Interdependence of Diachronic and Synchronic Analyses (Studies in Language Companion)

Interdependence of Diachronic and Synchronic Analyses Studies in Language Companion Series (SLCS) This series has bee...

Author: Folke Josephson | Ingmar Sohrman

365 downloads 1980 Views 3MB Size Report

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!

Report copyright / DMCA form

DOWNLOAD PDF

Interdependence of Diachronic and Synchronic Analyses

Studies in Language Companion Series (SLCS) This series has been established as a companion series to the periodical Studies in Language.

Editors Werner Abraham University of Vienna

Michael Noonan

University of Wisconsin, Milwaukee

Editorial Board Joan Bybee

Christian Lehmann

Ulrike Claudi

Robert E. Longacre

Bernard Comrie

Brian MacWhinney

University of New Mexico University of Cologne Max Planck Institute, Leipzig University of California, Santa Barbara

William Croft

University of New Mexico

Östen Dahl

University of Stockholm

Gerrit J. Dimmendaal University of Cologne

Ekkehard König

Free University of Berlin

University of Erfurt

University of Texas, Arlington Carnegie-Mellon University

Marianne Mithun

University of California, Santa Barbara

Edith Moravcsik

University of Wisconsin, Milwaukee

Masayoshi Shibatani

Rice University and Kobe University

Russell S. Tomlin

University of Oregon

Volume 103 Interdependence of Diachronic and Synchronic Analyses Edited by Folke Josephson and Ingmar Söhrman

Interdependence of Diachronic and Synchronic Analyses Edited by

Folke Josephson Ingmar Söhrman University of Gothenburg

John Benjamins Publishing Company Amsterdam / Philadelphia

8

TM

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 Interdependence of Diachronic and Synchronic Analyses / edited by Folke Josephson and Ingmar Söhrman. p. cm. (Studies in Language Companion Series, issn 0165-7763 ; v. 103) Includes bibliographical references and index. 1. Grammar, Comparative and general--Verb. 2. Historical linguistics. 3. Typology (Linguistics) I. Josephson, Folke. II. Söhrman, Ingmar. P281.F67 2008 415--dc22 isbn 978 90 272 0570 4 (Hb; alk. paper)

2008019878

© 2008 – 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 Contributors

vii

Introduction

1

Synchronic and diachronic evidence for parallels between noun phrases and sentences Jan Rijkhoff

13

The development of tense, mood, and aspect, in creole languages and the typology of affix order Peter Bakker

43

Aspectual oppositions from Proto-Indo-European to Latin Dag Haug

61

On the development of actionality, tense, and viewpoint from Early to Late Latin 73 Gerd V. M. Haverling Continuity and change: The history of two Greek tenses Eva-Carin Gerö & Hans Ruge

105

Actionality and aspect in Hittite Folke Josephson

131

Imperfectivity and complete events Atle Grønn

149

Predicative expressions of transition in Portuguese and Spanish: A cognitive approach to aspect, aktionsart and tense Ingmar Söhrman & Kåre Nilsson

167

The Old Nordic Middle Voice in the pre-literary period: Questions of grammaticalisation and cliticization Kjartan Ottosson

185



Interdependence of Diachronic and Synchronic Analyses

The relevance of tense and aspect in Semitic Languages: The case of Hebrew and Arabic Sven-Olof Dahlgren The verb phrase in the Kerebe language Christina Thornell Comparative TAM morphology in Niger-Congo: The case of persistive, and some other markers in Bantu Jouni Filip Maho

221

249

283

Indexicals as sources of case-markers in Australian languages William B. McGregor

299

Differential object marking in Sahidic Coptic Åke Engsheden

323

Index

345

Contributors Peter Bakker, University of Aarhus [email protected] Sven-Olof Dahlgren, University of Gothenburg [email protected] Åke Engsheden Uppsala University [email protected] Eva-Carin Gerö, Stockholm University [email protected] Atle Grønn, University of Oslo [email protected] Dag Haug, University of Oslo [email protected] Gerd V.M. Haverling, Uppsala University (formerly at University of Aarhus) [email protected] Folke Josephson, University of Gothenburg [email protected] William B. McGregor, University of Aarhus [email protected] Jouni Filip Maho, University of Gothenburg [email protected] Kåre Nilsson, University of Oslo [email protected] Kjartan Ottosson, University of Oslo [email protected] Jan Rijkhoff, University of Aarhus [email protected] Hans Ruge, Johannes-Gutenberg-Universität, Mainz

 Interdependence of Diachronic and Synchronic Analyses

Ingmar Söhrman, University of Gothenburg [email protected] Christina Thornell, University of Gothenburg [email protected]

Introduction The present volume contains the acta of a Colloquium on Diachronic Typology which was arranged in Gothenburg in November 2004 as a cooperation between the West Scandinavian universities of Gothenburg, Oslo and Aarhus. The project was based on the fundamental insight that a typological approach is indispensable in diachronic studies and that the diachronic dimension of continuity and change is an important complement to synchronic analysis. The matters treated were conceptual parallels and diachronic relations between grammatical categories in the clause (and in the NP) with respect to actionality, aspect and tense. Other topics were paths of grammaticalization and the syntax, semantics of clitics and verb semantics. Reflexive and Middle were discussed as well as the possibility of an interrelationship of reflexivity and aspect. The depth of the attested diachronic dimension was increased by well-established results of the reconstruction of ProtoIndo-European, Proto-Italic and Proto-Nordic and the typological approach was carried on to modern languages with the diachronic perspective in view. A measure of reconstruction of Bantu and Australian languages was suggested. As the antecedents of formal elements of Creole languages are generally well known a study of grammaticalization in those languages contributed to the diachronic dimension. In a theoretical article on synchronic and diachronic evidence for parallels between the layered structures of NPs and sentences Rijkhoff (this volume) discusses conceptual parallels and diachronic relations between grammatical categories in the clause (aspect, tense, mood) and in the NP and shows that they can have the same lexical modifier as well as the same aspectual modifier. There are three layers of modification: quality, quantity and location. As Rijkhoff sees it nominal and verbal aspects are qualifying modifiers whereas demonstratives and tense are localizing modifiers. Unidirectional semantic grammaticalization takes place by means of reanalysis and analogy as syntactic reanalysis/metonymy is followed by analogy/metaphor. Semantic change according to grammaticalization theory occurs by metaphor from NP to clause, from space to time and from concrete to abstract and also, by metonymy, from demonstrative to article and from progressive to tense. There is also grammaticalization from the semantically more complex to the less complex. In the clause, aspect markers may change to tense markers and tense markers to markers of mood. In the NP, collective markers can become plural markers and demonstratives can develop into definite articles. Examples are given of diachronic developments from space to time, i.e. collective > perfective, for which evidence from IE languages is cited. It involves qualifying



Interdependence of Diachronic and Synchronic Analyses

modifiers such as Germanic ge- and Greek sun- as well as quantifying modifiers as in the case of nominal plural > verbal plural. Bakker (this volume) discusses grammaticalization of lexical elements into markers of TAM in Creole and a typology of affix order that differs from the prototypical order where mood precedes tense and aspect (Bybee 1985) since the generalised affix order in Creole languages is tense-mood-aspect. In Creole languages tense connects an event in relation to a context, while mood has scope over the proposition and aspect over the verb (Givón 1982). Bakker argues that this could seem to be a more “natural” affix order, though on the other hand Creoles seem to be rather conservative. In most other respects the sources for the TAM elements are similar to those found in other languages. He demonstrates how the well known lexical sources of the grammaticalized elements of Creole languages can be used in the reconstruction of historical processes both in languages in general and in establishing the historical connection between different Creoles. One further advantage that Creoles have in the historical reconstruction of languages is that being fairly young linguistic varieties their TAM markers have uncontested etymologies, a fact that should make the typological analysis clearer, especially since they seem to contain a “minimal” grammatical system where TAM markers are expressed by a limited number of preverbal elements. This affix order as well as grammaticalization turn them into some kind of “test tube” languages. The function of the PIE verb suffix *-sk’e/o- has become a central problem in the study of aspect and tense in ancient IE. Cf. Jasanoff (2003: 133–134) who states that productive elaborations of *-sk’é/o- include the Latin inchoative presence in -ēscō, the Greek (Ionic) iterative imperfects in -σκε-, Hittite iteratives in -ske/a- and the Tocharian B causatives in -sk-. The development that led to this situation was analyzed by Ruijgh (2004: 49–50) who described the situation in the following way: IE had a stative suffix *-eh1. Ingressive -eh1-s- and inchoative -eh1-sk’é/ó- were formed by means of additional suffixation. The inchoative meaning was lost in Greek except for one verb aldēskei ‘it is becoming fullgrown’. The suffix -s formed a telic stem. Greek depso was the only remaining present of this kind. After the PIE period -s could become the morpheme of the aorist. Stative + telic gave ingressive meaning (cf. ekrátēsa ). Hittite -esk- seems to be a near equivalent to fientive -es. Verbs such as Greek gignosco and Latin gnosco refer to a process step by step leading up to a resultant state (cf. aldēskei, rubescit). baskō ‘I am going step by step’ explains the iterative meaning found in Hittite and Greek preterit forms like histaskon, staskon, formed from present and aorist stems. Latin and Greek are the topics of three contributions to this volume. Haug follows the changes of aspectual oppositions from prehistoric times to Latin with its perfectum and imperfectum. Reconstructed Early and Late Proto-IE and Proto-Italic are used to illustrate the development of the aspectual system. He further discusses the PIE background of Latin and Italic and observes that imperfective stems were obtained by grammaticalized derivation with several different suffixes. *-sk’e/o- was one of those. Haug states that PIE aspect was grammaticalized aktionsart as in modern Slavic, but was obtained by suffixation alone. Aspectual stems were in focus, and various stem

Introduction

formants had phasal meanings. The present stem became the imperfective at a time when a regular opposition between two preterits gave rise to the inflectional category of aspect. The coexistence of forms such as *g’n®h3‑ and *g’n®h3‑sk’e/o- gave rise to the category of aspect as derivation became grammaticalized. Sanskrit prcchati and Latin posco show that the consuetudinal and iterative meaning of the *-sk’e/o- form could cause its use as a characterized present. Haverling (this volume) analyses the development of actionality, tense and viewpoint from Early to Late Latin and how the earlier actional opposition which was expressed by the help of prefixes, was replaced by other means, such as the use of definite and indefinite articles or the viewpoint opposition between past tense and imperfect tense in the Romance languages. In an earlier study Haverling (2000) showed that the Latin unprefixed -sc-verbs designate activities and processes of on-going nature and gradual action. When the “inchoative-progressive” -sc-verbs were prefixed by conthey were dynamic and simultaneously telic. Gerö and Ruge study the continuity of function of imperfect and future throughout the history of Greek: The imperfect is understood as a marked imperfective form. The synthetic future is seen as aspectually neutral. The Latin viewpoint imperfect which describes a situation from within and can indicate progressivity is used for backgrounding (cf. also Haverling 2000 and this volume). The Greek imperfect is used for the same purpose (cf. Rijksbaron 1988). According to Gerö and Ruge (this volume) Greek imperfective is a marked form which is used in order to indicate a backgrounded event. The concept of interval is discussed by Gerö and Ruge with a reference to Cowper (1999: 218) who stresses that languages choose either moment (perfective) or interval. In the narrative style of Middle Hittite the present tense of progressive -ske/a -verbs was used in tense switching in order to express backgrounding and interval (cf. Dahlgren this volume and Josephson 2006 and this volume). In Josephson’s treatment of Hittite actionality and aspect synchrony is involved as well as five centuries of attested diachronic development of the language. In that contribution, as also in Haverling’s treatment of Latin actionality in the present volume, there is a discussion of the function of PIE *-sk’e/o- and of questions of imperfectivity, progressivity and distributivity. Josephson furthermore argues that tense as used in Hittite tense switching can express aspect, and that actionality and aspect can be expressed by some of the Wackernagel clitics. The nature of the process that caused a grammaticalization of the PIE *-sk’e/o-form can be understood by an analysis of form and function of the Hittite -ske/a-verbs. Josephson raises the question if a grammaticalization of the -ske/a- form had been completed in Hittite. The language did not possess an imperfect tense but had a general past tense and temporal distinctions were thus similar to those of Italic described by Haug. The preterit of a telic stem was functionally perfective by implicature as in PIE. Using the terminology of Johanson (2000) Josephson proposes that the Hittite -ske/a- verbs were +INTRA LOW FOCAL progressives with frequentative (consuetudinal) and iterative meaning. The employment of those verbs shows a remarkable





Interdependence of Diachronic and Synchronic Analyses

similarity to that of the -ske/o- verbs of Homeric Greek. In Hittite, however, there was a proliferation of present tense forms of -ske/a-verbs which clearly behaved as progressives. A combination of progressive form and present tense is found in tense switching, which is common in narrative texts of the Middle Hittite period (cf. Josephson 2006). This fact illustrates the statement of Haug (this volume) that “progressives should be found in the present tense”. Hittite verbs with suffixed -nn- have a similar, but not identical, meaning. They are not iterative. The action is seen as a whole and the -nn- form seems to be a +INTRA HIGH FOCAL progressive (cf. Josephson 2006). They can refer to ongoing motion as iyanna- ‘march’, ‘proceed’ or to violent or intensive activity as hattanna- (cf. Jasanoff 2003: 122). The action can be ingressive in verbs of motion as in the case of iyanna“start marching”. The -nn- verbs had probably been lexicalized. The suffix -ske/a- can be added to the -nn-verb. According to Jasanoff (2003: 193–194) PIE completive stems with -s had developed to anteriors, to witness the -s endings of Hittite. In PIE, the preterit of telic roots was understood as perfective by implicature. Cf. the similar view of Haug (this volume). This was in fact the situation in Hittite which did not have an aorist. For a different reason Hittite was thus similar to Italic, which had lost the PIE imperfect and whose “perfective past was the general past” as stated by Haug (this volume). Haug explains that, in Italic, expressions of temporal distinctions became necessary for secondary verbs. This caused an extension of the use of -s-completives to anteriors and then to perfectives in the case of all semantic types. The distribution of resultatives (completives) widened as new anteriors developed, to witness the Sabellian forms suffixed by -tt-. A new Latin imperfect developed from constructions such as *agents βwāt ‘he was an agent’ (with habitual meaning). Consequently, aspect could be reintroduced. The relation in Russian and other Slavic languages between “(aspectual) perfectivity” and “(lexical) telicity – an Aktionsart sometimes also referred to as ‘perfectivity’” (Abraham 2004: 106–107) has become a highly debated issue. In a contrastive study, Grønn (this volume) discusses viewpoint aspect in Russian and French and concentrates on the nature of imperfectivity and problems of markedness that he analyses formally using Optimality Theory which enables him to give an invariance definition of perfectivity to which he contrasts the different usages of unmarked imperfectivity in Russian while French imparfait is marked. He argues that the apparent similarities between imperfective viewpoint in the two languages are accidental. French imparfait thus gets a uniform interpretation and has the meaning of contemporaneity. Diachronically, there is a pressure towards a binary opposition in Russian that develops a derivational aspectual category Perfective/Imperfective, whereas French partly retains the original inflectional category. The use of Optimality Theory presented in Grønn’s analysis provides a better understanding of TAM in Russian and French at the same time as it points out that accidental similarity has to be considered in typological research, and that accidental parallelisms also add to our knowledge of universal typology.

Introduction

In a contrastive study of predicative expressions of verbs of transition in Portuguese and Spanish which have mostly been neglected in monolingual descriptions of Romance syntax Söhrman and Nilsson (this volume) are concerned with aspect, aktionsart and tense and how an act of transition is carried out in languages where Latin stare and essere have developed into two parallel copula verbs thus creating a system with individual norm and general norm (Falk 1979). The many verbal expressions of this act of semantic transition can be divided into two cognitive categories: resultative and transformational. Germanic languages (especially Scandinavian languages) tend to use very few verbs to express the semantic value of transition, while Romance and Slavic languages among others need a greater variety. These languages often use many verbs with an ingressive value. This is also sometimes the case in Scandinavian languages as for instance Norwegian mørkne with the ingressive suffix -ne. However, in the case of Romance languages both French and Italian turn out to be slightly closer to the Germanic usage of fewer verbs of transition as there are verbs such as devenir/e/. In Latin there are inchoative verbs of transition, but fieri played to a large extent the same role as Scandinavian bli. This means that there are considerable typological differences between languages of the same linguistic family. Although Spanish and Portuguese share many linguistic structures it turns out that Spanish has a more complex system and that this polysemantic field to a great extent also concerns aspect and to a lesser extent tense. The prefix con- plays an important role in Haverling’s research on viewpoint aspect and tense in Latin (Haverling 2000 and this volume). In Hittite, the clitics -kan and -san, which commonly occur in Wackernagel’s position, function in relation to the verb very much like Latin prefixed con- and Greek sun- (cf. Josephson this volume). The semantics of Gothic verbs prefixed by ga- confirm this use of Hittite -kan as both clitics show a tendency towards perfective meaning. Rijkhoff (this volume) also discusses this particle and stresses that Latin con- and Gothic ga- form a collective perfective and have a quantized realisation. Italic com has the meaning of coming together in one final point as in Osc.-Umbr. comenei and Latin comitium. Greek koinós has the sense of ‘gemeinsam’ (cf. Untermann 2000: 414). Sabellian com also functions as preposition and as postposition accompanied by the ablative case. Similarly, Hittite -kan frequently occurs with the ablative case, but is also commonly found with a locative case form. Hittite provides an important insight into the original deictic and directive functions of -kan and -san (Josephson 1995, 1997 and this volume). Differently from -kan, which occurs with locative and ablative, -san can be combined with allative or dative (Josephson 1995 and this volume). In later diachronic levels of Hittite, -san was gradually replaced by -kan. The opposite development occurred in Latin where the preverb con- in its aktionsart meaning was replaced by ad- (and in-), as observed by Haverling (2000). Umbrian ad had a meaning close to that of com (cf. Untermann 2000: 409). The Hittite clitic “sentence particles” indicate direction and location and also affect the actionality of the verb and possibly aspect (Josephson 1972 and later publications).





Interdependence of Diachronic and Synchronic Analyses

The fact that these clitics are mostly found in Wackernagel’s position adds one more typological dimension which has been discussed intensively in several recently published studies of Wackernagel clitics in Romance, Slavic and other languages (Gerlach 2002, Beukema & den Dikken 2000, Heggie & Ordóñez 2005). The great majority of the Wackernagel clitics in Slavic and Romance are pronouns whereas in Hittite several Wackernagel clitics have other functions (cf. Josephson this volume). The behaviour of Wackernagel clitics in Vedic, Avestan and Hittite was described by Hale (1987a, 1987b) who suggested that those that function as sentence clitics were the first to be placed in Wackernagel position and that pronominal clitics were moved into that position by a later process. Regarding the order of Wackernagel clitics in ancient IE languages Garrett (1996) observed that the system of Wackernagel Location clitics (WL) in Hittite is different from that of Greek and Sanskrit, which have syntactic clitics, and that it may have developed into a system of second position affixation. He says that there are phrasal affix systems in which second-position affixation could be sensitive to lexical featuring of their verb. Syntactically projected elements were diachronically reinterpreted as morphological affixes (Garrett 1996: 123). In Hittite WL was transformed into a phrasal affixation system, in which second-position affixation of that kind was sensitive to lexical features of their verb (Garrett 1996: 126). Garrett (1996: 102) made the important observation that “Wackernagel’s Law in Hittite targets direct objects, certain oblique elements, and unaccusative subjects but not unergative or transitive subjects”. Unaccusative subjects are underlying objects. In Hittite there was grammaticalization of WL. Grammaticalization allowed subject clitics to mark a distinction that otherwise surfaced only in the perfect construction with es- ‘be’ and har(k)- ‘have’. According to Legendre (2000: 219–220), who cites Klavans (1985), verbal clitics are best analyzed as phrasal affixes. The Romanian verbal clitics show clustering and rigid order which “results from the same violable alignment constraints that operate in South Slavic Wackernagel languages”. The approach of Legendre (2000), who favors Optimality Theory as do several other scholars in recent literature on clitics in Romance and Slavic languages, seems to be appropriate for explaining the behaviour of those Hittite clitics that belong to verbs, have a rigid order and are moved into a wellordered chain in Wackernagel position (cf. Josephson this volume). The necessity to know the order of elements in order to determine meaning was recently stressed by Heggie & Ordóñez (2005: 1–2). They refer to Hopper & Traugott (1993: 7) in regard to “clines” such as content item>grammatical word>clitic>inflectional affix. It should be observed that the Hittite clitics -kan, -san, -(a)sta -(a)pa and -an occupy the final position in the Wackernagel chain and that the function of those clitics when combined with local adverbs and case forms give rise to a comparison with the role of etymologically related local adverbs/preverbs in Germanic, Slavic, Latin, Sabellian and Greek. Of the five Hittite clitics -kan and -san are especially relevant to

Introduction

actionality and aspect and are obviously sensitive to the meaning of the verb. They are also found in positions outside of the Wackernagel chain (cf. Josephson this volume). In an article in which the productivity of a combination of philology and linguistic analysis is demonstrated in a most convincing way, Ottosson (this volume) shows that what he understands as the “regrammation” of the Old Nordic middle voice is a very early phenomenon which falls within the Proto-Nordic period. He shows that the reflexive element was an affix in a period preceding Old Nordic and proposes that the cliticisation is connected with case syncretism. Diachrony and reconstructive method are important in his treatment of Middle Voice semantics and grammaticalisation. Ottosson’ s study illustrates the problem of Reflexive and Middle in IE languages generally (cf. Josephson 2003 and this volume). The diachronic dimension of the problem of Clitic and Affix is treated by Ottosson with several references to earlier discussion. It should be added that this discussion was recently enriched by Heggie and Ordóñez (2005) in the introductory article to their volume in which they suggest a development word>clitic>affix and that diachronic development of a different kind was proposed by Garrett (1996) in his explanation of Wackernagel’s Law clitics in Hittite as a system of second-position affixation. In a recent review of Hetzron (1997) Shlomo Izre’el (2002) treats the question of tense and aspect in Hebrew with reference to the views of various scholars and summarizes the difficulties of the problem by citing from a lecture by Chaim Rabin: “Semitic has either aspects that express tenses or tenses that express aspects”. In his review, Izre’el illustrates the aspectual view of the question by a citation from Dahl (1985: 79): “The marked feature of a form would be aspectual, yet this form may have a regular implicature of tense especially when a perfective form tends to be used for the past tense”. The tense dominant view was recently advocated by Fehri (2004). The text-linguistically oriented approach of Nyberg (1952) which was “based on action versus state and description” was brought to attention by Eskhult (2005) who states that Nyberg, in spite of giving preference to the temporal comprehension of the verbal system, nevertheless adopts the subjective, socio-psychological aspectual view of J. Pedersen (1926): “The temporal relation pointed out, then depends on the situation that is narrated, or the speaker’s mental attitude towards his utterance”. The long and intense debate concerning tense and aspect in Hebrew and Arabic from its beginning to the present day is presented by Dahlgren in his contribution to the present volume. He suggests that Hebrew has a fundamental relative tense system based on foregrounding and backgrounding in preference to the commonly held opinion which advocates a fundamental aspectual system. Questions of markedness are discussed as well as deictic points of time reference in narrative prose. Dahlgren’s discussion of foregrounding, backgrounding and narrative tense switching in Hebrew has an affinity to questions discussed by Gerö & Ruge and Josephson (this volume). As a typological parallel it can be added that Berman and Neeman (1994: 292–295) treated tense shifting in modern Hebrew narration and proposed (p. 292) “that switching back and forth between present and past tense forms provides Hebrew





Interdependence of Diachronic and Synchronic Analyses

narration with means for signalling aspectual distinctions of a more global, discoursemotivated nature”. They observe (n.4) that “this differs markedly from reliance on formal distinctions of grammatical aspects marked in Biblical Hebrew between sequentially ordered foreground events in the developing narrative compared with background states or simultaneous and anterior events marked by different verb-forms”. The grammar of the Bantu language Kerebe was described by Hurel in 1909. The data of his work reflects the result of Catholic missionaries’ experience of the language during a long period at the end of the 19th century. Thornell (this volume) has verified the data by consulting a present-day mother tongue speaker of standard Kerebe. The TAM system of the Kerebe verb includes five moods, three tenses, three aspects, immediate, recent and remote past and a definite and an indefinite past. The speaker’s reference time and the order of events are taken into account as regards future time. There are subtle distinctions in the past perfective of conditionals. Subject concord markers and TAM markers precede the verbal base and a final vowel expresses aspect and mood. Aspect is intertwined with tense and the present tense expresses the narrative or the habitual. The three past tenses distinguish between recent and remote events. The complicated morphological features form the focus of the description. The lack of data in the earlier descriptions has been remedied by Thornell who promises further fieldwork that will take into account semantic aspects of the TAM system and have a focus on tones. In his article on the comparative TAM morphology of Niger-Congo languages Maho (this volume) treats of Persistive markers in Bantu. His treatment of “frequentative-progressive-persistive” actions is akin to the discussion of similar habitual and continuative functions of the IE -sk’e/o- suffix in Latin and Hittite in other contributions to the volume. Maho’s article has an important diachronic dimension, which is not uncommon in Bantu studies. He shows that present-day persistives were not originally an inflectional category, but an auxiliary construction coding imperfectives and/ or progressives. Maho compares different Bantu languages in order to show structural similarities in the field of Persistive markers. The paper also contains a discussion of problems of cognacy. McGregor, in his contribution to this volume, proposes a diachronic explanation for case markers in Australian languages. He understands them as having originated from indexical elements in apposition with a NP. They subsequently lost their independent status and deictic value. However, there are no proofs that the indexicals have ever passed through an intermediate state as definite articles, which in many languages is considered as a common grammaticalization target. Another interesting development in several Australian languages is that these indexicals (pronominal or spatialdeictic) developed into genitive and ergative and not accusative, although there are some languages where this happened. There is also a pragmatic element incorporated in the new case-markers, i.e. the status of the referent as definitite. The cognitive process involved is called semioticisation. Even if the indexical component is lost, the grammaticalization of the case-markers has led to the establishment of a new morphological,

Introduction

semantic and pragmatic component. From a typological point of view this shows an alternative to more widespread linguistic changes which opens a universal perspective of the consequences of grammaticalization. In order to discuss these changes McGregor deals with many problems of prehistoric reconstruction of Australian languages. Differential object marking in Sahidic Coptic is the subject of the contribution by Engsheden. The main factors that govern the distribution of forms are referentiality and thematicity in agreement with patterns that are attested cross-linguistically. Definiteness alone does not trigger the use of the marker n but referentiality by itself does not account for cases without it. No noun or pronoun without thematic weight has access to the marker. There is no link to animacy. An historical explanation is attempted which implies that the word order change to SVO that took place in the Old Kingdom in the case of the progressive was followed by the disappearance of the copula, which made the present underspecified. n- subsequently spread to other tenses and SVO became the only possible word order. Differential object marking arose in response to word order changes in order to ensure a a correct distinction between thematic object and thematic subject. Differential object marking operates only where there is a an opposition ref/non-ref or them/non-them and where there is a connection with high transitivy. Where transitivity is low a pronominal infinitive is used with the object directly attached. In this volume the comparative analysis of IE languages gives rise to universal typological questions, which are further illustrated by studies on similar problems in Semitic, Hamitic, Bantu, Australian and Creole languages. The book shows the value of a diachronic dimension in providing a better comprehension of linguistic similarities.

References Abraham, W. 2004. VP-internal subjects as ‘unaccusatives’: Burzio’s ‘object account’ vs. the ‘perfectivity account’. In A. ter Meulen & W. Abraham (eds.), The Composition of Meaning, From lexeme to discourse, 83–111. Amsterdam: John Benjamins. Berman, R. A. & Neeman, Y. 1994. Development of linguistic forms: Hebrew. In R.A. Berman & D.I. Slobin (eds.), Relating Events in Narrative, 285–328. Hillsdale NJ: Lawrence Erlbaum Associates. Beukema, F. & den Dikken, M. (eds.). 2000. Clitic Phenomena in European Languages. Amsterdam: John Benjamins. Bybee, J. L. 1985. Morphology. A study of the relation between meaning and form. Amsterdam: John Benjamins. Cowper, E. 1999. Grammatical aspect in English. The Linguistic Review 16: 205–226. Dahl, Ö. 1985. Tense and Aspect Systems. Oxford: Blackwell. Eskhult, M. 2005. A forthcoming English translation of H.S. Nyberg’s Hebreisk grammatik. Orientalia Suecana LIV: 49–55. Falk, J. 1979. Visión de ‘norma general’ versus visión de ’norma individual’. Studia Neophilologica 51: 275–293.





