thank Joop van der Horst and the anonymous reviewers for helpful comments on earlier drafts of this paper, and Robert Cloutier for checking the language. Remaining errors are solely the responsibility of the author.
242 Freek Van de Velde
such as Latin. Turkish or Russian. which seem to do without articles and have bare nouns fulfilling the function of NP/DPs. A solution to save the universality of the D-slot would be to assume a covert determiner, one that does not crop up in the surface structure but which can be 'cryptotypically' recovered. The latter point of discussion is not entirely unconnected to the former. The DP approach naturally leads to a more universal conception of the determiner slot than an NP approach, where the determiner is seen as merely a modifier of the NP. Still, there is no necessary entailment One could argue that not all languages have determiners, but that as soon as a language does acquire them, they become the head of the whole nominal complex. With regard to the history of the English language, some have argued that Old English lacks a determiner projection (e.g. Yamamoto 1989; Osawa 2007), whereas others have maintained that it was there already (e.g. Wood 2007). In this paper, I do not want to weigh the arguments for or against a universal determiner slot, but rather that the elements that can occupy the determiner slot do not remain constant over time: some elements may shift from the adjective category to the determiner category. Extension of the determiner category by elements that formerly functioned as adjectives or nouns is discussed by Spamer (1979), Roberts & Roussou (2003:ch.4), Denison (2006) and Van de Velde (2009b) for English, by Plank (1992) for German and by Van de Velde (2009a, 2010) for Dutch. In the functionalist approach, the process has been linked with leftward movement due to grammaticalisation and subjectification (see Adamson 2000; Davidse et al. 2008). A good example is the history of the attributive possessive in English. Following Lyons (1986, 1999) in his distinction between adjectival genitive (AG) and determiner genitive (DG) languages, English can be argued to have developed determiner possessives out of adjectival possessives. 1 The combination of a demonstrative with a possessor, as attested in Old English, see (3), is no longer possible -or at least only marginally possible -in Present-day English. Assuming that determiners cannot be stacked (see below), these data show that the possessive was not a determiner yet in Old English. The pattern was gradually ousted in Modern English. (Bede 282.23, cited in Wood 2007)
(3) Old English
in pis
user circlice
steer
in this.ACC.N our ecclesiastical history.N "in (..this) our ecclesiastical history" In order to decide whether an element occupies the determiner slot or whether it is an adjectival modifier of the noun, a set of criteria is needed. What are the diagnostics of determiners? Semantically, determiners are concerned with deixis and definiteness, but semantics alone will not be enough to decide the syntactic status. Not all elements that are associated with definiteness qualify as full-fledged determiners. In older stages Some scholars reject the idea that the pattern in (3) shows that Old English is a DG language. Indeed. the whole AG/DG distinction remains a matter of debate.
1.
Anaphoric adjectives becoming determiners 243
of Germanic, for example, ordinary adjectives could signal definiteness by their inflectional ending. Weak inflection was sufficient to make the NP definite (see Traugott 1992: 171 ). This does not imply that these adjectives were determiners. ln fact, they are even more closely linked to the noun than strongly inflected adjectives (see Spamer 1979; Fischer 2001: 258), which were neutral with regard to definiteness (see Quirk & Wrenn 1969: 68; Traugott 1992: 173). One way to explain this would be to assume that in this case definiteness is pragmatically inferred, rather than syntactically encoded. Are there any formal indications to distinguish determiners from adjectives? In Present-day English, the literature provides the tollowing criteria: (i) mutual exclusivity; (ii) obligatoriness; (iii) linear order vis-a-vis numerals and other adjectives; and (iv) exclusion from the predicate position (see a.o. Quirk et al. 1985: 253-264. 402403, 1238-1239; Bernstein 2001; Longobardi 2001; Huddleston & Pullum 2002: 354358, 368-399, 538-540 esp.). The same criteria apply to Present-day Dutch (see Van de Velde 2009a: 252-258). First, as was already briefly touched upan above, determiners are considered to be mutually exclusive. They cannot be 'stacked'. This means that if we encounter two elements in a row that are potentially involved in definiteness, say a demonstrative and a possessive, as in ( 3), only one of them can be a determiner. The idea of mutual exclusivity seems to fly in the face of patterns like all the men (BNC) or such a proposal ( BN C), in which a definite and an indefinite article are preceded by all and such. 2 Since the articles themselves are generally considered to be prototypical determiners (Huddleston & Pullum 2002: 538; Alexiadou et al. 2007: 52, 55ff.), the BNC examples in question display stacking, as all and such seem to be determiners as well, at least in cases like all men (BNC) or such proposals (BNC). 3 In most reference grammars the issue is resolved by putting all and such in the former examples into a slot for so-called 'predeterminers' (see Huddleston & Pullum 2002: 433-436). The second criterion for determiners is that they are obligatory to some extent When NPs are used as an argument of the verb, they have to be preceded by a determiner. So-called bare NPs cannot normally be used as arguments, see (4-5 ). Still, there 2. M a matter of principle, all examples in this article have been taken from text corpora. The following corpora have been used: British National Corpus (BN C), the 38 million word corpus ofiNL (INL38), the internet (queried by Google), the corpus De Gids (GDS) and the Literom corpus (LTR). The latter two corpora are not commonly used in linguistics, and they will be properly introduced below. Details on the former can be found ln the references at the end of thls article. 3· The such example is in fact more complicated. M it is indefinite (as opposed to the example with all), it could be analysed as having a 'zero determiner': [NP 0 such proposals]. Still, although not frequently, bare such can also occur ln front of singular count nouns, e.g. and in case such person be convicted (... ) (BNC), suggesting that it is a determiner there. And to complicate matters even more, there are patterns like no such person (BNC), where such follows the determiner. The intricacies of such will not be discussed ln this article. The reader is referred to Bolinger (1972), Mackenzie (1997), Wood (2002), and Splnillo (2003) for further discussion.
244 Freek Van de Velde
are notable exceptions, the most obvious of which are indefinite plurals and non-count nouns in Present-day English and Dutch, as in (6-7).4 Other exceptions are languagespecific. For Dutch an overview is given in Haeseryn et al. (1997: 195-223). (4) He stopped the car (5) *He stopped car (6) They deliver cars. (7) Don't eat rice or pasta today.
(BNC) (BNC) (BNC)
The third criterion for determiners is their linear order with respect to numerals and adjectives. Apart from what Huddleston & Pullum (2002: 436-439) have called peripheral modifiers (focussing adverbs and the like) determiners constitute the left edge of the NP. Numerals and adjectives are enclosed between the noun and the determiner, some exceptions like the 'so-tall-a-man' construction notwithstanding. From this we can derive negative evidence for determiner status: if a certain element, say a possessive, can follow an adjective, it may be concluded that the possessive is not a determiner in this language stage. Note that in Early Modern English, possessives could still be preceded by adjectives, see (8), a pattern that is ungrammatical nowadays. Indeed, there is convergence in the other criteria suggesting that the possessive only reached determiner status during the Modern English period (see Van de Velde 2009b). (8) Early Modern English
(Wood 2007: 174, AD 1590-1598)
sweet my child The fourth criterion is exclusion from the predicate position: determiners are excluded from the predicate position in predicational copular clauses, see (9-10). 5 (9) the results are good (10) *the results are some/the/my/ ...
(BNC)
A number of provisos have to be made with regard to this criterion. First, it must be noted that possessives do occur in predicational copular clauses, but only if they are marked with additional morphology (my-mine, your-yours etc.). Second, the criterion is only a necessary condition, not a sufficient one: some adjectives are excluded from predicational copular clauses, but not because they are determiners.
4· It is generally assumed that we have a zero (unexpressed, covert) deter miner in these cases. Positing zero categories can be objected to on general methodological grounds, though, so it ls perhaps safer to say that these cases are simply contexts without an obligatory determiner. In thls article, the criterion will only be used to assess determinerhood in definite NPs. 5· See e.g. Declerck (1988) for the distinction between 'predicational' and 'speclficational' copular clauses. Note that determiners are only excluded from the former, not from the latter, as shown by the following example with a demonstrative as a speclficational predicate: So the lessons are the.se (BNC).
Anaphoric adjectives becoming determiners 245 These four criteria do not exhaust the proposals in the literature. Huddleston & Pullum (2002: 539) observe that determiners can be used in a partitive construction. (11) some examples (12) some of the example (13) good examples (14) *good of the examples
(BNC) (BNC) (BNC)
However, this criterion is neither necessary (articles and possessives do not occur in the partitive construction), nor sufficient (see Seppanen et al. 2002 on the 'so-tall-of-aman-as-he-is' construction). An obvious question about these tour criteria is how it can be independently established that they are characteristics of determinerhood. Indeed, this is often left implicit in the literature cited above. With regard to mutual exclusivity, the danger of circularity is lurking ifwe say that (obsolescent) patterns with proximal demonstrative+ possessive show that the possessive is not a determiner yet. Still, it can be observed that mutual exclusivity evolves together with the other criteria (see Van de Velde 2010), which may be taken as an indication that it is at least related to the other criteria. Second, there is diachronic unidirectionality in that mutual exclusivity is of later date than combinability. This fits well with the idea that grammaticaVfunctional categories like auxiliaries and determiners unidirectionally grammaticalise out of descriptive categories like main verbs and adjectives. The criterion of mutual exclusivity would be much more suspect if mutual exclusivity was historically prior to mutual combinability. So while there is no direct semantic evidence that links mutual exclusivity to determinerhood, there is circumstantial evidence: grammaticalisation of undisputed determiners seems to go hand in hand with the emergence of mutual exclusivity. With regard to obligatoriness, the idea is that the determiner is needed to 'anchor' the noun phrases in the discourse. This idea is upheld both in generative accounts (see e.g. Bernstein 2001) and in functional accounts (see e.g. Langacker 1991). In the present article, I take a cautious stance by assuming that languages can, but need not grammaticalise this anchoring function, but that if a language grammaticalises it, it has to be obligatorily expressed in certain contexts. Note that it need not be obligatory in all contexts. This is by the way the case for all kinds of 'obligatory' categories: in languages that have them, agreement phenomena, double negation, nominal classifiers etc. have exception contexts too. The point is that the exceptions are consistent. That is, a singular count noun, probably the most typical instance of a (non-pronominal) NP, is a context where the determiner is obligatory in English (and in Dutch). The obligatoriness criterion would be more problematic if we could not clearly delineate the contexts where a determiner is lacking, or if obligatoriness could only be observed in the least prototypical NPs. The third criterion, relative order vis-a-vis adjectives, is a consequence of scope, with determiners taking a broader scope than regular adjectives, and the fourth
246 Freek Van de Velde
criterion, loss of predicative use. can be linked to the loss of their descriptive content (see Van de Velde 2009a: 255-256). Moreover, there is unidirectionality here as well: (unmarked) possessives may at a certain point stop occurring in copular clauses, but not the other way around. ln the remainder of this article, these criteria will be applied to the category of 'anaphoric adjectives' in Dutch, in order to show that they undergo a slow and gradual shift from adjective to determiner.
2.
Anaphoric adjectives
The label 'anaphoric adjectives' is conveniently used here as shorthand for a whole group of adjectives that fulfil a discourse-deictic (or textual-deictic) function (see Lyons 1977: 667-668 and Levinson 1983: 85-89 for these terms). 6 1hey occur in other languages as well, see for example Perridon (1989: 207) on Swedish and Mortelm.ans (2006) on French. In Dutch the prototypical member of this group is voornoemd ("aforementioned"). Other members are vermeld and gemeld (both meaning "mentioned"), bedoeld ("intended") engezegd ("said"). Examples are provided in (15-19).
heer bezigde onheuse taal en als ambtenaar (15) Voornoemde Aforementioned man used inappropriate language and as civil.servant in functie heb ik hem scherp terechtgewezen (INL38) in office have I him sharply reprimanded "The aforementioned man used inappropriate language and as a civil servant in office. I severely reprimanded him." (16) In vermelde brief heb ik voorgesteld ook voor de ftetsers op de Randweg In mentioned letter have I proposed also for the cyclists on the Randweg en de Leonard Sptingerlaan verkeerslichten aa.n te brengen (INL38) and the Leonard Springerlaan traffic.lights PART to install "In the aforementioned letter I have proposed to install traffic lights at Randweg and Leonard Springerlaan for the cyclists as well." (17) de overige bepalingen van gemelde wet(...) the other stipulations of mentioned law "the other stipulations of the mentioned law"
(INL38)
( 18) De naam van bedoeld werktuig is 'zege'. is '7£ge The name of intended tool "The name of the intended tool is 'zege":
(INL38)
6. Both Lyons and Levinson distinguish between discourse (or textual) deixis and anaphora. The issue will not be pursued here.
Anaphoric adjectives becoming determiners 247 (19)
Wethouder H. Jansen van Onderwijs vindt dat het met gezegde principiele Alderman H. Jansen of Eduction thinks that it with said
opstelling wel meel'a.lt.
principled (1NL38)
attitude well was.nottoo.bad "Alderman H. Jansen of Education thinks the said principled attitude was not too bad after all" Apart from the prototypical voornoemd, all the adjectives just mentioned need not be used discourse-deictically, but can occur in other contexts as well. In the legalese example (20), vermelde does not signal that the "instructions" (gebruiksaa.nwijzingen) have been mentioned earlier in the text, but rather literally means that the instructions are explicitly mentioned on the package of a certain product. Likewise in (21), where bedoeld does not mean that the referent has been mentioned earlier in the text, but simply means "the word that the aphasia patient intends to use". (20) Deze aanduidingen moeten letterlijk worden 1'ermeld, met dien These indications must literally be mentioned, on the
verstande
dat, wat betreft
de vermelde gebruiksaanwijzingen,
understanding that, what concerns the mentioned instructions,
niet alle aangege'l'en toepassingen behoe'l'en te worden vermeld.