Interdependence of Diachronic and Synchronic Analyses Fehri, A. 2004. Temporal/aspectual interaction. In J. Guéron & J. Lecarme (eds.), The Syntax of Time, 235–257. Cambridge MA: The MIT Press. Garrett, A. 1996. Wackernagel’s law and unaccusativity in Hittite. In A.L. Halpern, & A.M. Zwicki (eds.), Approaching Second, 85–133. Stanford CA: GSLI. Gerlach, B. 2002. Clitics between Syntax and Lexicon. Amsterdam: John Benjamins. Givón, T. 1982. Tense-aspect-modality: The Creole prototype and beyond. In P.J.Hopper (ed.), Tense-Aspect: Between semantics & pragmatics, 115–163. Amsterdam: John Benjamins. Hale, M. R. 1987a. Studies in the Comparative Syntax of the Oldest Indo-Iranian Languages. PhD dissertation, Harvard University. Hale, M. R. 1987b. Notes on Wackernagel’s law in the language of the Rigveda. In C. Watkins (ed.), Studies in the Memory of Warren Cowgill (1929–1985). Untersuchungen zur indogermanischen Sprach- und Kulturwissenschaft, Vol. 3: 38–50. Berlin: de Gruyter. Haverling, G. 2000. On Sco-verbs, Prefixes and Semantic Functions. A Study in the Development of Prefixed and Unprefixed Verbs from Early to Late Latin [Studia Graeca et Latina Gotho burgensia LXIV]. Göteborg. Hetzron, R. (ed.), 1997. The Semitic Languages. London: Routledge. Heggie, L. & Ordóñez, F. (eds.). 2005. Clitic and Affix Combinations. Amsterdam: John Benjamins. Hopper, B. & Traugott, E. 1993. Grammaticalization. Cambridge: CUP. Hurel, E. 1909 La langue kikerewe: Essai de grammaire. Mitteilungen des Seminars für Orientalische Sprachen 12: 1–113. Izre’el, S. 2002. Review of R. Hetzron: The Semitic Languages. Israel Oriental Studies XX: Semitic Linguistics – The state of the art at the turn of the 21st century 501–510. Jasanoff, J. H. 2003. Hittite and the Indo-European Verb. Oxford: OUP. Johanson, L. 2000. Viewpoint operators in European Languages. In Ö. Dahl (ed.), Tense and Aspect in the Languages of Europe, 27–187. Berlin: Mouton de Gruyter. Josephson, F. 1972. The Function of the Sentence Particles in Old and Middle Hittite. Uppsala: Skriv Service. Josephson, F. 1995. Directionality in Hittite. In W. Smoczyński (ed.), Kuryłowicz Memorial Volume. Part One, 165–176. Cracow: Universitas. Josephson, F. 1997. Review of J. Boley: The Sentence Particles and the Place Words in Old and Middle Hittite. Bibliotheca Orientalis LIV(1/2): 155–160. Josephson, F. 2003. The Hittite reflexive construction in a typological perspective. In B.L.M. Bauer & G.-J. Pinault (eds.), Language in Time and Space. A Festschrift for Werner Winter on the occasion of his 80th birthday, 211–232. Berlin: Mouton de Gruyter. Josephson, F. 2006. Début d’un emploi créatif de la langue Hittite. In G.-J. Pinault & D. Petit (eds.), La Langue poétique indo-européenne. Actes du colloque de travail de la Société des Études Indo-européennes (Indogermanische Gesellschaft/Society for Indo-European Studies). Paris, 22–24 octobre 2003. [Collection linguistique de la Société de linguistique de Paris 91],141–156. Leuven: Peeters. Klavans, J. L. 1985. The independence of of syntax and phonology in cliticization. Language 61: 91–120. Legendre, G. 2000. Positioning Romanian verbal clitics at PF. An optimality-theoretic analysis. In B. Gerlach & J. Grijzenhout (eds.), Clitics in Phonology, Morphology and Syntax, 219–254. Amsterdam: John Benjamins.

Introduction Nishida, C. 1994. The Spanish reflexive clitic pronoun se as a situation class marker. Linguistics 32: 425–38. Nyberg, H.S. 1952. Hebreisk Grammatik. Stockholm: Almquist & Wiksell. Pedersen, J. 1926. Hebraeisk grammatik. København: V. Pios Boghandel. Rijksbaron, A. 1988. The discourse function of the imperfect. In A. Rijksbaron, H.A. Mulder & C.G. Wacker (eds.), In the Footsteps of Raphael Kühner. Proceedings from the International Commemoration of the 150th anniversary of the publication of Raphael Kühner’s Ausführliche Grammatik der griechischen Sprache, II. Theil: Syntaxe. Amsterdam 1986, 237–254. Amsterdam: Gieben. Ruijgh, C.J. 2004. The Stative Value of the PIE Verbal Suffix *-éh1-. In J.H.W. Penney (ed.), IndoEuropean Perspectives. Studies in honour of Anna Morpurgo Davies, 48–64. Oxford: OUP. Untermann, J. 2000. Wörterbuch des Oskisch-Umbrischen. Heidelberg: C.Winter.



Synchronic and diachronic evidence for parallels between noun phrases and sentences Jan Rijkhoff It has been shown that, up to a point, noun phrases and clauses have the same underlying structure: they share the same kind of ‘layered’ organization and accommodate the same kind of semantic modifier categories (Rijkhoff 2008a). This article presents synchronic and diachronic evidence to substantiate the claim that there are parallels between the layered analysis of the noun phrase and the clause. It will be argued that the layered NP/clause model is supported by two largely unidirectional historical developments in language, one metaphorical in nature (‘from space to time’) the other metonymic (‘from inner to outer layer’).

1. Introduction1 Both formal and functional linguists have developed models in which clauses and noun phrases (henceforth NPs) can at least partly be analyzed in a similar fashion. But whereas purely formal approaches to grammar have concentrated on similarities between the (underlying) syntactic structure of the NP and the clause (e.g. Jackendoff 1977, Abney 1987), functional theories such as Functional Grammar or Functional Discourse Grammar have investigated parallels between the underlying semantic structure of the NP and clause (e.g. Rijkhoff 1988, 1992, 2002, 2008a-b). This article has two main goals: [a] to present a revised and expanded version of the layered (semantic) NP/clause model presented in Rijkhoff (2002), and [b] to give synchronic and diachronic evidence in favour of this particular model. 1. Abbreviations: 1 = first person, 3 = third person, A = adjective, ABS = absolutive, ASP = aspect, C1 = noun class 1, CLF = classifier, CN = connector, DEF = definite, dem = demonstrative, DET = determiner, DS = different subject, ES = ergative suffix, ERG = ergative, EXH = exhortative, F = feminine, FUT = future tense, GENR = general aspect-tense-mood marker, IMP = imperative, IMPF = imperfective, INDEF = indefinite, INF = infinitive, INGR = ingressive, IRR = irrealis, LOC = locative, N = noun, NEG = negative, NP = noun phrase, num = numeral, PAST = past tense, PL = plural, POSS = possessive, PROH = prohibitive, PRT = participle, Rel. cl = relative clause, SG = singular, SIM = simultaneous, SS = same subject, SUPP = support verb, V = verb.



Jan Rijkhoff

2. Parallels between the semantic (layered) representation of the NP and the clause There are at least three functional approaches to grammar that make use of so-called layered structures in the representation of linguistic constructions: Halliday’s Systemic Functional Grammar (Halliday 2004), Van Valin’s Role and Reference Grammar (Van Valin and LaPolla 1997), and Dik’s Functional Grammar (Dik 1997) as well as its successor Functional Discourse Grammar (Hengeveld and Mackenzie 2006, 2008). Here I will only be concerned with the layering in Functional Grammar and Functional Discourse Grammar, henceforth collectively referred to as F(D)G.2 Layering was introduced in ‘standard’ FG in 1989, when Dik adopted Hengeveld’s (1988) ideas concerning the scoping of categories such as illocution (interrogative, declarative etc.), mood, tense, and aspect. Inspired by work in other functional approaches to grammar, Hengeveld (1989) proposed a new model of the clause, which consisted of hierarchically ordered layers, each accommodating its own set of grammatical and lexical modifiers, called operators and satellites (i.e. non-arguments or adjuncts) respectively in F[D]G. His model only applied to clauses, but subsequently Rijkhoff (1988, 1990) also proposed a layered model of the NP, arguing that – up to a point – the underlying semantic structure of NPs and sentences is basically identical. The most recent version of this NP/clause model (Rijkhoff 2008 a-b) is represented in Figure 1 below, in which operators and satellites are distributed over two levels in the FDG model: the Interpersonal Level and the Representational Level. Grammatical modifier categories of the NP (e.g. definiteness, number) are symbolized by ω/Ω; grammatical modifier categories of the clause (e.g. mood, tense, aspect) are symbolized by π/Π. Lexical modifiers in the NP are represented by the symbol τ/Τ (adjectives, relative clauses, etc.) and in the clause by σ/Σ (adverbs, adverbials); subscripts 0 to 6 indicate the layer at which modifiers are specified. The layers are organized hierarchically, which means that modifiers that are specified in the highest layer of the clause (illocutionary modifiers Π6, Σ6) have the widest scope (the whole message), whereas the scope of clausal modifiers represented at the lowest layer (classifying modifiers π0, σ0), which indicate what kind of event is being referred to, is restricted to the main predicate (typically a verb). By definition, members of a grammatical modifier category (i.e. operators) constitute a smallish, closed set of items, capturing only a limited number of crucial, absolute distinctions (Dik 1997: 160). For example, if Definiteness is a grammatical category in a language, the only two available choices are +Definite or –Definite (Indefinite). Qualifying operators (ω1, π1) are not deemed to exist (Rijkhoff 2008a). Since qualities are typically gradable rather than absolute properties, they can only be expressed through lexical modifiers: a house can be ‘rather bigA’ but not ‘rather singular’ or ‘rather definite’. 2. See Butler (2003: 239-246) for a comparison of layering in Systemic Functional Grammar, Role and Reference Grammar and Functional (Discourse) Grammar.

Synchronic and diachronic evidence for parallels between noun phrases and sentences

INTERPERSONAL LEVEL (‘language as exchange’). At the Interpersonal Level, modifiers are concerned with the interpersonal status of four kinds of entities in the World of Discourse: [i] clauses (or rather the messages contained in the clauses), [ii] propositions, [iii] events and [iv] things. MODIFIERS IN THE CLAUSE 6. Illocutionary (Π6, Σ6) S informs A about the illocutionary status of message Ei (Decl, Int, Imp).

MODIFIERS IN THE NP 4. Discourse-Ref. (Ω4, Τ4)

5. Propositional (Π5, Σ5) A is informed about S’s personal assessment of / attitude towards proposition Xi as regards the probability, possibility or desirability of the actual occurrence of event ei. 4. Discourse-Ref. (Π4, Σ4)

scope increase

REPRESENTATIONAL LEVEL (‘language as carrier of content’). At the Representational Level, modifiers specify properties of spatiotemporal entities (things xi, events ei) in the World of Discourse in terms of the notions Kind (Class), Quality, Quantity, and Location. 3. 2. 1. 0.

Localizing Quantifying Qualifying Classifying

(ω3, τ3) (ω2, τ2) (τ1) (ω0, τ0)

3. Localizing (π3, σ3) 2. Quantifying (π2, σ2) (σ1) 1. Qualifying 0. Classifying (π0, σ0)

Figure 1. Layers of modification in the NP and in the clause (Discourse-Ref. = Discourse-Referential)

The classifying layer (‘Kind’) contains the head of the construction (noun, verb) and accommodates modifiers (classifying modifiers) that only relate to the property that is designated by the head, such as nominal or verbal aspect markers. The quality layer accommodates qualifying modifiers, which specify more or less inherent properties of the entity as designated by material in the kind layer, typically qualifying adjectives (if a language has them), and adverb(ial)s of manner, speed or duration. The quantity layer contains the quality layer and accommodates quantifying modifiers, which express number distinctions (singular, plural) or specify the cardinality of the referent. In its turn the quantity layer is contained by the location layer, which accommodates localizing modifiers such as demonstratives, tense markers, or adverb(ial)s such as ‘in the garden’ or ‘yesterday’.





Jan Rijkhoff

those

ω3

two

DOGN

ω2

Quality

Quantity

Location

black

in the garden

τ1

τ3

NP operators

NP satellites

Figure 2. Simplified representation layers of modification in the NP ‘those two black dogs

in the garden’.

In the NP ‘those two black dogs in the garden’ the qualifying modifier ‘black’ only relates to the noun ‘dog’, not the quantity or the location. The quantifying modifier ‘two’ specifies the number of black dog entities, not the number of locations; and both ‘those’ and ‘in the garden’ specify the location of the dog entities with all their qualitative and quantitative properties (Figure 2; notice that this example has no classifying or discourse-referential modifiers).3 It is important to emphasize there is no one-to-one correlation between the form and the function of a modifier; this holds in particular for lexical modifiers (σ/Σ, τ/Τ). For example, an adnominal prepositional phrase (PP) can occur as a localizing satellite (τ3), a quantifying satellite (τ2), or as a qualifying satellite (τ1) in the NP (Rijkhoff forthcoming).

2.1

Two levels: Interpersonal and representational

Figure 1 shows that modifiers in the NP and in the clause can be divided into two major subcategories: interpersonal and representational (or descriptive) modifiers (Hengeveld and Mackenzie 2008). This division reflects the double function of NPs and clauses (Rijkhoff 2002: 231): they provide a physical description of a thing or an event and at the same time they are the constructions used to refer to entities in the world of discourse created by the interlocutors. The two levels are briefly discussed in turn below.

3. The reader is referred to Dik (1997) and Hengeveld and Mackenzie (2008) for an account of the way the layered underlying structure is turned into an actual linguistic expression in FG and FDG respectively.

Synchronic and diachronic evidence for parallels between noun phrases and sentences 

2.1.1 Modifiers at the Interpersonal Level The Interpersonal Level is regarded as that part of the grammar that is concerned with ‘language as exchange’ or ‘language as communicational process’ (Butler 2003: 111; Halliday 2004: 61) and has three layers of modification, accommodating – discourse-referential modifiers for things or events (i.e. first and second order entities), which are specified at the fourth layer; – proposition modifiers for propositions (i.e. third order entities), which are specified at the fifth layer; – illocutionary modifiers for clauses (or rather: the message contained in the clause, a fourth order entity), which are specified at the sixth layer. Grammatical and lexical modifiers (operators and satellites, resp.) at the Interpersonal Level in Figure 1 inform the Addressee about a. the existential status of a spatial or temporal entity (thing xi and event ei) in the world of discourse (more on this below); b. the modal status of proposition Xi, in particular the speaker’s mental or cognitive attitude towards the proposition; at this level we find, for example, grammatical and lexical expressions of subjective modality and evidentiality (Dik 1997: 295–299); c. the illocutionary status of the clause; some relevant operator categories are Declarative, Interrogative, Imperative, Exclamative (on illocutionary satellites, see Dik 1997: 300–307). Since this paper is concerned with parallels between NP and clauses, I will only be concerned with modifiers at the Interpersonal level that are specified at the fourth layer, the so-called ‘discourse-referential modifiers’ in Figure 1 (notice that I confine myself to NPs used to refer to concrete objects or ‘things’, so NPs such as ‘last week’s meeting’, ‘her true love’ or ‘your bad answer’ will be ignored here; see also footnote 5). Discourse-referential modifiers relate to the status of the referent (thing, event) in the world of discourse (Rijkhoff and Seibt 2005; Rijkhoff 2008 a-b) in that they specify whether or not an entity occupies a certain spatio-temporal region in the discourse world. The relevant grammatical modifier categories are Definite/Indefinite and Specific/Non-specific in the NP and Realis/Irrealis in the clause. Adverbs and adjectives such as ‘actually’ and ‘same’ are instances of discourse-referential satellites. The relation between discourse-referential operators is particularly interesting because it is symmetrical in one way and anti-symmetrical in another (with a special role for Specificity).4 The grammatical categories Realis and Definite are similar because they both indicate that the entity, i.e. the event or thing referred to by the speaker, 4. In order to explain symmetrical and anti-symmetrical relations between (ir)realis and (in) definiteness, we need to distinguish between specific and non-specific indefinite reference. Consider the following examples, which show that there are two ways to continue ‘Max wants a dog’ (Karttunen 1976).



Jan Rijkhoff

(already) has a certain location (is ‘grounded’) in the shared world of discourse. By contrast, Irrealis and (non-specific) Indefinite signal that the entity referred to by the speaker does not have a proper location (is not ‘grounded’) in the world of discourse – at least not yet. There are thus two symmetries, one linking the grammatical categories Realis and Definite, the other linking Irrealis and Non-specific Indefinite (Figure 3). NOUN PHRASE (thing)

OCCURRENCE IN WORLD OF DISCOURSE

CLAUSE (event)

DEFINITE

thing or event (already) has a location in the discourse world, i.e. the entity is ‘grounded’

REALIS

NON-SPECIFIC INDEFINITE

thing or event does not have a location in the discourse world (yet), i.e. the entity is not ‘grounded’

IRREALIS

Figure 3. Symmetry between Definite/Realis and Non-specific Indefinite/Irrealis

The idea that irrealis and non-specific indefinite reference are closely related is supported by the fact that in some languages, such as Jacaltec (Mayan), the same marker is used to express these notions (variation is due to vowel harmony):5

Jacaltec (Craig 1977: 93):

(1) (2)

Way-oj ab naj sleep-OJ EXH CLF/he ‘Would that he slept!’ [‑oj: exhortative mood] X–Ø-’oc heb ix say-a’ hun-uj munlabel ASP-ABS.3-start PL woman look_for-FUT a-OJ pot ‘The women started looking for a pot.’ [‑oj marks non-specific reference]

Anti-symmetry between (in)definiteness and (ir)realis has to do with the number of ways that an entity can be definite/indefinite (thing) or actual/non-actual (event). Here

a. Specific-indefinite reference: Max wants a dog. It is black b Nonspecific-indefinite reference: Max wants a dog. It must be black.

The difference is due to the fact that in a. reference is made to a specific dog (‘Max wants a certain dog, which is known to be black’), whereas in b. the speaker does not refer to any particular dog (English has no special article for specific or non-specific reference unlike, for example, many Polynesian or sub-Saharan African languages). 5. For another example, see Du Feu (1987, 1989, 1996) on Rapanui he, which used to mark both ‘indefinite action’ (–tense) and nonspecificity. Nordlinger and Sadler (2004) discuss aspect/ tense/mood marking on nominals.

Synchronic and diachronic evidence for parallels between noun phrases and sentences 

Definite aligns with Irrealis in that both definite things (referents of definite noun phrases) and non-actual events (referents of irrealis clauses) can occur in the world of discourse for many different reasons. Thus, the referent of an NP can be definite, for example, because – it has been mentioned before (anaphoric use)

(3) I just bought a book and a calendar. Surprisingly, the book was much cheaper than the calendar.

– it is available in physical context (situational or deictic use)

(4) Now tell me – what do you see on the monitor?

Similarly, there are many reasons why an event is non-actual, as is shown in these examples from the Papuan language Amele (Roberts 1990: 371–372; see also Elliott 2000).

Amele

(5) Ho bu-busal-eb age qo-qag-an pig SIM-run.out-3SG.DS.IRR 3PL hit-3PL-FUT ‘They will kill the pig as it runs out.’ (6) Ho bu-busal-eb age qu-ig-a pig SIM-run.out-3SG.DS.IRR 3PL hit-3PL-IMP Kill the pig as it runs out!’ (7) Ho bu-busal-eb cain qu-wain pig SIM-run.out-3SG.DS.IRR PROH hit-NEG.F.3PL ‘Don’t kill the pig as it runs out!’ By contrast, there is basically one reason why an NP is (specific) indefinite: because the entity (thing) designated by the indefinite NP has not been properly introduced in the world of discourse – hence the speaker assumes that the addressee does not know (yet) what particular thing is being referred to. There is also only one reason why a sentence is in the realis mood: because the entity (event) designated by the sentence is real, i.e. it has happened (or is happening). Thus anti-symmetry between (in)definiteness and (ir)realis is due to the fact that a. referents of definite NPs and referents of irrealis clauses are part of the world of discourse for many different reasons; b. referents of specific indefinite NPs and referents of realis clauses exist in the world of discourse for one and the same reason: they ‘ground themselves’ in the world of discourse when being referred to for the first time (Figure 4).



Jan Rijkhoff

NOUN PHRASE (thing)

NUMBER OF REASONS FOR AN ENTITY TO BE CLAUSE MARKED AS (IN)DEFINITE OR (IR)REAL (event)

DEFINITE

many

one (grounds itself)

REALIS

SPECIFIC INDEFINITE

one (grounds itself)

many

IRREALIS

Figure 4. Anti-symmetry between Definite/Realis and Specific Indefinite/Irrealis

2.1.2 Modifiers at the Representational Level The Representational Level deals with language ‘as carrier of content’ (Butler 2003: 111). Descriptive modifiers are concerned with properties of the referent of the NP (thing) or clause (event) in terms of the notions Class or Kind (what kind it is), Quality (how it is), Quantity (how much/many it is) and Location (where it is).

2.1.2.1 Localizing modifiers Localizing modifiers specify locative properties of the entity (thing, event) as defined by material in the quantity layer. Grammatical expressions of the notion ‘location’ such as adnominal demonstratives are referred to as localizing operators; a lexical expression such as the adnominal NP ‘on the wall’ in ‘the picture on the wall’ is an example of a localizing satellite. – localizing operators ω3 and π3 Adnominal demonstratives (ω3) are grammatical manifestations of the notion location in the NP. As is the case with other modifier categories, they do not occur in all languages (see e.g. Derbyshire (1979: 131) on Hixkaryana), and in some languages they require the presence of a classifier or a definite article.

Mandarin Chinese (Li and Thompson 1989: 105)

(8) nèi-tiáo niú DEM-CLF cow ‘that cow’ Since tense is “[..] grammaticalised expression of location in time” (Comrie 1985: 9), inflectional tense markers (π3) are obviously instances of localizing operators at the level of the clause. – localizing satellites τ3 and σ3 Recall that there is no one-to-one relationship between the form of a modifier and its place in the layered representation of a linguistic construction. Thus, a localizing satellite in the NP can take the form of an adnominal prepositional phrase such as ‘on this carpet’, as in:

Synchronic and diachronic evidence for parallels between noun phrases and sentences 

(9) The stain on this carpet was difficult to remove.

Restrictive relative clauses and possessive modifiers (‘genitives’) are also typically used as localizing satellites (Rijkhoff 2002: 173–178, 194–211; on the relationship between possession and location, see e.g. Clark 1978 and Heine 1997). (10) Could you give me the book that is lying on that tableRel.cl? (11) I’d like to talk to the father of that boyPoss.NP. It is essential for the localizing satellite to provide a referential anchor for the addressee, i.e. the localizing satellite must contain a reference to an entity that is easily identifiable for the addressee and which makes it possible to locate the referent of the matrix NP. In the examples above the referential anchors are: ‘this carpet’, ‘that table’ and ‘that boy’. They enable the hearer to locate (and identify) the referents of the embedding matrix NPs: ‘the stain’, ‘the book’, and ‘the father’ respectively. Referential anchors also play an important role in presupposition (Rijkhoff 2002: 176, Rijkhoff forthcoming).

2.1.2.2 Quantifying modifiers Quantifying modifiers specify quantitative properties of the referent (thing or event) and relate to all the material contained in the quality layer. Grammatical expressions of the notion ‘quantity’ are quantifying operators; lexical expressions are called quantifying satellites (Rijkhoff 2002: chapter 5). – quantifying operators ω2 and π2 In many languages number distinctions are either optional or altogether absent in the NP (Rijkhoff 2002: 106–119, 146–155), but if nominal number is a relevant category in some language it is commonly expressed by some grammatical element like a nominal affix:

Dutch

(12) fiets-en bicycle-PL ‘bicycles’ In the absence of clear lexical properties, adnominal cardinal numerals are often categorized as ‘form words’ rather than ‘content words’ in many languages (notice that the atoms of a numeral system constitute a closed class; Greenberg 1978b: 256):

Hungarian (Moravcsik 1994: 8)

(13) két lány two girl ‘two girls’ Some languages also employ grammatical means to indicate how often an event takes place. When reference is made to a single occurrence this is called semelfactive aspect



Jan Rijkhoff

and in the case of multiple occurrences this is called iterative, repetitive, or frequentative aspect. This example from Hidatsa (Amerind) contains such a number marker at the level of the clause (glossed as ‘iterative mood’ in the source grammar):

Hidatsa (Matthews 1965: 158)

(14) Wí i hírawe ksa c woman she sleep INGR ITER.MOOD ‘The woman kept falling asleep.’ – quantifying satellites τ2 and σ2 In quite a few languages across the globe cardinal numerals are categorized as lexical elements or they appear as predicates (Rijkhoff 2002: 168–172). For example, in Krongo numerals are categorized as verbs, which appear in the imperfective when they modify a noun (τ2):

Krongo (Reh 1985: 252)

(15) nóo-còorì nk-óotòonò PL-house CN.PL-IMPF:be_three ‘three houses’ In Samoan the numeral appears as the head of a special kind of relative clause introduced by the general tense-aspect-mood marker [GENR] e if the NP has specific reference.

Samoan (Mosel and Hovdhaugen 1992: 318)

(16) Sa fau=sia e Tagaloaalagi fale e tolu ... PAST build=ES ERG Tagaloaalagi house GENR three ... ‘Tagaloaalagi built three houses...’ Adverbs and adverbials are typically employed as lexical modifiers to specify how often an event occurs (σ2), as in: (17) Every once in a while/Sometimes/Rarely our cat catches a mouse.

2.1.2.3 Qualifying modifiers Qualifying modifiers specify more or less inherent properties (qualities) of the referent. Lexical expressions of the notion Quality are referred to as qualifying satellites; recall that there are no qualifying operators (section 2). – qualifying satellites τ1 and σ1 Qualifying satellites are lexical modifiers specifying notions such as size, weight, color, age, and value in the case of things and manner, speed or duration in the case of events. They are typically expressed by adjectives (in the NP) or adverbs (in the clause) – if a language has these lexical categories.

Synchronic and diachronic evidence for parallels between noun phrases and sentences 

(18) a beautifulAdj songN (19) She sangV beautifullyAdv Adjectives and adverbs are, however, not attested in every language. If a language lacks a distinct class of adjectives, it will usually employ qualifying NPs or relative clauses instead, as in the English paraphrases ‘the man with richness’ or ‘the man who is rich’. Speakers of Galela, for example, use a kind of relative clause (headed by a stative verb) and Hausa speakers employ an adnominal NP (headed by an abstract noun). Notice that the first syllable of the attributive verbal predicate in question is reduplicated in Galela, yielding the participial form.

Galela (van Baarda 1908: 35)

(20) awi dohu i lalamo his foot it be_big:PRT ‘his big foot’

Hausa (Schachter 1985: 15)

(21) mutum mai alheri / arzaki / hankali person with kindness / prosperity / intelligence ‘a kind/prosperous/intelligent person’ Wambon is one of the languages that, apart from one or two exceptions, has no distinct class of adverbs. This language employs medial verb constructions to express qualitative notions at the level of the clause (the verb matetmo is derived from the adjective matet ‘good’).

Wambon (de Vries 1989: 49)

(22) Jakhov-e matet-mo ka-lembo? they-CN good-SUPP.SS go-3PL.PAST ‘Did they travel well?’

2.1.2.4 Classifying modifiers Classifying modifiers further specify the kind of entity denoted by the noun or verb and appear at the innermost layer of modification, i.e. between the head constituent and the layer that accommodates qualifying modifiers. – classifying operators ω0 and π0 Classifying operators are grammatical modifiers that (further) specify what kind of spatial or temporal entity is being referred to. Just as verbs are coded for a particular Aktionsart (‘mode of action’: activity, achievement, state etc.), nouns are lexically coded for a particular Seinsart (‘mode of being’: singular object, mass, set, collective etc.; Rijkhoff 1992, 2002). And just as verbal aspects like ‘perfective’ and ‘imperfective’ modify the Aktionsart (or lexicalized aspectuality; Dik 1997: 224f.), nominal aspect



Jan Rijkhoff

markers can change the Seinsart of a noun (Rijkhoff 2002: 100–121). For example, nouns in Oromo are lexically coded for a Seinsart that I have called ‘set’ (meaning that Oromo nouns are transnumeral and can be in a direct construction with a numeral), but when they are provided with a (what I call) collective or an individual aspect marker, they designate a special kind of set, viz. a collective set or a singleton set (with just one member) respectively:

Oromo (Stroomer 1987: 76–77, 84–85)

(23) farda ‘horse/horses’ (unmarked set) vs. fardoollee ‘horses’ (collective set) (24) nama ‘man/men’

(unmarked set) vs. namica ‘a/the man’ (singleton set)

– classifying satellites τ0 and σ0 Classifying satellites in the NP are lexical items that further specify what kind of entity is being denoted by the head noun, for example: annual in ‘annual report’, presidential in ‘presidential election’, electric in ‘electric train’, social in ‘social security’, of sin in ‘house of sin’ (Rijkhoff forthcoming). They typically appear adjacent to the head noun and differ from qualifying satellites in that they do not specify an objective property of the entity (round table, red car) or the speaker’s subjective attitude towards the entity (interesting dissertation, beautiful picture), but rather a particular subclass of the entity in question. In English classifying adjectives differ from qualifying adjectives in that they usually do not admit intensifiers, comparison, or predicative position (Quirk et al. 1985: 1339; on non-predicative adjectives, cf. also e.g. Farsi 1968, Levi 1973; see also Giegerich 2005 on associative adjectives): (25) an electric train

vs. * a very electric train

[intensifier]

(26) a medical examination vs. * a more medical examination [comparison] (27) the presidential election vs. * the election is presidential

[predicative position]

At the level of the clause, so-called ‘stripped nouns’ are good examples of classifying satellites (Miner 1986, 1989; Gerds 1998). Stripped nouns are rather similar to incorporated nouns, but the crucial difference is that a stripped noun is a separate word (according to phonological criteria such as stress placement), which must appear next to the verb. Thus, in Kusaiean adverbs can appear between verb and object (28a) but not between verb and stripped noun (28b).

Kusaiean (Gerds 1998: 94; original example in K. Lee 1975)

(28) a. Sah el twem upac mitmit sac Sah he sharpen diligently knife the ‘Sah is sharpening the knife diligently.’

Synchronic and diachronic evidence for parallels between noun phrases and sentences 

With stripped noun mitmit ‘knife’:

(28) b. Sah el twetwe mitmit upac Sah he sharpen knife diligently ‘Sah is diligently knife-sharpening.’ In these examples a distinction is made between sharpening in general (28a) and a certain kind of sharpening, namely knife-sharpening (28b).

3. Synchronic and diachronic evidence for parallels between the layered structure of NPs and clauses Some synchronic and diachronic evidence for the layered analysis of clausal structures was presented in Hengeveld (1989) and several subsequent publications.6 This section is specifically concerned with evidence in support of parallels between the layered organization of NPs and clauses, the most recent version of which is shown in Figure 1.

3.1

Synchronic evidence

3.1.1 Morpho-syntactic parallels When we consider the relative order of demonstrative (dem), numeral (num), qualifying adjective (A) and noun (N) in languages across the globe, we find that ordering patterns typically reflect the layered (semantic) organization of the noun phrase presented above (Rijkhoff 2004).7 (29) dem num A N Alamblak, Dutch, Georgian, Hungarian, Kayardild, Ket, Nama Hottentot, Imbabura Quechua, Pipil, Tamil, Turkish dem num N A Burushaski, Guaraní (also e.g. French and other Romance languages) 6. For evidence in favour of the layered model of the clause, see Dik et al. (1990) and other articles in Nuyts et al. (eds.) 1990; see also e.g. Dik and Hengeveld (1991), Cuvalay-Haak (1997), Crevels (1999). For an early discussion of variables and scoping in FG, see Vet (1986). 7. Notice that we are only concerned with integral, non-complex NPs that are used to refer to concrete objects (see also section 2.1.1). This excludes, for example –

complex NPs, containing embedded modifiers such as relative clauses (recall that in many languages adjectival or numerical notions are expressed in the form of an embedded modifier);

–

NPs containing bound modifiers (sometimes modifiers are expressed as clitics, affixes, incorporated forms or part of a compound);

–

appositional forms of modification (in certain languages some or all modifiers may or must be in an appositional relation with the structure containing the head noun; Rijkhoff 2002: 19-23).



Jan Rijkhoff

dem A N num dem N A num num A N dem num N A dem A N num dem N A num dem

Zande Bambara Berbice Dutch Creole, Bislama, Sranan Basque, Hmong Njua Sango Oromo, Fa d’Ambu, Nubi

In all patterns the localizing modifier (dem) is in the periphery and the qualifying adjective (A) immediately precedes of follows the noun, as in the following abstract schema: (30) dem num A N A num dem From a logical point of view there are 16 additional ordering possibilities (N dem A num, num A dem N, etc.), but very few of these patterns have been attested so far. Languages that are deemed to use such ‘non–iconic’ NP internal ordering patterns are discussed in Rijkhoff (2002: 19–23, 273–276, 324–332; 2004), where it is argued that in none of these cases we are actually dealing with simple, integral NPs in which modifiers are expressed as free forms. For example, if adjectives do not appear next to the noun in some language, this is probably because they are actually verbs or nouns heading an embedded phrasal modifier (relative clause or noun phrase; cf. Heine 1980: 182, Dryer 1992: 96). It may be hypothesized that at least one of the following statements is true for the handful of languages that are assumed to display a ‘non-iconic’ word order pattern in the NP with respect to the ordering of demonstrative, numeral, adjective and noun. – so-called adjectives are better categorized as verbs or nouns, i.e. adnominal relative clauses or NPs, turning the NP into a non-simplex construction (examples (20) and (21));8 – numerals are expressed as phrasal modifiers, also turning the NP into a syntactically complex construction (examples (15) and (16)); – modifiers are expressed as bound rather than free elements, which means their expression is a matter of morphology rather than syntax; – modifiers are in apposition (rather than fully integrated constituents). For example, in the Australian language Kalkatungu “there are in fact no noun phrases, but [..] where an argument is represented by more than one word we have nominals in parallel or in apposition. [...] Each word is a constituent of the clause [...]” (Blake 1983: 145; see also Blake 2001); – modifiers are assigned a special pragmatic function like Focus, indicating we are dealing with a marked pattern (Rijkhoff 2002: 272–273). A similar tendency to iconically adhere to the layered organization (albeit less strictly) is attested in ordering patterns at the level of the clause. In English, for example, 8. Apart from languages without modifying adjectives, there are also languages without adnominal demonstratives or numerals as well as languages in which a certain adnominal modifiers never seem to co-occur in the same NP (Rijkhoff 2002: 329f.; Derbyshire 1979: 132; Everett 2005).

Synchronic and diachronic evidence for parallels between noun phrases and sentences 

different kinds of temporal satellites tend to occur in the order time duration (‘for a short while’ = qualifying satellite), time frequency (‘every day or so’ = quantifying satellite) and time position (‘in January’ = localizing satellite), as in this example (Quirk et al. 1985: 551): (31) I was there for a short while every day or so in January Quality (duration) Quantity (frequency) Location (in time) In her study on morphology, Bybee (1985: 196) investigated the ordering of inflectional morphemes relative to the verb in a sample of fifty languages. She found that aspect (π0) occurs closest to the stem, followed by tense (π3), and then by mood (Π4/ Π5). The only exception to this ordering she found in her 50-language sample is in Ojibwa, where the dubitative suffix precedes the preterit suffix. 3.1.2 Isomorphism I: NPs and clauses sharing the same lexical modifier Languages across the globe employ various kinds of parts-of-speech systems (Rijkhoff 2007). Figure 5 shows that there are both qualitative and quantitative differences between the parts-of-speech systems. Languages not only differ in the number of major word classes they employ, but also with respect to the degree of flexibility displayed by the members of a certain word class. For example, Samoan is deemed to have a single, extremely flexible class of contentives (Mosel and Hovdhaugen 1992: 73–74, 77), whereas Hungarian has four distinct lexical word classes: verbs, noun, adjectives and adverbs. For a detailed discussion of parts-of-speech systems, including the intermediate types (not indicated in Figure 5), the reader is referred to Hengeveld et al. (2004) and Hengeveld and Rijkhoff (2005).

Flexible

Contentive

Type 1 Type 2 Type 3 Type 4

Rigid

Type 5 Type 6 Type 7

Non-verb

Verb Verb Verb

Modifier

Noun Noun

Verb

Adjective Noun

Verb

Adverb Adjective Noun

Verb

Figure 5. Parts-of-speech systems (based on Hengeveld 1992; adverb = manner adverb).

In the present context, languages of Type 3 are of special interest, as in these languages nouns and verbs are modified by members of the same word class (called Modifier in



Jan Rijkhoff

Figure 5). Let us first consider an example from English, where manner adverbs are often derived from adjectives by adding the derivational suffix -ly. (32) She recorded a beautifulA songN (33) She sang beautiful -lyAdverb By contrast, languages of Type 3 such as Ngiti have a single word class (Modifier) whose members are all flexible in that they can all be used to modify nouns as well as verbs (Kutsch Lojenga 1994: 336): [In Ngiti] there is no morphological nor a clear syntactic distinction between a class of adjectives and a class of adverbs in Ngiti. The functional term modifiers is therefore used [...] to cover a fairly large grammatical class of words, containing about 150 items, which are neither nouns nor verbs and which all have a modifying function in relation to different constituents.

In other words, in Ngiti and other languages of Type 3 we find that the same element is used as a quality satellite (σ1, τ1) in the NP and in the clause. The same true for Dutch:

Dutch

(34) Ze zong een mooiτ1 lied she sang a beautiful song ‘She sang a beautiful song.’ (35) Ze zong mooi σ1 she sang beautiful(ly) ‘She sang beautifully.’ 3.1.3 Isomorphism II: NPs and clauses sharing the same grammatical modifier In quite a few languages, we also find the same grammatical modifier being used in the clause and in the NP. For example, several studies have remarked on phonological similarities between markers of nominal and verbal plurality, both quantifying operators (Frajzyngier 1977, 1997; Mithun 1988; Newman 1990; Gil 1993). More will be said about this in section on 3.2 on diachronic evidence for layering. It was already mentioned in section 2.1.1 that in Jacaltec discourse-referential operators in the clause (Irrealis) and in the NP (Non-specific Indefinite) are formally identical. Fongbe (a Kwa language mainly spoken in Benin) is an example of a language

Synchronic and diachronic evidence for parallels between noun phrases and sentences 

that employs the same marker for realis and definite reference (glossed as DET, as in the original publication):9

Fongbe (Lefebvre 1998: 94, 99; see also Lefebvre and Brousseau 2002)

(36) N 2ú àsfn f I eat crab DET ‘I ate the crab (in question/that we know of).’ (37) Jan wá f John arrive DET ‘Actually, John arrived.’ Next we will consider some historical evidence in support of the hypothesis that each layer in the underlying structure of the NP has its counterpart in the underlying structure of the clause.

3.2

Diachronic evidence

According to the model in Figure 1, NPs and clauses can be analyzed in a similar fashion in that they share the same kind of layered organization and accommodate the same kind of modifier categories. For some categories there is evidence that clearly points to some diachronic connection between linguistic forms or constructions that are used to talk about spatial entities and those that are used for temporal entities (cf. also the examples from Jacaltec and Fongbe above).10 This section consists of three parts. Section 3.2.2 discusses historical connections between linguistic material relating to spatial and temporal entities (i.e. across linguistic constructions), section 3.2.3 is concerned with diachronic relations between modifier categories in the NP and in the clause (i.e. within linguistic constructions). First, however, historical relationships within and across the spatial and the temporal domain are considered from the perspective of grammaticalization (section 3.2.1), a special form of language change which crucially involves metaphor (a cognitive process based on analogy by transfer of features) and metonymy (a form of reanalysis based on associative relations). It will be argued that historical connections between linguistic material across the space-time boundary (‘from space to time’), discussed in section 3.2.2, are largely due to metaphorical processes, whereas historical relations between grammatical modifiers inside the NP or the clause (‘from inner to outer layer’), the topic of section 3.2.3, are metonymic in nature.

9. The same goes for Haitian Creole (Fongbe is one of the substratum languages of Haitian Creole). 10. Note that this is not the same as saying that there is a diachronic relationship between NPs and clauses (see also below).



Jan Rijkhoff

3.2.1 Grammaticalization, metaphor, and metonymy Grammaticalization is defined as “the change whereby lexical items and constructions come in certain linguistic contexts to serve grammatical functions and, once grammaticalized, continue to develop new grammatical functions” (Hopper and Traugott 2003: xv). To give an example, the French preposition chez ‘with, at’ is historically derived from the (Latin) noun casa ‘house’. The two most significant mechanisms that lead to grammaticalization are reanalysis and analogy (Hopper and Traugott 2003: chapter 3). Reanalysis is a covert, syntagmatic phenomenon and involves the morphological, syntactic, and semantic modification of forms and constructions: “the hearer understands a form to have a structure and a meaning that are different from those of the speaker” (Hopper and Traugott 2003: 50). For example, at some point in the history of French certain nouns were added to reinforce the negative element ne ‘not’ (as in ‘I did not walk/take a[nother] step’, ‘I did not speak a[nother] word’) but eventually these nouns were reanalyzed as being part of a larger negative construction. Subsequently one form (pas, an erstwhile noun meaning ‘step’) ousted all other variants and came to be used with verbs that had nothing to do with moving on foot (cf. Je ne sais pas ‘I don’t know’), regularizing part of the grammar. This last phase, where pas appears with any kind of verb, exemplifies the second mechanism: analogy. Analogy is an overt, paradigmatic phenomenon and refers to “the attraction of extant forms to already existing constructions” (ibid. 64). Reanalysis and analogy are closely associated with two cognitive strategies: metonymy and metaphor respectively (Heine et al. 1991). Metaphor is discontinuous and involves a transfer of features from one conceptual domain to another, more abstract conceptual domain (Heine 2003: 586; see section 3.2.2). This is shown in the following hierarchy of conceptual domains (Heine et al. 1991: 48) (38) Person > Object > Activity > Space > Time > Quality This conceptual shift toward more abstract categories is illustrated by the following examples from Ewe (Heine et al. 1991: 66 – my numbering): In sentence (39), for example, megbé may denote both a body part (‘back’) and the ‘back part’ of an inanimate item, and in (40) it may mean either the ‘back part’ of an object or the ‘space behind’ that object. Sentence (41) again may have either a spatial or a temporal meaning. Finally, sentence (42) has both a temporal and a qualitative significance:

Ewe (Heine et al. 1991: 66)

(39)

megbé keke-á2é le é-sí back broad-INDEF be 3SG.POSS-hand a. ‘He has a broad back.’ b. ‘Its back is broad.’

Synchronic and diachronic evidence for parallels between noun phrases and sentences

(40)

dzra xf-á pé megbé 2ó prepare house-DEF of back ready a. ‘Prepare the back wall of the house!’ b. ‘Prepare the place behind the house!’

(41)

é-le megbé ná-m 3SG-be behind to-1SG a. ‘He is behind me.’ (spatially) b. ‘He is late.’ (= he could not keep pace with me)

(42)

é-tsí megbé 3SG-stay behind a. ‘He remained behind/is late.’ b. ‘He is backward/dull.’

Thus, one could regard the change from ‘collective’ to ‘perfective’ (see section 3.2.2.1 below), or the change from demonstrative pronoun to tense marker (i.e. from spatial to temporal deixis; see section 3.2.2.3), as instances of the metaphorical change from space to time. Notice, however, that we only have good evidence of markers of temporal notions having developed from markers for spatial notions, which is not the same as saying that grammatical modifiers at the level of the clause have developed out of grammatical modifiers in the NP. This is most clearly seen in the case of the localizing operators discussed in section 3.2.2.3, where it is mentioned that tense markers in Panare did not developed from adnominal demonstratives, but rather from free demonstrative pronouns. Section 3.2.3 is more directly concerned with the other cognitive phenomenon that plays an important role in grammaticalization: metonymy. Metonymy is the result of a continuous process in which a certain context invites a specific (typically more abstract) interpretation of some linguistic form or construction. For example, in older forms of English the equivalent of the subordinating conjunction ‘since’ used to have a purely temporal meaning (‘from the time that’), but it gradually acquired another, causal sense (‘because’), as in certain contexts one could interpret the event described in the matrix clause as a consequence of the event described in the subordinate clause introduced by ‘since’ (Hopper and Traugott 2003: 82–83), as in: (43) Since his wife left him, he buys his lunch at MacDonald’s. In this example ‘since’ can be interpreted as just a temporal subordinator, but it is not difficult to see a causal connection between the two events referred to in (43).11 This kind of process is also responsible for the fact that historically a grammatical modifier

11. Furthermore, there is nothing in the context that tells the addressee that the causal sense is not what the speaker had in mind when the sentence is uttered. This kind of reasoning is called ‘abduction’, an important component of language change (Hopper and Traugott 2003: 43).





Jan Rijkhoff

represented at some lower, more central layer in NP/clause model in Figure 1 tends to develop into a modifier at a higher, more peripheral layer (i.e. with a wider scope).12 Thus the layered model seems to be supported by two kinds of historical developments. The historical relationships across the NP-clause boundary described in section 3.2.2 are mainly metaphorical in nature, in that elements to talk about entities that belong to one dimension (space) end up being used to also talk about entities of another dimension (time). The metonymic dimension of grammaticalization is best exemplified by the more or less continuous change ‘from center to periphery’ in the layered representation of linguistic structures (section 3.2.3). Changes along the two dimensions seem to be largely unidirectional (from space to time, from inner to outer layer). Some attention is also given to counterexamples to unidirectionality. 3.2.2 Historical relations between modifiers across linguistic constructions

3.2.2.1 Qualifying modifiers Nearly a century ago Von Garnier (1909) already investigated historical connections between markers of collectivity and perfectivity (section 2.1.2.4).13 According to Leonid Kulikov (to whom this reference is due), the development from collective to perfective is attested in many Indo-European languages.14 Van Garnier’s article contains evidence from e.g. Classical Greek, Latin, and Sanskrit. Kulikov states that languages from the Slavic branch of Indo-European (Old Church Slavic, Russian) went through the same historical process. We also find traces of this development in the modern Germanic languages. For example, the Germanic prefix ge- still has a collectivizing meaning in modern Dutch nouns like gebroeders ‘(collection of) brothers’ and gebergte ‘(collection of) mountains’ (in modern German Gebrüder and Gebirge, respectively). At some point in the history of Germanic this prefix came to be used with certain imperfective verbs to express the notion of completeness (i.e. perfectivity) and ultimately it became associated with the past participle form of the verb (Kirk 1923: 65), as in:

Dutch

(44) Hij heeft net een lied voor ons ge-zongen he has just a song for us sung (past participle of ‘sing’) ‘He just sang a song for us’

12. Thus, the cases discussed here would all be instances of scope increase; cf. Song (2005) for a recent discussion on scope increase and scope decrease in grammaticalization. 13. The notions ‘collective’ and ‘perfective’ are both characterized by the feature of boundedness (Rijkhoff 2002: 53-59). 14. See Kulikov’s message to the Discussion List for The Association for Linguistic Typology, 31 March 1998 (Subject: collective/perfective). The message can be found at: http://listserv.linguistlist.org/cgi-bin/wa?A2=ind9803e&L=lingtyp&D=1&F=&S=&P=606

Synchronic and diachronic evidence for parallels between noun phrases and sentences 

3.2.2.2 Quantifying modifiers It was already mentioned in section 3.1.3 that in many languages there is a strong resemblance between markers of plurality in the NP and in the clause, i.e. between quantifying modifiers. Whereas in the case of qualifying operators (section 3.2.2.1) there seems to be general agreement on the direction of change (from collective to perfective, i.e. from space to time), in the case of quantifying operators such as plural markers there is no consensus whether the change was from the spatial to the temporal domain, or rather the other way around. According to Newman (1990: 118), verbal plural markers in Chadic developed from the nominal plural markers. His proposal finds initial support by a large body of cognitive studies, which all seem to point to the importance of the “time-as-space” metaphor in human cognition (see also e.g. Haspelmath 1997).15 In fact, Lakoff (1993: 218) has argued that our metaphorical understanding of time in terms of space is biologically determined. Nevertheless Frajzyngier (1977) claims that verbal plural markers developed into nominal plural markers in the Chadic languages, and Mithun (1988) postulates the same development for a number of North American languages. There is, of course, also the possibility that number distinctions have their origin outside the spatial or temporal domain (Lehmann 1982). Frajzyngier (1997) argues that both nominal and verbal plural markers developed from the same source: a set of deictics, determiners and anaphors. In the case of quantifying modifiers, there is perhaps another reason why it is difficult to decide on a preferred direction of change. If it is true that some numerical concepts are processed outside the language-based counting system, this may be reflected in the synchronic and diachronic behaviour of numerical expressions.16 I am not aware of cognitive research on the processing of number distinctions in the clause (π2) and in the NP (ω2), but a recent study investigating the relationship between language and arithmetic in speakers of Mundurukú, an Amazonian language with very few number words, suggests that there is “a distinction between a nonverbal system of number approximation and a language-based counting system for exact number and arithmetic” (Pica et al. 2004: 499).

15. Gil (1993: 281) claims that there is asymmetry between nominal and verbal quantification: “.., while nominal quantifiers can show up on the verb, …, verbal quantifiers can never show up on a noun.” 16. Also, intuitively the cognitive distance between operators in the clause and in the NP seems to be shorter for quantifying operators (both things and events can be Singular or Plural) than for e.g. qualifying operators: although both things and events can occur ‘collectively’, the notion of ‘completeness’ or ‘perfectivity’ seems to belong much more to the temporal domain of events (see also below on the role of spatiotemporal ambiguity of certain deictic elements in grammaticalization processes).



Jan Rijkhoff

3.2.2.3 Localizing modifiers In their article on deixis, Anderson and Keenan (1985: 297) noted that in most languages “the spatial expressions are imported directly into the temporal domain by means of metaphorical representation of time as a spatial dimension [..]” as in ‘this week’ or ‘that evening’. Some languages have gone further in that erstwhile demonstratives ultimately developed into tense markers. In Panare (a Cariban language), for example, two tense-marking auxiliaries are derived etymologically from demonstrative pronouns (Gildea 1993: 53): The original pronouns differed in spatial deixis, one marking proximate (‘this’), the other distal (‘that’). They came to be required between predicate noun and subject in predicate nominal clauses, and thus evolved into copulas. As copulas, the deixis of the pronouns shifted to time, with proximate becoming present or immediate future and distal becoming past (but also sometimes interpretable as distant future). These copulas then evolved further to become tense auxiliaries for a new generation of main clause verbs.

3.2.3 Diachronic relations between modifiers in the NP and between modifiers in the clause Both in the clause and in the NP we find that a grammatical modifier at some inner layer may turn into an operator at some outer layer, again adhering to the layered organization outlined in Figure 1. More specifically, what we see is a tendency for operators in the clause and in the NP to widen their semantic scope as they move toward a more peripheral layer (scope increase). There are also cases of operators that seem to move toward the head of the NP or clause (scope decrease), but as far as I am aware this involves hybrid categories such as the perfect, which combines tense and aspect features and consequently may develop in either direction (see also note 4). We will first present some examples of the grammaticalization of clausal operators and then turn to the grammaticalization of operators in the NP.

3.2.3.1 Grammaticalization of operators in the clause As was already hypothesized by Hengeveld (1989: 142), clausal operators tend to develop diachronically from an inner to an outer layer.17 Thus we see that in the course of time certain aspect markers (π0) may change into tense markers (π3). In their turn certain tense markers (π3) may eventually change into mood markers (Π5). – from aspect to tense According to the World Lexicon of Grammaticalization (Heine and Kuteva 2002: 231) there is a “general process whereby verbal aspect markers may be further grammaticalized to tense markers (see also Comrie 1976: 99–101; Bybee 1985: 196; Bybee and Dahl 17. Hengeveld (1989: 142): “Hypothesis 2: Diachronic developments in the field of operators tend to follow the direction π1 > π2 > π3 > π4.”

Synchronic and diachronic evidence for parallels between noun phrases and sentences 

1989: 56–57) …”. This reference work lists various studies that describe the change from Continuous or Progressive (a special type of imperfective) to Present or simply to a more general Imperfective, for which Bybee et al. (1994: 141) offer the following rationale: Since both present and imperfective meaning include the possibility of describing a situation as progressive, it is plausible to suppose that the more specific progressive grams may undergo development into either a present (in cases where the progressive was restricted to the present) or an imperfective (in cases where no temporal restrictions were in effect).

We also find that in many languages forms or constructions that were first used to express the perfect are later (also) used to indicate past tense. This is currently happening in e.g. Dutch and German, where the perfect construction (‘have’ + past participle) is now increasingly used to talk about events in the past:

Dutch

(45) Hij heeft tien jaar geleden alles verkocht he has ten year ago everything sold ‘He sold everything ten years ago.’ In other languages (e.g. French), the perfect construction is now used to express a perfective meaning (Bybee et al. 1994; Heine and Kuteva 2003: 232). As was already mentioned above (section 3.2.3), the fact that the perfect may develop into a past tense or a perfective aspect has probably to do with the somewhat hybrid character of the perfect (cf. Bybee et al. 1994: 86). It is tense-like in that it involves a reference point, which is typical for deictic categories: the present perfect is used for a past event that has ‘current relevance’. It is aspect-like in that it refers to an internal property of the event: its end point. Since it combines features of temporal and aspectual operators it may further grammaticalize either as a true tense or as a true aspect. – from tense to mood The change from Tense to Mood is clearly exemplified by future tense markers (Bybee et al. 1994: 205ff., 347–8), which often develop into markers of epistemic modality (possibility, probability). Below is an example from Swahili (Heine and Kuteva 2002: 143)

Swahili

(46) A-tu-ku-ja C1-FUT-INF-come ‘He will come.’ (47) A-tu-ku-wa nyumba-ni sasa C1-FUT-INF-be house-LOC now ‘He will be at home by now.’



Jan Rijkhoff

3.2.3.2 Grammaticalization of operators in the NP In the NP, too, there is a tendency for operators to widen their scope in the course of time, ‘moving’ from a lower/inner to a higher/outer layer in Figure 1. Thus we see that collective markers (ω0) may become plural markers (ω2), and that adnominal demonstratives (ω3) may eventually change into definiteness or specificity markers (Ω4). – from collective to plural It is well known that plural markers often originate as collective markers (Frajzyngier 1997: 237). In other words, what we see here is that over time a classifying operator (ω0) may turn into a quantifying operator (ω2). The diachronic relation between Collective and Plural has been observed in many languages and language families across the globe, such as Ket (a language isolate), the Kartvelian languages, the Mesoamerican languages, and the Semitic languages (Rijkhoff 2002: 117). The grammaticalization process can currently be observed in many languages, including Mandarin Chinese and Hmong. The suffix –men is often regarded as a plural marker in Mandarin Chinese. It behaves, however, differently when compared to plural markers in the western European languages: it is only used occasionally and its occurrence is still restricted to human nouns when reference is made to groups of people taken collectively: háizi-men ‘(a certain) group of children’ (Norman 1988: 159). In Hmong Njua the collective classifier cov replaces all other classifiers to express plurality. Ratliff (1991: 696, 699) writes that cov still has a collective meaning (it probably derives from a word meaning ‘bunches or clusters of fruit’) and is on its way to become a real plural marker (see also Bisang 1996).

Hmong Njua (Harriehausen 1990: 115, 117)

(48) kuv yuav cov tsev 1SG buy PL house ‘I buy (the) houses.’ – from demonstrative to definite article The development from adnominal demonstrative to definite article (and beyond) is also well documented is many languages. The most elaborate study on this subject is probably still Greenberg 1978a (but see also Diessel 1999 and references therein on the grammaticalization of demonstratives in general). According to Greenberg’s theory, four stages can be recognized in the process. Initially there is an adnominal demonstrative pronoun (Stage 0), which may or may not be marked for gender or class. In the following stage (Stage I), the demonstrative has lost its deictic function and has in fact turned into a definite article. The grammaticalization process need not stop here, for in Stage II the article indicates specificness, rather than definiteness. Ultimately, in Stage III, the element may become a gender/class marker, but only if the erstwhile attributive

Synchronic and diachronic evidence for parallels between noun phrases and sentences 

demonstrative pronoun was marked for gender or class. Otherwise the erstwhile demonstrative/article ends up as a noun marker, “a mere sign of nominality on the large majority of common nouns” (Greenberg 1978a: 69; see also Rijkhoff 2002: 92–96).18 The transition from demonstrative to definite article is clearly observed in Berbice Dutch Creole, where di is used both as a definiteness marker and a proximate demonstrative. The former precedes and the latter follows the noun:

Berbice Dutch Creole (Kouwenberg 1991: 111)

(49) di wari di the house this ‘this house’

4. Conclusion In this paper, I have presented evidence in support of an updated version of the layered NP/clause model presented in Rijkhoff (2002). The synchronic evidence (section 3.1) included typological data from syntax, morphology and parts-of-speech systems (instances of isomorphism among lexical and grammatical modifiers that are represented at the same kind of layer in the NP and in the clause). The diachronic evidence for parallels between the layered structure of the clause and the NP consisted of two parts (section 3.2). The first part dealt with historical relations between modifiers that appear in the same layer in the clause and the NP (section 3.2.2). It was shown that there is a diachronic connection between [a] elements marking the notions Collective (space) and Perfective (time), [b] expressions for spatial and temporal deixis, and [c] number markers for things and events. In the case of [a] and [b] there appears to be a clear preference for the metaphorical transition from space to time (cf. the hierarchy of conceptual domains in (38)), which does not necessarily mean, however, that modifiers in the NP become clause-level modifiers. For example, Gildea has argued that tense markers in Panare have developed from free rather than adnominal demonstrative pronouns. As a matter of fact, certain modifiers in the NP are also thought to have developed from independent forms or constructions (see e.g. Greenberg (1975) on numeral classifiers; Lehmann (1982) and Moravcsik (1997) on demonstratives). With respect to number markers in the clause and in the NP there is no consensus on the direction of change (from space to time or vice versa), which may have something to do with the non-linguistic nature of certain arithmetic notions or procedures (see note 16). The diachronic connection between NP and clause internal modifiers (section 3.2.3) also points to a strong preference for unidirectional change, both in the NP and 18. On the special semantics of elements in the final stages of grammaticalization, see Greenberg (1991).



Jan Rijkhoff

in the clause: from an inner to an outer layer. Since outer layer modifiers have inner layer modifiers in their scope, this implies that these cases of grammaticalization are all instances of scope increase. This goes against the popular idea that there is a correlation between grammaticalization and structural scope decrease (Song 2005: 795). It was argued that counterexamples to the change from inner to outer layer are due to the special semantics of the source item. For example, since the perfect combines features of the categories Tense (deixis) and Aspect (end point), it may develop in either direction, i.e. as an aspect (‘inward’) or as a tense (‘outward’).