(INL38)
not all indicated applications need to be mentioned. "These indications have to be mentioned explicitly, on the understanding that with regard to the instructions mentioned not all indicated applications have to be mentioned:' (21) De afasiepatient moet zich daa.rvoor extra inspannen en zelfs dan The aphasia. patient must himself for. that extra give.effort and even then lukt het vaak niet het bedoelde woord te zeggen (INL38) succeeds it often not the intended word to say "The aphasia patient has to make an extra effort and even then often does not succeed in coming up with the intended word:'
The examples in (15-19) all seem to 'lack' a proper determiner. The anaphoric adjective introduces a singular countNP, and would normally need an article in these contexts. So instead of 1'oornoemde heer, one would expect de voornoemde heer. But rather than assuming that the article has been 'dropped' or is 'missing' in these examples, one could also apply criteria (i) ('mutual exclusivity') and (ii) ('obligatoriness') to them and conclude that the anaphoric adjectives themselves occupy the determiner slot here (see also Roose 1964: 35). The anaphoric adjective then precludes the use of the article. This makes sense from a semantic perspective as well: the textual-deictic function of the anaphoric adjectives squares to the deictic function of the more typical determiners. In some instances, we have additional evidence that the anaphoric adjectives are determiners, rather than adjectives with article-drop. This evidence comes from the morphology. As a rule, attributive adjectives in (Standard) Dutch are always inflected,
248 Freek Van de Velde
by means of a schwa-ending unless the NP is indefinite, singular and neuter. Inflecting determiners like possessive pronouns lack the inflection not only in indefinite, but in definite single neuters as well. (22) onz-e klein-e tuin our.INFL small.INFL garden "our small garden"
[+def; +Sg; +masc]
klein-e huis (23) ons our.UNINFL small.INFL house "our small house"
[+def; +Sg; +neutr]
(24) een klein huis a small. UNINFL house "a small house"
[-def; +Sg; +neutr]
On the basis of their discourse-deictic properties, the anaphoric adjectives signal definiteness. If they were regular adjectives, they would have the inflectional schwa when preceding a singular neuter, like kleine in (23). If, on the other hand, they function as determiners, they can remain uninflected before a singular neuter, like ons in (23). This is indeed what is found. ln (18) the anaphoric adjective bedoeld is uninflected as it precedes a singular neuter noun werktuig. If the examples (15-19) were merely instances of article omission. then the inflectional pattern in (18) remains unexplained. 7•8 Another argument for the idea that the anaphoric adjectives can occupy the determiner slot is that they can all precede numerals. By criterion (iii) mentioned above, this lends credence to an analysis as determiners. An internet example obtained via Google is given in (25). (25) (... ) indien de gebruikerzich niet houdt aan de voorwaarden in if the user REFL not abides by the conditions in
voornoemde
twee paragrafen.
aforementioned two paragraphs Kif the user does not abide by the conditions in the aforementioned two paragraphs:' 7·
I credit this observation to Fred Weerman
One anonymous reviewer notes that a Google search yields nwnerous instances of het voor("the aforementioned UNINFL"), in which the amphoric adjective has determiner inflection but Is still preceded by a definite article. A doser look at the internet examples reveals, however, that the majority of them comes from Belgian websites, and in Belgian Dutch, the inflectional schwa on adjectives is dropped in definite singular neuters as well, so that the pattern Is perfectly norrn.al. The reviewer furthermore notes that a following adjective can remain uninflected after uninflected voonwemd, as in the Google example voonwemd groot balkon ("aforementioned UNINFL big.UNINFL balcony"). This phenomenon Is also observed, however, with other determiners, see De Rooij (1980) on ons bruin paard ("our. UNINFL brown. UNINFL horse"). 8.
J~oemd
Anaphoric adjectives becoming determiners 249 If anaphoric adjectives function as determiners, they are not expected to occur after a copula (criterion iv, above). And indeed, in their discourse-deictic uses, the adjectives under consideration are awkward in this construction. The proviso 'in their discoursedeictic uses' is important: the adjectives can be used as the predicate in their literal sense, exemplified in (21-22). (26) ??de voorwaa.rden zijn voornoemd!bedoeld!vermeld!... "the conditions are aforementioned/intended/mentioned/ ..."
To sum up, criteria (i)-(iv) seem to suggest that anaphoric adjectives can reasonably be regarded as determiners. The picture is not as neat as suggested, however. The idea that anaphoric adjectives function as determiners is complicated by the observation that they do not systema.tically do so. There is variation when it comes to whether or not the adjective is preceded by an article. In (27 -28) the anaphoric adjective is preceded by a definite article. (27) Deze aanwijzingen mogen niet itt strijd
These instructions may
zijn met de voornoemde
not in conflict be with the aforementioned
tekst.
(INL38)
text "These instructions must not contradict the aforementioned text:' (28) In de periode waarin de bedoelde onderzoeken werden verricht,
In the period wherein the intended investigations were
was
conducted was
aids nog niet onder de menselijke populatie gei'dentificeerd.
(INL38)
AIDS yet not under the human population identified "In the period in which the intended investigations were conducted, AIDS was not yet identified in the human population."
It appears that anaphoric adjectives can function alternatively as determiners and as adjectives. This kind of variation is nothing to be surprised about In English too, adnominal elements can occupy different slots: all can be a determiner and a predeterminer; many can be a predeterminer (many a book), a determiner (many (of the) books), and an adjective (the ma.ny books). What is interesting, however, is that the variation can be explained diachronically. There are indications that anaphoric adjectives in Dutch are diachronically shifting from the adjective category to the determiner category. The variation in Present-day Dutch could then be due to the fact that this shift is fairly recent and has not been completed (yet). In a way, the Dutch anaphoric adjectives show the same kind of variation as the Early Modem English and Early Modem Dutch possessives, which were then also in the middle of a transition from adjective to determiner. The intuition that the Dutch anaphoric adjectives drift towards the determiner slot, can be corroborated by a corpus study, the results of which will be presented in the next section.
250
Freek Van de Velde 3·
Variation and change in Late Modem Dutch
In what follows, a closer look will be taken at the variation and change in the syntactic behaviour of a number of Dutch anaphoric adjectives, namely those adjectives that have been mentioned above: voornoemd, bedoeld, gezegd, gemeld, vermeld. Two corpora will be queried to assess the changes in the syntactic context in which these five adjectives occur. 9 The first is a compilation of the 1841-1930 issues of the literary and cultural journal De Gids, comprising about 68 million words. The second is a selection of literary reviews (Literom) from the period 1931-1990, totalling around 50 million words. 10 Together, these corpora span a period of 150 years of Late Modern Dutch, from 1841 to 1990. The total period has been divided in 5 periods of30 years each. & said, the anaphoric adjectives are regularly used as determiners, as signalled by the absence of an extra article, as in ( 15), or by their ability to precede numerals, as in (29 ). (29) Volgens een zeer matige scha.tting zijn van bedoelde 500 plaatsen, Following a very moderate estimation are of intended 500 seats 300 elk £BOO(...) waard (GDS, 1876) 300 each £ 800 worth "According to a very moderate estimation 300 of the intended 500 seats are worth £ 800 each:' But today, there is still variation in their use, as shown in (27-28), where voornoemd and bedoeld have adjectival syntax. The variation is brought out clearly in the 'minimal pair' in (30-31), extracted from the internet via a Google search. (30) in voornoemde drie gevallen in atorementioned three cases "in the three aforementioned cases" 9· For practical reasons, only anaphoric adjectives in NPs inside PPs have been queried: by restricting the search string to a combination of one of the adjectives van (""of'), in ("in"), op ("olf), voor ("for"), aan ("td'), met ("with") + de!het ("the") or 0 +the anaphoric adjective, the verbal use of the past participles that are homonymous (or rather polysemous) with the anaphoric adjectives, e.g. gezegd, did not infiate the number with spurious hits. This does not mean that all the hits were useful The hits were manually checked to make sure that only the anaphoric uses of the adjectives were retained. Still, an WIWelcome consequence of the limited query is that the results presented in this paper only apply to a specific syntactic context. Additional research is needed to ascertain whether the results stand up to a study of other syntactic contexts as well 10. The size of the Literom corpus should be approached with some caution, as some literary reviews seem to have been selected more than once. In the results, this has been taken care of by eliminating the double hits. Furthermore. the corpus size is not equally distributed over the various decades. For the period 1931-1940, the total number of tokens is about 2 million, whereas for the period 1981-1990, the total number is about 20 million. As we will only be looking at proportions or normalised frequencies, this does not pose an immediate problem, but it is good to keep this in mind.
Anaphoric adjectives becoming determiners (31)
in de drie voornoemde
gevallen
in the three aforementioned cases "in the three aforementioned cases" The corpus data, however, reveal that there is a trend in Late Modern Dutch to put the anaphoric adjectives more and more in the determiner slot, rather than in the adjective slot. The following graph shows the proportion ofhits in which there is no preceding article ('determiner use') to the hits in which an article does precede voornoemd ('adjectival use'). 11 A statistical test was used to investigate whether the association between the period and the syntactic use of the anaphoric adjectives is significant The Chi-square test indicates significance to an alpha level of 0.01 (f = 22.2791, DF = 4, p = 0.0002). The Gamma test measures the correlation between the ordinal independent variable 'period' and the dependent variable 'syntactic us~ which as a two-value nominal variable can be treated as ordinal. Test results indicate a moderate, yet significant correlation (f = 0.2465, asymptotic standard error= 0.0507). The correlation is visible in the individual anaphoric adjectives as well, though for gezegd, gemeld and vermeld significance cannot be statistically guaranteed due to the limited number of observations. 12 100%
50%
0%
• adjective use • determiner use
18411870
18711900
19011930
19311960
169 71
19611990
163
89
18
43
114
77
16
51
Figure 1. Anaphoric voornoemd!ber:Weld!gezegd!gemeld!vermeld
u. In most of the cases, the criteria for determinerhood mentioned above are aligned Occasionally, however, conflicting results are obtained This happens when the anaphoric adjective is preceded by an article, but itself precedes a numeral. Such instances were left out of the quantitative results presented in this paper. The reason behind the conflicting results is that criterion (iii) is acquired earlier that criterion (i) and (ii), when adjectives become determiners (see also Van de Velde 2010).
This does not mean that all possible anaphoric adjectives necessarily move towards determiner status. An additional query on genoemd ("mentioned") (n = 1449), which can also be used as a determiner, shows no diachronic trend towards determiner status.
12.
251
252
Freek Van de Velde 4·
Theoretical ramifications and conclusions
An ali-or-nothing approach to the issue of what position anaphoric adjectives occupy in Present-day Dutch NPs is bound to fail. Taking attested corpus data into account, we do not obtain a clear-cut picture, but rather much variation: even on a fairly limited number of simple diagnostics for determinerhood (criterion i-iv, above), the anaphoric elements under consideration can function either as adjectives or as full-blown determiners. This is a problem for synchronic accounts of adnominal syntax that are not capable of coping with variational data. What is needed here is a historical, variationist perspective. As shown in Section 3, the data suggest a diachronic trend: anaphoric adjectives gradually shift to the determiner category in Late Modem Dutch. This process fits well with the idea expounded in Van de Velde (2009a, 2010) that the determiner is not a stable category in Dutch, but recruits various adjectival elements with deictic characteristics (see also Davidse et al 2008 for a semantic analysis) overtime. The shift towards determiner use has been argued to involve grammaticalisation (see e.g. Roberts & Roussou 2003, ch.4). Indeed, when demonstratives evolve into articles, or when adjectives like same and other acquire determinerlike features, they show characteristics of grammaticalisation as listed by Lehmann (1985) and Hopper (1991), like for example coalescence, paradigmaticisation and decategorialisation (see Breban 2010). The anaphoric adjectives undergo decategorialisation as well. They lose adjective features like the possibility to occur as a predicate. Still, the syntactic evolution of the anaphoric adjectives differs from other cases of grammaticalisation. Normally, grammaticalisation of an element boosts its token frequency, see for example Bybee (2003), but this is not the case with the anaphoric adjectives, which do not seem to become more popular in Present-day Dutch. On the contrary, the occurrence of anaphoric adjectives (in PPs) per million words decreases dramatically over time in the consulted corpora, as can be seen in Figure 2. Exactly how to understand this trend is not clear. One plausible explanation would be to assume a stylistic shift. Anaphoric adjectives occur predominantly in written, rather stilted registers. With the informalisation of written genres (see e.g. Biber & Finegan 1989 and, specifically for Dutch, Van den Toom 1997: 479-494), the anaphoric adjectives have fallen into disuse, to some extent, without, however, having an impact on the ongoing change in their syntax. The way to reconcile the grammaticalisation trend and the infrequency of anaphoric adjectives is to view the process as a case of grammaticalisation by analogical extension (see also Croft's (2000: 148-1 56) notion of intraference ). Such a process is different from the grammaticalisation of auxiliaries, because there is no 'specialisation' (Hopper 1991). Contrary to what is the case in the grammaticalisation of the article from, say, either the demonstrative, a negation particle or a perfective auxiliary, the anaphoric adjectives are not becoming the default expression for definiteness. Furthermore, analogical change in general is known to affect low-frequency elements as well (see Bybee 2007).
Anaphoric adjectives becoming determiners 253 12 10 8 6
4 2
0
• freq. I million
1841-
1871-
1901-
1931-
1961-
1870
1900
1930
1960
1990
11.6
10.8
7.7
4.8
2.0
tokens
Figure 2. Relative frequency of anaphoric voornoemdlbedoeld!gezegd!gemeldlvermeld The process is then comparable to other instances of grammaticalisation by analogical extension that do not entail increased token frequency. One example is the grammaticalisation of new complex prepositions of the form [eomplexP P + N + P] (see Hoffmann 2004). Some of these new forms are relatively frequent like in front of, in terms of, in spite of, in relation to and with respect to. Their grammaticalised use in the complex preposition has increased the token frequency of the noun. Other complex prepositions are, however, notably infrequent, like by dint of or in conformity with. Of course, one could object that the analogical extension of determiner properties to anaphoric adjectives is not really a case of grammaticalisation, and therefore that the decreasing frequency is no longer a theoretical problem. But given that the process does involve other characteristics of grammaticalisation, for example decategorialisation, the grammaticalisation account should not be immediately dismissed. The main concern of this article is not, however, the theoretical status of grammaticalisation, but rather to show how a corpus-based study can bring order to the chaos in the syntax of anaphoric adjectives.
References Abney, Steven P. 1987. The English Noun Phrase in its Sentential Aspects. PhD dissertation, MIT. Adamson, Sylvia M. 2000. A lovely little example. In Pathways of Change. Grammaticallzation in English, [Studies in Language Companion Series 53], Olga Fischer, Anette Rosenbach & Dieter Stein (eds), 39-66. Amsterdam: John Benjamins.