References Abney, S. P. 1987. The English Noun Phrase in its Sentential Aspect. PhD dissertation, MIT. Anderson, S. R. & Keenan, E. L. 1985. Deixis. In T. Shopen (ed.), Language Typology and Syntactic Description. Vol III: Grammatical categories and the lexicon, 259–308. Cambridge: CUP. Bisang, W. 1996. Areal typology and grammaticalization: Processes of grammaticalization based on nouns and verbs in East and mainland South East Asian languages. Studies in Language 20(3): 519–597. Blake, B. 1983. Structure and word order in Kalkatungu: The anatomy of a flat language. Australian Journal of Linguistics 3(2): 143–175. Blake, B. 2001. The noun phrase in Australian languages. In J. Simpson et al. (eds.), Forty Years On: Ken Hale and Australian languages [No. 512], 415–426. Canberra: Pacific Linguistics. Butler, C. S. 2003. Structure and Function: A guide to three major structural-functional theories. Part 1: Approaches to the simplex clause; Part 2: From clause to discourse and beyond. Amsterdam: John Benjamins. Bybee, J. L. 1985. Morphology: A study of the relation between meaning and form [Typological Studies in Language 9]. Amsterdam: John Benjamins. Bybee, J. L. & Dahl, Ö. 1989. The creation of tense and aspect systems in the languages of the world. Studies in Language 13(1): 51–103. Bybee, J., Perkins, R.D. & Pagliuca, W. 1994. The Evolution of Grammar: Tense, aspect, and modality in the languages of the world. Chicago IL: The University of Chicago Press. Clark, E. V. 1978. Locationals: Existential, locative, and possessive constructions. In J. H. Greenberg, C. A. Ferguson & E. A. Moravcsik (eds.), Universals of Human Language. Vol. 4: Syntax, 85–126. Stanford CA: Stanford University Press, Comrie, B. 1976. Aspect. Cambridge: CUP. Comrie, B. 1985. Tense. Cambridge: CUP. Craig, C. G. 1977. The Structure of Jacaltec. Austin TX: University of Texas Press. Crevels, M. 1999. Concession: A typological study. PhD dissertation, University of Amsterdam. Cuvalay, M. 1997. The Arabic Verb. A Functional Grammar approach to verbal expressions in Classical and Modern Arabic. Berlin/New York: Mouton de Gruyter. Derbyshire, D. C. 1979. Hixkaryana [Lingua Descriptive Studies 1]. Amsterdam: North-Holland. de Vries, L. 1989. Studies in Wambon and Kombai: Aspects of two Papuan languages of Irian Jaya. PhD dissertation, University of Amsterdam. Denning, K. & Kemmer, S. (eds). 1990. On Language: Selected writings of Joseph H. Greenberg. Stanford CA: Stanford University Press.

Synchronic and diachronic evidence for parallels between noun phrases and sentences  Diessel, H. 1999. Demonstratives: Form, function, and grammaticalization. Amsterdam: John Benjamins. Dik, S. C. 1997. The Theory of Functional Grammar (second revised edn, Kees Hengeveld ed.). Part 1: The structure of the clause. Part 2: Complex and derived constructions. Berlin: Mouton de Gruyter. Dik, S. C. & K. Hengeveld. 1991. The hierarchical structure of the clause and the typology of perception-verb complements. Linguistics 29, 231–259. Dik, S. C., Hengeveld, K., Vester, E. & Vet, C. 1990. The hierarchical structure of the clause and the typology of adverbial satellites. In J. Nuyts, A. M. Bolkestein & C. Vet (eds.), Layers and Levels of Representation in Language Theory: A functional view, 25–70. Amsterdam: John Benjamins. Dryer, M. S. 1992. The Greenbergian word order correlations. Language 68(1): 81–138. Du Feu, V. M. 1987. The determinants of the noun in Rapanui. Journal of the Polynesian Society 96: 473–495. Du Feu, V. M. 1989. Verbal parameters expressed in the NP in Rapanui. Paper presented at the Colloquium on NP Structure, University of Manchester, September 1989. Du Feu, V. 1996. Rapanui [Routledge Descriptive Grammars]. London: Routledge. Elliott, J. R. 2000. Realis and irrealis: Forms and concepts of the grammaticalisation of reality. Linguistic Typology 4(1): 55–90. Everett, D. L. 2005. Cultural constraints on grammar and cognition in Pirahã. Current Anthropology 46: 621–646. Farsi, A.A. 1968. Classification of adjectives. Language Learning 18: 45–60. Frajzyngier, Z. 1977. The plural in Chadic. In P. Newman and R. M. Newman (eds.), Papers in Chadic linguistics. Leiden: Afrika-Studiecentrum, 37–56. Frajzyngier, Z. 1997. Grammaticalization of number: From demonstratives to nominal and verbal plural. Linguistic Typology 1(2): 193–242. von Garnier, K. 1909. COM- als perfektierendes Praefix bei Plautus, SAM- im Rigveda, CYNbei Homer. Indogermanische Forschungen 25: 86–109. Gerds, D. B. 1998. Incorporation. In A. Spencer & A. M. Zwicky (eds.), The Handbook of Morphology, 84–100. Oxford: Blackwell. Giegerich, H. J. 2005. Associative adjectives in English and the lexicon-syntax interface. Journal of Linguistics 41(3): 571–591. Gil, D. 1993. Nominal and verbal quantification. Sprachtypologie und Universalienforschung (STUF) 46(4): 275–317. Gildea, S. 1993. The development of tense markers from demonstrative pronouns in Panare (Cariban). Studies in Language 17(1): 53–73. Greenberg, J. H. 1975. Dynamic aspects of word order in the numeral classifier. In C.N. Li (ed.), Word Order and Word Order Change, 29–45. Austin TX: University of Texas Press. (Reprinted in Denning & Kemmer (eds), 227–240). Greenberg, J. H. 1978a. How does a language acquire gender markers? In J. H. Greenberg, C.A. Ferguson & E.A. Moravcsik (eds.), Universals of Human Language. Vol. 3: Word structure, 48–82. Stanford CA: Stanford University Press. (Reprinted in Denning & Kemmer (eds), 241–270). Greenberg, J. H. 1978b. Generalizations about numeral systems. In J.H. Greenberg, C.A. Ferguson & E.A. Moravcsik (eds.), Universals of Human Language. Vol 3: Word Structure, 250–295.([Reprinted in Denning and Kemmer (eds), 271–309).



Jan Rijkhoff Greenberg, J. H. 1991. The last stages of grammatical elements; contrastive and expansive desemanticization. In E. Closs Traugott & B. Heine (eds.), Approaches to Grammaticalization, Vol. II [Typological Studies in Language 19:2], 301‑314. Amsterdam: John Benjamins, Halliday, M.A.K. 2004. An Introduction to Functional Grammar (3rd edn, revised by C.M.I.M. Matthiessen). London: Arnold. Harriehausen, B. 1990. Hmong Njua: Syntaktische Analyse einer gesprochenen Sprache mithilfe datenverarbeitungstechnischer Mittel und sprachvergleichende Beschreibung des südostasiatischen Sprachraumes [Linguistische Arbeiten 245]. Tübingen: Niemeyer. Harms, R.T. & Karttunen, F. (eds.). 1976. Papers from the Transatlantic Finnish Conference. Austin TX: Texas Linguistic Forum. Haspelmath, M. 1997. From Space to Time: Temporal adverbials in the world’s languages. München: Lincom. Heine, B. 1980. Determination in some East African languages. In G. Brettschneider & C. Lehmann (eds.), Wege zur Universalienforschung: Sprachwissenschaftliche Beiträge zum 60. Geburtstag von Hansjakob Seiler, 180–186.Tübingen: Narr. Heine, B. 1997. Cognitive Foundations of Grammar. Oxford: OUP. Heine, B. 2003. Grammaticalization. In B. D. Joseph & R. D. Janda (eds.), The Handbook of Historical Linguistics, 575‑601. Oxford: Blackwell. Heine, B., Claudi, U. & Hünnemeyer, F. 1991. Grammaticalization. Chicago IL: The University of Chicago Press. Heine, B. & Kuteva, T. 2002. World Lexicon of Grammaticalization. Cambridge: CUP. Hengeveld, K. 1988. Layers and operators [Working Papers in Functional Grammar 27]. Amsterdam: Dept. of Linguistics, University of Amsterdam. Hengeveld, K. 1989. Layers and operators in Functional Grammar. Journal of Linguistics 25(1): 127–157. Hengeveld, K. 1992. Non-verbal Predication: Theory, typology, diachrony. Berlin: Mouton de Gruyter. Hengeveld, K., Rijkhoff, J. & Siewierska, A. 2004. Parts-of-speech systems and word order. Journal of Linguistics 40(3), 527–570. Hengeveld, K. & Rijkhoff, J. 2005. Mundari as a flexible language. Linguistic Typology 9(3), 406–431. Hengeveld, K. & Mackenzie, J. L. 2006. Functional discourse grammar. In K. Brown (ed.), Encyclopedia of Language and Linguistics, 2nd edn, Vol. 4, 668–676. Oxford: Elsevier. Hengeveld, K. & Mackenzie, J. L. 2008. Functional Discourse Grammar. Oxford: OUP. Hopper, P. J. & Closs Traugott, E. 2003. Grammaticalization (2nd edn). Cambridge: CUP. Jackendoff, R. S. 1977. X-bar syntax. Cambridge MA: The MIT Press. Kirk, A. 1923. An Introduction to the Historical Study of New High German. Manchester: The University of Manchester Press. Kouwenberg, S. 1991. Berbice Dutch Creole: Grammar, texts, and vocabulary. PhD dissertation, University of Amsterdam. Kulikov, L. 1998. Collective/Perfective. LINGTYP – discussion list for the Association for Linguistic Typology, 31 March 1998. Kutsch Lojenga, C. 1994. Ngiti: A Central-Sudanic language of Zaire. Köln: Köppe. Lakoff, G. 1993. The contemporary theory of metaphor. In A. Ortony (ed.), Metaphor and Thought (2nd edn), 202–251. Cambridge: CUP. Lee, K. 1975. Kuseian Reference Grammar. Honolulu HI: University Press of Hawaii. Lehmann, C. 1982. Thoughts on grammaticalization. Arbeiten des Kölner Universalien-Projekts 48.

Synchronic and diachronic evidence for parallels between noun phrases and sentences  Levebvre, C. 1998. Multifunctionality and variation among grammars: The case of the determiner in Haitian and in Fongbe. Journal of Pidgin and Creole Languages 13(1): 93‑150. Levebvre, C. & Brousseau, A.-M. 2002. A Grammar of Fongbe [Mouton Grammar Library 25]. Berlin: Mouton de Gruyter. Levi, J.N. 1973. Where do all those other adjectives come from? In C. Corum, T. C. Smith-Stark & A. Weiser (eds), You Take the High Node and I’ll Take the Low Node. Papers from the 9th regional meeting of the Chicago Linguistic Society, April 13–15, 1973, 332‑345. Chicago IL: Chicago Linguistic Society Li, C.N. & Thompson, S. A. 1981 [1989]. Mandarin Chinese – A functional reference grammar. Berkeley CA: University of California Press. Matthews, G. H. 1965. Hidatsa Syntax. The Hague: Mouton. Miner, K. L. 1986. Noun stripping and loose incorporation in Zuni. International Journal of American Linguistics 52(3): 242‑254. Miner, K. L. 1989. A note on noun stripping. International Journal of American Linguistics 55: 476–7. Mithun, M. 1988. Lexical categories and the evolution of number marking. In M. Hammond & M. Noonan (eds.), Theoretical Morphology: Approaches in modern linguistics, 211–234. New York NY: Academic Press. Moravcsik, E. 1994. Inflectional morphology in the Hungarian noun phrase – a typological assessment. EuroTyp Working Papers VII(22), 1–58 (pages are numbered separately in each article). Moravcsik, E. 1997. Parts and wholes in the Hungarian noun phrase – a typological study. In B. Palek (ed.), Typology: Prototypes, item orderings and universals. Proceedings of LP ‘96, 307–324. Prague: Charles University Press. Mosel, U. & Hovdhaugen, E. 1992. Samoan Reference Grammar. Oslo: Universitetsforlaget. Newman, P. 1990. Nominal and Verbal Plurality in Chadic. Dordrecht: Foris. Nordlinger, R. & Sadler, L. 2004. Nominal tense in a crosslinguistic perspective. Language 80(4): 776–806. Norman, J. 1988. Chinese. Cambridge: CUP. Nuyts, J., Bolkestein, A.M. & Vet. C. (eds.). 1990. Layers and Levels of Representation in Language Theory: A functional view. Amsterdam: John Benjamins. Pica, P., Lemer, C., Izard, V. & Dehaene, S. 2004. Exact and approximate arithmetic in an Amazonian indigene group. Science 306: 499–503. Quirk, R. et al. 1985. A Comprehensive Grammar of the English Language. London: Longman. Ratliff, M. 1991. Cov, the underspecified noun, and syntactic flexibility in Hmong. Journal of the American Oriental Society 111(4): 694–703. Reh, M. 1985. Die Krongo-Sprache (Nìinò Mó-dì) – Beschreibung, Texte, Wörterverzeichniss. Berlin: Reimer. Rijkhoff, J. 1988. A typology of operators: Toward a unified analysis of terms and predications [Working Papers in Functional Grammar 29]. Amsterdam: Dept. of Linguistics, University of Amsterdam. Rijkhoff, J. 1990. Toward a unified analysis of terms and predications. In J. Nuyts, A. M. Bolkestein & C. Vet (eds.), Layers and Levels of Representation in Language Theory: A functional view, 165–191. Amsterdam: John Benjamins. Rijkhoff, J. 1992. The Noun Phrase: A typological study of its form and structure. PhD dissertation, University of Amsterdam. Rijkhoff, J. 2002. The Noun Phrase. Oxford: OUP. (a revised and expanded paperback edition was published in 2004).



Jan Rijkhoff Rijkhoff, J. 2003. When can a language have nouns and verbs? Acta Linguistica Hafniensia 35: 7–38. Rijkhoff, J. 2004. Iconic and non-iconic word order patterns: On symmetry in the NP and counter examples to Universal 20’. In H. Aertsen, M. Hannay & R. Lyall (eds.), Words in Their Places: A Festschrift for J. Lachlan Mackenzie, 169–180. Amsterdam: Free University, Faculty of Arts. Rijkhoff, J. 2007. Word classes. Language and Linguistics Compass 1(6): 709–726 [doi: 10.1111/j.1749–818X.2007.00030.x]. Rijkhoff, J. 2008a. Layers, levels and contexts in functional discourse grammar. In D. García Velasco & J. Rijkhoff (eds.), The Noun Phrase in Functional Discourse Grammar [Trends in Linguistics. Studies and Monographs 195], 63–115. Berlin: Mouton de Gruyter. Rijkhoff, J. 2008b. Layering and iconicity in the noun phrase: Descriptive and interpersonal modifiers. Linguistics 46(4). Rijkhoff, J. Forthcoming. On the co-variation between form and function of adnominal possessive modifiers in Dutch and English. In W. B. McGregor (ed.), The Expression of Possession [The Expression of Cognitive Categories 2]. Berlin: Mouton de Gruyter. Rijkhoff, J. & Seibt, J. 2005. Mood, definiteness and specificity: A linguistic and a philosophical account of their similarities and differences. Tidsskrift for Sprogforskning 3(2): 85–132. Roberts, J. R. 1990. Modality in Amele and other Papuan languages. Journal of Linguistics 26: 363–401. Schachter, P. 1985. Parts-of-speech systems. In T. Shopen (ed.), Language Typology and Syntactic Description. Vol I: Clause structure, 3–61. Cambridge: CUP. Song, J. J. 2005. Grammaticalization and structural scope increase: Possessive-classifier-based benefactive marking in Oceanic languages. Linguistics 43(4): 795–838. Stroomer, H. 1987. A Comparative Study of Three Southern Oromo Dialects in Kenya: Phonology, morphology and vocabulary [Cushitic Language Studies 6]. Hamburg: Buske. van Baarda, M.J. 1908. Leiddraad bij het Bestuderen van ‘t Galela’sch Dialekt, op het Eiland Halmaheira (Manual for the study of the Galela dialect, on the island of Halmahera). The Hague: Nijhoff. Van Valin, R. D., Jr. & LaPolla, R. J. 1997. Syntax: Structure, meaning and function. Cambridge: CUP. Vet, C. 1986. A pragmatic approach to tense in functional grammar [Working Papers in Functional Grammar 16]. Amsterdam: Dept. of Linguistics, University of Amsterdam.

The development of tense, mood, and aspect, in creole languages and the typology of affix order Peter Bakker Creole languages display systems of marking tense, mood and aspect (TAM) that are rather different from what is commonly found in non-creoles, but at the same time highly similar across creoles. In this chapter I will show some of these similarities and differences, and try to explain them with general cognitive principles, focusing on the order of these elements with respect to the verb stem. Further I will discuss the genesis and development of the form of some of these TAM markers from a typological point of view. Finally I will discuss their specific forms of these markers as more or less typical instances of grammaticalization. In creoles, being young languages, it is often easier to trace grammaticalization patterns, and this can help shed light on grammaticalization in general.

1. Introduction Creole languages display systems of marking tense, mood and aspect (TAM) that are rather different from what is commonly found in non-Creoles, but at the same time the systems are highly similar across Creoles. In this paper I will show some of the similarities and differences, and try to explain them with general cognitive principles, focusing on the order of these elements with respect to the verb stem. Further I will discuss the genesis and development of the form of some of these TAM markers from a typological point of view. I will also discuss some of their specific forms of these markers as more or less typical instances of grammaticalization. In Creoles, being young languages, it is often easier to trace grammaticalization patterns, and this can help shed light on grammaticalization in general. In a discussion of typology and diachrony, Creoles could play a much more prominent role than they have done hitherto. Creoles are new languages that inherited the lexicon from a colonial language, but which had to develop a grammatical system of its own in order to become a complete language. Therefore Creoles show accelerated processes of grammaticalization.



Peter Bakker

I will first deal with Creoles (section 2). Next I will deal with the grammaticalization paths of the TAM markers (3). I will discuss both form and function of the markers for one Creole, Saramaccan (4). As Creoles are relatively young languages, almost all markers have uncontested etymologies (5). After that I will briefly discuss a typological approach to affix order, and apply this to Creoles (6). Even though the status of Creole TAM elements is a matter of debate, ranging from affixes to clitics and separate particles, their ordering can be expected to be relevant to theories of morpheme ordering. In the next section (7) I will argue that the etymology of Creole TAM markers can be used for historical reconstructions. In my conclusions I state that the etymologies of TAM markers can be used in the reconstruction of earlier stages, and population migrations.

2. Creole languages Creole languages are new or young languages that came into being in situation of contact between groups of people who had no language in common. The creators of these languages did not have a common means of communication. They used (mostly) a subset of the lexicon of another language to create a new strategy of communication, and the results have systematized and have become Creoles. Creole languages are spoken in societies of former slavery and indentured labor and in colonization situations. Usually there was a small group of colonizers whose languages were imperfectly learned by the colonized. Partly due to lack of access to the language, partly due to a lack of necessity and probably also as an act of resistance to linguistic assimilation, only portions of the language – especially the most prominent part, the vocabulary – were transmitted to the members of the new societies: the rest had to be filled in by the new speakers of the language. In Creole languages, the lexicon can often be attributed unambiguously to a pre-existing language, while the grammatical system is often totally unlike that language, or any other language known from the contact situation. As the transmitted lexicon usually covers only a part of the lexicon of the colonial language, a number of derivational processes had to be developed spontaneously by its speakers to expand the lexicon (Hancock 1977), and only a limited lexical pool was available as sources for grammaticalization. These two examples illustrate how different Creoles can be grammatically from their lexifiers: (1) dee o-tei faka tjoko unu (Saramaccan, English/Portuguese lexifier Creole) they WILL-take knife stab you(s) They will stab you with a knife (They go take knife stab you-guys) (2) Ou té monté vin fisi You PAST climb come Furcy You came to Furcy

(Haitian Creole, French lexifier) (été monter venir Furcy)

Tense, mood, and aspect, in creole languages and the typology of affix order 

Due to these unusual circumstances of genesis, connected with colonialism and geographical expansion, all languages identified as Creoles are fairly young: all known Creole languages are younger than 600 years old. Almost all Creole languages are associated with European colonialism (Portuguese, Dutch, French, Spanish and English). Apart from these, two non-European languages became the bases of Creoles: Arabic and Malay, also as a result of geographical expansion of colonial powers. These seven languages are sometimes called the lexifiers, or the “bases” of Creoles. Thus, there are seven (not six, not eight) Creole lexifiers in the world. There are Arabic-based or Arabic-lexifier Creoles in Africa, such as Juba and KiNubi in Kenya (Heine 1982), Sudan (Luffin 2005) and Uganda (Wellens 2005). These Creoles are all associated with the Arabic military and trade expansion through East Africa. There are three documented Dutch-lexifier Creoles: Negerhollands from the Virgin Islands (Van Rossem & Van Der Voort 1996) and Berbice Dutch and Skepi (Essequibo) Dutch both from Guyana (Kouwenberg 1994, Robertson 1983). Afrikaans in South Africa is not a Creole, but it has some traits in common (partial loss of lexifier grammar) with Creoles, and is sometimes called a Creoloid. There are dozens of French-based Creoles in the world, in four main regions: the Indian Ocean (e.g. Mauritius, Seychelles), the Caribbean (Guadeloupe, Martinique, Antilles, etc.), the Guyanas (including Venezuela), North America (Louisiana) and one isolated Creole called Tayo in New Caledonia. English-lexifier Creoles are found in West Africa, the Caribbean and Guyanas, the coastal southern part of North America, the east coast of Central America and in Northern Australia (among Aborigines) and the Pacific (New Guinea, Vanuatu, Solomons Islands, Hawaii). Spanish-lexifier Creoles are mostly found in the Phillippines, one in the Caribbean (Papiamentu, with a strong Portuguese component) and one in South America (Palenquero in Columbia). Finally, the oldest Creoles are Portuguese-based and spoken in the Gulf of Guinea, Upper Guinea, India, (formerly) Indonesia and one in the Caribbean (Papiamentu; it is indeed contested whether its lexicon is Spanish or Portuguese). All Creoles are spoken in tropical or subtropical regions. An overview of these Creoles (and pidgins as well) can be found in Holm (1988–9). There are several textbooks on Creoles, among others Arends, Muysken, Smith (1994), Sebba (1987) and Holm (2000). Creoles came into being under a limited set of social circumstances, and it has been argued that some of these social circumstances had influence on the structure. One type of Creole is called plantation Creoles. These came into being in situations where (tropical) agriculture is practised, for which massive numbers of laborers were imported from a diversity of regions. In the beginning these were slaves from West Africa in the Atlantic context, later voluntary workers from India, at least in the West Indies and the Indian Ocean. These plantation Creoles have been argued to be structurally maximally different from the lexifier languages. The most radically different plantation Creoles are those spoken in areas where the lexifier language ceased to be spoken after only a few years, such as the English-based Creoles of Suriname, a longterm Dutch colony (now independent), and the French Creole of Saint Lucia, etc.



Peter Bakker

Other Creoles can be called fort Creoles, since they developed around trading forts, for instance in West Africa, where local inhabitants had quite intensive contacts with the European traders, and learned to speak quite a bit of their languages, and started to use it among themselves. Such Creoles are found in West Africa (English-lexifier). Finally, there are Creoles that have been spoken as non-native language for an extended period, before becoming the primary language of a community, and these have been called nativized pidgins, i.e. developed from limited communication systems used between groups with no common language, which at some point in time became primary languages in a society. Examples of nativized pidgins can be found in Nigeria and the Pacific, in both areas with an English base, and Grand Ronde Chinook Wawa, based on Amerindian languages. An alternative term for this group is Pidgincreoles.

3. Creoles and grammaticalization It must have transpired that Creoles did not evolve gradually from another language, but instead came into being fairly suddenly. This, however, is controversial. Some creolists claim that it took Creoles only one generation to become full-fledged languages (Bickerton 1981, 1984), whereas others hold the opposite extreme position and would claim as much as 200 years. Jacques Arends, Robert Chaudenson, Salikoko Mufwene and John Singler are among those who defend a slow genesis, be it with a range of different arguments. When compared to normal development and evolution of whole grammatical systems, however, even 200 years would be very quickly. Grammaticalization of individual traits may take less time, but generally not complete restructuring of languages (for exceptions, see Bakker 2000). In that sense, Creoles may shed light on the development of grammatical systems in an obvious way: Creole grammars seem to contain a kind of “minimal” grammatical system, without many of the unnecessary increments that are the results of many generations of usage and change (cf. McWhorter 2005). Creoles may not show many unique traits in their grammaticalization patterns, but in any case the speed in which a whole grammatical system develops and grammaticalizes may be unique for this category of languages. Due to the younger age of these languages, it is often easy to point to the lexical sources of morphologized elements in the lexifier. In fact, the source item and the grammaticalized element as a rule coexist in the Creole, leaving no doubt about the source and path of grammaticalization. Thus, Creoles provide interesting and unique windows on grammaticalization, making it easier to reconstruct such patterns for Creoles than for other languages, where one would need much deeper historical knowledge of the languages involved, as well as additional information about related languages. Normally, grammaticalization can only be studied by making use of historical and comparative evidence, but in Creoles these aspects can often be studied

Tense, mood, and aspect, in creole languages and the typology of affix order 

synchronically in one Creole, as both the source morphemes and the grammaticalized forms continue to coexist, with distinct meanings. Finally – and that is one of the points I will draw attention to in this paper – one can also use the grammaticalization patterns of the TAM markers for historical reconstruction. If languages (or rather their speakers) choose exactly the same source words for the expression of these grammatical categories, it is quite reasonable to assume a historical connection between these languages. When neighboring Creoles sometimes make use of completely different source words, this suggests an independent genesis of the TAM system hence an independent creolization/nativization process.

4. Case study: Grammaticalization in Saramaccan It is typical for grammaticalization processes that lexical forms evolve into grammatical forms by a formal reduction and semantic broadening, often with a change in word class. There are other types of grammaticalization (for example focusing strategies that change word order patterns), but this is the one I will discuss here. An example is the use of the word for “man” (manu) which was reduced to mã and used as an agentive suffix, hence also applicable to women. My examples will come from Saramaccan, one of the five Creole languages spoken in Suriname. Speakers descend from maroons who escaped from plantations between the late 1600s and around 1790, and who established their own societies in the rain forest. The lexicon of Saramaccan is based on English and Portuguese, with also a significant component from Dutch and a few West African languages, mostly Kikongo and Gbe. The English component is closely connected to Sranan, the Creole of the plantations, and the lingua franca of Suriname. I will illustrate the grammaticalization process by providing source lexical words and the related grammatical form, showing some formal rules of reduction and the semantic changes. Saramaccan phonotactics is relatively simple: it only allows for CV and V syllables. Further it is relevant to know that there are tones (almost exclusively lexical; but see Ham 1999) in the language and that there is distinctive vowel length (for a more sophisticated overview see Voorhoeve 1961 and Smith 1987). The formal processes of truncation is relatively predictable in Saramaccan, and can be summarized in four rules: (A) The first rule indicates loss of the final syllable (C1V1C2V2 > C1V1) (3) sabi > sa (to know > can) lobi > lo (to love > habitually do) fasi > fa (manner > way; nominal suffix)



Peter Bakker

manu > ma (man > person doing something, “agentive”) abi > a (have) kamia > ka (place > where; relativizer) disi > di (demonstrative > definite article)

(B) The second rule indicates loss of the second syllable combined with a lengthening of the vowel of the first syllable (C1V1C2V2 > C1V1V1): (4) taki > taa (say > that; complementizer) kaba > kaa (finish > already; aspect) * demV > dee (them > definite article plural)

~ (C) The third rule indicates the loss of nasalization (C1V1 > C1V1): (the final -n is used in writing to indicate nasalization of the preceding vowel) (5)

ten > te man > ma wan > wa tan > ta

(time > period; nominal suffix) (man > person doing X; nominal suffix) (one > something being X; nominal suffix) (stand > nonpunctual aspect; preverbal element)

(D) The fourth and last rule indicates the loss of the first consonant in monosyllabic words (C1V1 > V1) (6) Go > o fu > u

(to go > future, irrealis tense/aspect) (for > possessive, complementizer)

This is not an exhaustive listing of the forms, but it probably covers all the formal processes observed in Saramaccan. None of these formal or semantic processes are surprising from a typological point of view. Heine & Kuteva (2002) list all these semantic changes in their global overview of grammaticalization processes. This brief overview just illustrates that grammaticalization in Creoles can be compared without any reservation with cases of grammaticalization in non-Creoles. However, there are two possible exceptions: the speed of the process may be greater in Creoles, undoubtedly because of the communicative demands in the contact situation, and both the source word and the reduced form continue to coexist, with their separate meanings.

5. Sources of TAM markers in Creoles Creole tense-mood-aspect marking or TAM marking is the subject of an ongoing debate in Creolistics (see e.g. Singler 1990). This debate has focused on a number of questions. Before discussing those, I have to show in what way TAM is expressed in prototypical Creoles, especially those that developed on plantations. The presumed

Tense, mood, and aspect, in creole languages and the typology of affix order 

similarities can be summarized as follows for the bulk of the world’s Creoles, even though many may show a few exceptional patterns, and/or additional forms: – tense, mood, and aspect are expressed by a limited number of preverbal elements – the core of these elements express very similar meanings across Creoles – these elements are ordered in the same way relative to the verbal root: first negation, then tense, then mood, then aspect, then the verb. – these elements can combine in pairs, or in a trio, and the meanings of these two or three combined elements are highly similar across Creoles. Even though all four of these points have been contested for individual Creoles, and also for Creoles as a class, probably most creolists would agree that the TAM systems do constitute a very typical pattern in Creoles. One could even argue that they display a shared typological trait. This does not mean that the preverbal expression of TAM constitutes a trait unique to Creoles: one can also find preverbal TAM markers in West African languages and Sinitic languages, for instance, but the particular semantic categories and their ordering are not encountered in non-Creoles. There are a number of debates on different facets of TAM in the creolist literature. First, when creolists started to compare different Creoles, they noticed a number of striking similarities between Creoles, regardless of their geographical location or their lexifiers. The reason for these perceived resemblances were embedded in the discussion of Creole origins, and the explanation has been sought in a wide range of hypotheses: African origins/substrate languages (Alleyne 1980), neurology (Bickerton 1984), pragmatics (Givon 1982), and a historical pidgin source (Whinnom 1965). The question is still central in the debate on Creoles (cf. Parkvall 2000). Another point of debate is the question whether Creoles are really so similar in the expression of tense, mood and aspect. There are quite a few Creoles for which the TAM system is different – sometimes considerably – from the more prototypical forms, for instance Papiamentu, Modern Sri Lanka Portuguese, Berbice Dutch (Kouwenberg & Robertson 1988), Grand Ronde Chinook Wawa (Grant 1996), some of the Portuguese Creoles of India such as (Clements 1996), Tayo French on New Caledonia (Siegel, Sandeman & Corne 2000). Some of these exceptional Creoles are different because they inherited more of the grammaticalizer language’s TAM morphemes (often suffixes; in the Indian Creoles from Portuguese, in Berbice Dutch from the African language Ijo), or they were quite young yet when they were recorded (Tayo, Chinook), or they were restructured, or developed further, from a more typical Creole original system (Sri Lanka). Plus, in many cases Creoles show a much more extensive system of marking TAM than the prototypical three-member set. Especially in the realm of mood, a whole range of additional markers can be found. A third point of debate focuses on the point whether this TAM system is also found in pidgins or not. TAM-systems presumed similar to those found in Creoles have been identified in pidgins as well, such as Tok Pisin and West African Pidgin English. These objections have been dismissed by some, by pointing out that these



Peter Bakker

pidgins have expanded their use beyond communication between different language groups, and hence cannot be called pidgins anymore (Bakker 2003, 2006). A fourth point of debate is the question of the grammatical status of the morphemes expressing tense, mood and aspect. Some have called them particles, others have called them clitics, auxiliaries or affixes. It is by no means necessary that all elements in one Creole have the same status, and it is also likely that this can differ from Creole to Creole. This diversity may also constitute differences in degree of development from lexical elements to affixes (lexical word/phrase > auxiliary > clitic > bound morpheme/affix) in individual Creoles. It is not the right place here to provide a solution or to provide arguments for one or the other position. I focus on grammaticalization patterns. Table 1 lists TAM markers for a selection of Creoles with different lexifiers, and some of their lexical origins. In the next section I will discuss some etymologies, and their implications for language history.

6. Reconstruction and genetic connections It has been claimed by Bickerton (1981, 1984) that Creoles have such similar TAM systems because of a universal, indeed neural, predisposition in expressing these categories. Many alternative theories have been proposed. What has been less discussed is the actual concrete source of these elements. They constitute prime examples of grammaticalization: a lexical element (often a full verb in the lexifier language) has been reduced in form, altered its meaning and widened its semantic scope, in the process of becoming grammatical elements. In this section I will look at cross-linguistic grammaticalization patterns and see which lexical sources Creoles made to grammaticalize tense-mood-aspect. This can be used for historical reconstruction. Two adjacent and lexically closely related Creoles of Surinam, Sranan and Saramaccan, have used different lexical sources for most of their TAM particles, strongly suggesting two independent processes of creolization for these two languages. There are many potential lexical sources for TAM elements. See Heine and Kuteva (2002) for a listing. Some groups of Creoles with the same lexifier language show a striking similarity, or in some cases a dissimilarity, to the particular forms in other Creoles. These correspondences and discrepancies have hardly been used in the discussion on the relatedness of Creoles. One line of discussion focuses on the monogenesis of Creoles. Some have argued in the past for monogenesis, by claiming that all Creoles ultimately derive from one putative pidgin or Creole with a Portuguese base (e.g. Whinnom 1965). This theory must be repudiated for many reasons. For arguments against this, see Bakker (2002) and Huber & Baker (2001), among others. There is another line of thinking that argues that all, or the bulk of, Atlantic Creoles with the same lexifiers, and all Pacific English Pidgincreoles and Creoles go back to one source language, e.g. Hancock (1987) for English, Smith (1987) for Atlantic

Tense, mood, and aspect, in creole languages and the typology of affix order

English and Atlantic Portuguese, Valdman (1978) for French Creoles, and Keesing (1988) and Baker (1993) for Pacific English. Many arguments have been given implicitly and explicitly, such as the presence of identical words from languages other than the lexifier, the presence of identical sound changes and similar grammatical structures, as well as historical arguments about population movements. The form of the TAM markers has not played a central role in his discussion. If one can prove that two Creoles are lexically related because they share these exceptional words or particular sound changes, but they do not share the same inventory of TAM markers, then one can assume that these two Creoles crystallized (creolized, nativized) independently from one another. In other words, if Creoles do not show the same choice of lexical elements to develop into TAM elements, then they creolized independently. On the other hand, if Creoles have used the same elements, it cannot be taken as proof: some elements may lend themselves naturally to grammaticalize into specific TAM elements. One could formulate an exception for cases where the source of grammaticalization is cross-linguistically rare. McWhorter (1995), for instance, has argued that the grammaticalization of a deictic element “there” into a copula, is cross-linguistically quite exceptional. The fact that all Atlantic English Creoles show this particular source, suggests that these Creoles must have historical connections, perhaps even a single origin. One can extend this argumentation to exceptional grammaticalization in the TAM domain. For instance, the preverbal element ba marking a.o. future in Pacific Englishes goes back to the adverbial collocation by and by, which is too peculiar to have developed independently in different places in the Pacific. When we compare a number of core TAM markers (only those sometimes called, after Bickerton, anterior tense, irrealis mood, non-punctual aspect (all preverbal) and the sentence-final completive or perfective marker in different Creoles, we find groupings like the following for the French Creoles: French Atlantic (A), Indian Ocean (I) and Pacific (P) Creoles: tense été Te Ti

(Haitian, A) (Mauritian, I) (Petites Antilles, Guyane, A) Te (Tayo, New Caledonia, P) (Te)

mood va, pour ava/pu a/pu

aspect après ap pe

ke va

ka ãtrãde

aspect NONP PERF fin fin, fek fin ?? fini

The element te/ti is from French été ‘been’, the elements ap and pe are from the preposition après ‘after’, after the now defunct expression être après V, meaning ‘to be V-ing’. The modal elements are a(va) are from forms of the verb avoir ‘to have’, and fin is from the verb finir ‘to end, to finish’. The Tayo aspect marker is from French être en train de ‘to be V-ing’. The Petites Antilles forms ke and ka are less obvious. Based on the modal and nonpunctual aspect markers, these four Creoles seem to have grammaticalized





Peter Bakker

their TAM systems independently. The almost completely distinct forms for Tayo of New Caledonia suggests an independent genesis for this Creole, which is also in keeping with the known facts (Ehrhart 1992, 1993), whereas the Creoles of the Petites Antilles and Guyane are too similar to assume an independent genesis. When we look at a couple of English Creoles from around the world, a similar picture emerges. Lexical sources are given in italics under the forms. Atlantic (A) and Pacific (P) English Creoles:

Sranan (A) Saramaccan (A): Nigeria (A) Hawaiian (P) Tok Pisin (P)

tense

mood

aspect

aspect

Ben been Bi Have? Bin Been Wen Been bin Been

sa shall o go/will go go go go Ba byandby

e there ta stand de there stay stay

kaba acabar (Port.) kaa acabar (Port.) don done pau finish (Haw.) pinis finish

The similarities in sources between Nigerian Pidgin and Sranan, especially the typologically very rare source “there” to become an aspectual marker, suggests a common origin for these Creoles, and other Atlantic Creoles with similar systems. Several authors have indeed suggested common origins, or at least historical connections between them, for most of the English-lexifier Creoles of the Atlantic (e.g. Hancock 1987, McWhorter 1995). Even though some markers apparently are an obvious source (French été and English been), there is clearly enough choice so that Creole speakers choose different markers. The fact that Sranan and Saramaccan share large parts of their English and some Portuguese vocabulary, but not their TAM markers, suggests that these languages constitute different cases of creolization. An independent genesis of the TAM systems of Hawaiian and Tok Pisin from each other and the other Creoles is obvious. Also Sranan and Saramaccan seem independent developments, but the connection between Nigerian pidgin and the others is less clear. All Atlantic Creoles display TAM systems similar, and most likely historically related, to one another (Hancock 1987, Huber & Parkvall 1999). The Creole system also appears surprisingly conservative. Creoles like Fa d’Ambu of Annobon and some other Atlantic Portuguese Creoles have developed for six centuries, but the core of the system is still present (Bakker, Post & van der Voort 1994). Sri Lanka Malay and Sri Lanka Portuguese (Bakker 2006) developed from SVO and modifier-N to SOV and N-modifier language, both changing their typological makeup

Tense, mood, and aspect, in creole languages and the typology of affix order 

completely both in terms of constituent ordering and morphological type, from isolating to agglutinating (Bakker in press). Despite all these radical changes, including the development of agglutinative morphology in the noun phrase with plural and case suffixes and postpositions, the TAM preverbal elements did not become postverbal. There is no doubt that these two Creoles changed their typological makeup under the influence of agglutinative, SOV, N-modifier Tamil, but the Creoles did not shift to suffixes to mark TAM. This preservation of preverbal TAM-VERB ordering suggests that this system is somewhat more robust and natural than others. In the next section we will see that this ordering is rather different from ordering in non-Creole languages.

7. Creole TAM and the typology of affix order I have mentioned the fact that creolists disagree about the grammatical status of TAM markers, all the way from lexical elements to affixes. Here, I will treat them as if they were morphologically connected to the verb. This allows us to compare Creole TAM marking with a typological model for the morphological expression of semantic categories in the verb Bybee (1985). Considering the strict ordering of elements in Creoles, this can be justified: if these elements would develop via cliticization to verbal affixes, their order is unlikely to be different. Bybee (1985) is a typological study of the order of verbal affixes. She deals both with the orders relevant to the stem, and assimilatory effects on adjacent affixes. Grammatical concepts can be expressed in a range of ways: lexical (e.g. suppletive stems), derivational, inflectional and as free grammatical morphemes or in syntactic configurations. The first three of these would be morphological processes, and she uses the principles of relevance and (lexical) generality to explain the order of the affixes with regards to the stems, taking the semantic categories as points of departure. Bybee (1985) used a sample of fifty areally and genetically unrelated languages in order to find generalities in the order of affixes relative to the stem, and whether these affixes are derivational or inflectional. Her hypotheses are based on relevance and (lexical) generality. Relevance is defined as follows: ‘a meaning element is relevant to another meaning if the semantic content of the first directly affects the semantic content of the second’ (Bybee 1985: 13). Hypothesis 1 is: the more relevant a category is to a verb stem, the less often it is expressed lexically. Hypothesis 2 is: the more relevant a category is to a verb stem, the more often it is expressed inflectionally. Bybee also hypothesized that, the more relevant a semantic category is, the closer it will appear to the verb stem. She investigated this by looking at a range of semantic categories most commonly expressed in verbs. She looks at the following syntactic



Peter Bakker

categories in a non-biased variety sample (cf. Rijkhoff & Bakker 1998), and she predicts, on the basis of the concepts of relevance and generality that affixes would appear in this order, relative to the stem: Valence Voice Aspect Tense Mood Number Person Person (object) Gender

number of participants (causative, applicative...) perspective (passive, middle...) internal temporal perspective (durative, inchoative...) situation with respect to a point in time (past, future...) way speaker presents truth of proposition (irrealis, ability, potentiality...) concord in number with one or more arguments concord in person with one or more arguments concord in lexical class with one or more arguments

We will limit the discussion here to tense, mood and aspect. From her figure 2, it appears that most languages express aspect and mood morphologically (74 and 68 %), and half of the languages of her sample expressed tense with bound morphemes. She also investigated the relative orders of the tense, mood and aspect morphemes, relative to one another and relative to the stem. She found the following orders (p. 34–35): tense-aspect-STEM 8 out of 8 relevant languages mood-aspect-STEM 10 out of 10 relevant languages mood-tense-STEM 8 out of 9 (exception: Ojibwe) If we generalize these, we get the following order, with only one exceptional language: mood-tense-aspect-STEM

Figure 1. Derivational and inflectional categories marked on verbs (from Bybee 1985)

Tense, mood, and aspect, in creole languages and the typology of affix order 

Bybee explains this order by referring to the relevance of these semantic categories to the action expressed in the verb stem: the more relevant, the closer to the stem. There is only one exception in her 50-language sample, which deviates from this. Creoles, however, universally have a different order: tense-mood-aspect-STEM Are there no exceptions among Creoles to this generalization? Yes. Some Creoles (also) have postverbal morphemes, invariably inherited from the lexical source languages (e.g. Portuguese in India, Ijo in Berbice Dutch). Many Creoles express completive aspect with a sentence-final particle (fini, done, kaba, pau, see above), especially Atlantic Creoles. Some Creoles cannot combine all three elements. One can say, however, that IF a Creole language can combine preverbal TAM elements, they will invariably appear in this particular order. In that sense, Creoles form a collective exception to Bybee’s typological generalization. Why do Creoles differ from other natural languages? Givon (1982) has attributed Creole TAM order to the differences in scope of the different categories: aspect has scope over the action only (the verb), mood has scope over the whole proposition (the sentence), and tense places the event in relation to other events (discourse). This seems like a reasonable explanation. Somehow, the Creole system seems more “natural” Based on Bybee’s definition of relevance, one would have to show that the semantic content of either tense or mood more directly affects the semantic content of the verbal stem. Perhaps neither are relevant enough: tense places the action in a difference time, whereas mood places the action in a different reality. Both relate to the actual realization of the action, rather than to the action or the nature of the action itself. For Creoles the scope of the elements may have played a crucial role in the grammaticalization process. The naturalness of the Creole system is corroborated by the fact that it is so conservative.

8. Conclusion Creole languages, being young languages, are prime sources of interest for studies of diachronic typology. In this paper I have discussed several aspects of the grammaticalization of lexical elements into markers of tense, mood and aspect in Creoles. Most of the sources of these markers are similar to what one finds in non-Creole languages, and also the process of formal reduction seem to be non-exceptional. Two phenomena, however, are highly relevant in the discussion of diachronic typology. One is the order of TAM elements relative to the verb, which differs typologically from non-Creoles. The semantic scope of these elements probably determined their exceptional order in Creoles. Second, one can use the lexical sources for the TAM elements and their degree of cross-linguistic rarity, in the reconstruction of historical processes. Thus one



Peter Bakker

can show that, for instance, the two Surinamese Creoles Sranan and Saramaccan developed independently as Creoles from the same lexical sources.

References Alleyne, M. C. 1980. Comparative Afro-American. Ann Arbor MI: Karoma. Arends, J., Muysken, P. & Smith, N. (eds.). 1994. Pidgins and Creoles: An introduction. Amsterdam: John Benjamins. Bakker, P. 1993. Australian influence on Melanesian Pidgin English. Te Red 36: 3–67. Bakker, P. 2000. Rapid language change: Creolization, intertwining, convergence. In C. Renfrew, C.A. McMahon & R.L. Trask (eds.), Time Depth in Historical Linguistics, 585–620. Cambridge: McDonald Institute for Archaeological Research (Papers in the Prehistory of Languages). Bakker, P. 2002. Some future challenges for Pidgin and Creole studies. In G. Gilbert (ed.), Pidgin and Creole Linguistics in the 21st Century, 69–92. Frankfurt: Peter Lang. Bakker, P. 2003. Pidgin inflectional morphology and its implications for Creole morphology. In I. Plag (ed.), Yearbook of Morphology 2002. Special section on Pidgins and Creoles, 3–33. Dordrecht: Kluwer. Bakker, P. 2006. The Sri Lanka Sprachbund: The Newcomers Portuguese and Malay. In Y. Matras, A. McMahon & N. Vincent (eds.), Linguistic Areas. Convergence in historical and typological perspective, 135–159. Houndmill: Palgrave MacMillan. Bakker, P. In press. Pidgins versus Pidgincreoles and Creoles. In S. Kouwenberg & J.V. Singler (eds.), A Handbook of Pidgins and Creoles. Oxford: Wiley-Blackwell. Bakker, P., Post, M. & van der Voort, H. 1994. TMA particles and auxiliaries. In J.P. Arends, P. Muysken & N. Smith, Pidgins and Creoles. An introduction, 247–258. Amsterdam: John Benjamins. Bickerton, D. 1981. Roots of Language. Ann Arbor MI: Karoma. Bickerton, D. 1984. The language bioprogram hypothesis. Behavioral and Brain Sciences 7: 173–221. Bybee, J. L.1985. Morphology: A study of the relation between meaning and form. Amsterdam: John Benjamins. Clements, J. C. 1996. The Genesis of a Language: The formation and development of Korlai Portuguese. Amsterdam: John Benjamins. Ehrhart, S. 1992. The development of the preverbal markers in St-Louis Creole: The formation of a TMA system? Language Sciences 14: 233–247. Ehrhart, S. 1993. Le créole français de St-Louis (le tayo) en Nouvelle-Calédonie. Hamburg: Buske. Givón, T. 1982. Tense-aspect-modality: The Creole prototype and beyond. In P.J. Hopper (ed.), Tense-Aspect: Between semantics & pragmatics, 115–163. Amsterdam: John Benjamins. Grant, A. 1996. The evolution of functional categories in Grand Ronde Chinook jargon: Ethnolinguistic and grammatical considerations. In P. Baker & A. Syea (eds.), Changing Meanings, Changing Functions, 225–242. London: Westminster University Press. Ham, W. H. 1999. Tone sandhi in Saramaccan: A case of substrate transfer? Journal of Pidgin and Creole Languages 14(1): 45–91. Hancock, I. 1977. Lexical expansion within a closed system. In B.G. Blount & M. Sanches (eds.), Sociocultural Dimensions of Language Change, 161–171. New York NY: Academic Press.

Tense, mood, and aspect, in creole languages and the typology of affix order  Hancock, I. 1987. A preliminary classification of the Anglophone Atlantic Creoles, with syntactic data from thirty-three representative dialects. In G.G. Gilbert (ed.), Pidgin and Creole Languages, 264–334. Honolulu HI: University of Hawaii. Heine, B. 1982. The Nubi Language of Kibera – an Arabic Creole. Berlin: Dietrich. Heine, B. & Kuteva, T. 2002. World Lexicon of Grammaticalization. Cambridge: CUP. Holm, J. 1988–9. Pidgins and Creoles, Vol. 1: Theory and structure. Vol. 2: Reference survey. Cambridge: CUP. Holm, J. 2000. An introduction to pidgins and creoles. Cambridge: CUP. Huber, M. & Baker, P. 2001. Atlantic, Pacific, and world-wide features in English-lexicon contact languages. English World-Wide 22(2): 157–208. Huber, M. & Parkvall, M. (eds.). 1999. Spreading the Word. The issue of diffusion among the Atlantic Creoles. London: Battlebridge. Keesing, R.M. 1988. Melanesian Pidgin and the Oceanic Substrate. Stanford CA: University Press. Kouwenberg, S. 1994. A Grammar of Berbice Dutch Creole. Berlin: Mouton de Gruyter. Kouwenberg, S. & Robertson, I.E. 1988. The marking of tense, mood and aspect in the Berbice Dutch Creole language. In N. Boretzky, W. Enninger & T. Stolz (eds.), Beiträge zum 4. Essener Kolloquium über “Sprachkontakt, Sprachwandel, Sprachwechsel, Sprachtod”, 151–174. Bochum: Universitätsverlag Dr. N. Brockmeyer. Luffin, X. 2005. Un créole arabe. Le kinubi de Mombasa, Kenya [Lincom Studies in Pidgin and Creole Linguistics]. München: Lincom. McWhorter, J. 1995. Sisters under the skin: A case for genetic relationship between the Atlantic English-based Creoles. Journal of Pidgin and Creole Languages 10(2),: 289–333. (reprinted in McWhorter 2005: 199–224). McWhorter, J. 2005. Defining Creole. Oxford: OUP. Parkvall, M. 2000. Out of Africa. African Influences in Atlantic Creole Languages. London: Battlebridge. Rijkhoff, J. & Bakker, D. 1998. Language sampling. Linguistic Typology 2(3): 263–314. Robertson, I. 1983. The Dutch linguistic legacy in Guyana – Berbice and Skepi Dutch. Carib 3: 11–23. Rossem, C. van & van der Voort, H. 1996. Die Creol Taal: 250 Years of Negerhollands Texts. Amsterdam: Amsterdam University Press. Sebba, M. 1997. Contact Languages: Pidgins and Creoles. London: Macmillan Seuren, P. A. M. 1998. Western Linguistics: An historical introduction. Oxford: Blackwell. Siegel, J., B. Sandeman & C. Corne 2000. Predicting substrate influence: Tense-modality-aspect marking in Tayo. In J. Siegel (ed.), Processes of Language Contact. Montreal: Fides. Singler, J. V. (ed.) 1990. Pidgin and Creole Tense-Mood-Aspect Systems. Amsterdam: John Benjamins. Smith, N. 1987. The Genesis of the Creole Languages of Surinam. PhD dissertation, University of Amsterdam. Valdman, A. 1978. Le créole: Structure, statut et origine. Paris: Editions Klincksieck. Voorhoeve, J. 1961. Le ton et la grammaire dans le Saramaccan. Word 17: 146–163. Wellens, I. 2005. The Nubi Language of Uganda An Arabic creole in Africa [Studies in Semitic Languages and Linguistics 45]. Leiden: E.J. Brill. Whinnom, K. 1965. The origin of the European-based Creoles and Pidgins. Orbis 14:2: 509–527.

té F. été ‘was/been’

kan(a) A. k~na 'was/been'

>kan

ta(va) P. estava? ‘was/been’

bi P. aber? ‘have’

Juba Ar.

Nubi Ar.

Sao Tome P

Fa d’Ambu P

ske

ka

bi

bi A. bidd 'want'

avá F. va ‘go(es)’

‘shall’

‘was/been’

Haitian F

sa E. shall

ben E. been

Sranan E

o E. go ‘go’

bi P. abi ‘have’

IRR

Saram. E/P

ANT

xa

ska

gi

gi/ge A. ga'ad 'sit'

ap F. après >after=

LOC.

e E. there?

tá E. stand ‘stand’

NONP

za

za /ja P. ja

ka=las

gibeel ?

‘finish’

kba P. acabar

káá P. acabar ‘finish’

COMPL

Table 1. A list of some TAM morphemes, with their lexical sources.

dwe?

‘may,can’

kan D. kan?

sa E. shall ‘shall’

EPIST

Pu F. Pour >For=

‘for’

fu E. for

fu E. for ‘for’

DEONT

bagder

fouti?

???

man ???

sa E.shall ‘shall’

PART

must

musu E. must

musu E. must must

OBLI

‘want’

wani E. want

wani E. want ‘want’

VOLO

 Peter Bakker

Berbice D

Negerhollands D

wa D. was ‘was/been’

D. had?

ANT

ma EI ??

a D. zal?

IRR

COMPL

-arε (SF) EI ??

fama EI ‘finish’

sa loo D. lopen, P. logo

NONP

kan D ‘able’

Kaa P. acabar

EPIST

fi E? D? ‘for’

DEONT

timi EI ‘?’

fo D. voor

PART

hafi/u E? D? ‘must’

OBLI

suku D. ‘search’

VOLO

Tense, mood, and aspect, in creole languages and the typology of affix order 

Aspectual oppositions from Proto-Indo-European to Latin Dag Haug In this paper, I attempt to trace the evolution of the aspect system from IndoEuropean times to the historically attested Latin system. The focus is not on the study of single changes, but on providing an overview which shows what diachronic typology can contribute to the history of the Latin aspect system.

Introduction In this paper, I will attempt to trace the evolution of the aspect system from Indo-European times to the historically attested Latin system. The focus is on the aspectual stems and temporal categories will be involved only to the extent that this is necessary for the understanding of aspect: in particular I will not discuss the striking expansion of the system of relative tenses in Latin, or the creation of the future tense. This will be a general overview with a focus on what diachronic typology can contribute to the history of the Latin aspect system. Many of the changes mentioned here demand further study, which cannot be undertaken here. A number of terminological issues will be raised, since bringing together Latin historical linguistics and general, diachronic typology requires a harmonization of the terminology: in keeping with the generalist aspirations of this volume, I have relinquished many traditional terms of Indo-European and Latin grammar in favour of modern terminology. Adopting a simplification which is common in historical linguistics, I will analyse the evolution as if it happened in discrete stages: in this case, I will operate with four linguistic stages and thus three sets of changes that happened in passing from one to the next. The first two stages will be what I call early and late PIE: the grammaticalization of aspect which probably happened at some time during the Indo-European period will serve to define these two stages. I arrive at the hypothesis of such a change by using the method of internal reconstruction: therefore, the conclusions of this first part of the paper are less certain than the later ones. Also, note that changes that we infer from internal reconstruction are not datable in relation to each other except in very fortunate circumstances: from the reconstructed PIE system we can form many more or less probable hypotheses about earlier changes in different subparts of the



Dag Haug

linguistic system, but we cannot reconstruct stages of the whole language. For example, a popular account (or family of accounts) of the reconstructable PIE ablaut system traces the zero grade back to apocope in unaccented syllables.1 But even if we believe both in this account of the ablaut system and in the hypothesis below of the rise of aspect as a grammatical category, it will be impossible to date these changes relative to each other. It follows that my definition of early and late PIE in this paper is contextually determined and not transferable to other probable or possible changes in the prehistory of PIE. In this respect, the two later stages assumed here, Proto-Italic and Latin, are fundamentally different. Proto-Italic is the more or less homogenous language which gave rise to the later Italic languages: it is the common parent of Latin and Faliscan on the one hand, and of the Sabellic languages Oscan, Umbrian and South Picene on the other. It is a proto-language reconstructable by traditional means. Latin, on the other hand, is directly attested, but is of course itself a changing entity. In this study, I am concerned with the classical system where there is one aspectual opposition, namely that between the perfectum and the imperfectum in the past tense. Thus it will appear that the stages reconstructed here have very different status, but in this context we do not need to bother too much about this, since the focus is here on the changes to be reconstructed, and not so much on the different synchronic systems.

1. Early and Late Proto-Indo-European 1.1

From aktionsart to aspect

There is general agreement that from a PIE root, one could form up to three different stems, which are traditionally labeled present stem, aorist stem and perfect stem. It is also generally agreed that the semantic difference between these three stems has to do with aspect, although this is not clear from the traditional terminology: “present stem” could lead one to think of a temporal category, but this is clearly wrong – both a present and a past can be formed on the present stem, and such past forms are clearly imperfective. We see this in Greek, which according to the communis opinio is the language that best preserves the function of the aspectual stems: a 3. sg. past like ἔλυε, formed on the ‘present stem’ λυ- has all the functions associated with imperfectivity (habituality, iterativity, progressivity). The corresponding present tense forms (e.g. λύει) are of course essentially imperfective (Comrie 1976: 66). In other words, imperfectivity is the property shared by formations from this stem, which should therefore be called the imperfective stem.

1. Most Indo-Europeanists believe in some version of this theory: for example, they refrain from positing accented zero grades unless the comparative evidence is absolutely compelling.

Aspectual oppositions from Proto-Indo-European to Latin 

The past tense form ἔλυσε is based on a stem λυσ- which clearly has perfective functions (complete action). But unlike ‘present stem’, the term ‘aorist stem’ does not lead to confusion with temporal categories. Nevertheless I have opted for ‘perfective stem’ to have a term more clearly opposed to ‘imperfective stem’. We will return to the so-called perfect stem later, but first we will have a look at the morphology of the perfective and the imperfective stems. According to traditional reconstruction, they can be formed by means of a variety of suffixes: 2 1 *gwem- ’come’ -> perfective stem -> imperfective stem

2 *dewk- ‘lead’

-> perfective stem

*gwem-Ø- *gwm-sk’e- ® -ye- *gwm ® * d uk -s-

*(e-)gwem-Ø-t3 *(e-)gwm-sk’e-t ® -ye-t *(e-)gwm ® *(e-) d uk -s-t

-> imperfective stem

*deuk-Ø-

*(e-)deuk- Ø-t

3 *g’neh3- ‘get to know’ -> perfective stem -> imperfective stem 4 *h1es-’be’

no perfective stem -> imperfective stem

5 * sneh1- ‘spin’

-> perfective stem -> imperfect stem

*g’neh3-Ø- *(e-) g’neh3- Ø-t *g’n h3-sk’e- *(e-)g’n h3-sk’e-t ® ® *h1es-Ø-

*(e-)h1es- Ø-t

* sn h1-s- *sneh1-Ø-

*(e-) sn h1-s-t *(e-)sneh1- Ø-t

These examples show some of the suffixes used in forming perfective and imperfective stems. There is an s-suffix with which perfective stems are formed and a variety of suffixes forming imperfective stems: in the above examples we find -sk’e/o- and -ye/o-, but there are numerous others (-neu-/-nu-, -e-/-o- etc.) which are used only for imperfective stems. On the other hand, we also find a zero morpheme (symbolized by -Ø- above) which can form both imperfective and perfective stems: for example, *e-g’neh3-t (> Gr. ἔγνω) is an aorist, but *e-h1es-t (> Dor. Gr. ᾖς ‘he was’) is an imperfect. As Karl Hoffmann (1970: 29ff.) made clear, this points to a period before the aspect-stems were fully grammaticalized in PIE: the roots differed in aktionsart, but there were no aspectual oppositions. At this time, the various stem formants probably had phasal meanings like iterativity, inchoativity, habituality, totality etc. But at some time their semantics developed into more general aspectual meanings like perfectivity or imperfectivity. They became associated with the root formations and formed pairs of stems in 2. A note on traditional IE transcription: y and w are the approximants IPA [j] and [ʋ]. A diacritic ring below (m-) marks syllabicity, IPA [m̩]. k’ and g’ are palatals, IPA [c] and [ɟ]. h1, h2, h3 ® (‘laryngeals’) are back consonants of uncertain phonetic quality, maybe the fricatives [ç], [x], [xw]. 3. It seems that past tense was expressed not only through the endings, but also by the prefixed augment e-. It is a matter of dispute whether this element was obligatory or not. At any rate, it has not left any traces in Italic, and we will not discuss it further here.