254 Freek Van de Velde
Alexiadou. Artemis, Haegem.an. Liliane & Stavrou, Melita. 2007. Nou11 Phrase in the Generative Perspective. Berlin: Mouton de Gruyter. Bernstein, Judy B. 2001. The DP hypothesis: Identifying clausal properties in the nominal domain. In Baltin & Collins, 536-561. Biber, Douglas & Finegan. Edward 1989. Drift and the evolution of English style: A history of three genres. InLanguage65(3): 487-517. BNC: Davies, Mark 2004-. BYU-BNC: The British National Corpus. Bolinger, Dwight. 1972. Degree Words. The Hague: Mouton. Breban, Tine. 2010. E11glish Adjectives of Comparison: Lexical and Grammatlcalized Uses. Berlin: Mouton de Gruyter. Bybee, Joan. 2003. Mechanisms of change in grammaticalization: The role of frequency. In The Handbook of Historical Linguistics, Brian D. Joseph & Richard D. Janda (eds), 602-623. Oxford: Blackwell Bybee, Joan. 2007. Diachronic linguistics. In The Oxford Handbook of Cognitive Li11guistlcs, Dirk Geeraerts & Hubert Cuyckens (eds), 945-987. Oxford: OUP. Croft, William. 2000. Rtplaining Language Change. A11 Evolutlo11ary Approach. Harlow: Longman. Davidse, Kristin. Breban, Tine & Van linden, An. 2008. The development of secondary deictic meanings by adjectives in the English NP. English Language and Linguistics 12(3): 475-503. Declerck, Renaat 1988. Studies on Copular Sentences, Clefts and Pseudo-clefts. Leuven: Leuven University Press. Denison, David. 2006. Category change and gradience in the determiner system. In The Handbook of the History of English, Ans van Kemenade & Bettelou Los (eds), 279-304. Oxford: mackweiL Escribano, Jose Luis Gonzalez. 2006. NPs just as NPs. Language Sciences 28(6): 529-579. Fischer, Olga. 2001. The position of the adjective in (Old) English from an iconic perspective. In The Motivated Sign [Iconicity in Language and Literature I I], Olga Fischer & Max Nanny (eds), 249-276. Amsterdam: John Benjamins. Haeseryn, Walter, Romijn, Kirsten, Geerts, Guido, de Rooij, Jaap & van den Toorn, Maarten C. 1997. Algemene Nederlandse Spraakkunst, 2nd edn. Groningen: Marttnus Nijhoff. Hewson, John. 1991. Determiners as heads. Cognitive Linguistics 2(4): 317-337. Hoffmann, Sebastian. 2004. Are low-frequency complex prepositions grammaticalized? On the limits of corpus data - and the importance of intuition. In Corpus Approaches to Grammatlcalization In English [Studies in Corpus Linguistics 13], Hans Lindquist & Christian Mair (eds), 171-210. Amsterdam: John Benjamins. Hopper, Paul J. 1991. 0 n some principles of grammaticization. In Approaches to Grammatlcallzation, I: Focus on Theoretical and Methodological Issues ['J)'pologtcal Studies in Language 19], Elizabeth C. Traugott & Bernd Heine (eds), 17-35. Amsterdam: John Benjamins. Huddleston, Rodney D. & Pullum, Geoffrey K. 2002. The Cambridge Grammar of the English Language. Cambridge: CUP. Hudson, Richard. 2004. Are determiners heads? Functions ofLanguage 11 (1 ): 7-42. INL38: Kruyt, J.G. & Dutllh, M.F.W. 1997. A 38 million words Dutch text corpus and its users. Lexikos 7: 229-244. Langacker, Ronald. 1991. Foundations of Cognitive Grammar, Vol 2: De.scriptive Application. Stanford CA: Stanford University Press. Lehmann, Christian 1985. Grammaticalization: Synchronic Variation and Diachronic Change. Linguae Stile 20(3): 303-318. Levinson, Stephen C. 1983. Pragmatics. Cambridge: CUP.
Anaphoric adjectives becoming determiners 255 Longobardi, Giuseppe. 2001. The structure ofDPs: Some principles, parameters and problems. In BalUn & Collins, 562-603. Lyons, Christopher. 1986. The syntax of English genitive constructions. Joun1al ofLinguistics 22: 123-143. Lyons, Christopher. 1999. De.finitene.ss. Cambridge: CUP. Lyons, John. 1977. Semantics. Vols. 1-2. Cambridge: CUP. Mackenzie, J. Lachlan. 1997. Grammar, discourse and knowledge: The use of such. In Studies i11 English Language a11d Teaching, Jan Aarts, Inge de Mtinnink & Herman Wekker (eds), 85-105. Amsterdam: Rodop!. Mortelmans Jesse. 2006. Ledit vs the demonstratives in Middle French: contexts of use? Langue Franfalse 152: 70-83. Osawa, Fuyo. 2007. The emergence ofDP from a perspective of ontogeny and phylogeny: Correlation between DP, TP and aspect in Old English and first language acquisition. In NomiJ~al Determinatio11: 'Ijpology, Context Constraints, and Historical Emerge11ce [Studies in Language Companion Series 89], Elisabeth Stark, Elisabeth Leiss & Werner Abraham (eds ), 311-337. Amsterdam: John Benjamins. Perridon, Harry. 1989. Reference, Definiteness and the Noun Phrase in Swedish. PhD dissertation, University of Amsterdam. Plank, Frans. 1992. Possessives and the distinction between determiners and modifiers (with special reference to German). Journal ofLinguistics 28(2): 453-468. Quirk, Randolph, Greenbaum, Sidney, Leech, Geoffrey & Crystal, David. 1985. A Comprehensive Grammar of the English Language, 2nd edn. London: Longman. Quirk, Randolph & Wrenn, Charles Leslie. (1969) [1955]. An Old English Grammar. London: Methuen & Co. Roberts, Ian & Roussou, Anna 2003. Syntactic Change. A Minimalist Approach to Grammaticallzation. Cambridge: CUP. de Rooij,Jaap. 1980. Ons bruin(e) paard. Taal & Tongval32(1-2): 3-25 and 32(3-4): 109-129. Roose, Herman. 1964. Het Probleem van de Woordsoorten in het Bijzonder van het Bijwoord In het Nederlands. The Hague: Mouton. Seppanen, Aimo, Granath, Solveig & Danielson, Lars. 2002. The construction ~djP - a(n) noun' in present-day English syntax. Leuvense Bijdragen 91 (3-4): 97-136. Spamer, James B. 1979. The development of the definite article in English: A case study of syntactic change. Glossa 13: 241-250. Spinillo, Martangela. 2003. 0 n such. English Language and Linguistics 7(2): 195-210. Traugott, Elizabeth C. 1992. Syntax. In The Cambridge History of the English Language, I: The Beginnings to 1066, Richard M. Hogg (ed.), 168-289. Cambridge: CUP. van den Toorn, Maarten C. 1997. Nieuwnederlands (circa 1920-nu). In Geschledenls van de Nederlandse Taal, Maarten C. van den Toorn, W. J. J. Pijnenburg, J. A. van Leuvensteijn & Johannes M. van der Horst (eds ), 479-562. Amsterdam: Amsterdam University Press. Van de Velde, Freek. 2009a. De Nominale Constituent. Structuur en Geschiedenis. Leuven: Leuven University Press. Van de Velde, Freek. 2009b. Do we need the category of postdeterminers in the NP? 'Iransactions of the Philological Society 107(3): 293-321. Van de Velde, Freek. 2010. The emergence of the determiner in the Dutch NP. Linguistics48(2): 263-299. Van Eynde, Frank. 2006. NP-internal agreement and the structure of the noun phrase. Journal of Linguistics 42(1 ): 139-186.
256 Freek Van de Velde
Van Langendonck, Willy. 1994. Determiners as heads? Cognitive Li11gulstlcs 5(3): 243-259. Wood, Johanna L. 2002. Much about such. Studia Linguistica 56(1): 91-115. Wood, Johanna L. 2007. Is there a DP in Old English? In HIStorical Linguistics 2005 [Current Issues in Linguisic Theory 284], Joseph C. Salmons & Shannon Dubenion-Smith, 167-187. Amsterdam: John Benjamins. Yamamoto, Kelko. 1989. The historical development of determiners: A parametric approach. English Linguistics 6: 1-17.
FromNto D* Charting the time course of the internal rise of French n-words Viviane Deprez Rutgers University, L2C2, CNRS
Based on a thorough review of their feature make up and on novel diachronic data on their modification properties, this paper maps out the internal syntactic structure of French n-words and the historical time course oftheir internal rise within a nominal projection. It charts out precise steps that directly relate internal structural changes to corresponding changes in concord properties. Indirect modification with autre as in rien dautre is shown to be an innovation of modem French that serves to provide a distinctive signature of the final morphosyntactic change of n-words into negative quantificational expressions. The evidence offered supports the hypothesis (Deprez 2000; Condoravdi & Kiparsky 2006) that it was change in the internal structure of n -words that determined the current properties of negative concord, not changes in the sentential negation marker, contra Zeijlstra (2004 and following work).
1.
Introduction
It has been proposed that negative concord systems be distinguished on the basis of (1) the internal syntax of their n-words and 2) the semantic interface that these syntactic structures imply (Deprez 1997, 2000; Lohndal & Haegeman (2010). In particular, for languages like Haitian Creole (HC) Deprez (1997, 2000) proposed that n-words occupy a low structural position in a nominal constituent and showed that they manifest syntactic and semantic properties relating to those ofbare nominals. Negative concord, in these cases, involves an operator variable system, with negation binding a bare
*
For discussion of various points in this paper, we wish to thank all the participants in the Grammattcalization Network to which parts of this paper were presented and in particular, Pierre Larrivee, Richard Ingham, France Martineau, Agnes Jager, Johan van der Auwera and Eric Haeberli We also thank the reviewers of this paper for their valuable remarks and the editors for their invaluable help. All remaining errors are our own.
258
Viviane Deprez nominal variable. For languages such as French, in contrast, n-words occupy a high position in their DP, behaving syntactically like determiners with the semantics of quantificational expressions. Negative concord, in these cases, involves resumptive quantification (May 1989; Deprez 2000; De Swart & Sag 2002). The proposal was motivated by the observation that although in both languages, n-words share the appearance of 'bare' nominals, pesonn Knobody" in HC and personne "nobody" in French, the general licensing conditions on bare- N radically differ, bare-N occurring freely as arguments in HC, but not in contemporary French. Ifbare Ns have a null determiner, this difference implies that null-Ds are fully licensed in H C, but not in French, leaving for French the structure (lb) in which the 'bare' n-word occupies a determiner position. 1 (1) a.
b.
[0 p D 0 (NP pesonn]]
[0 ppersonne [NP1]
(Haitian Creole) (Contemporary French)
Following Deprez (2000), we regard these two systems as the two polar extremes of negative relations between which micro-variation can span. Empirical support for this approach to negative concord requires a detailed analysis of the internal structure of n-words and evidence of robust correlations between this inner syntax and distinctive properties of the concord relation. This paper has two goals: first, to provide a detailed internal analysis for standard contemporary European-French n-words, focusing on their feature composition and their modification possibilities and second, to offer an in depth exploration of the changes that have affected this internal structure in history and of the correlation that these changes present with those of the negative relations they participate in. The paper is structured as follows. Section 2 sums up and expands the distributional evidence for the complex D-structure of French n-words. Next, we turn to their feature composition and modification possibilities and we take stock of what these mean for their internal structure. In the second part of the paper, we turn to historical changes in feature composition and modification possibilities, focusing on novel evidence provided by modification with autre, which are particularly revealing for the evolution of French n-words through internal structural changes.
2.
2.1
Looking inside N -words Distributional evidence
As is well known, the lexeme personne still has a double life in contemporary French as a common count noun meaning "person" and as ann-word with a strong negative The idea that French n-words are determiners is also independently proposed in Sleeman's (1996) dissertation. The similarities and differences between this for mal proposal and the classic grammatical tradition that considers n-words as indefinite pronouns (Gougenheim 1951; Grevisse 1980, Nyrop 1930, among others) are discussed in depth in this work.
1.
From N to D 259
interpretation. A careful comparison of the two can thus provide interesting grounds to determine the minimal factors that enter in their distinction. Given that the frequency of use of each of these items is roughly the same, questions arise as to how speakers know which is which. 2 The answer turns out to be relatively simple. Personne has a (positive) nominal interpretation whenever it co-occurs with a determiner in an argument position (2a). Conversely, it is interpreted as ann-word whenever it occurs bare, in an argument position, with no determiner, as in (2b ). (2) a. J'ai vu une!cette!la personne. "I have seen alone/this/the person:'
b. ]'ai vu personne. "I have seen no one:' This complementary distribution between the co-presence of overt determiners and then-word interpretation of personne is quite general, apart from a few interesting cases discussed below, and reveals a competition for the same syntactic position that clearly supports placing then-word personne in a determiner position. This structural hypothesis, in turn, provides a straightforward explanation as to how a seemingly bare nominal personne succeeds in meeting the stringent French requirement for obligatory determination: personne requires no determiner whenever it is, itself, a determiner. Worth stressing, moreover, is the fact that complementary distribution with D is far more reliable a cue to the interpretation of personne than its co-occurrence in a negative context with either a weak negative marker like ne or a strong one like pas. The presence/absence of ne, as is known, changes nothing in then-word interpretation of a bare personne, nor can ne or pas coerce the nominal interpretation of a determined personne to negation. Indeed, the semantic distinction between the negated nominal in (3a) and then-word in (3b), although subtle, is nonetheless patent Consider a scenario in which aliens that look like humans would have invaded our world and would be looking for remaining humans they call 'persons' for final annihilation. Then, if an alien hunter was looking for humans in a crowded alien party and came up empty handed, he could truthfully utter (3a), but surely not (3b). (3) a.
Je n'ai pas vu de personne(s). "I did not see any person:'
b. Je n'ai vu personne "I saw nobody." Only in (3a), but not in (3b), does the lexical meaning remain prevalent. as expected if personne is indeed a nominal predicate in the tormer, but not in the latter. ( 3a) is yet of further interest for the peculiar nominal structure in [de NP] it features. Kayne (1977) 2. Our assertion, here, is based on a rough estimate of occurrences ofpersom1e in the Elicop corpus. In each of the files of this corpus that we considered and tallied, the count of negative personne came to about the same number as the count of the positive uses.
26o Viviane Deprez
argued that this construction sports a remnant null determiner, an analysis supported by the fact that it manifests a characteristic subject/object asymmetry (4), like other null D structures. (4) a.
*]e ne crois pas que [de personnes] seront invitees. "I don't think that persons will be invited."
a
b. ]e ne crois pas qu'il sera invite [de personne(s)] son anniversaire. "I don't think that there will be invited persons to his birthday:' [de NP] constructions also require an overt c-commanding operator, here negation, which is expected if their remnant null D is a variable that must appropriately be bound for interpretation. If this much is correct, it provides strong contrastive evidence that militates against contemporary bare n-words containing a null D, given that their distribution manifests no comparable subject/object asymmetry and no comparable overt operator-binding requirement There is also empirical evidence that they must be in D. These are provided, on the one hand, by the interesting exceptions to the complementary distribution with D. and, on the other hand, by the modification facts discussed in §2.2.3 Turning first to the exceptions, note that (5) features a 'bare' personne that is necessarily interpreted as a positive nominal. (5)
Vous (?? ne) recevrez deux livres pa.r personne. "You (not) will receive two books per person:'
Despite bareness, personne in (5) cannot be coerced into a negative, even with ne. Yet, as (6) shows, PP constructions do not always have this effect: (6) Ce livre (n? a ete envoye par personne. "This book was (not) sent by no one." So why and how do (5) and (6) differ? Repiacingpersonne by other nominals, we quickly note that while (6/7a) displays the properties of a regular PP, taking all sorts ofnominals as complements, this is not true of ( 5/7b). In (7b ), the range of allowable complements is very narrow, in effect, strictly limited to nominals with no determiners: (7) a.
Ce livre a ete em'oye par les!(tous) mes!mon!un ami(s). KThis book was sent by the (all) my/a friend(s):'
b. Vous recevrez deux li1'res par *le!*mon!*un enfant. "You will receive two books per the/my/a child." As (7b) reveals, the presence of overt determiners is simply excluded in this construction. Yet, as (8) shows, adjectival modification is allowed, ruling out an analysis of (5) as compounding, or N to P incorporation.