Dag Haug

aspectual opposition. The pattern of this evolution is quite clear and understandable: roots whose meaning was prototypically telic4 used their root formation as a perfective stem and generalized another, secondary formation as the imperfective stem: *g’neh3‘get to know, understand’ kept *(e-)g’neh3-t as the aorist whereas the stem *g’n h3‑sk’e/o-, ® – of the which historically speaking is certainly a derivation – and not a flexional form root *g’neh3-, was grammaticalized as the imperfective form. It is only from the time when a regular opposition between the two preterits *(e-)g’n h3-sk’e-t and *(e‑)g’neh3‑t ® had arisen that we can speak of an inflectional category of aspect. Root formations from prototypically atelic roots, however, were used as imperfective stems and a perfective stem could be created with the s-suffix: thus, the root preterit *(e‑)sneh1-t from the root *sneh1- ‘spin’ would take on the function of an imperfective past as soon as it was in opposition to the sigmatic formation *e-sn h1-s-t (> Gr. ἔνησε with some morphological changes). In other cases, the perfective partner was suppletive: the preterit *é-bhweh2-t (> Ved. ábhūt with new ablaut) from the root *-bhweh2- ‘become’ could probably be used in a suppletive aspectual opposition with the preterit *(e-)h1es-t from *h1es- ‘be’ already in PIE. Remember that the mechanism behind this grammaticalization of aspect is one which is often seen in semantic change, namely the strengthening of implicatures (Bybee et al. 1994: 285ff.). Cross-linguistically, we observe that when overt aspectual marking is lacking, speakers tend to interpret atelic descriptions imperfectively and telic descriptions perfectively: for example, the sentence Als wir in Nijmegen eintrafen, regnete es is likely to be understood as an ‘Inzidenzschema’ (the event of arriving happening during the imperfectively viewed raining event) whereas Als wir in Nijmegen eintrafen, regnete es eine Stunde lang is more likely to be understood as a sequence of events (Bohnemeyer/Swift 2004: 268). These interpretations are inferences that can be cancelled by the choice of connective, but semantic change typically involves the conventionalization of such implicatures. In late PIE, then, aspect was aktionsart grammaticalized much as in the modern Slavic languages. There are two differences to be noted: first, the expressions of aktionsart that were grammaticalized as aspectual markers in Slavic were both prefixal and suffixal, whereas for Indo-European we can only reconstruct suffixes. There is no indication that the later preverbs of the daughter languages could fuse with the verb in Indo-European, so that even though such elements could certainly influence the aktionsart of the verb phrase in which they occurred, they were still free elements which could hardly be grammaticalized as expressions of aspect.

4. I use this term to avoid speaking of telic or atelic roots, since aktionsart is not determined by the verbal root alone, but can be influenced by other elements of the verb phrase, such as negation, directional complements and other arguments. Still, some verbal roots (e.g. ‘come’) are more likely than others to be found in telic verb phrases and so can be said to be prototypically telic even if they can be used atelically (e.g. ‘nobody came’, a stative predicate).

Aspectual oppositions from Proto-Indo-European to Latin 

The other difference between the grammaticalization process aktionsart -> aspect in Slavic and in Indo-European is that in Slavic the present of perfective verbs came to be used as a regular future. A similar use is often found with telic verbs in languages without grammatical aspect such as Norwegian, though with less consistency: Jeg spiser (I eat) is likely to be interpreted as a present progressive, whereas Jeg spiser opp kaka (I eat up the cake) is more likely to receive a future interpretation (see Faarlund et al. 1997: 571). On typological grounds, it seems unlikely that telic verbs could not form a present in PIE, so one might legitimately ask the question of where forms like **gwem-ti (i.e. a later perfective stem + present tense endings) disappeared. A final point which should be mentioned here is that the early PIE system proposed here is not reconstructable in the traditional sense: root presents have an ablauting paradigm 3. sg. *h1és-ti, 3. pl. *h1s-énti, but if we assume that this paradigm reflects an older **h1és-ti, 3. pl. **h1es-énti, we cannot tell whether the assumed phonetic development (apocope) happened before or after the formation was grammaticalized as an imperfective form.

1.2

The resultative

If we now go on to the so-called perfect stem, the first thing to be noted is that it is lexically restricted: obviously, not all roots could form a perfect stem. If we look at the distribution of the perfect in the Lexikon der Indogermanischen Verben (Rix et al. 2001), we see that it is almost always found in combination with a root aorist, i.e. it is formed from telic roots. The same lexical restriction basically applies in Homeric Greek (Haug 2004a: 400). Furthermore, it has long been observed that the so-called perfect tends to be intransitive, although this must be qualified. An obviously archaic example like * g’e-g’noh3- ‘has come to know, knows’ is transitive. But its direct object does not have the thematic role of patient; its object does not undergo any change of state. Rather, with a verb such as *g’neh3- ‘get to know’ it is the subject that undergoes a change of state. This is obviously also the case with intransitive telic verbs, such as θνῄσκει ‘is dying’ → perfect τέθνηκε ‘is dead’. In sum, then, the perfect is restricted to telic verbs which express a change of state in the subject. From these distributional facts alone, we could probably conclude that the PIE perfect, as well as the early Greek perfect, is not a perfect gram at all (because such a gram typically has unrestricted distribution or at least occurs with all non-stative predicates, see Bybee et al. 1994: 74), but rather what Nedjalkov calls a subject resultative (Nedjalkov 1988: 9). Again, there is an unfortunate terminological problem: Wackernagel, one of the most influential scholars to have worked on the Greek and Indo-European perfects (cf. Wackernagel 1904), coined the term ‘Resultativperfekt’ for transitive perfects referring to events whose effects still present as a state of the object. In other words, Wackernagel reserved the term ‘resultativperfekt’ for object resultatives.



Dag Haug

Intuitively, however, ἄνδρα ἀπέκτονε ‘he has killed a man’ expresses no less a resultant state than τέθνηκε ‘he has died’, only in the first case the resultant state applies to the object, whereas in the second it applies to the subject. It makes little sense to call only the second one a resultative, but this terminology persists in Classical and IndoEuropean studies. Nevertheless, I will use ‘resultative’ indiscriminately about subject and object resultatives here. Furthermore, to avoid any confusion between ‘perfective’ and ‘perfect’, I shall use ‘anterior’ for the latter, as in Bybee et al. 1994.

1.3

The late PIE system

Summing up, we can illustrate the late PIE system in the following tables: In addition, the telic root *g’neh3- is able to form a subject resultative g’e-g’noh3-e with the meaning ‘he knows’. The picture is somewhat complicated by the secondary verbs; i.e. the denominatives, statives and causative-iteratives which were later to make up the first, second and fourth conjugation in Latin. These are stem formations which do not belong directly to any verbal root and originally they formed only one stem, as we can still observe in Vedic. Morphologically, this stem is formed like an imperfective stem (with suffixes like -ye/o- etc.) but this does not necessarily mean that they were imperfectiva tantum; probably they had biaspectual function. Table 1. 3sg forms from the atelic root *sneh1- ‘spin’: Imperfective stem *sneh1Active Medium

Present *sneh1- ti

*sneh1- toi

Past *(e-)sneh1- t

*(e-)sneh1- to

Perfective stem *sn h1-sPresent – –

Past *(e-)sn h1-s-t

*(e-)sn h1-s-to

Table 2. 3sg forms from the telic root *g’neh3- ‘get to know’ Imperfective stem *g’neh3-sk’eActive Medium

Present *g’n®h3-sk’e-ti *g’n®h3-sk’e--toi

Past *(e-)g’n®h3-sk’e--t *(e-)g’n®h3-sk’e-- to

Perfective stem *g’neh3Present – –

Past *(e)-g’neh3-t *(e-)g’neh3-to

Aspectual oppositions from Proto-Indo-European to Latin 

2. Proto-Italic 2.1

The three inherited stems

We now come to the Proto-Italic stage, the common system which later developed into Latin on one side and Sabellian on the other. As has long been recognized, the Sabellian languages exhibit the same sort of division of verbal present stems into 4 conjugations as in Latin. This is a matter of morphological structuring which will not be pursued further here. More important for our issue is that there is a remarkable agreement between the present stems of verbs in the Italic languages (cf. Meiser 2003: 62): of 62 present stems attested in both Latin and Sabellian, 49 show the same formations. For example, Osc. acum is a simple thematic present just as Lat. ago, Umb. hondu shows the same nasal infix as Lat. fundo (as opposed to the thematic Germanic *geuta-), Umb. habetu is an ē-verb like Lat. habeo. In contrast, Meiser (2003: 72ff.) draws attention to the fact that in the perfect system, the situation is completely different: although some agreement is found (e.g. Lat. dedi and Osc. deded), there are more cases of disagreement where Latin continues an aoristic formation whereas an old perfect is found in Sabellian or vice versa, e.g. Latin dixi, finxi as opposed to Umb. dersicurent, Osc. fifikus or conversely, Lat. sedi vs. Umbr. sesust. From this situation we must conclude that although the aorist and the perfect have merged into one category in both Sabellian and Latin, the opposition must still have been intact in Proto-Italic. This means that there must have been two more or less independent mergers, but this is not surprising, since the evolution of anteriors to perfective or simple pasts is typologically very common (Bybee et al. 1994: 81ff.; for some views on how this change happened in Greek, see Haug 2008). But unlike in modern French and German where the periphrastic perfect is ousting the old simple past, we are clearly dealing with a situation of semantic overlap in Italic, since the result is a morphological syncretism and not the replacement of one category by another. In other words, we may assume that Proto-Italic, just like Proto-Indo-European, maintained a system of three verbal stems which morphologically continued the imperfective, the perfective and the resultative stem. However, there is reason to believe that this morphological continuity conceals semantic change. First of all, Italic, like Celtic and Germanic, lacks clear traces of the Indo-European imperfect. There is indeed nothing to suggest that Proto-Italic had such a form. The important implication is that in a certain period the morphological continuation of the IE perfective past must have been the only past tense form: thus, it must have served as a general past tense. In other words, whereas the morphology of primary verbs continued the IE forms of the present of the imperfective stem and the past of the perfective stem, the relevant semantic distinction can no longer have been aspectual but only temporal. However, expression of temporal distinctions was necessary not only in the primary verbs, but also in the secondary verbs which could not form an



Dag Haug

aorist. The obvious solution to this problem was to extend the use of the s-morpheme, which was probably productive already by the late IE period; remnants of such forms are probably found in the so-called ‘contracted’ forms of the perfect, such as amasti, amastis, amarunt etc.5 Another change connected to the collapse of the distinction between primary and secondary verbs concerns the Indo-European resultative which must have become an anterior by the proto-Italic period. Typological studies suggest that this is a necessary stage between the resultative semantics clearly reconstructable for Indo-European and the perfective past semantics that we find in Latin and Sabellian. The extension of this gram to all verbs through the creation of ‘weak’ preterits (see below) also suggests that we are now no longer dealing with a resultative which typically has restricted distribution: “As completives or resultatives develop into anteriors and further into perfectives, they must generalize to occur with verbs of all semantic types.” (Bybee et al. 1994: 74) But the Indo-European resultative was not only semantically restricted, it was also morphologically restricted to primary verbs. In other words, when its distribution widened, a morphological means was required to form perfects from secondary verbs. The comparative evidence seems to point to a variety of periphrastic expressions. This concurs with the typological fact that anterior grams tend to be expressed periphrastically. In Oscan we find a perfect in -tt- which probably comes from a periphrastic expression involving the -to-participle although the details are unclear. The common Sabellian type in -f- is unexplained. But the once mysterious Latin type in -u- has been given a reasonable explanation by Rix (1992). Although not all details may be convincing, it seems obvious that the recurring syllable -uis- should be connected with the perfect participle active in -wos-, as Rix suggested. In all likelihood, Proto-Italic conserved the three-way opposition of stems inherited from Indo-European in the primary verbs, but extended the system to all verbs. The present stem is still used in the same way and the morphology is identical. The old perfective stem, restricted to past tense already in PIE, became an expression of general past tense after the loss of the imperfect: therefore, it was necessary to form such stems even from secondary verbs, and such means were provided by the -s-suffix. Finally, the resultative changed its semantics to an anterior and a way of expressing this category was created through periphrases with various participles.

2.2

The new imperfect

The value of the old perfective past was changed through the loss of the imperfect. However, there is evidence that what we call the imperfect in Latin grammar – or at 5. That these forms were probably never contracted from the longer forms amauisti etc. has been pointed out a number of times in the literature, see for example Meiser (1998:207). In the 4th conjugation, however, the short forms can result from the loss of -v- between identical vowels and subsequent contraction.

Aspectual oppositions from Proto-Indo-European to Latin 

least a predecessor of it – was created already in Proto-Italic times, since we find a form fufans (which would have been Latin *fubant) in Oscan. So we need to pose the question as to how this form should be analyzed. Meiser (1998: 197) suggested that the origin of fufans is a reduplicated pluperfect (transposed into Indo-European it would be *bhe‑bhwoh2-nt) in which the reduplication syllable was reinterpreted as the root and thus -fā- (Proto-Italic *-βā-) as a tense suffix. But this hypothesis seems to be at odds with what typology tells us about the creation of new verbal categories: while reinterpretation of this kind can clearly affect the morphological expression of a category, it is less obvious that it can lead to the creation of a new category. It is hard to see how -ba- could be reinterpreted as an expression of imperfective aspect and past tense as long as no imperfective category existed. Typically, new categories develop from periphrastic constructions through a process where words are grammaticalized as shown by Bybee et al.(1994:passim). An explanation of the imperfect as a periphrastic formation was in fact suggested a long time ago by Franz Skutsch (1903) [1914] who considered the first parts of amābam, monē-bam, agē-bam and audiē-bam to be remnants of the present participles amāns, monēns, agēns, audiēns. The origin of the Italic imperfect should therefore be sought in periphrastic constructions like *agents βwāt (where the latter form is a reflex of the PIE root aorist *(e-)bhweh2-t). Phonetically there can be no serious objection to this explanation,6 although strictly speaking we cannot prove such a cluster reduction since there are no parallels at hand. Meiser (1998: 197) also raised a semantic objection which seems more serious at first sight. He claimed that the root aorist *βwāt cannot be the origin of the imperfect morpheme -bat because of its perfective semantics. But the force of this argument vanishes as soon as it is realized that in the period after the loss of the PIE imperfect (probably very early since it is a common West-IE feature) and the creation of the new form, there were no possible aspectual oppositions. There are therefore no objections to Skutsch’s simple explanation of the Latin imperfective from a periphrasis like *agents βwāt. We can of course have no direct knowledge of the semantics of such a periphrasis, but it is possible to make some conjectures based on typological data. First of all, a periphrastic construction signals a young form in the grammaticalization cycle – therefore, a general meaning like imperfective aspect is unlikely. We should rather expect a more specific meaning within the domain of imperfective aspect, such as progressivity or habituality. 6. Meiser claims that one would expect the nasal to be attested in Old Latin, but what applies to intervocalic *-ns- groups does not necessarily apply to -nsC- groups which could behave as -ns at word end, where the nasal seems to have disappeared early, leaving a lengthened vowel. Only two forms do not fit Skutsch’s explanation, namely dăbam (because of the short a) and ībam (because of iēns). However, these can be analogical formations based on the infinitive dăre and īre. There is a certain tendency for the Latin present infinitive to serve as a basis for the imperfective stems, see Haug (2004b:236n3) for some examples in the imperatives. The same analogy can account for the by-forms in the fourth conjugation such as audībam.



Dag Haug

Superficially, the construction *agents βwāt resembles the modern English progressive was doing, but the origin of the latter construction is probably a locative construction ‘I am a-reading’, “where a represents the preposition on, and the form in -ing is not the participle but the noun” (Jespersen 1949: 168). Bybee et al. (1994: 135) even go so far as to claim that non-locative progressives are very rare or even non-existent. Furthermore, if Proto-Italic had a periphrastical construction of progressive meaning, we should perhaps expect to find traces of it even in the present tense, since progressives are usually not temporally restricted. Habituals, however, are often confined to past tense as typological studies show. So although the matter may never be settled, it seems more probable that *agents βwāt expressed habituality than progressivity.7

3. From Proto-Italic to Latin The further evolution of the tense/aspect-system seems to have taken place within the individual languages, since there are differences of detail between Latin and Sabellian, and even between the Sabellian languages themselves. For example, all Italic languages seem to have lost the distinction between perfect and perfective past: but not always in the same way, as we can gather from the distinction between Latin dixi (‘I said’, an old s-aorist) and Umbrian dersicurent (‘they said’, an old perfect). In the Sabellian languages so few of these past forms are attested that we cannot know if there were any principles behind the choice. In Latin, on the other hand, certain principles can be established; for example, s-aorists were kept only in verbs ending with a stop, nasal, -s or -h, i.e. in verbs where the morpheme would undergo phonological change; see Meiser 2003 for the entire problem. But the most important change from Proto-Italic to Latin was the further grammaticalization of the periphrastic construction *βingens βwāt into a general past tense. Through this evolution, the category of aspect, which had been lost probably since the early days of the Western Indo-European dialect continuum, was reintroduced in the Latin system. Before the advent of large-scale typological studies of temporal and aspectual categories (such as Dahl 1985, Bybee et al. 1994), it would seem strange to claim that the Latin imperfect, which is so similar in its functions to that of imperfects in other Indo-European languages, does not directly continue the semantics of the Indo-European imperfect: but we now know that imperfective categories all over the world have a lot in common. The Latin imperfect is simply a manifestation of the cross-linguistically attested Past Imperfective gram. Also its origins are typical: it

7. In this respect, it is also interesting to notice that the Old English construction of participle + auxiliary verb sometimes expressed habituality as in the example (Ælfric, Lives, I.52-55) cited by Bybee et al. (1994:135): Sume syndan creopende on eorðan mid eallum lichoman, swa swa wurmas doð ‘Some creep on the earth with their whole body, as worms do’.

Aspectual oppositions from Proto-Indo-European to Latin 

started out as a periphrastic construction with a specific meaning, but through univerbation and semantic bleaching, it became a morpheme with a more general meaning. At this point we have reached the state that obtains in Latin; we have an aspectual opposition in the past between a form which is specifically imperfective and another form which is ambiguous between perfective past and anterior. The evolutionary pattern as been reconstructed here is in accord with what diachronic typology tells us, and moreover, repeats itself in a similar way: even in Classical Latin, we observe the emergence of a new, resultative construction with habere + perfect participle (cognitum habeo ‘I have understood’); this construction was grammaticalized as an anterior, giving us the three term system (imperfective: perfective: anterior) of Romance, which is again being simplified to a two term system in spoken Modern French. Once again, a single category (the passé composé) expresses perfective past and anterior, but this time there is no morphological mixture as in Latin, where some verbs use the old aorist and some verbs the old resultative; rather, we observe the ousting of one category by another with the old perfective past (passé simple) losing ground. What we see in Modern French seems a more ‘normal’ type of change: in the nominal domain, for example, when there is casual syncretism, we most often see one case taking over the functions of another. There can be a period of morphological indeterminacy, but eventually, the morphology of one case wins and is generalized throughout (at least within one inflectional class). Consider for example the Greek syncretism of the locative in -i and the dative in -ei: in Mycenaean times, there was vacillation, but eventually -i was generalized to all athematic nouns. We did not get a split, with some nouns taking -i and others -ei, and for good reasons, most of us would think: but exactly such a thing happened in the Latin perfect, and more research is needed into the conditions which led to such a peculiar evolution.

References Bohnemeyer, J. & Swift, M. 2004. Event realization and aspectual interpretation. Linguistics and Philosophy 27(3): 263–296. Bybee, J. et al. 1994. The Evolution of Grammar. Chicago IL: University of Chicago Press. Comrie, B. 1976. Aspect. Cambridge: CUP. Dahl, Ö. 1985. Tense and Aspect Systems. Oxford: Blackwell. Faarlund, J. T. et al. 1997. Norsk referansegrammatikk. Oslo. Haug, D. 2004a. Aristotle’s kinesis/energeia-test and the semantics of the Greek perfect. Linguistics 42(2):387–418. Haug, D. 2004b. On unaccented short vowels in Sabellian and the morphology of the Italic 2nd conjugation. Indogermanische Forschungen 109: 235–249. Haug, D. 2008. From resultatives to anteriors in Ancient Greek: On the role of paradigmaticity in semantic change. In T. Eythorsson (ed.), Grammatical Change and Linguistic Theory. The Rosendal papers. Amsterdam: John Benjamins.



Dag Haug Hoffmann, K. 1970. Das Kategoriensystem des indogermanischen Verbums. Münchener Studien zur Sprachwissenschaft 28:19–41 (= Hoffmann, K. Aufsätze zur Indoiranistik II:523–540). Jespersen, O. 1949 A Modern English Grammar on Historical Principles. Part IV, Syntax. 3rd Vol, Time and tense. Copenhagen: Munksgaard. Meiser, G. 1998. Historische Laut- und Formenlehre der lateinischen Sprache, Darmstadt: Wissenschaftliches Buchgesellschaft. Meiser, G. 2003. Veni Vidi Vici – Die Vorgeschichte des lateinischen Perfektsystems, München: C.H. Beck. Nedjalkov, V. (ed.). 1988. Typology of Resultative Constructions, Amsterdam: John Benjamins. Rix, H. 1992. Zur Entstehung des lateinischen Perfektparadigmas. In O. Panagl & T. Krisch (eds.), Latein und Indogermanisch, 221–240. Innsbruck: Institut für Sprachwissenschaft. Rix, H. et al. 2001. Lexikon der indogermanischen Verben. Wiesbaden: Reichert. Skutsch, F. 1903 [1914]. Su alcune forme del verbo latino. In W. Kroll (ed.), Kleine Schriften von Franz Skutsch, 283–296. Leipzig. Wackernagel, J. 1904. Studien zum griechischen Perfektum. PhD dissertation, University of Göttingen (= Wackernagel, J. Kleine Schriften II:1000–1021).

On the development of actionality, tense, and viewpoint from Early to Late Latin Gerd V. M. Haverling In Early and Classical Latin, verbal affixes indicate actional oppositions, such as non-dynamic vs. dynamic, atelic vs. telic, or the focus on the beginning or the end of a development or change. The tense system emphasises the relative sequence of events and sometimes it even distinguishes the normal past from the anterior past. There is a viewpoint opposition in all the forms of actionality and even in the indication of permanent states in the past. In Late Latin, several of the verbal affixes lose their actional functions and several previously transitive verbs become intransitive. The sensitivity to the relative sequence of events diminishes and there are major changes in the functions of the tense forms and a tendency to use the imperfective viewpoint, where Classical Latin distinguished the perfective viewpoint from the imperfective one.1

Introduction There are huge differences between Indo-European languages as regards the way tense, actionality and viewpoint are expressed. These differences show that such categories are often subject to change. In the development from Early to Late Latin, we can follow such changes over a period of about a thousand years. The earliest Latin texts which are long enough to allow us to study the development of syntax are from the late 3rd century BCE. The literary norms for prose and poetry were defined in the Classical period in the 1st century BCE and influence the written language during the following centuries. However, in the texts from around 200 CE we can observe some of the fundamental linguistic changes that were taking place even in the literary texts. In the later centuries, there is a growing gap between the language actually spoken in everyday life and the language written by the educated 1. The present article is a brief overview of my chapter on Actionality, Tense and Viewpoint for New Perspectives on Historical Syntax of Latin (= Haverling, forthcoming) and of the mono graphy which I intend to publish on the subject. I refer to these works as well as to my previous work on the subject (especially Haverling 2000) for a fuller presentation of the facts and of the vast bibliography on this complex problem in Latin.



Gerd V. M. Haverling

elite. The literary form of Late Latin is at times very conservative, at times less, and changes with the development of the spoken language during these centuries. In the 5th century the Western empire collapsed and was replaced by a number of Germanic kingdoms. In the late 6th century the Roman school system and the Roman society ceased to function. During the next 150–200 years very few people learned to read and write properly and the language developed very fast. Around 800 CE the interest in the great Roman literature is renewed and there is a growing awareness of the difference between the literary norm and the spoken language, which paves the way for the Romance languages and for Medieval Latin.

1. Actionality In Early and Classical Latin, verbal suffixes and infixes indicate various kinds of stative and dynamic actionality and agency, while prefixes indicate telicity, change of state and focus on the beginning or the end of a development or change. The suffixes indicate such things as stativity, dynamicity, causativity, intransitivity or iterativity (1):2 (1)

Some functions of the suffixes: – stativity: e.g. -ēre (e.g. caleo ‘I am warm’) – dynamicity: e.g. -scere (e.g. calesco ‘warm up’) – causativity: e.g. -ēre and -are (e.g. augeo ‘make grow, increase’, inuetero ‘allow to become old’) – intransitivity: e.g. -scere (e.g. augesco ‘increase, grow’, inueterasco ‘grow old’) – iterativity: e.g. -tare, -itare etc. (e.g. uentito ‘come often’) We have verbal families in which there is a stative verb, a suffixed verb that indicates a process or gradual change and a prefixed verb that indicates a change of state (2). In some cases there is no unprefixed dynamic verb; this is because the change indicated by those verbal families could hardly be seen as atelic processes.3 The system met with in Latin much resembles the one found in other Indo-European languages, such as for instance Gothic, but it differs from Gothic in the sense that Latin needs a dynamic suffix or infix to create a verb indicating change of state from a stative verb:4

2. See, e.g., Baldi (2002, pp. 371–376); cf. Haverling (forthcoming, section 2). 3.

See Haverling (2000, pp. 43 ff., 137 ff., 249 ff.; forthcoming, sections 2.1.3 & 2.1.4).

4. See Haverling (2000, pp. 252–267; 2003b, p. 117; forthcoming, sections 2, 2.1.6). – Cf. Latin dormio ‘sleep’ and condormisco ‘fall asleep’ and Gothic slepan ‘sleep’ and gaslepan ‘fall asleep’: on ga- see, e.g., Josephson (1977) and Lloyd (1979).

On the development of actionality, tense, and viewpoint from Early to Late Latin 

(2) Affixes and actionality in the infectum: –dyn. / state +dyn., –tel. / activity caleo ‘I am warm’ calesco ‘become warmer, am warming up’ taceo ‘I am silent’ dormio ‘I am asleep’ sedeo ‘I am sitting’ sido ‘sink gradually to a lower level’

+dyn, +tel. (+/–mom.) concalesco ‘become warm’ conticesco ‘stop talking’ condormisco ‘fall asleep’ consido ‘sit down’

In the perfectum we have a corresponding system, where unprefixed perfect tense forms of stative verbs have a stative function and where prefixed perfect tense forms indicate a change of state. However, in poetry we find unprefixed such verbs in a dynamic function (3):5

(3) Prefixes and actionality in the perfectum: –dyn. / state +dyn., change of state; poetic. calui ‘I was warm’ calui ‘became warm’ tacui ‘I was silent’ dormiui ‘I was asleep’ sedi ‘I sat, was sitting’ sedi ‘sat down’

+dyn, change of state; prosaic concalui ‘became warm’ conticui ‘stopped talking’ condormiui ‘fell asleep’ considi ‘sat down’

The different prefixes indicate focus on different parts of the process or change and sometimes they indicate whether a telic action is initio-transformative or fini-transformative (4).6 The prefix in- emphasises the beginning of the change and indicates that a change starts to take place and that the process continues for a certain amount of time, whereas ex- emphasises the end and tells us that the process has been brought to its final conclusion. There is a somewhat similar relationship between the prefixes ad- and con-, the prefix ad- emphasises the gradual character of the change, whereas con- just emphasises the change as such.7 The poets and certain prose writers who use poetic elements in their prose sometimes employ unprefixed verbs in the function of the prefixed ones:8

(4) Unprefixed verbs and the prefixes in- and ex+dyn., –tel. aresco ‘I am drying’

5.

beginning, graduality inaresco ‘start becoming dry, become somewhat dry’

end, change as such exaresco ‘become totally dry’

See Haverling (2000, p. 209 ff.; forthcoming, section 2.1.4).

6. See Johanson (2000, pp. 58–63). 7. See Haverling (2000, pp. 121 ff., 252 ff., 276 ff., 292 ff., 327 ff.; forthcoming, section 2.1.7). 8. See Haverling (2000, e.g. pp. 143 ff., 167; 2003b, p. 117 f.; forthcoming, section 2.1.4).



Gerd V. M. Haverling albesco ‘grow white, pale, I am growing white’

(suesco ‘get used to’; Tacitus around 100 CE, poetic character) nosco ‘I am studying, trying to find out about’

inalbesco ‘start turning white, become somewhat white’

exalbesco ‘turn white or pale (in the face because of fear)’

insuesco ‘start getting used to, get somewhat used to’

consuesco ‘get used to’

adsuesco ‘get used to gradually’ agnosco ‘get to know gradually, identify’

cognosco ‘get to know’

It is, however, a complicated system, in which most prefixes have several different functions. They may, for instance, have a spatial function, as ad- in aggredior ‘go towards, against, assault’, as well as an actional one, as ad- in adsuesco ‘get used to gradually’. The very important prefix con- is met with in four somewhat different functions (5):9

(5) The functions of con– function verb with con completive 1 comedo ‘eat up’ (focus on the end) concalesco ‘become warm’ completive 2 (focus on the beginning)

conticesco ‘start being silent, i.e. stop talking’

intensive

commoror ‘stay (on), remain, delay, wait’ condormio ‘I am fast asleep’ congredior ‘get together, meet’

‘with, together’

conspicio ‘catch sight of ’

interrelates with: adedo ‘eat away, eat into, nibble, wear down, exhaust’ incalesco ‘start warming up, get somewhat warm’

adspicio ‘notice with the eyes, look at, witness’

digredior ‘go in different directions, depart from’

In certain cases, however, the prefixes overlap the functions of one another; the functions of re– and con– are very close in rescisco and cognosco ‘get to know of, find out about’ (6):10

9. See García-Hernández (1980; 1995) and Haverling (2000, e.g. pp. 252–272, 276–287; 2003b, p. 119 f.; forthcoming, section 2.1.7); cf. also Rosén (1992). 10. See Haverling (2000, p. 365; 2003b, p. 119; forthcoming, section 2.1.7).