3·
See also Sleeman ( 1996) for further independent evidence.
From N to D 261 (8)
Vous distribuerez un cadeau par gros client fortune. "You will distribute one present per big wealthy customer."
Constructions like (5) appear to select bare singular NPs, banning any nominal with a complex determiner structure.4 This recalls Pereltsvaig's (2006) observation about some Russian prepositions that strictly select for a 'small nominaf with no DP projection. Returning to n-words, we can now take stock of what this data entails about their structure and interpretation. If the n-word interpretation of personne were compatible with a bare NP structure, its exclusion from (5) would be mysterious. If it requires a complex D-structure, as defended here, the facts are entirely as predicted; a complex D-structure fails to meet the selection restrictions of this peculiar PP construction. The tacts in (5) exclude analyzing bare personne as an N-level pro-noun (Nishigauchi 1990), making the clear point that not all apparent 'bareness' can be treated equal, and providing limpid empirical evidence that the bare nominal n-words of contemporary French must beD-like structures.5 Let us now briefly see how these observations about personne extend to other nominal French n-words and, in particular, to rien. Although rien no longer has a comparable common nominal counterpart in contemporary French, it is nevertheless not hard to find evidence that the above structural conclusions extend to it. First, the complementary distribution with determiners is still observable to some extent tor rien in examples like (9), and, as expected, the presence of a determiner is required to allow it:
4 That the complement of such prepositions must be singular is demonstrated by (i) below, where we see that a phonetically distinct form of the plural is unacceptable: (i)
Vous mettrez un ballot db:voine par cheval!*chevaux "You will put one straw pUe per horse/'"horses."
5·
A reviewer objected that the exclusion of then-word In (5) could have a semantic basis,
barring negative terms for some (pre-theoretic intuitive) semantic reason. Note, however, that it is not just negative quantifiers that are excluded in (5) but all positive ones as well Thus as (i)
shows, chacun "each" is equally excluded, even though the sentence is pretty much synonymous to (ii) and thus not semantically odd in any obvious ways. Moreover (iii) may be semantically odd, though it is far from impossible, and yet there is no doubt that personne, there, is the n-word and not the nominal. Such examples show quite dearly that a purely semantic explanation would simply not suffice. (i) '"Vous leur donnerez un livre par chacun.
you them will-give a book per each (ii) Vous leur donnerez un livre (a) chacun.
"You will give a book (to) each of them."
a
(iii) Vous (ne) donnerez un livre personne.
"You will (not) give a book to nobody."
262
Viviane Deprez (9) Il est tres sensible. Le plus petit rien/un rien le fait sursauter. 6 "He is very sensitive. The merest thing makes him jump:'
Moreover, although ( 10) unlike ( 5) is strictly speaking not fully acceptable, it nonetheless significantly improves with a pre-nominal modifier. (10)
Vous utiliserez un sac par *rienl?petit rien "You will use one bag per nothing/little thing."
( 10) illustrates the fact that the nominal equivalent of rien can still be coerced if the nominal structure is made obvious by a pre-nominal adjective that forces rien to be construed in a low nominal position. Remarkably, rien loses its negative import there. Hence, the reasoning applied to (5) above extends to (10), suggesting that rien, just like personne, sports a complex determiner structure that constrains its interpretation. The tests used above clearly do not apply to the French nominal n-word aucun. For a thorough discussion of the structure of aucun see Deprez & Martineau (2004).
2.2
The features and modification of n -words
The evidence provided in Section 2.1 tor a complex D structure tor French n-words relied on distributional factors. In this section, we offer a feature comparison between n -words and their nominal counterparts and compare the modification possibilities that each allows. Turning to features, (lla) shows that the nominal personne is feminine and varies in number, as reflected by determiner or adjectival agreement and by co-referring pronouns. In (llb) in contrast, then-word triggers no adjectival agreement and has a corresponding singular and masculine bound pronoun. (11) a.
Les personnes intelligentes pensent quelles nbnt pas toujours raison. "Intelligent persons think they.FEM are not always right"
b. Personne d'intelligent (ne) pense qu'il a toujours raison. "No one intelligent thinks he is always right" In these respects, n-words behave like the existential quantifier quelqu'un, also unmarked for gender and number. Note, however, that despite number invariability,
6. M noted by a reviewer, rien with a positive meaning is strange with a singular definite determiner, if no scalar adjective precedes it. Compare: ??le rlen "the thing" vs. le petit rien "the small thing': We are not convinced, however that this strangeness is significant What seems to be happening here is that with a definite determiner, le rlen is easily and preferably Interpreted as a synonym to le neant "the nothingness~ This seems to only happen with a definite determiner, presumably because only this determiner allows for the kind generic interpretation required for this type of reading. With a demonstrative, which cannot have a generic interpretation, ce rien "this thing" is perfectly acceptable with a positive meaning.
From N to D 263
n-words seem semantically plural as shown in (12) where personne is occurring with plural subject requiring verbs? (12)
Personne ne se rassemble plus ici. "No one gathers here anymore:'
How do these facts support the view that n-words occupy a high D-like position in their functional structure? First, the lack of gender and number features makes nwords characteristically un-nominal. Moreover, syntactic invariability and semantic plurality are characteristic of a certain type of D-elements in French such as mass terms, the impersonal pronoun on "one" and the existential quantifier qu.elqu 'u.n "someone" or the bare quantificational pronoun question words qui "who" and quoi "what': These similarities make the case that n-words share features with invariable pronouns and quantifiers, but not with nouns, hence supporting the view that they are merged in aD position. Further support arises from their modification properties. Observe first in (13) that pre-nominal adjectives are excluded with n-words, while they are possible and stackable with their nominal counterpart (13) ]'ai vu *autre charmante petite personne/une autre charmante petite personne. "I have seen *other charming small nobody/another small charming small person."
In the spirit of Cinque's cartographic approach, the lack of pre-nominal modification is a further indication that these n-words occupy a position higher than NP in the DP structure. But it is not just pre-nominal modification that is excluded. More generally n-words like personne and rien do not allow direct adjectival modification; they require a specific modification construction with de as in (14). As (14) shows, this is again a property n -words share with existential quantifiers like quelqu'un or bare question words like qui or quoi. 8 7· Likewise, as a reviewer points out, reference to the n-word personne in non-commanding relations involves preferably plural pronouns. (i)
Personne ne pense fa et ils ont bien ralson/*et il a raison. "No one thinks that and they are right/*he is right."
Notably, this is also true of quantifiers like toutle monde and of some collective nouns: (ii) Ce groupe d~dultes, toutle monde, pense fa et lis ont bien raison/?? et II a bien raison.
"This group of adults, everyone, believes that and they are rlghtmand he is right." This illustrates the known but curious fact that number and gender of co-referring unbound (e-type) pronouns can be governed by semantic rather than syntactic features. Clearly, however, what the relations between morphological and semantic features are is far too complex an issue
to be addressed here. Thus, in the text, loss of number and gender features is understood to refer strictly to formal syntactic features. 8. Indirect modification of this type is not possible with universal quantifiers, a fact that strongly suggests that n-words are like existential/indefinite quantifiers or numerals in their
264
Viviane Deprez (14) personne/rien/quelqu'un/qui/quoi *(d)' interessant no one/nothing/someone/who/what *(of) interesting
Kayne (1994) observes that indirect de modification is more constrained than regular modification, in that it disallows stacking as in (15): (15) *]'ai rencontre personnelquelqu'un/qui de charmant de petit. I have met no one/ someone/who/ of charming of small
He argues that this anti- stacking constraint goes against a traditional adjunction structure for these modification constructions and proposes, following Huot (1981), a structure akin to relative clause modification, with de as a complementizer-like element. as in (16):9 (16) D 0
(DP/PP
quelqu.'un/pe,.sonne (de (IP intelligent I0 t]]] someone/no one
of
intelligent
In (16), then-word or quantifier raises from a position internal to the clause to Spec
de, in a fashion paralleling that of predicate inversion in DP (Den Dikken 1998). Kayne following Huot (1981) notes one intriguing exception to the stacking constraint that occurs with the modifier a.ut,.e "other": (17) personnelrienlquelqu'un/qui/quoi d'autre d'interessant "no one/nothing/someone/who/what else (of) interesting"
D'autre modification is also possible with quelqu'un or a question word, but crucially not with the nominal personne, for which modification by autre must be strictly prenominal and direct: 10 (18) a.
une autre personne "another person"
b. *une personne d'autre a person of other The modifier autre is peculiar in other respects that are of particular interest here. First, like a few other pre-nominal adjectives in French (di1'ers, differents "several"), a.ut,.e is sometimes able to play the role of a determiner and/or a pronoun. (19) a. Auh-es temps, a.ut,.es moeu,.s, autre histoire! "Other times, other stories!"
syntactic nature, i.e. quantificattonal indefinites (Deprez 2000) and not like universal quantifiers, contra Giannakidou (2000). 9·
For distinct proposals on the structure of this indirect modification see Azoulay-Vicente
(1985) and Hulk & Verheugd (1994), among others.
to. Post-nominal autre is also margl.nally possible but with a meaning closer to that of different "different": une personne autre.
From N to D 2.65 b. A dautres! "To others!"
Clearly as well, it must always be the highest pre-nominal modifier in any noun phrase, as it can never be preceded by any other pre-nominal adjective: (20) a.
*une petite autre voiture a
small other car
b. une autre petite voiture an other small car In these characteristics, autre differs from regular adjectives, appearing instead to share properties with numerals. 11 In fact, like numerals, autre is able to directly modify strong pronouns as in nous autres (us other) comparable to nous deux (us two). Furthermore, while indirect de-modification with regular adjectives is sometimes possible with noun phrases as in (21), indirect modification with dautre is always excluded: (21) Il y avait plusieurs voitu.res de bleues!*dautres. there were several cars of blue/*of otherlZ In sum, indirect modification with d'autre is only possible with bare n-words, bare existential quantifiers and question words, never with other nominal expressions. This clearly gives it a peculiar status, shared by no other adjectival modifiers. Huot (1981) suggests that autre involves quantity not noun modification, grouping in effect modification by d'autre with comparative degree modifiers like de plus "more", or de moins "less" (quelqu'un/quoi de plus/de moim; "someone/what more/less"). She proposes the structure in (22): (22)
N"
N'
~
spec!
Quantite quelqu'un
~
de
autre [ ]
u. Confirming this evolution. it is notable that in some French based Creoles (particularly in Guadeloupe Creole) zot "other" is used as a plural marker. 11. Indirect 'de modification' is also possible with so-called quantitative en- pronominalization AB (i) shows, for regular adjectives both direct and de- modification are possible. Here again. however, indirect modification with d~utre is excluded: (i) /~11 ai TROIS (de) petites I* d~utres, pas quatre I of-them have three of small /*of other, not four
266
Viviane Deprez In a current model, a possible re-interpretation of her insight would assume that diwtre involves NumP and not NP modificationP We have seen above Kayne's (1994) proposal for indirect de-modification with regular adjectives. Transposing Huofs suggestion, we propose, to account for the stacking effect, that autre is a NumP modifier that can move with it out of the clausal constituent of the indirect adjectival de-modification and be stranded in Spec DP, while then-word moves yet higher in the DP structure as in (23a), in a fashion akin to CP recursion. (23) a.
[pppersonnei [DPIPP de [ti autre \]k [de
no one b.
[ep[Qu~
who c.
[1t
of
else
of
hPsympathique Infl ~]]]] sympathetic
dautre]] [c as-tu [IP inviM de sympathique]]] of else
[epQuik as-tu
have you invited of sympathetic
invite [DPIPP [1t dautre] [de sympathique \]]]
who have you invited
of else
of sympathetic
d. /e nai rienk VU [DPIPP [tt dautre] [d'interessant ~]] I NEG have nothing seen of other of interesting That dautre can indeed move along with a quantity expression (i.e. be pied piped) is confirmed by (23b ), in which di:I.Utre has moved along with a wh- term to Spec CP. That movement out of the projection containing the modifier is also possible is confirmed by (23c), in which dautre is optionally stranded by the moved wh-term. (23d) replicates this structure with the moved n-word rien. 14 (23a) accounts, on the one hand, for the exceptional stacking possibility with dautre and on the other hand, for the restricted nature of the elements that allow it. Only elements moving out of NumP to a higher position in DP, here termed FP, manifest this DP recursion possibility. To sum up, we have shown that the modification properties of n-words are very different from those of regular nouns. The facts reviewed above provide evidence for the quantificational nature of n-words and for the proposal that they occupy a high position within DP. We have spent some time detailing the peculiar facts of da.utre modification here because, as discussed below, there are, in this respect, revealing changes in the internal structure of n-words that are important in delimiting steps in their historical evolution.
13. Throughout, we assume NumP to be the projection that houses numerals and weak quantifiers, not grammatical number. Number could be a separate PlP (PlW'al Phrase) projection as in Heycock & Zamparelli (2005). 14. CW'iously and interestingly, stranding, in this case is obligatory. We do not at this point
have an explanation for this fact.
From N to D 267
2.3
Mapping the structure
The above discussion provided much empirical evidence of the complex internal structure of n-words, but remained imprecise as to the detailed mapping of their internal structure. In this section, adopting Zamparelli's (1995) proposal to subdivide the DP into the three quantificational zones SDP, PDP, and KIP, we flesh out this internal structure. 15 Only weak quantifiers housed in PDP license ne cliticization, because only they are close enough to license a trace in NP. The Italian quantifier qualcuno "someone" is ambiguous in this respect. When compatible with ne cliticization, qualcuno is an equivalent of"some". When incompatible with it, qualcuno is restricted to human reference and is a pronominal meaning "someone". To account for this ambiguity, Zamparelli proposes that qualcuno is merged as a weak quantifier in PDP/NumP that can move or re-merge higher as a strong quantifier in SDP/DP. From this higher position, a Q cannot license the trace of partitive ne. Transposing this approach to French, we note that the ambiguity observed for qua leu no is replicated with quelqu'un "someone" vs quelques-uns "some': Like the invariable qualcuno, invariable quelqu'un means "someone" and fails to allow en cliticization and combination with a true partitive (24a-b). The variable quelques-uns, in contrast, means "some" and is compatible with en cliticization and with a partitive (24c-d). (24) a.
"1en
a.i
vu
quelqu'un.
I of-themhave seen someone b. *]' ai vu quelqu'un de mes amis. I have seen someone of my friends
c. fen ai vu quelques-uns. I of-them have seen some ones "I have seen some."
d. fai vu quelques-uns de mes amis. I have seen some ones of my friends "I have seen some of my friends." Let me hence suggest that, in similarity with qualcuno + ne, variable quelques-uns is merged in PDP/NumP, while invariable quelqu'un, in contrast, is directly merged to SDP/DP. Since, as seen above, then-words per.sonneand rien behave like the invariable existential quantifier quelqu'un, we have analogical grounds to posit the same structure tor these contemporary n-words. This motivates the structure in (25) tor the French n-words:
15. Respectively: SDP = strong determiner phrase, PDP predicative determiner phrase, KIP Kind Phrase.
268
Viviane Deprez (25)
SDP/DP
quelqu'un personne/rien quelques-uns trois "three"
KIP/NP
(25) proposes that in contemporary standard French, n-words are merged directly in SDP/DP. Note that as expected in this proposal n-words are incompatible with true partitivity and with quantitative en cliticization (26): (26) a.