On the development of actionality, tense, and viewpoint from Early to Late Latin 

(6) a.

Plaut. Asin. 743 ne uxor resciscat metuit so that notconj neg wifenom gets to knowpr subj (he) fearspr ind ‘he is afraid of his wife finding out’

b. Plaut. Men. 428–429 eadem ignorabitur,/ ne uxor cognoscat itnom will be ignoredfut spl so that notconj neg wifenom gets to knowpr subj

te habere, si in uia conspexerit youacc havepr inf ifconj in the streetprep+abl will have caught sight offut ex ‘thus it will look different, so that my wife will not find out that you have it on you, if she notices it on the street’

Classical Latin differs from Classical Greek in the sense that it uses actional forms to express semantic differences often indicated by the viewpoint opposition in Greek (7). In Latin the difference between a state in the past, such as in ‘I was silent, did not speak’, and the change into it, as in ‘I stopped talking’, is indicated by lexical means; but in Greek this difference may be indicated by the opposition between the imperfect and the aorist (cf. ex. 41 and 42). Another example is the difference between the atelic and ‘conative’ ‘tried to persuade’ and the telic ‘persuaded, did persuade’, which Classical Latin indicates with the help of prefixation, whereas Greek may express this difference with the help of the opposition between the imperfect and the aorist. In Russian we have here the opposition between the imperfective and the perfective aspect:11 (7) Classical Latin actionality vs. Classical Greek viewpoint a. ‘was silent, did not speak’ vs. ‘stopped talking’ – Latin, unprefixed vs. prefixed: tacui / tacebam ‘was silent, did not speak’ (pf. / impf.) – conticui ‘stopped talking’ – Greek, imperfect vs. aorist: ἐσίγησα / ἐσίγων ‘was silent, did not speak’ (aor. / impf.) – ἐσίγησα ‘stopped talking’ (aor.) b. ‘tried to persuade’ vs. ‘did persuade’ – Latin, unprefixed vs. prefixed: suasi / suadebam ‘tried to persuade’ (pf. / impf.) – persuasi ‘persuaded’ (pf.) – Greek, imperfect vs. aorist: ἔπειθον ‘tried to persuade’ (impf.) – ἔπεισα ‘persuaded’ (aor.) – Russian, imperfective vs. perfective verb: ugovarival ‘tried to persuade’ – ugovoril ‘persuaded’ In Late Latin there are major changes in the actional system as well as in the system of tense and viewpoint. In the actional system several of the suffixes, infixes and prefixes 11. See Haverling (2001, p. 361; 2006; forthcoming, section 4.1.6); on Russian see e.g. Comrie (1978, p. 19) and on Classical Greek e.g. Rijksbaron (1994, pp. 16 f., 20 f.); cf. Hedin (2000, p. 247).



Gerd V. M. Haverling

lose their original actional function. The stative suffix -ēre loses its function and in several cases previous verbs are replaced by a form of esse ‘be’ and an adjective. In the texts this is reflected in a decline in the use of some old stative verbs.12 The dynamic sco-suffix loses its function too and sco-verbs may occur in a stative sense and the suffix may be added to dynamic verbs that did not take it before. Eventually the suffix became an element in the conjugation of certain verbs in the Romance paradigms (8):13

(8) suffixes lose their functions in Late Latin

a. the stative suffix -ēre is losing ground i. calet ‘it is hot’ > calidum (or caldum) est ii. senet ‘he is old’ > senex est b. the dynamic suffix -scere remains frequent but is often without function i. Cl. Latin lateo ‘I am hiding’ vs. delitesco ‘vanish, disappear, go into hiding’ > Late Latin lateo = delitesco ii. -scere is added to already dynamic verbs, e.g. fulgeo > fulgesco iii. traces of -scere in Romance: cf. It. finire – finisco and Fr. finir – finissais The opposition between atelicity and telicity expressed by unprefixed and prefixed verbs is lost. In Classical Latin persuadeo means ‘persuade’ and the unprefixed suadeo ‘try to persuade’ (ex. 7b), but in Late Latin the two verbs may replace each other so that suadeo means ‘persuade’ and persuadeo ‘try to persuade’ (9a–b):14

(9) a.

Late Lat. persuadere = Cl. Lat. suadere, Vulg. Ruth 1.18

noluit nec ultra ad suos pf ind conj neg adv did not want and not further to her relativesprep+acc reditum persuadere returnacc try to persuadepr inf ‘she did not want to go on trying to persuade her to return to her family’

b. Late Lat. suadere = Cl. Lat. persuadere, Vulg. Act. 21.14 et cum ei suadere non possemus conj conj dat pr inf and when him persuade notneg we couldimpf subj quieuimus we were calmpf ind ‘and when he would not be persuaded, we ceased’ In Early and Classical Latin there is an opposition between the stative verbs and verbs which indicate a change into that state. This is the difference between taceo ‘I am silent’ 12. See Lehmann (1995); cf. Haverling (2007; forthcoming, section 2.2.1). 13. See e.g. Haverling (2000, pp. 53, 65; forthcoming, section 2.2.1). 14. See Haverling (2000, pp. 206–207; 2003b, p. 113 f.; forthcoming, section 2.2.4).

On the development of actionality, tense, and viewpoint from Early to Late Latin 

and conticesco ‘stop talking’ (ex. 2); there is a corresponding difference between the unprefixed and the prefixed perfect tense forms (ex. 3 and 7a). In Late Latin, however, this system is blurred and the unprefixed form takes over the function of the prefixed form. As a result of this development the old prefixed form no longer is needed and becomes rare in the later texts (10a–c):15 (10) a. 3rd c. CE, Ulp. Dig. 48.19.6

qui tam diu conticuerunt whonom soadv for long timeadv have been silentpf ind ‘who for so long time have kept silent about’

b.

3rd c. CE, Lact. Inst. 4.27.14 in perpetuum conticescet for everprep+acc will keep silentfut spl 3sg ‘will keep silent for ever’

c.

4th c. CE, Vulg. Act. 15.13 postquam tacuerunt, respondit Jacobus whenconj they stopped talkingpf ind answeredpf ind 3sg Jacobnom ‘when they had stopped talking, Jacob answered’

The loss of the opposition between panem edo ‘I eat bread’ and panem comedo ‘I eat up the bread’ is likely to be one of the elements which triggered the development of definite and indefinite articles and the introduction of partitive expressions in Romance. In, for instance, French this opposition is expressed by the one between a partitive expression (manger du pain) and the one with a definite or indefinite article (manger le pain / un pain). There are certainly traces of the development towards the Romance system of definite and indefinite articles in vulgar Late Latin, but there was no systematic use of such devices in Latin.16 There seems in fact to be no Latin example of an article clearly expressing the telicity once expressed by the prefix: i.e. there seems to be no example of panem comedo ‘I eat up a / the bread’ being replaced by for instance *panem unum comedo or *panem illum comedo. What we do have, however, are some examples of partitive expressions with the prepositions de, ex or ab, which underline the atelicity once indicated by the unprefixed verb (11a–c). We have such examples with bibo ‘drink (some) of ’, comedo ‘eat (some) of ’ and manduco ‘eat some of ’. But the problem with these examples is that many of them occur in texts translated from Greek and often seem to correspond to a similar expression in Greek, or in texts written by Greeks in Latin. Anyway these partitive expressions were probably acceptable in Late

15. See Haverling (2000, pp. 166, 205 f., 224–225, 255, 265 f., 420; 2005; forthcoming, section 2.2.3). 16. See, e.g., Szantyr (1965, § 106) and Väänänen (1981, § 275).



Gerd V. M. Haverling

Latin (there are no such examples in earlier Latin), but there was no systematic use of them to make the difference between ‘eat up’ and ‘eat some of ’ clear:17 (11) a.

4th c. CE, Vulg. Ier. 51.8 de uino eius biberunt gentes of wineprep+abl hergen have drunkenpf ind 3sg the nationsnom ‘the nations have drunken of her wine’

b.

4th c. CE, Vulg. 1 Reg. 13.28 non comedit leo de cadauere neg pf ind 3sg nom not has eaten the lion of the carcassprep+abl ‘the lion has not eaten of the carcass’

c.

4th c. CE, NT I Cor. 9.7 & Vulg. 1 Cor. 9.7 τίς φυτεύει ἀμπελῶνα καὶ τὸν καρπὸν αὐτοῦ οὐκ whonom plantspr ind a vineacc andconj the fruitacc itsgen notneg quis plantat uineam et fructum eius non nom pr ind acc conj who plants a vine and the fruitacc itsgen notneg

ἐσθίει; ἢ τίς eatspr ind? Or whonom edit, quis eatspr ind, whonom

ἐκ τοῦ γάλακτος τῆς ποίμνης οὐκ ἐσθίει; of the milkprep+gen of the flockgen notneg eatspr ind? de lacte gregis non manducat? of the milkprep+gen of the flockgen notneg eatspr ind? ‘Who plants a vineyard without eating any of its fruit? Who tends a flock without getting some of the milk?’

ποιμαίνει nourishespr ind pascit nourishespr ind

ποίμνην a flockacc gregem a flockacc

καὶ andconj et andconj

Two prefixes, in- and ad-, which both put the focus on the beginning and on the gradual character of a change in Classical Latin, are still productive in Late Latin. Verbs prefixed with these prefixes now tend to replace as well unprefixed as prefixed scoverbs. Ad- now occurs in a great variety of contexts, sometimes with an unclear

17. See Haverling (2003b, p. 129 ff.; forthcoming, section 2.2.4). – In early Germanic, such as Gothic, there is an interaction between the actional markers on the verbs and the cases, i.e. atelic verbs are combined with the partitive genitive and telic verbs with the accusative; this system was eventually replaced by the use of articles: see Abraham (1997), Philippi (1997) and Leiss (2000); cf. also e.g. Comrie (1989, p. 127) and Croft (1990, p. 177).

On the development of actionality, tense, and viewpoint from Early to Late Latin 

semantic function, but in some cases it seems to replace the function of the sco-suffix or of other prefixes (12a–c):18 (12) a. ad- is added to old verbs: i. ca. 200 BCE: garrio ‘chatter’ ii. 5th c. CE: aggarrio ‘chatter’ b. ad- is replacing old unprefixed sco-verb: i. 2nd c. BCE: uesperascit ‘it is approaching evening’ ii. ca. 400 CE: aduesperat ‘it gets towards evening’ c.

ad- is replacing old prefixed sco-verb: i. 2nd c. BCE: condormisco ‘fall asleep’ ii. 1st c. BCE: obdormisco ‘fall asleep’ iii. 5th c. CE: addormio ‘fall asleep’ iv. 6th c. CE: addormisco ‘fall asleep’ v. Standard Italian: addormentarsi

The prefix in- occurs in a number of new intransitive so-called parasynthetical verbs which correspond to old sco-verbs (13a–c). In Late Latin the new transitive verbs become a lot more common than the old sco-verbs, while the new intransitive verbs remain very rare. A great number of the instances of the transitive verbs are, however, in the passive voice and this fact may hide a rather extensive use of the new intransitive verbs in the language of everyday conversation:19 (13) a. crassescere > incrassare ‘become thick, fat, thicken’ i. 1st c. CE: crassesco ‘grow thicker, thicken’ ii. 4th c. CE: incrassesco ‘grow fatter’ iii. 3rd c. CE: incrasso tr. ‘make fat(ter)’ – incrasso intr. ‘became fat, grow fat(ter)’ b. pinguescere > impinguare ‘become thick, fat, thicken’ i. 1st c. BCE: pinguesco ‘grow fatter’ ii. 4th c. CE: impinguesco ‘grow thick, fat’ iii. 4th c. CE: impinguo tr. ‘make fat’ – impinguo intr. ‘become fat, grow fat’ c.

in- is replacing old prefixed sco-verb i. 2nd c. BCE: condormisco ‘fall asleep’ ii. 1st c. BCE: obdormisco ‘fall asleep’ iii. 4th c. CE: indormisco ‘fall asleep’

18. See Haverling (2000, p. 284 ff.; 2003a, p. 164 f.; 2003b, p. 124 f.; forthcoming, section 2.2.5). 19. See Haverling (2000, pp. 153 ff., 302 ff., 313 f.; 2003a, p. 168 f.; 2003b, p. 123 ff.; forthcoming, sections 2.2.2, 2.2.5); cf. Feltenius (1977, e.g. pp. 96 ff., 103, 109); on these so-called parasynthetical verbs see e.g. Crocco Galèas & Iacobini (1993).



Gerd V. M. Haverling

iv. 4th c. CE: indormio ‘fall asleep’ v. French: s’endormir The intransitive function of the sco-verbs is in Late Latin sometimes replaced by the passive voice of a transitive verb or even by the intransitive use of a previously intransitive verb (14a); and several previous sco-verbs are replaced by reflexive expressions in Late Latin and in Romance (14b):20 (14) a. change within the same verbal family: i. crassesco intr. > incrassor pass. > incrasso intr. > It. incrassare, Fr. engraisser, s’engraisser ii. Lat. calesco > It. riscaldarsi and Fr. s’échauffer b. change to another verbal family: i. Lat. proficiscor > se or sibi uadere > It. andarsene and Fr. s’en aller The changes in the actional system take place gradually and over a long period of time, but in the texts from the late 2nd and the early 3rd centuries CE, it is quite clear that the old system does not work as it used to in Classical Latin (cf. ex. 10). In the earlier periods those who spoke and wrote Latin chose between the different verbs with various affixes according to the actional content expressed by each of these verbs, but in the later periods they chose between the old verbs according to their traditional pragmatic functions rather than according to the way they were formed and new forms were created to replace some of the old ones. There are probably several different reasons for these changes in Latin. One relevant factor is probably the extremely complex character of the Early and Classical Latin actional system, where the most important prefixes involved all combine various functions and where several different prefixes sometimes overlap the functions of one another.21

2. Tense The Latin tense system is the result of the blurring of the Proto-Indo-European perfect and aorist and the Latin verb is strictly divided into two stems, the infectum and the perfectum. As a result formerly aspectual oppositions are reinterpreted as temporal.22 There are synthetic forms in the active voice that indicate past, present and future in both the infectum and the perfectum. In the passive voice there are periphrastic forms in the perfectum (15). The Latin tense system is fully developed only in the indicative: 20. See e.g. Szantyr (1965, § 164); cf. also García-Hernández (1998) and Haverling (2003a, p. 169; 2005; forthcoming, section 2.2.2). 21. Cf. Hopper & Traugott (2003, p.p. 102 ff., 116 ff.) and Traugott & Dasher (2002, p. 11 ff.). 22. See e.g. Szantyr (1965, § 178), and Baldi (2002, pp. 366–367, 377 ff., 398 f.); cf. Haverling (forthcoming, section 3).

On the development of actionality, tense, and viewpoint from Early to Late Latin 

(15) Tenses in the indicative in Classical Latin Infectum Present Imperfect Future

Active uoco uocabam uocabo

Passive uocor uocabar uocabor

Perfectum Perfect Pluperfect Future perfect

Active uocaui uocaueram uocauero

Passive uocatus sum uocatus eram uocatus ero

The problematic form in this new system is the Latin perfect tense form which is used both as a past tense form (16a) and as a perfect, indicating Current Relevance (16b):23 (16) a.

as past tense: Lat. ueni = Gr. aorist ἀφικόμην, Cic. Fam. 11.5.1 Romam autem ueni a. d. V Id. Dec. to Romeacc howeverconj I camepf ind on the 9th of Decemberadverbial ‘I came to Rome on the 9th of December’

b. as perfect (Current Relevance): Lat. ueni (perueni) = Gr. perf. ἀφῖγμαι, Plaut. Merc. 939 nunc perueni Chalcidem nowadv I have come topf ind to Chalcisacc ‘now I have reached Chalcis’ As opposed to the Greek perfect in Homeric (17a) or even in later times (18a), the synthetic Latin perfect tense form cannot indicate Resultativity. In that function we either have verbs in the present tense (17b) or periphrastic expressions with esse ‘be’, if the voice is passive (18b), and habere ‘have’, if it is active (18c):24 (17) a. Hom. Od. 19.222–223 ἤδη γὰρ οἱ ἐεικοστὸν ἔτος ἐστὶν,/ ἐξ οὗ nowadv forconj for himdat twentieth yearacc ispr ind, sinceprep+gen

κεῖθεν ἔβη καὶ ἐμῆς ἀπελήλυθε from hereadv he wentaor ind andconj from mygen is gonepf ind πάτρης from countrygen ‘it is now the twentieth year since he went from there and is gone from my country’

b.

Cic. Att. 13.31.2 Quoniam etiamnum abes Sinceconj stilladv you are awaypr ind 2sg ‘Since you are still away’

23. See e.g. Pinkster (1990, p. 230 ff.); cf. Haverling (2002, p. 155 f.; forthcoming, section 3). – On past tenses and perfects see also Comrie (1978, p. 52) and Bybee & Perkins & Pagliuca (1994, p. 61 f.). 24. See e.g. Bybee & Perkins & Pagliuca (1994, p. 68 f.) and Johanson (2000, pp. 112 ff., 114); on Latin cf. Szantyr (1965, § 178.c.b) and Haverling (2002, p. 156; forthcoming, section 3.1.3).



Gerd V. M. Haverling

(18) a.

Plutarch. Caes. 32.8 “ἀνερρίφθω κύβος!” may be castpf pass imperat dienom ‘“Let the die be cast!”’

b.

Suet. Iul. 33 ‘Iacta alea est’ pf ptc nom cast die ispr ind ‘“The die is cast”’

c.

Cic. Fam. 3.10.7 Nam si me uirum bonum... in maximis rebus because ifconj meacc good manacc in greatest thingsprep+abl perspectum habes pf ptc acc seen through you havepr ind 2sg ‘If you have satisfied yourself that I am an honourable man in affairs of the greatest gravity’

The Latin perfect indicates that a past situation does not hold in the present. Latin uses the present tense and not the perfect with durative temporal adverbials when the situation in question is regarded as still lasting, i.e. to indicate ‘anterior continuing’ (19a); we have a similar usage in Greek (19b). There is a corresponding use of the imperfect tense in Classical Latin (ex. 46b).25 The Latin perfect indicates that a situation does not hold in the present and the perfect tense forms fuit ‘has been’ and uixerunt ‘they have lived’ indicate that something or somebody does not exist any more (20a–b); in this function the Latin perfect corresponds to the Greek aorist (20c):26 (19) a.

Cic. Verr. II 4.38 Is Lilybaei multos iam annos habitat nom loc acc adv acc he at Lilybaeum for many already for years is livingpr ind ‘he has now been living at Lilybaeum for many years’

b.

Aristoph. Plut. 437 Πενία μὲν οὖν, ἣ σφῷν ξυνοικῶ povertynom in factadv, whonom for you twodat live withpr ind 1sg πόλλ’ ἔτη for many yearsacc ‘I am Poverty, who have been your mate for many a year!’

25. See e.g. Szantyr (1965, § 170 Zus. a) and Smyth (1956, § 1885); and cf. Comrie (1978, p. 60) and Bybee & Perkins & Pagliuca (1994, p. 62); cf. Haverling (2002, p. 158; forthcoming, section 3.1.1). 26. See e.g. Haverling (2002, p. 161 f.; forthcoming, section 3.1.2); cf. Johanson (2000, p. 81 f.).

On the development of actionality, tense, and viewpoint from Early to Late Latin 

(20) a.

Verg. Aen. 2.325 fuimus Troes, fuit Ilium we have beenpf ind Trojansnom, has beenpf ind Iliumnom ‘We Trojans are not, Ilium is not’

b.

Cic. Rep. 6.16 qui iam uixerunt whonom alreadyadv have livedpf ind ‘who have completed their earthly lives’

c.

Plutarch. Cic. 22.4 “ἔζησαν” εἶπεν they have livedaor ind he saidaor ind ‘he said: “They have lived”’

In the periphrastic expressions in the passive voice of the perfectum, there is an opposition between the perfect or pluperfect with sum ‘I am’ and eram ‘was’ and the anterior situation indicated by the periphrasis with fui ‘was’ and fueram ‘had been’. This system is, however, never quite consistent and already in Early Latin there are signs of the fact that the difference between the pluperfect and the expressions with fueram was not always clear (21a–b):27 (21) a. Cic. De orat. 1.187 omnia fere quae sunt conclusa nunc all the thingsnom roughlyadv thatnom arepr ind includedpf ptc nom nowadv

artibus, dispersa et dissipata quondam abl pf ptc nom conj pf ptc nom in the arts , dispersed and scattered onceadv

fuerunt werepf ind ‘nearly all elements, now forming the content of arts, were once without order or correlation’

b. Caes. Ciu. 1.41.1 pons, qui fuerat tempestate nom nom plupf ind the bridge which had been by the stormabl

interruptus, paene erat refectus interruptedpf ptc nom, almostadv wasimpf ind reconstructedpf ptc nom ‘the bridge, the construction of which had been interrupted by the storm, was almost reconstructed’

The tense system met with in Early and Classical Latin puts much emphasis on the relative sequence of events. We find this in the way absolute-relative tenses are used for 27. See e.g. Szantyr (1965, § 179 Zus. b); cf. Haverling (forthcoming, section 3.3.3).



Gerd V. M. Haverling

instance in certain temporal clauses (22a–c). This is the background of the famous Latin system of the sequence of tenses:28 (22) a. priusquam ‘before’ + fut. ex., Cic. Flacc. 51 nihil tamen contra disputabo priusquam nothingacc anywayadv againstadv I shall arguefut spl beforeconj

dixerit he will have spokenfut ex ‘I shall not argue against it before he has spoken’

b. cum ‘when’ + plupf. subj., Cic. Fam. 13.29.4 ibi eum Caesar cum uidisset, nihil adv acc nom conj impf subj there him Caesar when had seen , nothingacc

aspere … dixit harshlyadv … he saidpf ind ‘as Caesar had seen him there, he did not utter a harsh word’

c. dum ‘as long as’ + simple fut., Cic. Fam. 6.3.4 Mihi tu, tui, tua omnia for medat younom, your peoplenom, all your thingsnom

maximae curae sunt et, dum dat pr ind conj for the greatest concern are and , as long asconj

uiuam, erunt I shall livefut spl will befut spl ‘I care for you and your family and all that concerns you and shall do so as long as I live’

In Early Latin there are some elements which seem to reflect the Proto-Indo-European aorist. There are a few examples of the so-called ‘sigmatic future’, which sometimes seems to have an aspectual function and sometimes indicates absolute-relative tense (23a–c); and there is an apparently ‘perfective’ use of the perfect infinitive, especially in some old prohibitive formulas (24):29 (23) a. Plaut. Rud. 578 Tu istaec mihi dato; exarescent younom those thingsacc to medat givefut imperat they get dryfut spl

28. See e.g. Szantyr (1965, §§ 180.b, 297.II) and Pinkster (1990, p. 226 f.); cf. Comrie (1985, pp. 64 f., 69 ff., 74), cf. Haverling (forthcoming, sections 3.2.2, 3.2.3, 3.3.2). 29. See Rix (1998, pp. 621 ff., 624); Haverling (forthcoming, sections 3.2.2, 3.2.4).

On the development of actionality, tense, and viewpoint from Early to Late Latin 

faxo I shall do itsigm fut ‘Give me those things; I shall see to it that they get dry’

b. Pacuv. Trag. 325 ap. Non. p. 74.30 Haud sinam quidquam profari priusquam notneg I shall letfut spl anythingacc pronouncepr inf beforeconj

accepso quod peto sigm fut I shall have received whatacc I demandpr ind ‘No, I’ll not let you speak a single word, before I have received what I demand’

c.

Plaut. Asin. 839 ne dixis istuc notconj neg you shall saysigm fut 2sg that of yoursacc ‘Do not say that!’

(24)

CIL I2.581.3 neiquis eorum Sacanal (sc. Bacchanal) habuise uelet no-onenom of themgen pl Bacchic worshipacc have hadpf inf may wantpr subj ‘let none of them have in mind to have a place for Bacchic worship’

Some such elements remain in Classical Latin, for instance in the use of the perfect subjunctive in prohibitions, where the perfect subjunctive often points to a particular situation (25a) as opposed to the present subjunctive that points to a general principle (25b), and in expressions indicating potentiality, where there often seems to be an overlap between the perfect and the present subjunctive and sometimes even the simple future (26a–c):30 (25) a. Cic. Parad. 41 tu posse te dicito, quoniam quidem younom 2sg be able topr inf youacc sayfut imperat, sinceconj indeedadv

potes, debere ne dixeris, you canpr ind 2sg, be obliged topr inf notconj neg you saypf subj 2sg,

quoniam nihil quisquam debet sinceconj nothingacc anyonenom mustpr ind 3sg ‘Say that you can, since you can, but do not say that you must, because nobody must anything’

b. Cic. Cato 33 Isto bono utare, dum adsit, cum that blessingabl usepr subj 2sg, as long asconj is therepr subj, whenconj 30. See e.g. Szantyr (1965, § 186.III Zus. b); cf. Haverling (2002, p. 159; forthcoming, section 3.2.4).



Gerd V. M. Haverling

absit, ne requiras is gonepr subj, notconj neg you pine forpr subj 2sg ‘you must enjoy that blessing so long at it is there, when it is gone you must not pine for it’

(26) a.

perf. subj.: Cic. Cato 8 sed fortasse dixerit quispiam conj adv pf subj but perhaps will say someonenom ‘but perhaps someone will say that’

b.

pres. subj.: Liv. 5.52.5 Forsitan aliquis dicat perhapsadv someonenom may saypr subj ‘Perhaps someone may say that’

c.

simple fut.: Cic. Sull. 84 dicet fortasse quispiam will sayfut spl perhapsadv someonenom ‘perhaps someone will say’

The system met with in Early and Classical Latin thus puts a strong emphasis on tense and on the relative sequence of events, but it is not entirely without contradictions. There also seems to be some difference between the various levels of style: the future perfect seems, for instance, to hold a particularly strong position in the colloquial language.31 In Late Latin some of the former tense forms are replaced by others and there is a shift between the functions of some of the tense forms. The Classical system distinguished between the anterior past and the normal past, but in colloquial Late Latin this system is lost and in the passive voice there is a tendency to use periphrastic forms in the infectum too, so that we get a system where the synthetic present tense form is replaced by a periphrastic one (27a) and where the old perfect is replaced by the former anterior perfect (27b); related to this development is the resultative function in expressions like iacta alea est ‘the die is cast’ in Classical Latin (ex. 18b). There are, however, notable differences between different kinds of texts in this respect; the tense system met with in the literary texts from the later centuries is closer to the system met with in Classical Latin than the system met with in the more vulgar texts:32 (27) a.

Cl. Lat. audiatur > Late Latin auditus sit, Peregr. 36.3, ut... gemitus populi omnis auditus sit so thatconj the sighnom of the whole peoplegen heardpf ptc nom ispr subj ‘so that... the sigh of the whole people is heard’

31. See Haverling (forthcoming, section 3.2.4). 32. See, e.g., Szantyr (1965, §§ 170 Zus. d, 179 Zus. b, 211.a); cf. Haverling (2002, p. 156 ff.; 2005; forthcoming, sections 3.1.5, 3.3.5).

On the development of actionality, tense, and viewpoint from Early to Late Latin 

b.

Cl. Lat. condita est > Late Latin condita fuit, Justin. 13.7.1 Cyrene autem condita fuit ab Aristaeo Cyrenenom howeverconj foundedpf ptc nom waspf ind by Aristaeusprep+abl ‘Cyrene was founded by Aristaeus’

In Classical Latin the active periphrastic expression with habeo ‘have’ and the past participle expressed Resultativity (ex. 18c), but in Late Latin this construction gradually loses its resultative character and may indicate Current Relevance; we have undisputable examples of this development in texts from the 6th c. CE. In the earlier examples the participle is always agreeing with the direct object (28a), but in the 6th century translation of a medical text this is not always the case (28b):33 (28) a. 6th c.: Cl. Lat. inuitauisti > Late Latin inuitatum habes, Greg. Tur. Vit. patr. 3.1 ‘Ecce episcopum... inuitatum habes’ Lookadv the bishopacc invitedpf ptc acc you havepr ind 2sg ‘“And look you have invited the bishop”’ b.

6th c.: Oribas Syn. 7.48 omnia probatum habemus everythingneutr pl acc examinedneutr sg acc we havepr ind ‘we have examined everything’

Several new periphrastic expressions are used to indicate the future. The most successful of these is the one with habeo ‘have’ and the infinitive which gradually replaces the simple future (e.g. laudabo > laudare habeo ‘I will praise’). There are numerous examples from the 3rd century onwards (29a). In the colloquial language of around 600 CE, the construction developed into a new synthetic future, as indicated by the pun on the name of the city Daras which may now also mean “you shall give” (from dare habes = Cl. Lat. dabis; 29b). The new construction was probably quite common in the colloquial language, but it is avoided in the more literary texts. We therefore find more examples of it in the less elegant translations of the Bible (30a) than in the official Versio Vulgata (30b):34 (29) a.

3rd c., Porphyr. Hor. Epist. 2.1.17 oriturum (esse): nasci habere about to be bornfut ptc acc to bepr inf be bornpr inf to havepr inf ‘be about to be born’

33. See, e.g., Szantyr (1965, §§ 178.c.b, 181). – This change follows a pattern met with in many languages in the world: see Bybee & Perkins & Pagliuca (1994, p. 105); cf. Haverling (forthcoming, section 3.1.5). 34. See, e.g., Szantyr (1965, § 175); cf. Haverling (forthcoming, section 3.2.5).