H]e ne I
NEG
connais personne de tes know
no one
amis
of your friends
b. *]e ntm connais personne I NEG of-them know no one (26) shows that n-words cannot license a null N/NP category; they are merged too high in the DP structure. Strictly speaking then, (lb) above should be amended so as to not contain a null NP. If so, n-words are intransitive Ds that license no null NP complements. (25) accounts for all then-word properties detailed above, as it gives them the highest position in DP and makes the claim that they behave essentially as strong existential quantifiers. Expected on this view is their invariability, given that they are merged in SDP, above the DP layer, where number agreement is syntactically negotiated. Also expected are their modification possibilities, given that they behave essentially as quantity expressions. For indirect 'de modificatio~ we have adopted Kayne's (1994) structure to which we have added a layer to allow modification by di:mtre as in (23a) above. On this view, d'autre modification, and the lack of partitive structure, can be taken as evidence of an additional movement step towards the highestlayerofSDP/DP, possibly recursive, in similarity with CP recursion. In contrast, the noun personne is merged in N P and acquires number features like other French NPs, possibly through movement to PIP ( Heycock & Zamparelli 2005 ). 16
16. A reviewer asks whether we assume that there are one or two lexical items personne. Such a question only arises if the lexicon is conceived as a 1ist' of preset words, a view we do not adopt A lexical item in our view is a set of semantic, formal and phonological features, sometimes as-
sodated with a concept, that is built compositionally (Marant7, to appear). Assume, as is current nowadays, that number features are an independent head that contributes to making a common noun by composing with a concept to derive an atomized representation. Then our assumptions that n-words are directly merged above the number node has the consequence that then-word persom1e does not have the atomicity linked to common nouns, while the noun persom1e does. Evidence for this is provided by the fact that they appear semantically plural (or more exactly mass-like). This is crucial to their respective (im)possible association with determiners. In short, the set offeatures that are eventually spelled out as personne = n-word, does not correspond to a predicate [Ax [person x]] associated with number, but minimally to a set of formal features 3rd person, human, quantificational.
From N to D 269 For completeness, our study of the structure of contemporary French nominal nexpressions requires that aucun N "no N" be considered. As this expression and the historical evolution of its internal syntax are discussed in great detail in Deprez & Martineau (2004), we remain brief here. In contrast with personne and rien, aucun N clearly contains both a determiner element and an obligatory N, so that the status of a.ucun as a determiner is not in doubt. Unlike pet-sonne and rien, however, this n-word expression allows both quantitative en cliticization, and true partitive structures: (27) Je (n) ai vu aucun de tes enfants. Je nlm ai vu aucun. "I have not seen any of your children. 1have seen none of them."
In this respect, aucun seems to behave both like a numeral in French and like the strong quantifier chacun, which allows partitive structures and varies in gender, but notably, not in number. (28) Aucune idee na/*Aucunes idees nbnt ete acceptee(s). 17 "No idea was/*no ideas were accepted:'
To account for these differences as well as for the similarities between aucun and the other n-words, we propose that the determiner part aucun be merged in PDP and moved to SDP as in (29). In contrast to bare n-words, aucun is a transitive D that can license a null N, as expected if it is merged in PDP. (29)
SDP/DP
'l~· ~un
KIP
Thus, like numerals, aucun participates in a strong/weak alternation, but unlike numerals, it does not allow a preceding determiner in contemporary French (les trois, vs *les aucuns).This indicates thatre-merge to the higher layer of the DP is obligatory for this n-expression like with rien!personne. This was not always true in the history of French. Expressions like les aucuns am is "the some friends" were commonly found up to the 16th century (Martineau & Deprez 2004). This, and the fact that aucun is now invariable in number, motivates moving this n-determiner to the highest D position.
17. One of the reviewers objected that plural aucun is still acceptable for him/her. The prediction of our approach is that such a speaker should also accept aucun in NPI non-negative contexts, which would indicate that for him/her aucun is still able to stay in PDP. As discussed In Martineau & Deprez (2004), the decline of nwnber variability historically correlates with an NPI -like interpretation. For speakers like ourselves, and for our consultants, (28) is judged infelicitous with a liaison (*[okynZide] vs. [okyNide]) and the plural auxiliary, which shows that plurality is clearly disfavored. Expressions like *aucunZanimaux "no animals" vs aucunNanimal "no animal" are unacceptable.
270
Viviane Deprez To sum up. on the basis of their distributional, feature, and modification properties, we have argued that all nominal French n-words occupy the highest quantificational position in the DP and as such behave syntactically like determiners and strong quantifiers. Of particular interest moreover is the fact that this high position correlates with the loss of number variation. All the nominal French n-words are invariable in number, and consequently always trigger singular or more exactly unmarked agreement. It is these combined properties that account, in our view, for their possibility to ambiguously partake in the complex polyadic resumptive quantification that produces negative concord readings, as well as in the scopal quantification that produces double negative readings (Deprez 1999, 2000; Deprez & Martineau 2004; May 1989). As will be seen in Section 3, loss of feature variability historically starts the loss of nominal status and the development of a pronominal and quantificational status.
The internal evolution of French n-words
3·
Having discussed the syntactic nature and structure of contemporary French n-words, it is now time to turn to their historical evolution to chart the course of their internal change as well as to identify the micro-parametric nature of this evolution. We offer here a detailed examination of the evolution of rien and personne focusing on changes in their feature make up and changes in their modification capacities, centering on autre for reasons of space. Our study is based on data mined in the electronic databases Frantext and TFA. 3.1
Changing features
Let us first consider feature changes. Both personne and rien started out as feminine and pluralizable nouns, and both ended up on their n-word interpretation as singular and masculine, or more exactly, as expressions unmarked for number and gender. Charting the course of changes of this nature in historical corpuses, however, is not easy largely because. for French, the best cues to determine the gender and number make up of nominal expressions involve only indirect inspection of their dependent satellites. In Old and Middle French rien, which comes from the Latin noun res "thin(, was a feminine noun that could refer to inanimate or animate entities. (30) is an example from Chretien de Troyes, 12th century, of animate nominal rien preceded by a feminine definite determiner. (30) Quant la rien voit que il plus ainme. When the thing/the person sees that he most loves Chretien, de Troyes, [1240], Chevalier de la Charrette
Examples with a clearly feminine determiner as la rien could be found up to the middle of the 16th century in the Cent Nouvelles Nouvelles, as shown in (31):
From N to D 271
(31) elle voit d'elle tloigner la rien en ce monde she sees ofher leaving the thing in this world
dont
Ia presence plus luy pla.it
of-which the presence most her pleases de La Roche [1550), Les Cent Nouvelles Nouvelles This appears to be the last example of this sort in Frantext. which suggests that rien became unmarked for gender around the middle of the 16th century. Confirming this, it is around this time that the first examples of rien with a clear masculine determiner are found, as in (32), an example from the poet Ronsard from 1552: (32) ... elle auroit cognoissance Qu'un rien quon ne voit pas, fait souvent un grand "she would have knowledge that a thing that one cannot see often makes a large ..." Ronsard, Pierre de [1552], Les Amours (Garnier, Paris) That Jien becomes invariable for gender does not mean, however, that it completely loses its nominal character. Indeed, rien was still found with a (masculine) determiner and a positive meaning up to almost the end of the classic period, as attested by (33). Interestingly, however, rien is then strictly inanimate in reference: (33) pour moi, le rien qui m'arrive d'heureux, cest toujours... for me the thing that me happens of happy is always Goncourt. Edmond et Jules [ 1896] . Thus, the loss of gender and animacy reference represents only a first step in the grammaticalization of this expression. 18 Turning to number, we face the same difficulty in determining with precision when exactly rien stopped allowing number inflection, as it can still vary even nowadays when it is a noun clearly accompanied by a determiner (cf. des riens). To get some idea nonetheless, we searched our databases for examples of rien overtly marked for number but with no determiner. The last example where riens occurs as a clear bare argument occurs in 1559 in Frantext (34) LeRoy ne luy respondit riens, sinon: m'asseurez-vous que vous li:wez "The king did not answer things, but: comfort me that you have it..." Marguerite de Navarre [1559], EHeptameron
Based on this indirect clue, we tentatively suggest that it is around the end of the 16th century that bare rien effectively ceases to be variable in number. But again, rien does not immediately cease to have a nominal character, since it can be found as an invariable bare nominal expression without a determiner with a positive meaning in forclusive contexts up to the classic period: 18. & pointed out by a reviewer, the picture of gender loss for rien is further complicated by the fact that rien could be used as a masculine noun synonymous with le 11eant "the nothingness" (Gougenheim 1951: 84) beginning in the 16th century.
272 Viviane Deprez
( 35) Diable m'emporte si j'entend rien en m~decine (Med. III.l)
"The devil take me if I understand anything of medicine:' Moliere [1673, 1674], Le ma.la.de imaginaire19 19. A revlewer objected that s!he could utter this sentence in hls/her present day idiolect with-
out a 'trace of archalsni This, in contrast, feels strongly archaic in our own speech, a feeling fully replicated with 10 non-linguist sclentific friends and relatives we poled informally. The preferred form here in our idiolect ls with quoique ce solt "whatever it is': The claim that n-words can still be currently used as positive items in negative polarity contexts ls common in the French conservative/normative grammatical literature on n-words, and it is quite dear that authors and speakers differ quite widely in their (social) acceptance of such examples. Acceptance can also vary depending on the type of polarity context considered For instance, even in our own idiolect, a non- negative reading ofpersonne and rien is fully available in exceptive contexts with sans as in (i):
II est dlfficile de vtvre sans personne pour vous aider. It ls difficult to leave without anyone to help you. Notably, however, these contexts are peculiar in other respects. First, they continue to allow nominal expressions without determiners as in (ii), in contrast to standard verbal argument contexts, where such bare arguments are excluded. (ii) n est dlfficile de vtvre sans argent et encore plus sans amis. It is difficult to live without money and even more, without friends. Second, these contexts also allow, in an elevated style, a post-nominal aucun (cf sans ralso11 aucune "without any reasons") a construction that was common in the 16th/17th centurles (see Martineau & Deprez 2004 for discussion) but which is now excluded in more standard contexts il nb; aucune raison/*ralso11 aucu11e "He has no reason" nowadays. Given these possibilities, our approach correctly predicts that in such contexts the polarity readings of n-words are possible (since remnant null D is licensed) just as they were in argument positions in classic French. The exact reason why this should be so remains, however, to be elucidated. Note that this clearly reveals that the landscape of polarity licensing is complex, more so than generally thought, and probably not uniform, contrary to what many semantic approaches to the phenomenon often presuppose. So there might well be divergence in acceptability in distinct polarity contexts for distinct speakers and for distinct reasons. Here again a detailed analysis would be required but this is beyond the scope of the present paper. In any event, what is surely the case, here as well as in concord phenomena, is that the flexibility of a micro-parametric approach is necessary to allow an empirically correct account of this diversity. It is largely on the basis of this widespread diversity, which is influenced by differing normative attitudes towards spoken vs. written language, as well as by the type of polarity contexts considered, that we advocate a micro-parametric approach to the analysis of negative dependencles, concord and polarity relations included that allows for variation within languages, within dialects and idiolects and not just for cross-linguistic variation As defended in our recent writings (see in particular Deprez. to appear, for an extensive discussion), the micro-parametric approach based on the internal structure of n-words we defend contrasts in this respect with many popular current formal Minimalist approaches to negative concord (Zeijlstra 2004, among others) that are strongly macro-parametric in spirit, and hence presuppose a language-internal uniformity across n-expressions that is rarely observed and often problematic. (i)
From N to D 273 Evidence confirming the enduring nominal quality of rien is based on modification possibilities, to which we turn in Section 3.2. Before that, however, let us first detail the time course of feature change for then-word personne. Turning to personne the exact time course of the loss of its features is even more difficult to compile from corpus searches given the continuous current use of its nominal counterpart. We have nonetheless endeavored to chart an approximate course by looking at examples of bare personne triggering teminine agreement with an adjective, a past participle or a coreferring pronoun. In Old French, bare personne was clearly feminine as witnessed by (36) from the Vie de St Alexis (11th) with a co-referring feminine pronoun. (36) Car la sainte Ecriture si nous dit et tesmoigne, Quau monde na personne, pour tant qu~lle se joigne ...(in Sarre 2003) "Because the Saint Writing here tells us and witnesses, That in the world there isn't (a) person for as much as she takes ..." The capacity to trigger feminine agreement for bare personne seems to have lasted, with some amount of flux, up to the 17th century, as witnessed by (37) with a past participle agreement from l'Astree (1607-1625). (37) Je ne puis dire avec veritt que jamais personne ne fut plys aymee que moi "I cannot tell with truthfulness that ever person was more loved than me:' (H. Durfe, Astree, p. 536, (cited in Sarre 2003) Yet. as early as the mid 14th century, some examples lack agreement: (38) ]amais noU1'elle ntm seroit a personne 1'i1'ant never news NEG of-it would be to person alive.MASC
(Cent Nouvelles Nouvelles, 387,73) More generally, it seems that after the Renaissance, bare personne had largely lost its capacity to be marked for gender and, thus, came to trigger default masculine agreement and co-reference, as shown in (39) from Pascal (1670), with pronominal reference: ( 39) Personne na dassurance - hors Ia foi, s'il veille ou s'il dort anyone does not have assurance, besides firlth, of whether he wakes or sleeps (Pensees, 164, cited in Sarre 2003) In sum, our corpus searches reveal that both rien and personne seem to have lost their characteristic nominal features of gender first. and number second, respectively about the middle or by the end of the 16th century, although, according to Grevisse (1980: 650), some flux lasted up to the 17th century.
3.2
Modification
This section focuses on historical changes in modification possibilities. We argue that the striking changes in modification possibilities, to our knowledge discussed here for
274 Viviane Deprez
the first time, provide a revealing window into the evolving internal structure of nwords. Going back to its earlier occurrence, rien was modifiable by a number of prenominal adjectival quantifiers like nu.lle "no': tou.te "every': which all attest of its feminine gender. In Frantext, the last example of nulle riens is found in the 15th century in Christine de Pisan's work:
une riens (40) Ne deU.Stfaire nulle riens, Toutefois entreulx Not need to do no things, yet between them a thing Christine, de Pisan, ca 1364-ca. 1431. [1402], Le Livre du chemin de lone estude. There are furthermore 94 occurrences of Toute rien in the TFA database. Our searches in this database revealed that this feminine modification was possible up to 1381. In Frantext, the last example of this modification is found in 1550, again in the text of the Cent Nouvelles Nouvelles. (41) este bien court tenu; et sur toute rien luy estoit et fut was very narrowly held and on every thing to him is and was
defendu
le mestier
forbidden the protession De La Roche [1550], Les Cent Nouvelles Nouvelles It is not easy to look for regular modifiers in a corpus, there are of course too many possibilities. This is why we chose to focus our searches on modification by autre, a strategy that allowed us to compare through different times and turned out to be particularly revealing in its evolution. ln some early examples, we see plainly that autre began as a prenominal modifier of rien as in (42). This is clearly comparable to today's modification of regular nouns. (42) Nous ne demandons autre riens Que no us y mettre. "We are not asking for other things than to put ourselves to it" Anonymous (p. 164) [c. 1376], Miracle de sainte Bautheuch The last example of this kind, given here in (43), occurs in 1606. Rien in (43) is clearly masculine, plausibly, invariable in gender, but still positive and here, preceded by a determiner. (43) Dieu qui de rien fit tout, et qui de tout encore peut faire un autre rien, "God who out of nothing made everything and who out of everything can yet make another thing." Bertaut J. [1606], Recueil Qq. Vers Amoureux Taking the pre-nominal position of autre to be indicative of the nature of rien, examples like (43-44) suggest that its nominal character, as indicated by its positioning in the N projection, was maintained somewhat beyond the loss of its gender and number features, up to the beginning of the 17th century. Thus, the interesting fact that no example of this sort occurred past this date can be taken to mark the end of a historical phase.
From N to D 275 From 1606 on, throughout the classic period and up to the early 20th century, our data indicate that autre has resolutely become a post-nominal modifier of rien as in (44).
me donner rien autre chose (44) elle me dit a.dieu, et se retira sans she said goodbye and left without giving me anything other qu'une hague... than a ring (Lesage A.- R. [17 32], Histoire de Gil Bias de Sant. -173 2 (In Romanciers dulSE s.,Ed.) Despite the change of position, modification by autre remains direct at this time, never making use of a linking de. This modification remains seemingly close to regular post-nominal adjectival structures. However, recall that as seen above, modification by autre of regular nouns is strictly pre-nominal up to contemporary French. Thus, in effect, we must conclude that post-nominal direct modification by autre is not equivalent to a regular case of post-nominal adjectival modification. Instead, the contrastive position of autre with rien, as opposed to regular nominals, must be taken to indicate that it is rien that no longer behaves like a regular noun in the language. Taking the position of a.uh-e to mark the upper limit of NP internal modification, as suggested by its high contemporary position in a modification stack, examples like (44) imply that by the 17th century, rien had definitely left the nominal domain for the NumP domain. That is, after a period of transition where autre could occur alternatively as pre- or post-nominal modifier, i.e. where the movement of rien from the nominal domain to the NumP domain was optional, rien came to be directly merged in NumP, accounting for the obligatory post-nominal position of autre. In this, n-words were similar to current numerals, which also display direct post-nominal modification with autre as in trois autres "three other ones". This form of direct post-nominal modification, in turn, lasts through the classic period and is found rather late, up to 1925, in the writings of Claude!. Of interest is the tact that in Claudefs writings as well as throughout the classic period, characteristic NPI uses of n-words in forclusive contexts are commonly found. As shown by the following table reproduced from Deprez & Martineau (2004), positive uses and polarity context uses (excluding negation) of n-words amounted to about 30% during the 17th and 20% during the 18th century. Against the backdrop of this previous study that made use of the same data source, we can see quite distinctly that direct post-nominal modification with autre appears to be a characteristic feature of the classic period (17th-18th) that coincides with the
Table 1. frequency of aucun in positive, polarity and negative contexts periods 17th century 19th century
positive contexts
polarity contexts
negative contexts
3.5% (7/200)
27.5%
69%
20.5%
79%
0%
276 Viviane Deprez
rising use of n-words as NPI (Deprez & Martineau 2004)20• In this regard, it is offurther interest to note that this direct modification pattern with autre is also sometimes found in contemporary Quebec French, a construction noted as an apparent 'Canadianism' in Clas (1976). (45) shows one such example, found in an online corpus: (45) ... la censure nest rien autre chose que lexamen ... censure is nothing else than the examination... §bilan.usherbrooke.ca/bilan/pages/collaborations/8580.html As evidenced in Martineau & Deprez (2004), Quebec French also contrasts with contemporary continental French in commonly allowing the co-occurrence of the negation pas with n-words as well as common N PI uses of these expressions. In short, there is both diachronic and comparative evidence of the correlated occurrence of direct post-nominal modification by autre and of the use of n-words as polarity items.21 The next pattern of autre modification observed in our diachronic corpus is that of indirect modification with de. Notably, this indirect modification has a peculiar historical evolution. In Frantext, it is first found in the 12th century in Chretien de Troyes, but very sporadically. After this, rather curiously, this pattern disappears entirely from the corpus until1843, where we see it re-surface after a seven-century gap, in the writings of Honore de Balzac. ( 46) Ld, madame; ne 1'0zts jaut-il rien dautre? "So madam, don't you need anything else?" Balzac [1843], La Rabouilleuse
The use of this indirect pattern with autre remained fairly infrequent until about 1890, when it finally reached about equal status with the previously discussed post-nominal direct modification rien autre and rien autre chose. Yet it is not until 1909 with the writings of Andre Gide that the indirect de modification pattern comes to firmly dominate and ultimately win over the direct modification pattern. Worth stressing here is the fact that this is an interesting discovery, as our research serves to establish, we believe for the first time. that the use of the modern indirect modification pattern with d' autre as in rien d'a.uh-e is not firmly established until essentially the beginning of the 20th century. Referring back again to the findings of Deprez & Martineau (2004), this appears to quite strikingly coindde with the rise of the strongly negative interpretation ofn-words. For further empirical evidence, let us now consider the modification possibilities with bare personne, concentrating again for the sake of brevity on the modification of bare person ne with autre. Here, of course, precaution must be used to avoid confusion 20. For some speakers, and in some contexts, NPI readings have not completely disappeared from the modern language. For some discussion, see fn. 19. 21. For a more detailed discussion of the NPI uses of n-words both in classic French and in Quebec French, see Deprez & Martineau (2004) and for modern French see fu. 19.
From N to D 277
in considering only examples in which personne occurs without a co-occurring determiner, lest we include in our observations irrelevant contemporary uses of the nominal form. As for rien, we observe that the first occurrences of the modifier a.utn:: with a bare personne are pre-nominal as in (47). In Frantext, the last example ofbare personne modified by a pre-nominal autre is found in the 17th century.
autre personne qu'un sonneur de cloches. (47) on n~ntend one did not hear an other person than a bell ringer Deimier P. de [1610], EAcademie de EArt Poetique (Paris,]. de Bordeaulx. 1610) After this date, the prenominal modifier autre is only found in the Det + personne combination, which is the nominal construction still currently in use. After this date, as well, with a bare personne, the modifier autre occurs strictly in post-nominal position, as in the following characteristic example: (48) pour le sens commun de croire quelle nest aperfue par for the common sense of believing that she is not seen by
personne autre any other person Rousseau, ].-]. [ 1776], Rousseau luge de Jean-Jacques With bare personne, this type of modification is found up to the early 20th century. Again, it is here as well quite striking to discover that the current form of indirect modification with de i.e. personne d'autre arose only quite recently. Searching Frantext tor indirect modification with d'autre, we observe again that it first occurred in Chretien de Troyes in the 12th century. But, as above, the pattern disappears entirely for a period of eight centuries to finally resurface only in 1902 in the writings of Adam Paul. (49) quand elle rentrait, personne di:l.utre ne se derangeant pour elle? when she was getting back. no one else bothering tor her Adam, PauL 1862-1920. [1902], .Lknfant d'Austerlitz (Paris, Ollendorff, 1902.) Thus we see here again that, significantly, indirect modification with di:l.utre does not occur with bare personne until the very beginning of the 20th century. 3·3
Theoretical implications and conclusions
The previous section has provided novel detailed evidence of the internal evolution of the n-words rien and personne, attempting to establish, on the basis of data from extensive historical corpuses, the exact time course of their feature loss on the one hand, and related structural changes based on modification changes on the other hand. Let us now take stock of the observed facts to unfold their signification and theoretical consequences. Summarizing and highlighting the central facts, we have observed three distinct stages in the evolution of these n-expressions. In the first stage, n-words are essentially
278 Viviane Deprez nominal in nature, as evidenced by the use of a determiner, their gender and number features as well as by the pre-nominal position of the modifier autre. By and large, we have dated the loss of features to the middle or end of the 16th century and the disappearance of the pre-nominal modification pattern by the beginning of the 17th century. We concluded that, structurally, during this period, the n-expressions occurred within the NP layer of their nominal structure. In the second stage, then-words have come to gradually lose their nominal features and to become invariable both in number and gender. Modification by autre is now post-nominal but, crucially, it remains direct, without the help of the linking functional de. We have suggested that at this stage, which essentially spans the classic period, these n-words have climbed a first step within the nominal structure, reaching NumP, so that they are now directly merged in this position, accounting for the strictly post-nominal position of autre. The third stage corresponds in our data with the appearance of the indirect de modification. As discussed in Section 2.2, this type of modification with d'autre, corresponds to quantity or degree modification, like de plus, or de mains, and indicates that n-words have now taken on a quantificational nature. Following Zamparelli (1995) we have assumed that this could correspond syntactically to yet another step up the structural ladder of the DP structure, reaching in Zamparelli's terms the level of strong quantification or SDP. Putting all these assumptions together, the following structure sums up the evolution of then-word that our conjoined empirical evidence has unearthed. (SO)
SDP/DP Modem French rien/personne
Rien/personne
On this view, it is visible that the evolution of n-words corresponds to a gradual step by step movement up the D P structure. Of particular interest are the parallels that can be observed between the historical course of this internal evolution and the findings that Deprez & Martineau (2004) and Martineau & Deprez (2004) report on the historical evolution of the meaning of n-words. Focusing on the interpretation of aucun these authors have distinguished 3 periods that largely correspond to the ones we have distinguished here. In the first period, corresponding to Old and Middle French, the interpretation of n-words seems largely context-independent and positive. Beginning with the 16th century and into the classic period, however, the interpretation of nwords becomes context-dependent. As Martineau & Deprez (2004) argue, n-words at this period manifest characteristic features of N PI. They are licensed only in restricted contexts, i.e. the so-called forclusive contexts. They can also be licensed at long distance, i.e. in cross-clausal domains. Moreover, they are not licensed in subject position and they are fully compatible with negation, including the negation pas, which is dominant at the time as the sole exponent of sentential negative meaning and can co-occur
From N to D 279
with them, sometimes even in conjunction with ne. 22 Finally. in the last period, which corresponds to the birth of Modern French at the beginning of the 20th century, nwords come to take on a negative meaning of their own, now occurring alone, i.e. without the co-presence of the sentential negative marker pas. This evolution is well known. What had not been put together up to now, however, is the parallelism that the meaning evolution of n-words shows with the changes that affect their internal structure. This is what this paper hopes to have now accomplished by coupling the historical evolution of the meaning contribution of n-words at the sentential level with a careful study of their internal changes.
References Azoulay- Vicente, Avlgall. 1985. Les Tours Comportant frxpression Ue' + Adjectif. Geneva: LibrairieDroz. Clas, Andre. 1976. Materiaux pour Utude du Franfais au Canada. Neologismes-canadianismes, Vol 1. Mont:real: Universite de Montreal. Condoravd!, Cleo & Kiparsky. Paul 2006. Jespersen's cycle: The argument phase. Handout of talk presented at the Stanford Semfest. March 2005. Deprez. Viviane. 1997. Two types of negative concord Probus 9(2): 103-142. Deprez. Viviane. 1999. The roots of negative concord in French and Creoles. In Language Creation and Language Change: Creole, Diachrony and Development, Michel DeGraff (ed.), 375-428. Cambridge MA: The MIT Press. Deprez, Viviane. 2000. Parallel (a)symmetries and the structure of negative expressions. Natural Language and Linguistic Theory 18(2): 253-342. Deprez, Viviane. To appear. Atoms of negation. Ms. L2C2/Rutgers university. (To appear in Ingham, Richard & Larrivee, Pierre (eds). Berlin: Mouton de Gruyter). Deprez, Viviane & Martineau, France. 2004. Micro-parametric variation and negative concord. In Contemporary Approaches to Romance Linguistics [Current Issues in Linguistic Theory 258], Julie Auger, Clancy Clements & Barbara Vance (eds), 139-158. Amsterdam: John Benjamins. den Dikken, Marcel1998. Predicate Inversion in DP. In Possessors, Predicates and Movement in the Determiner Phrase [Linguistik AktueWLinguistics Today 22], Artemis Alexiadou & Chris Wilder (eds), 177-214. Amsterdam: John Benjamins. Giannakidou, Anastasia. 2000. Negative ... Concord? Natural Language and Linguistic Theory 18:457-523. Gougenheim, Georges. 1951. Grammalre de la langue franfalse du seizieme siede. Lyon: lAC. Grevisse, Maurice. 1980. Lebon usage. Paris: Duculot. Heycock. Caroline & Zamparell!, Roberto 2005. Friends and colleagues: Coordination, plurality, and the structure ofDP. Natural Language Semantics 13:201-270. Huot, Helene. 1981. Constructions infinitives du franfais. Le subordonnant 1ie'. Geneve: Droz.
22.
See Martineau & Deprez (2004) for concrete examples and relevant statistical computations.
28o Viviane Deprez
Hulk, Aafke & Verheugd, Els. 1994. Accord et operateurs nuls dans les projections adjectivales. Revue Quebtcoise de Linguistique 23: 17-46. Kayne, Richard 1977. Syntaxe du fran fa is. Paris: Seu!l Kayne, Richard 1994. TheAntisymmetryofSyntax. Cambridge MA: The MIT Press. Lohndal, Terje & Haegeman, Liliane. 2010. Negative concord and (multiple) agree: A case study of West Flemish. Linguistic Inquiry 41(2): 181-211. Marantz. Alec. To appea.t Phases and words. Martineau, France & Deprez. Viviane. 2004. Pas rlen/pas aucun en franc;ais dassiqwf. Variation dialectale et historique. Langue Franfalse 143: 33-47. May, Robert. 1989. Interpreting logical form. Linguistics and Philosophy 12: 387-435. Nishigauchi, Taisuke. 1990. Quantification In the Theory of Grammar. Dordrecht: Kluwer. Nyrop, Kristoffer. 1930. Grammalre I-nstorlque de la Langue Franfalse, V: Syntaxe. Copenhagen: Nordisk Forlag. Pereltsvaig, Asya. 2006. Small nomi.nals. Natural Language & Linguistic Theory 24:433-500. Sarre, Nguisaly. 2003. Di.achronie des pronoms indefinis abase nominale du moyen f~ais au fran~is dassique: Les remanences d'un emplot nominal atravers un emploi pronominal. Revue de LlnguisNque Romane 265-266: 117-136. Sleeman, Petra. 1996. Licensing Empty Nouns in French. The Hague: HAG. de Swart, Henriette & Sag, Ivan. 2002. Negation and negative concord in Romance . Linguistics and Philosophy 25: 373-417. Zamparelli, Roberto. 1995. Layers in the Determiner Phrase. PhD dissertation, University of Rochester. Zetjlstra, Hedde. 2004. Sentential Negation and Negative Concord PhD dissertation, University of Amsterdam.
Electronic data bases ARTFL-FRANTEXT comprising more than 2,900 French language texts spanning from the 12th through the 20th centuries TFA (Textes de F~ais Ancten 10th-13th centuries) ELICOP
Index A adje...""tival article see pre-adjectival article adje...""tival inflection 3, 12., 76, 111-113, 116-118, 120, 121, 12.4, 2.43 adjectival modifier/modification 28, 36, 37, 112, 116, 236, 2.42., 2.60, 263, 265, 2.75 adjective passim anaphoric adjectives 2.41,
anaphoric: 30, 144, 145, 176, 179, 180, 187, 2.04-206, 230 Ancient Greek 13, 227 animacy 226, 231, 237, 271 appositive 13, 230, 232, 234. 235 Aramaic 161, 173 aspect 3,4, 26,29,32-38 imperfective aspect 4, 33, 34 perfective aspect 3,4, 33, 34.36 Atlantic c:reo les 142, 154
2.46-253 attributive adjective 8, 11, 13,
B
14. 91, 96, 97· 99· 101, 104. 112., 113, 195. 20J, 217, 2J4, 235. 242, 2.47 gradable adjective 90, 105, 106 post-nominal adjective 11, 12, 14.2%275 predicative adjective 11, 12,
bare noun 15, 27, ]2, 13 5, 136, 138, 139· 143· 150, 184-186, 189, 242,258,260 bare NP So, 85,141-143, 150,152, 164, 165, 243. 261 bare singular count noun 163165,172 Berbice Dutch 150,152
91, 96, 101, 104, 195· 203, 217 pre-nominal adje...""tive 11, 12, 14. 121, 202-2.04, 212, 217, 218, 262-2.65 strong adjective 3, 8, 12.-14, 45. 46, 107, 11J, 117, 118, uo, 121, 137. 243. weak adjective 3, 8-10, 13, 14. 16, 45. 46, 48, 107, 111-113, 117, 118, 120, 121, 137· 2.43 Agree 35, 76, 116, 128-130, 132, 193, 195-197, 199, 201-203, 210, 212, 218, 2.19 Local Agree 193, 195, 197, 218,219 Long Distance Agree 197, 199· 210, 218, 219 agreement 64, ]6, 89, 91, 94, 101-103, 105, 111, 114. 12.4, 129, 132, 143· 182,187,188, 196,213, 2.45· 262, 268, 270, 273 long-distance agreement 187, 188
c Celtic 2, 3, 175, 187, 189, 190 classifier 30, 121, 123 conceptual individuation 229 copula 61, 63, 64, 66, 67, 70-73, 76--78, SJ-85, 96, 2.17, 249 creole 141-143, 145, 148-154. 206, 236, 257. 258, 265
D Danish 7, 9. 11, 16, 89. 91, 93, 94, 96--98, 10o-10S, 107, 108, 112, 113, 115-117, 119, 120, 123, 124,136 definite article 3, s-s, 1o, 13, 14, 16, 37. ¢, 90, 93. 112, 11J, 116, 119, 131, 135-137. 139. 160-167, 171,175-190, 193-218, 2.48, 2.49 definiteness passim double definiteness 5, 9,111, 112, 114. 118, 119, 122, 131, 132,136
definiteness valuation 193, 195. 197. 199. 201, 2.10, 213, 216,218 definite reference 138, 159, 161, 1]0
degree expressions 89-91 demonstrative passim distal demonstrative 5, 9.14, 160' 169,175 post-nominal dem(onstra:tive) 197,198, 203, 204. 207, 211, 213, 218 pre-nominal demonstrative 205, 206 determiner passim definite determiner 5-7,31, 32, 45· 132,141-154.167,194, 201,215, 262,270 indefinite determiner 45, 141, 143.215 discontinuous 4. 57-59, 63, ]8,228 discourse reference 112, 12.4 Distributed Morphology 111, 121,130 Dutch 7, 12, 15, 16, 5]-85, 96, 142.. 150,151,2.41-2.52
E enclitic 5, 177, 193, 195, 2.03, 204,218 English passim American English so-54, 108,237 British English 50-54. 162 present-day English (PDE) 97, 106, 164, 1]6, 178, 179, 182, 185-189, 22.5, 242-2.44
F familiarity 115-117, 144,160, 166,177 Faroese 5, 6, 16, 111, 112, 11.4. 116, 119, 122, 124, 132., 136
282
The Noun Phrase ln Romance and Germanic: Structure, Variation, and Change floating quantifier 4:1, 46, 48, 52,53 French 5, 7, 11, 128, 134, 135, 138' 139. 161, 197. 206, 246, 257-265,267-279 functional noun 167
G Gbe 141, 142, 144> 145, 149, 150, 152 generic 35, 48, 49, 53, 134. 135, 138,139. 143-145. 147· 148, 16), 184-186, :111, 230, :16:1 generic NPs 145, 147, 148 generic reference 134, 135, 138,139 genitival 7, 14-16, 211, 234 genitive 1, 6, 14-16, 27-30,32, 37.166,167,169-171, 182-184. 187,210, 211,:115, 219. 220, 223-238,242 adjectival genitive 169, 242 anchoring genitive 232 group genitive 16 inflectional genitive 209-213, 225,236 postnominal genitive 14, 15,187,:111,218, 224,228, :130-232, :134· :136, :137 prenominal genitive 15,166, 169, 200,201,:115-217,226, 2:18, 230, 231, 236 prepositional genitive 210 synthetic genitive 166, 169"-171 German 7, 12, 15, 16, 26, 28, 29, 31, 35. 37. 41,43-54.89, 91. 93. 94. 9?--101, 103-108, 165, 176,178 Germanic passim Gothic 2, 12-14 gradable noun 90, 105 grammaticalization 5, 16, 91, 105-107, 151,152,154,163-165, 169, 170, 226, 236, 237· 239· :140, 253-255. 257. :171 Greek 2, 6, 13, 169. 177, 227 Gungbe 144.145,150
H Haitian Creole 142, 257, 258 Hausa 150 Hungarian 49
N Icelandic 5-7, 9, 10, 15.16, 112, 113, 115-117,119,123,124. 136, 185, 187, 189 identifiability 116, 133, 134. 144, 160, 162,166, 170,171 inalienable possession 166, 171,:135 inddinite article 58--61, 65, 66, 69, 70. 74-76. 83, 89-94. 97. 99. 105, 161, 168, 170-172, 185, 187, 189, 190, 243 spurious indefinite article 59, 61, 66, 69, 75· 76, 83 infinitive 26--37 interpretable feature 128,194 interrogative DP 17, 41, 43, 44. 49.52-54 Italian 2, 6, 11, 41, 42,49-51, 139, 169, 178, 224, :135. 267
negation 245, 252, 257, 259, 260, 275· 276, 278 negative concord 257, 258, 270, 272 nominal copula 61, 63, 64, 66, 67, 70-73, 76-78, 83--85 nominalizations 25-29, 31-33, 35. 36--39. 211, 232 nominalized infinitive 17, 39, 40 Norwegian 5-7, 9, 16, 111, 112, 114-119,122, 124. 132,136 noun-noun constructions 237 NPI 269, 275, 276, 278 null determiner/D see zero determiner null morpheme 129 number marking 130, 143, 160 n-words 257, 258, 260-26), 265-270,:17:1,274-279
0
Jamaican Creole 141-154 L Latin :1-6,12-15,177, 189,203, :104, 2:13, :124. 226-237. 242, 270 linking pronouns 236
Old English 4, 12, 13, 160, 163, 173, 175-190,224. 236, 242 Old French 235, 273 Old High German 13 Old Norse 5, 8, 9, 13 OldRomanian 193,194.199. :101, 218, 219
M
p
mass nouns 33, 6o, 121, 123, 184,187 Mauritian Creole 206 micro-diversity 57, 59, 81 micro -variation 17, 25, 38, 65, 258 Middle English 106, 16), 164, 171, 176, 224, 225 Middle French 270, 278 morphosyntactic definiteness 166, 171 movement 46, 47, 53, 54. 58, 6o, 62-64, 67, 72, 77, 78, So, 83, 84. 91. 92. 104. 123, 124. 129, 136,137.179.198, 199.:103,:106, 207, 213, 214. :119, 266, :168, :175.278 Head Movement 72, 84, 206 Long Distance Move(ment) 204, 212 multiple definites 202
parantetricvariation 178,219,279 partitive 49, 96, 135, 138, 139, 228, 235, 245· 26?-269 pied-piping 47, 48, 51, 54 plural 9, 31-33, 36, 37, 6o-62, 75, 76, 106, 117, 121-123, 128-131, 143.145. 146,184-186,188, 244· 261, 26J, 265, 266, 268-270 Portuguese 6,42 position passim postnominal position 5o 7, 12, 14. 4J, 92, 104. 19Q, 210, 275· :177,278 predicate position 59, 91, 104, 243.244 prenominal position 5, 7, 11, 1:1,14, 15, 27,37, 92, 234, 264, 274· 278 possessive 6-9, 14-16,32, 45. 96, 118, 145· 166, 167, 169"-171, 183, 184. 223, 225, 226, 228, 231, 234-237.242-245.248 weak possessive 6, 7
Index 2.83 pre-adjectival article 8-10, 14, 112, 114-117, 120, 121, 12J, 124 predetenniner 90, 249 predicate displacement 17, 5]'--61, 6J, 64. 6!}, 72, 73· 77· 78, 81,84,85 Predicate Fronting 63 Predicate Inversion 17, 59> 61, 63, 69. 74· 76. 79. 86, 264, 2.79 predicate raising 89, 91, 95, 97,100 proper name 13, 6o, 150,178, 18J, 194, 197· 2J3 Q
quantification 170, 258, :qo, 278 quantifier 17, 19,41--54. 139, 184, 262-2.64, :167, 269 strong quantifier 267, 269> 270,278 universal quantifier 17, 41-44, 47-54 weak quantifier 266, 267 Quebec French 276
R raising 83-85, 89, 91, 95, 97, too, 136, lSO, 198,199. 238 reanalysis 159, 163-165, 172, :104, 22.5, 226, 239 reduced relative clause 96 referential 114-117, us, 153,154, 163, 164. 169, 171, 184, 1% 211, 213,230
relational noun 166, 168, 170, 171 Romance passim Romanian 2, 5--8, 10, u, 25-27, 29-37· 177, 19J-219 Rumanian see Romanian
suffixed article 5, 7, 112, 114-116, 119-121, 124 superlative 10, 114, 147, 188, 234 Swedish 4, 5, 7--9, 16, 50, 51, 53, 54, 111-122, 124, 1J1, 132, 135. 1J6, 140· 246
s
T
SaramaJ:can 142 Scandinavian 5, 9, 65, 111-113, 119-122, 124, 128, 131-133. 135-137. 175. 177, 189, 204, 2.05, 218, 22J, 225 semantic change 91 Spanish 2, 5-7, 11, 25-27, 29-32, 34-37· .p, 49· so. 197 specificity 48, 111, 116, 133, 141-146, 149. lSO, 15:1,153. 167,204 specific reference 112, 116, 124 split pattern 58, 81, 83, 85 Sranan 141-154 stacking 243, 264. 266 stranding 42, 46-55, 266 structure of the DP 91, 92, m, 125 sub extraction 59. 69, 78. 81-85 subordination 235 Subset Principle 218 suffix J-j, 7, 8, 10, 13, JO, 6!}, 80, 120, 131, 1J2, 1J6, 161, 19J, 195· 203, 206, 207, 218
topicality 226
sa.
u uninterpretable feature uS uniquely identifiable 134. 164 uniqueness 117, 133, 144, 149, 160, 162, 166, 167, 170, 188
w weak definites 18, 147, 148, 152, 159.162, 164,165,167, 16!}, 171-173 wh-word 17, 44. 47, so, 51, 54. 58, 62-64. 66, 77· 82, 84. 85 word order 4-6, 13, 14o 42, 74. 77, 79, 91, 94, 95, 97, 98, 10J, 108, 120, 223-226, 228, 2JO, 2J5, 2J6, 238
z zero determiner 150,178,187,
188, 243. :158, 26o, 2.72.
Linguistik Aktuell/Linguistics Today A complete list of titles in this series can be found on the publishers' website. www.benjamins.wm 178 GALANI, Alenndn, Glyn lllCKS and George TSOULAS (eds.): Morphology and its Interfaces. Expected May 2011 177 WSKUP,Petr: Adverbials and the Phase ModeL Expected April2o11 176 YUASA,Etsnyo, Tista BAG em and Katharine P. BEALS (eds.): Pragmatics and Autolexical Grammar. Jn honor of Jerry Sadock xxv. 335 pp. + index. Expected April 2011 175 MEIBAUER, )llrg and Markus STEINBACH (eds.): Experimental Pragmatics/Semantics. .x, 238 pp. + index. Expected April 2011 174 LOMASHVIT.I, Lela: Complex Predicates. The syntax-morphology interfac.e. xi, 187 pp. + index. E:~.pected
April2ou 173 SAPP, Christopher D.: The Verbal Complex in Subordinate Clauses from Medieval to Modem Gem1an 2011. ix, 225 pp. 172 JUNG,Hakynng: The Syntax of the Be-Possessive. Parametric variation and surface diversities. x~ 264 pp. + index. Expected March 2011 171 SLEEMAN,Petn and Harry PERRIDON (eds.): The Noun Phrase in Romance and Gem1anic. Structure, variation, and change. 2011. vii. 283 pp. 170 HUNTER, Tim: Synta.."':ic Effects of Conjunctivist Semantics. Unifying movement and adjunction. xi, 179 pp. + index. Expected April2o11 169 SANCHEZ,Llliana: The Morphology and Syntax of Topic and Focus. Minimalist inquiries in the Quechua periphery. 2010. xiii, 242 pp. 168 FELDHAUSBN, Ingo: Sentential Form and Prosodic Structure of Calalan 2010. xiii, 285 pp. 167 MERCADO,Raphaei,EricPOTSDAM and Lisa deMena TRAVIS (eds.): Austronesian and Theoretical Linguistics. 2010. vii, 379 pp. 166 BRANDT, Patrick and Marco GARdA GARdA (eds. ): Transitivity. Form, Meaning. Ac-quisition, and Processing. 2010. vii, 308 pp. 165 BREUL, Cantea and Edward GOBBEL (eds.): Comparative and Contrastive Studies of Information Structure. 2010. xi~ 306 pp. 164 ZWART, Jan-Wo1tter and Mark de VRIES (eds.): Structure Preserved. Studies in syntax for Jan Kostet 201 o. xxiii, 395 pp. 163 KIZIAK, Tanja: Emaction Asymmetries. Experimental evidence from German. 2010. xvi, 27 3 pp. 162 BOTT,Oliver: The Processing of Events. 201o.xvii, 383 pp. 161 HAAN, Germen J. de: Studies in West Frisian Grammar. Edited by Jarich Hoekstra, Willem Visser and Golfe Jensma. 2010 . .x, 384 pp. 160 MAVROGIORGOS, Marlos: Clitics in Greek A minimalist aro>unt of proclisis and enclisis. 2010 . .x. 294PP· 159 BREITBARTif, Anne, Christopher LUCAS, Shela WATTS and David WILLIS (eds.): Continuity and Change in Grammar. 2010. viii, 359 pp. 158 DUGUINE, Maia, Susana HUIDOBRO and Nerea MADARIAGA (eds.): Argument Structure and Syntactic Relations. A cross-linguistic perspective. 2010. v~ 348 pp. 157 FISCHER, Snsann: Word-Order Change as a Source of Grammaticalisation. 2010. ix, 200 pp. 156 DI SCIULLO, Anna Maria and Virginia lllLL (eds.): Edges, Heads, and Projections. Interface properties. 2010. vii, 265 pp. 155 SATO, Yosuke: Minimalist Interfaces. Evidenc.e from Indonesian and Javanese. 2010. xiii, 159 pp. 154 HORNSTEIN, Norbert and Marla POLINSKY (eds.): Movement Theory of Control. 2010. vii, 330 pp. 153 CABREDO HOFHERR,Patricia and Ora MATUSHANSKY (eds.): Adjectives. Fom1al analyses in syntax and semantics. 2010. vii, 335 pp. 152 GALLEGO, Angel J.: Phase Theory. 2010. xii, 365 pp. 151 SUDHOFF, Stefan: Focus Particles in Gem1an Syntax, prosody, and infom1ation structure. 2010. xiii, 335 pp. 150 EVERAERT, Martin, Thm LENTZ, Hannah de MULDER, 0ystein NILSEN and Arjen ZONDERVAN (eds.): The Linguistics Enterprise. From knowledge of language to knowledge in linguistics. 2010. i:r, 379 pp. 149 AELBRECHT, LobJre: The Syntactic Licensing of Ellipsis. 2010. xii, 2 30 pp.
148 HOGEWEG, Lotte,IW.eo de HOOP and Andrej MALCHUKOV (eds.): Cross-linguistic Semantics of Tense. Aspect. and Modality. 2009. vii. 406 pp. 147 GHOMESHI, Jila,Deaaa PAUL and Martina WILTSCHKO (eds.): Determiners. Universals and variation. 2009. vii, 247 pp. 146 GEIDEREN,FJiyvan (ed.): Cyclical Change. 2009. viii, 329 pp. 145 WESTERGAARD, Marit: The Acquisition ofWord Orde& Micro-cues, information structure, and economy. 2009. .xii, 245 pp. 144 PUTNAM, Michael T. (eel.): Towards a Derivational Syntax. Survive-minimalism. 2009. .1, 269 pp. 143 ROTHMAYR,Antonia: The Structure of Stative Vems. 2009. xv, 216 pp. 142 NUNES, Jaim (eel.): Mininlalist Essays on Brazilian Portuguese Syntax. 2009. vi, 243 pp. 141 ALEXIADOU, Artemis, Jorge HANKAMER, Thomas McFADDEN, Justin NUGER and Florian SCHAFER (eds.): Advances in Comparative Germanic Syntax. 2009. xv, 395 pp. 140 ROEHRS, Dorian: Demonstratives and Definite Articles as Nominal Auxiliaries. 2009. .xii, 196 pp. 139 HICKS. Glyn: The Derivation of Anaphoric Relations. 2009. .xii, 309 pp. 138 SIDDIQI, Daniel: Syntax within the Word. Economy, allomorphy, and argument selection in Distributed Morphology. 2009. xii, 138 pp. 137 PFAU, Roland: Grammar as Processor. A Distributed Morphology account of spontaneous speech errors. 2009. .xiii, 372 pp. 136 KANDYBOWICZ, Jason: The Gran1mar of Repetition. Nupe grammar at the syntax-phonology interface. 2oo8. .xiii, 168 pp. 135 LEWIS, William D., Slmin KARIMI,Heidi HARLEY and Scott 0. FARRAR (eds.): Tinle and Again. Theoretical perspectives on formal linguistics. In honor of D. Terence Langendoen. 2009 . .xiv, 265 pp. 134 ARMON-LOTEM, Sharon, Gabi DANON and Susan D. ROTHSTEIN (eds.): Current Issues in Generative Hebrew Linguistics. 2008. vii, 39 3 pp. 133 MACDONALD, Jonathan E.: The Syntactic Nature of Inner Aspect A mininlalist perspective. 2008. xv. 241 pp. 132 WBERAUER, Theresa (ed.): The Limits of Syntactic Variation. 2008. vii, 521 pp. 131 DE CAT,
113 EYTH6RSSON, 1b6rballur (eeL): Grammatical Change and Linguistic Theory. The Rosendal papers. 2008. v~ 441 pp. 112 AXEL, Katrin: Studies on Old High German Syntax. Left sentence periphery, verb placement and verbsecond 2007• .xii, 364 pp. 111
110 109 1 o8 1 07 106 105 1 04 103 102 101 100 99 98 97 96 95 94 93 92 91 90 89 88 87 86 85 84 83 82 81 8o 79
EGUREN, Luis and Olga FERNANDEZ-sORIANO (eds.): Coreference, Modality, and Focus. Studies on the syntax-semantics interface. 2007 . .xii, 239 pp. ROTHSTEIN, Susan D. (ed.): Theoretical and Crosslinguistic Approacl!es to the Semantics of Aspect 2008. viii, 453 pp. CHOCANO, Gem a: Narrow Syntax and Phonological Form. Scrambling in the Germanic languages. 2007. X, 333 pp. REULAND, Eric, Tanmoy BHATTACHARYA and Giorgos SPATHAS (eds.): Argument Structure. 2007. xviii. 24 3 pp. CORVER, Norbert and Jalro NUNES (eds.): The Copy Theory of Movement 2007. vi. 388 pp. Dmffi,Nicole andYordanka KAVALOVA (eds.): Parentheticals. 2007. xii, 314 pp. HAUMANN,Dagmar: Adverb Licensing and Clause Structure in English. 2007. ix, 438 pp. JEONG, Yoangmi: Applicatives. Structure and interpretation from a mininlalist perspective. 2007. vii, 144 pp. WURFF, Wim van der (eeL): Imperative Clauses in Generative Grammar. Studies in honour of Frits Beukema. 2007. viii, 352 pp. BAYER, Josef, Tanmoy BHATTACHARYA and M. T. Hany BABU (eds.): Linguistic Theory and South Asian Languages. Essays in honour of K. A. Jayaseelan. 2007. x. 282 pp. KARIMI, Simin, Vida SAMHAN and Wendy K. WILKINS (eds.): Phrasal and Clausal Architecture. Syntactic derivation and interpretation. In honor of Joseph E. Emonds. 2007. vi, 424 pp. SCHWABE, Kerstin and Susanne WINKLER (eds.): On Information Structure, Meaning and Form. Generalizations across languages. 2007. vii, 570 pp. MARTINEZ-GIL,Femando and Sonia COLINA (eds.): Optimality-Theoretic Studies in Spanish Phonology. 2007. viii, 564 pp. PIRES,Acrisio: The Mininlalisl Syntax of Defective Domains. Gerunds and infinitives. 2006•.xiv, 188 pp. HARTMANN, Ju.tta M. and Uszl6 MOLNARFI (eds.): Comparative Studies in Germanic Syntax. From Afrikaans to Zurich German. 2006. vi, 332 pp. LYNGFELT,Benjamin and Thrgrim SOLSTAD (eds.): Demoting the Agent Passive, middle and other voice phenomena. 2006. x. 333 pp. VOGEI.EER, Swtlana and Liliane TASMOWSKI (eds.): Non-definiteness and Pluralily. 2006. vi, 358 pp. ARCHE, Maria J.: Individuals in Time. Tense, aspect and the individual!stage distinction. 2006 . .xiv, 281 pp. PROGOVAC, Ljijana, Kate PAESANI, Eugenia CASIELLES and FJlen BARTON (eds.): The Syntax of Nonsententials. Multidisciplinary perspectives. 2006. x. 372 pp. BOECKX, Cedric (ed.): Agreement Systems. 2006. ix, 346 pp. BOECKX, Cedric (ed.): Minimalist Essays. 2006. xvi, 399 pp. DALMI, GNte: The Role of Agreement in Non-Finite Predication. 2005. xv~ 222 pp. VELDE, Joha R. te: Deriving Coordinate Symmetries. A phase-based approaclJ integrating Select, Merge, Copy and Match. 2006. x, 385 pp. MOHR, Sabine: Clausal Architecture and Subject Positions. Impersonal constructions in the Germanic languages. 2005. viii, 207 pp. JULIEN, Marit: Nominal Phrases from a Scandinavian Perspective. 2005. xvi, 348 pp. COSTA, Joio and Maria Cristiaa FIGUEIREDO SILVA (eds.): Studies on Agreement 2006. vi, 285 pp. MIKKELSEN, Liae: Copular Clauses. Specification, predication and equation. 2005. viii, 210 pp. PAFEL, Jllrgen: Quantifier Scope in German. 2006. xvi, 312 pp. SCHWEIKERT, Walter: The Order of Prepositional Phrases in the Structure of the Clause. 2005. xii, 338 PP· QUINN, Heidi: The Distribution of Pronoun Case Forms in English. 2005 •.xii, 409 pp. FUSS, Eric: The Rise of Agreement A formal approaclJ to the syntax and grammaticalization of verbal inflection. 200 5· .xii, 336 pp. BURKHARDT SCHUMACHER, Petra: The Syntax-Discourse Interface. Representing and interpreting dependency. 2005. xi~ 259 pp. SCHMID, Tanja: Infinitival Syntax. Infinitivus Pro Participio as a repair strategy. 2005 . .xiv, 251 pp.
78 DIKKEN,Marcel den and O!ristina TORTORA (eels.): The Function of Function Words and Functional Categories. 200 5- vii. 292 pp. 77 OZTCRK, Ballaz: Case. Referentiality and Phrase Structure. 2005. x. 268 pp. 76 STAVROU, Melita and Arbon to TERZI (eels.): Advances in Greek Generative Syntax. In honor of Dimitra Theophanopoulou-Kontou. 2005. viii, 366 pp. 75 DI SCIULLO, Anna Maria (eel.): UG and External Systems. Language, brain and romputation. 2005. xviii, 398 pp. 74 HEGGIE,Lorie and Francisco ORD61ffiz (eds.): Clitic and Affix Combinations. Theoretical perspectives. 2005- viii, 390 pp. 73 CARNIE, Andrew, Heidi HARLEY and Sbela Ann DOOLEY (eds.): Verb First On the syntax of verbinitial languages. 200 5- xiv, 434 pp. 72 FUSS, Eric and Carola TRIPS (eds.): Diachronic Clues to Synchronic Grammar. 2004- viii. 228 pp. 71 GEIDEREN,FJlyvan: Gran1maticalization as Economy. 2004- xvi, 320 pp. 70 AUSTIN, Jennifer R., Stefan ENGELBERG and Gisa RAUH (eds.): Advel.'bials. The interplay between meaning, context, and syntactic structure. 2004- :x, 346 pp. 69 KISS, Katalin 1t and Henk 'Vlln RIEMSDIJK (eds.): Verb Clusters. A study of Hungarian, German and Dukh. 2004 vi, 514 pp. 68 BREUL, Canten: Focus Structure in Generative Gran1mar. An integrated syntactic, semantic and intonational approach. 2004. x. 432 pp. 67 MISESKA TOMIC, Olga (eel.): Balkan Syntax and Semantics. 2004- xvi, 499 pp. 66 GROHMANN, Kleanthes K.: Prolific Domains. On the Anti-Locality of movement dependencies. 2003. xvi. 372 pp. 65 MANNINEN, Salu IW.ena: Small Phrase Layers. A study of Finnish Manner Adverbials. 2003. .Iii, 275 pp. 64 BOECKX, Cedric and Kleanthes I<. GROHMANN (eels.): Multiple Wh-Fronting. 2003. x, 292 pp. 63 BOECKX, Cedric: Islands and Chains. Resumption as stranding. 2003. :xii, 224 pp. 62 CARNIE, Andrew, Heidi HARLEY and MaryAnn WILIJE (eels.): Formal Approw:hes to Function in Gran1ma1: In honor of Eloise Jelinek. 2003. .Iii, 378 pp. 61 SCHWABE, Kentin and Susanne WINKLER (eds.): The Interfaces. Deriving and interpreting omitted structures. 2003. v~ 403 pp. 6o TRIPS, Carola: From OV to VO in Early Middle English. 2002. xiv. 359 pp. 59 Dmffi, Nicole: Particle Verbs in English. Syntax, information structure and intonation. 2002. .Iii, 305 pp. 58 DI SCIULLO, Anna Maria (eel.): Asymmetry in Gran1ma1: Volume 2: Morphology. phonology, acquisition. 200 3· vi, 309 pp. 57 DI SCIULLO, Anna Maria (eel.): Asymmetry in Gran1ma1: Volume 1: Syntax and semantics. 2003. vi, 405 pp. 56 COENE, Martine and fie& D'HULST (eds.): From NP to DP. Volume 2: The expression of possession in noun phrases. 2003. x, 295 pp. 55 COENE, Martine and Yves D'HULST (eds.): From NP to DP. Volume 1: The syntax and semantics of noun phrases. 2003. vi, 362 pp. 54 BAPTISTA, Madr-e; The Syntax of Cape Verdean Creole. The Sotavento varieties. 2003. x.xii, 294 pp. (incl. CD-rom). 53 ZWART, Jan-Wouter and Werner ABRAHAM (eds.): Studies in Comparative Germanic Syntax. Proceedings from the 15th Workshop on Comparative Germanic Syntax (Groningen, May 26-27, 2ooo). 2002. .xiv, 407 pp. 52 SIMON, Hont J. and Heike WIESE (eds.): Pronouns - Gran1mar and Representation. 2002 . .xii, 294 pp. 51 GERLACH, Birgit: Clitics between Syntax and Le:xic.on. 2002. .xii, 282 pp. 5o STEINBACH, Markus: Middle Voice. A comparative study in the syntax-semantics interfu:.e of German. 2002. .xii, 340 pp. 49 ALEXIADOU, Ademis (ed.): Theoretical Approaches to Universals. 2002. viii, 319 pp. 48 ALEXIADOU,Ademis,Elena ANAGNOSTOPOULOU, SjefBARBIERS and Hans-Martin GARTNER (eds.): Dimensions of Movement From features to remnants. 2002. vi. 345 pp. 47 BARBIERS, Sjef,Frits BEUKEMA and Wim van derWURFF (eds.): Modality and its Interaction with the Vel.'bal System. 2002. x. 290 pp. 46 PANAGIOTIDIS, E. Pboevos: Pronouns, Clitics and Empty Nouns. 'Pronominality' and licensing in syntax. 2002. :x. 214 pp.