 Gerd V. M. Haverling

b. 7th c., Fredeg. Chron. 2.62.15 Et ille respondebat: ‘Non Andconj henom answeredimpf ind notneg

dabo’. I shall givefut spl

Iustinianus dicebat. ‘Daras’ Justiniannom saidimpf ind you shall givefut 2sg ‘And he answered: ‘I shall not give’. Iustinian said: “Daras” (= you shall give)’

(30) a.

Vet. Lat. Luc. 5 19.4 quia inde habebat transire sinceconj that wayadv he had toimpf ind passpr inf ‘for he was to pass that way’

b.

Vulg. Luc. 19.4 quia inde erat transiturus conj adv impf ind since that way (he) was about to be passingfut ptc nom ‘for he was to pass that way’

Classical Latin put strong emphasis on a detailed description of the relative sequence of events (cf. ex. 21 and 22). Late Latin is less concerned with that and we may now find the present subjunctive instead of the perfect future, the imperfect instead of the pluperfect and even the present indicative instead of the future in expressions like ‘as long as I live’ (31a–c):35 (31) a. priusquam + pres. subj., Vulg. Marc. 14.70 priusquam gallus cantet bis ter me beforeconj the cocknom singspr subj twiceadv three timesadv meacc

negabis you will denyfut spl 2sg ‘before the second cock-crow you will deny me three times’

b. cum + impf. subj., Vulg. Luc. 13.12 quam cum uideret Iesus uocauit ad se acc conj impf subj nom her when saw Jesus calledpf ind to himselfprep+acc

et ait illi... andconj sayspr ind to herdat ‘And when Jesus saw her, he called her and said to her …’

c. quamdiu ‘as long as’ + pres., Aug. In Psalm. 36 Enarr. Serm. 1.11 faciam quidquid uolo quamdiu uiuo, fut spl acc pr ind conj I will do whatever I want to as long as I livepr ind,

35. See Szantyr (1965, §§ 180, 297.II.B); cf. Haverling (forthcoming, sections 3.2.5, 3.3.5).

On the development of actionality, tense, and viewpoint from Early to Late Latin 

postea non ero adv ´ afterwards notneg I will befut spl ‘I will do whatever I want to as long as I live, afterwards I will not be’

Late Latin has not, however, lost the sensitivity to the relative sequence of events, but rather got a new kind of sensitivity to these things. We now find the pluperfect where Classical Latin had had the perfect, for instance with the conjunction postquam ‘after’ or with the adverb paene ‘almost’ (32a–b); the pluperfect now also tends to replace the imperfect in indications of anterior continuing in the past (ex. 49).36 The perfect subjunctive met with in prohibitions and expressions indicating potentiality in Classical Latin (ex. 25 and 26) is often replaced by the present subjunctive in Late Latin (33a– b):37 (32) a. Vulg. Gen. 17.1 postquam uero nonaginta et nouem annorum esse afterconj howeverconj ninety-ninenum of yearsgen to bepr inf

coeperat, apparuit ei Dominus plupf ind pf ind dat had begun appeared to him the Lordnom ‘When he was ninety-nine years old the Lord appeared to him’

b.

Hier. Tract. p. 514 l. 247 quod paene obliti fueramus dicere whatacc almostadv (we) had forgottenplupf ind to saypr inf ‘what we had almost forgotten to say’

(33) a. in prohibitions: Aug. Spec. 23 et ne dixeris: ‘quomodo potui?’... et andconj notconj neg you saypf subj 2sg: ‘howadv I couldpf ind 1sg?’ andconj

ne dicas: ‘miseratio dei magna est...’ notconj neg you saypr subj 2sg: ‘compassionnom of Godgen greatnom ispr ind’ ‘and do not say “how could I?”... and do not say: “God’s compassion is great...”’

b.

potentiality: Ambr. Expos. CXVIII psalm. 20.7 fortasse dicat aliquis perhapsadv may saypr subj someonenom ‘perhaps someone will say’

Signs of ongoing changes in the tense system are met with in the texts from the late 2nd and the early 3rd century CE, when we get the first examples of the new construction with habere and the infinitive indicating future and a growing tendency to use the periphrastic form with esse and the past participle instead of the synthetic passive 36. See, e.g., Szantyr (1965, §§ 183.b, 322.a); cf. Haverling (forthcoming, sections 3.3.2, 3.3.5). 37. See Szantyr (1965, § 186.III); cf. Haverling (forthcoming, section 3.2.5).



Gerd V. M. Haverling

present met with in Classical Latin. Some 200 years later, in the late 4th century and early 5th century, the passive perfect and pluperfect tense forms are regularly created with fui and fueram in the more vulgar texts. After two more centuries, in the late 6th and early 7th centuries, we have the earliest undisputable examples of the new perfect tense form with habere + participle and the earliest evidence for the new synthetic future created from the habere + infinitive construction.

3. Viewpoint In Classical Latin we find an opposition between the perfect and the imperfect in all the forms of actionality and there are similar oppositions between the past tense forms and the imperfect tense forms in for instance Greek and in the Romance languages too. In several other languages a state or an atelic activity is always presented in the imperfective viewpoint and in Latin the use of the imperfect tense is more common when the actionality is non-dynamic or atelic than when it is telic or momentaneous. The perfect gives us an overview of a past situation or action and the imperfect offers us a description of a situation from within and may among other things indicate progressivity (34a). In certain cases the Latin viewpoint opposition resembles systems met with in many other languages (34b–d):38 (34) a. Latin impf. (background) – pf. (event), Cic. Fam. 9.7.1 Cenabam apud Seium, cum utrique nostrum I was diningimpf ind with Seiusprep+acc, whenconj to bothdat of usgen

redditae sunt a te litterae pf ind pass prep+abl were given from you lettersnom ‘I was dining in Seius’ home, when we both got letters from you’

b. Greek impf. (background) – aorist (event), Xenoph. Anab. 1.10.10 ἐν ᾧ δὲ ταῦτα ἐβουλεύοντο , In thatprep+dat butconj those thingsacc they were discussingimpf ind,

καὶ δὴ βασιλεὺς παραμειψάμενος andconj thenadv the kingnom having putaor ptc med nom

εἰς τὸ αὐτὸ σχῆμα κατέστησεν ἀντίαν prep+acc to the same pattern put uppaor ind oppositeacc

38. See, e.g., Comrie (1978, p. 16 ff.), Bybee & Perkins & Pagliuca (1994, p. 133 ff.), Smith (1997, pp. 74 f., 79, 89, 150), Giorgi & Pianesi (1997, pp. 175, 177) and Johanson (2000, pp. 80–81); cf., e.g., Pinkster (1990, p. 227 ff., 237 ff.), Oldsjö (2001, pp. 177 ff., 205 ff., 227 ff., 242 ff.) and Haverling (2001, p. 355 ff.; forthcoming, section 4.1); on Greek cf. e.g. Smyth (1956, § 1896) and Rijksbaron (1994, pp. 12–16, 28 ff.).

On the development of actionality, tense, and viewpoint from Early to Late Latin 

τὴν φάλαγγα the line of battleacc ‘But while they were taking counsel about this matter, the king changed his line of battle to the same form as theirs’ c. Russian imperfective – perfective My pisali pis’mo, kogda on Wenom were writingpast ind impfv a letteracc whenconj henom

pozvonil calledpast ind perfv ‘We were writing a letter when he called’

d. Romance – Italian, progressive periphrasis vs. passato remoto Ieri Gianni stava raggiungendo la vetta, yesterdayadv Gianni he stoodimpf ind reachingger (= progr perifr) the top,

quando un violento temporale glie- lo impedì conj dat when a violent storm for him it preventedpass rem ‘Yesterday Gianni was reaching the top, when a violent storm prevented him’

The Latin perfect, as well as the Greek aorist, is used to describe a series of events or situations (35a). The aorist is, as a matter of fact, used in the Greek translation of one of the most famous of all quotations with the Latin perfect used in this manner, Caesar’s brag after the Pontic victory (35b):39 (35) a.

Lat. perf., Suet. Iul. 37.2 ‘ueni, uidi, uici’ I camepf ind, I sawpf ind, I conqueredpf ind ‘“I came, I saw, I conquered”’

b.

Greek aorist, Plutarch. Caes. 50.3 “ἦλθον, εἶδον, ἐνίκησα” I cameaor ind, I sawaor ind, I conqueredaor ind ‘“I came, I saw, I conquered”’

In certain cases, however, the viewpoint opposition met with in Classical Latin does not resemble systems met with in many other languages. This regards not least the treatment of viewpoint when the actionality is stative. We have, for instance, a viewpoint opposition in the descriptions of permanent states in the past, such as the identity of someone’s parent (36a–b) or the location of a building (37a–b); this is not the

39. See Smyth (1956, § 1927); cf. Pinkster (1990, p. 237 f.); cf. Haverling (forthcoming, section 4.1.1).



Gerd V. M. Haverling

case in Ancient Greek (36c, 37c) or in Modern Romance languages such as French and Italian (36d, 37d):40 (36) a.

Latin – identity of parent: overview (perf.), Nep. Them. 1.2 Pater eius Neocles generosus fuit Fathernom hisgen Neoclesnom well-bornnom waspf ind ‘His father Neocles was of high birth’

b. Latin – identity of parent: background description (imperf.), Sen. Dial. 5.18.3 Modo C. Caesar Sex. Papinium, cui pater Recentlyadv C. Caesarnom Sex. Papiniumacc, of whomdat the fathernom

erat consularis,... fagellis cecidit... wasimpf ind of consular ranknom with a whipabl floggedpf ind ‘Recently C. Caesar flogged Sextus Papinius, whose father was of consular rank’

c. Greek impf., Plutarch. Them. 1.1 πατρὸς γὰρ ἦν

fathergen

becauseconj

wasimpf ind 3sg

of he ‘Because his father was Neocles’

Νεοκλέους of Neoclesgen

d. Romance imperf. – French, Guillemin (1923, p. 14) Son père Néoclès était de grand famille

His fathernom Neoclesnom wasimpf of great family ‘His father Neocles was of high birth’

(37) a. Latin – location of building or sim.: overview (perf.) Cic. Dom. 101 in Vacci pratis domus fuit M. Vacci, inprep Vaccus’gen meadsabl housenom wasimpf ind of M. Vaccusgen,

quae publicata est et euersa whichnom was confiscatedpf ind andconj (was) razedpf ind pass ‘On Vaccus’ Meads stood the house of Marcus Vaccus, which was confiscated and razed’

b. Latin – location of building or sim. (imperf.), Caes. Gall. 7.68–69 Caesar (Alesiam)... circumuallare instituit. Caesarnom (Alesiaacc) build a wall aroundpr inf act decidedpf ind act.

Ipsum erat oppidum in colle summo... Thatnom wasimpf ind townnom on a hillprep+abl most highabl ‘Caesar decided to build a wall around Alesia. The town of Alesia was set atop a hill’

40. See Haverling (2001, p. 357 f.; 2002, p. 163 f.; forthcoming, section 4.1.4). This development was not yet completed in Early Romance: see Fleischman (1990, p. 42 ff.).

On the development of actionality, tense, and viewpoint from Early to Late Latin 

c. Greek impf., Hom. Od. 17.208 ἀμφὶ δ̉̉ ἄρ̉ αἰγείρων ὑδατοτρεφέων ἦν aroundADV but thereconj poplars bred in watergen wasimpf ind

ἄλσος a grove of poplarsnom ‘around was a grove of poplars, that grew by the waters’

d.

Romance imperf. – Italian, (Squartini 1995, p. 119) La casa si trovava in campagna The housenom wasimpf ind in the countryside ‘The house was in the countryside’

However, when the past state is not permanent, the viewpoint opposition in Latin (38a–b) corresponds better to the systems met with in Romance (39a–b) and Greek (40a–b):41 (38) a.

Latin – overview (perf.), Nep. Them. 6.1 Magnus hoc bello Themistocles fuit Greatnom in this warabl Themistoclesnom waspf ind ‘Themistocles was great in this war’

b.

Latin – background description (imperf.), Nep. Dat. 2.1–3 Pari se uirtute postea praebuit with equalabl himselfacc with valourabl thereafteradv showedpf ind (sc. Datames) ... Erat eo tempore Thuys dynastes (Datamesnom) … wasimpf ind at that timeabl Thuysnom princenom

Paphlagoniae of Paphlagoniagen ‘Datames later showed himself equally valiant... At that time there was a prince of Paphlagonia called Thuy’

(39) a.

Romance – French, overview (passé simple), Guillemin (1923, p. 18) S’ il fut grand dans la guerre nom pass spl ind If he was greatnom during the war ‘If he was great during the war’

b. Romance – French, background description (imparfait), Guillemin (1923, p. 75) Il y avait à cette époque un prince de Paphlagonie ... Thuys There wasimpf ind then a prince of Paphlagonia … Thuys ‘At that time there was a prince of Paphlagonia called Thuy’

41. See Haverling (2001, p. 359 f.; 2002, p. 161; forthcoming, section 4.1.4).



Gerd V. M. Haverling

(40) a.

Greek – overview (aorist), Thuc. Hist. 5.5.1 καὶ ἐγένετο Μεσσήνη Λοκρῶν τινὰ χρόνον andconj wasaor ind Messenenom of the Locriansgen for some timeacc ‘and Messene was in the hands of the Locrians for some time’

b. Greek – background description (imperf.), Xenoph. Anab. 6.6.5 ἐτύγχανε δὲ τò στράτευμα ἔξω ὄν impf ind conj nom adv happened to because the army outside beingpr ptc

ὅτε ἀφίκετο whenconj he cameaor ind ‘the army happened to be outside, when he came’

Classical Latin has an actional system that distinguishes states from changes into states and the difference between ‘I was silent, did not speak’ and ‘I fell silent, stopped talking’ is expressed by the opposition between an unprefixed (tacui) and a prefixed (conticui) perfect tense form (ex. 3 and 7a). This pattern pervades the whole verbal system and means that the perfect tense form habui ‘I had, have had’ is not used in the sense ‘I got, acquired in Classical Latin (41a–d). Instead we find other verbs, for instance genui, if it is a child that someone has got, or accepi, if it is a letter. Classical Greek (42) and Modern Romance (43a–b) may use forms of the same verb in both functions:42 (41) a. Latin – overview (perf.), Liv. 41.23.10 duos, ut scitis, habuit filios Philippus acc conj pr ind pf ind 3sg acc two , as you know , had sons Philipnom ‘As you know, Philip had two sons’ b. Latin – background description (imperf.), Liv. 24.48.13 Carthaginienses... ad Galam in parte altera the Carthaginiansnom to Galaprep+acc in another partprep+abl

Numidiae... regnantem legatos mittunt. Filium of Numidiagen reigningpr ptc acc legatesacc they send.pr ind A sonacc

Gala Masinissam habebat septem decem annos natum... nom Gala Masinissaacc hadimpf ind seventeennum yearsacc bornacc ‘the Carthaginians... sent legates to Gala, who reigned in another part of Numidia... Gala had a son Masinissa, seventeen years old’

c.

Latin – change into a state (other vb. perf.), Liv. 1.34.2 uxore ibi ducta duos filios genuit a wifeabl thereadv marriedpf ptc abl two sonsacc procreatedpf ind 3sg ‘he married there and begot (got, had) two sons’

42. See Haverling (2000, pp. 55 f., 242; 2002, p. 162 f.; forthcoming, section 4.1.5); cf. Comrie (1978, p. 19) and Smith (1997, p. 70).

On the development of actionality, tense, and viewpoint from Early to Late Latin 

d.

Latin – change into a state (other vb. perf.), Cic. Att. 9.14.1 eodem die... litteras accepi a Q. Pedio that same dayabl a letteracc I gotpf ind from Q. Pediusprep+abl ‘that same day … I got a letter from Q. Pedius’

(42) Greek – change into a state (aorist) vs. overview (imperf.), Plato Protag. 321d τὴν μὲν οὖν περὶ τὸν βίον σοφίαν ἄνθρωπος ταύτῃ thedef art in factadv about lifeprep+acc wisdomacc mannom in this waydat ἔσχεν τὴν δὲ πολιτικὴν οὐκ εἶχεν, ἦν γὰρ aor ind acc neg impf ind impf ind acquired , the civic one not had , was becauseconj

παρὰ τῷ Διί with Zeus prep+dat ‘although man acquired in this way the wisdom of everyday life, civic wisdom he had not, since this was in the possession of Zeus’ (43) Romance – state vs. change into it a. Italian: aveva un bambino (imperfetto) = ‘had a child’ vs. ebbe un bambino (passato remoto) = ‘begot/bore (had) a child’ / alt. ha avuto un bambino (passato prossimo) = ‘begot/bore (had) a child’ b. French: elle avait un enfant (imparfait) = ‘she bore (had) a child’ vs. elle eut un enfant (passé simple) = ‘she bore (had) a child’ / alt. elle a eu un enfant (passé composé) = ‘she bore (had) a child’ In Classical Latin there is a similar viewpoint opposition in the indication of plurioccasional events in the past; for instance the iterative verb uentitare ‘come often, repeatedly’ may occur in the perfect as well as in the imperfect.43 The Latin imperfect tense form may indicate progressivity (ex. 34). When it is combined with telic or momentaneous actionality, the action is often described as interrupted before reaching its goal. In Latin we encounter a momentaneous verb like uenire ‘come’ in the phrases meaning ‘they were on their way but turned away’ (44). When found with a negation the imperfect tense sometimes indicates what could not happen or what someone would not do (45a–b):44 (44) the ‘imperfective paradox’, Cic. Lig. 24 ueniebatis in Africam... prohibiti estis you were comingimpf ind 2pl to Africaprep+acc you were prevented frompf ind pass 2pl in prouincia uestra pedem ponere prep+abl in your province footacc putpr inf ‘you were on your way to Africa but you were not allowed to enter the province’ 43. See Haverling (forthcoming, section 4.1.2). 44. See Hedin (2000, pp. 230 f., 238 ff., 246–250); cf. Smith (1997, pp. 63, 75 ff., 83); cf. e.g. Szantyr (1965, §§ 168.d, 176 Zus. a), Pinkster (1983, pp. 301–303; 1990, p. 228) and Haverling (forthcoming, 4.1.6).



Gerd V. M. Haverling

(45) a.

with a negation, Petron. 51.1 phialam uitream, quae non frangebatur a cup of glassacc, whichnom notneg could be crushedimpf ind ‘a glass cup that was unbreakable’

b. with a negation, Liv. 34.4.10 aurum et purpuram data et oblata acc conj acc pf ptc acc conj gold and purple given and offeredpf ptc acc

ultro non accipiebant evenadv notneg would acceptimpf ind 3pl ‘they would not accept gold or purple voluntarily offered to them’

Sometimes the effect thus created between telicity and imperfectivity is rendered by expressions like ‘tried to’ in English translations and in the school grammars this usage is sometimes defined as ‘conative’. It should, however, be emphasised that Latin does not use this effect to the same extent as for instance Classical Greek and Russian. In Latin the difference between ‘tried to persuade’ and ‘did persuade’ is rendered by the actional opposition between an unprefixed and a prefixed verb (ex. 7b). Greek may in a similar case use the opposition between the imperfect tense and the aorist and Russian the one between the imperfective and the perfective aspects.45 The perfective viewpoint is generally preferred when stative and atelic expressions occur with durative temporal adverbials indicating for how long time a past situation lasted (46a); the imperfect tense with such adverbials usually indicates ‘anterior continuing’, thus implying that this situation was still going on in the past moment referred to (46b; cf. the corresponding use of the present tense in ex. 19a):46 (46) a. perfect – overview, Suet. Aug. 72.1 ac per annos amplius quadraginta eodem cubiculo conj prep+acc adv and for years more than fortynum in the same bedroomabl

hieme et aestate mansit in the winterabl andconj in the summerabl he stayedpf ind ‘for more than 40 years he remained in the same bedroom winter as summer’

b. imperfect – anterior continuing, Nep. Att. 12.3 qui complures annos studio ductus philosophiae nom acc abl pf ptc nom who for many years by interest driven of philosophygen

habitabat habebat -que he was livingimpf ind at Athensabl he hadimpf ind andconj

45. See Haverling (2001, p. 361; forthcoming, 4.1.6); cf. Szantyr (1965, § 168.d), Pinkster (1983, p. 303) and Oldsjö (2001, p. 239). 46. See Smith (1997, p. 115 ff.) and Johanson (2000, pp. 79 ff., 148); cf. Haverling (2001, p. 359 f.; 2002, p. 160 f.; forthcoming, section 4.1.3).

On the development of actionality, tense, and viewpoint from Early to Late Latin 

in Italia pretiosas possessiones prep+abl in Italy valuable possessionsacc ‘who for many years had been living at Athens because of his interest in philosophy and who had valuable possessions in Italy’

In Late Latin the opposition between prefixed and unprefixed verbs is blurred and unprefixed perfect tense forms that previously described past states may now indicate the change into that state (ex. 10). In Early and Classical Latin a perfect tense form like habui ‘had’ did not indicate change of state (ex. 41), but in Late Latin it may occur in such a function and have the sense ‘got’ and be used for instance of somebody who got a child (47). As a result the imperfect tense now tends to replace the perfect in the function of giving an overview of a past situation; this tendency is stronger in the less literary texts, for instance in the less literary translations of the Bible, than in more elegant and literary texts and in the Vulgate translation of the Bible:47 (47) a. perfect – change of state, filium habuit ‘begot (had) a son’, 6th c. CE, Joseph. Antiq. 1.3.4 (86) Μαθουσάλας … Λάμεχον υἱὸν ἔσχε Matusalasnom Lamech sonacc hadaor ind, Matusalas... Lamech filium habuit, Matusalasnom Lamech sonacc hadpf ind, περἰ ἔτη γεγονὼς prep+acc about years having becomepf ptc nom cum esset annorum whenconj he wasimpf subj of yearsgen

ἑπτὰ καὶ ὀγδοήκοντα καὶ ἑκατόν eighty-seven and a hundrednum octoginta septem et centum eighty-seven and a hundrednum ‘when Matusalas... was 187 years old he begot (had) his son Lamech’

b.

imperfect – past state, Joseph. Antiq. 2.13.1 (277) καὶ παραλαβὼν τὴν Σαπφώραν ἥν γεγαμήκει … andconj takingaor ptc Sefforaacc whomacc he had marriedplupf ind

sumens -que pr ptc nom taking andconj

Sefforam Sefforaacc

quam habebat uxorem... acc impf ind whom he had a wifeacc

καὶ τοὺς ἐξ αὐτῆς andconj thedef art from herprep+gen et filios quos habebat andconj the childrenacc whoacc he hadimpf ind

παῖδας … childrenacc ex ea... from herprep+abl

47. See Haverling (2000, p. 242; 2003b, p. 120 f.; 2005; forthcoming, section 4.2.1); on the change of state often indicated by stative past tense forms cf. Johanson (2000, p. 81 f.).

 Gerd V. M. Haverling

ὥρμησεν εἰς τὴν Αἴγυπτον he wentaor ind to Egyptprep+acc

tetendit ad Aegyptum wentpf ind to Egyptprep+acc ‘he took Seffora... whom he had for wife, and the children he had with her ...with him and went to Egypt’

The circumstance that the imperfect may now also be used to give an overview of a past situation is now seen in a growing number of cases in which the imperfect tense occurs with durative temporal adverbials without indicating ‘anterior continuing’. We have such examples in texts from the 4th century CE, when the phenomenon seems to have been more frequent in the colloquial language (48a); in the corresponding cases in Greek we have the aorist (48b). As a result of this development the pluperfect tends to replace the imperfect in the function of indicating ‘anterior continuing’, for instance in an 8th c. passage (49; cf. ex. 33):48 (48) a.

overview of past situation – imperfect, ca. 400 CE, Vulg. Luc. 1.24 et occultabat se mensibus quinque... andconj hidimpf ind herselfacc for five monthsabl ‘and for five months she hid herself ’

b.

overview of past situation – aorist in Greek, NT Luc. 1.24 καὶ περιέκρυβεν ἑαυτὴν μῆνας πέντε … conj and hidaoe ind 3sg herselfacc for five monthsacc ‘and for five months she hid herself ’

(49) ‘anterior continuing’ in the past – pluperfect, 8th c. CE: VSPE 4.10.8 illo benigne reddente atque illa cum gaudio heabl graciouslyadv returningpr ptc abl andconj sheabl with joyprep+abl

quae diu desiderauerat recipiente whatacc for a long timeadv had wantedplupf ind receivingpr ptc abl ‘as he was graciously returning and she joyfully receiving what she had so long desired’

Towards the end of the Late Latin period we find an increasing number of apparently progressive periphrases, seemingly replacing the imperfect tense in that function (50a–b). The earlier examples of this development are somewhat problematic and often met with in Christian texts translated from Greek and possibily influenced by Semitic, such as in the translations of the Bible. However, in the later centuries we find an increasing number of examples written directly in Latin. We can compare the use of

48. See Haverling (2001, p. 365; forthcoming, section 4.2.2).

On the development of actionality, tense, and viewpoint from Early to Late Latin 

the imperfect tense in Pliny around 100 CE (50a) to the use of the periphrasis in a text from the 8th c. (50b):49 (50) a. Cl. Lat. – imperfect, ca. 100 CE: Plin. Epist. 6.16.8 Egrediebatur domo; accipit codicillos He was leavingimpf ind the houseabl; he receivespr ind messageacc

Rectinae of Rectinagen ‘As he was leaving the house he got a message from Rectina’

(50b) Late Latin – periphrasis, 8th c. CE: Vita Hugb. p. 485 l. 12 Ipse uero egrediens erat ab ecclesia Ipsenom butconj leavingpr ptc nom wasimpf ind from the churchprep+abl, circumdatus plebe surroundedpf ptc nom by the peopleabl ‘he was leaving the church, surrounded by the people’ In texts from the late 4th century and early 5th century, we find the indications of a change in the relationship between the perfect and the imperfect tenses, when the imperfect tense is sometimes used with durative temporal adverbials without indicating anterior continuing and when one and the same passage in the Bible may be translated differently by different translators. Around the same period we have a growing number of periphrastic expressions with esse and the present participle that indicate imperfectivity, especially in the Christian texts translated from Greek or Hebrew, and a couple of centuries later the periphrastic expressions that indicate imperfectivity and progressivity become frequent even in texts written directly in Latin.

4. Summary We can thus distinguish several different stages in the development of the verbal system from Early to Late Latin. In the earliest preserved literary texts from the late 3rd century BCE to Classical Latin a couple of centuries later, we have a verbal system in which the opposition between the Proto-Indo-European aorist and perfect has been blurred and which puts strong emphasis on actional oppositions and on the expression of the relative sequence of events and situations. In the earliest preserved texts we encounter, however, some traces of the earlier opposition between the aorist and the perfect, which disappear in Classical Latin.

49. See Pinkster (1998, p. 230); cf. Eklund (1970; e.g. p. 74) and Amenta (2003); on Italian see Giorgi & Pianesi (1997, p. 175); cf. Haverling (forthcoming, section 4.2.4).

 Gerd V. M. Haverling

Towards the end of the 2nd and the beginning of the 3rd centuries, we have the first signs of the changes in the actional system and of the changes in the tense system. A couple of centuries later there are signs of the ongoing changes in the viewpoint opposition between the perfect and the imperfect tenses as well as of the changes in the way the relative sequence of events and situations is indicated. At the end of the Late Latin period actional oppositions are often indicated more according to the rules met with in modern Romance languages than according to the ones found in Early Latin: This change is connected with changes in the viewpoint system as well as with other syntactic changes, such as the development of indefinite and definite articles in Romance. The changes in the actional system are met with in all sorts of texts from the later periods, whereas the changes in the tense and viewpoint systems are more evident in the less elegant and literary texts.

References Abraham, W. 1997. Case, aspect and referentiality. In van Kemenade & Vincent, 29–61. Amenta, L. 2003. Perifrasi aspettuali in Greco e in Latino. Origini e grammaticalizzazioni [Materiali linguistici, Università di Pavia]. Milan: FrancoAngeli. Baldi, P. 2002. The Foundations of Latin. (2nd edn). Berlin: Mouton de Gruyter. Bybee, J. L., R. Perkins & W.Pagliuca 1994. The Evolution of Grammar: Tense, aspect and modality in the languages of the world. Chicago IL: University of Chicago Press. Comrie, B. 1978. Aspect. An introduction to the study of verbal aspect and related problems [Cambridge Textbooks in Linguistics]. Cambridge: CUP. (Repr. with corr. 1978; 1st ed. 1976). Comrie, B. 1985. Tense [Cambridge Textbooks in Linguistics]. Cambridge: CUP. Comrie, B. 1989. Language Universals and Linguistic Typology. (2nd edn). Oxford: Blackwell. Crocco Galèas, G. & Iacobini, C. 1993. Lo sviluppo del tipo verbale parasintetico in latino: I prefissi. AD-, IN‑, EX-, Quaderni Patavini di Linguistica 12: 31–68. Croft, W. 1990. Typology and Universals [Cambridge Textbooks in Linguistics], Cambridge: CUP. Dahl, Ö. (ed.). 2000. Tense and Aspect in the Languages of Europe, Berlin: Mouton de Gruyter. Eklund, S. 1970. The Periphrastic, Completive and Finite Use of the Present Participle in Latin (With Special Regard to Translations of Christian Texts in Greek up to 600 AD) [Studia Latina Upsaliensia 5]. Uppsala: Acta universitatis Upsaliensis. Feltenius, L. 1977. Intransitivizations in Latin [Studia Latina Upsaliensia 9]. Uppsala: Acta universitatis Upsaliensis. Fleischman, S. 1990. Tense and Narrativity: From medieval performance to modern fiction [Texas Linguistics Series]. Austin TX: University of Texas Press. García-Hernández, B. 1980. Semántica estructural y lexemática del verbo. Reus: Ediciones Avesta. García-Hernández, B. 1995. Synonymie et analyse fonctionelle dans le système préverbal latin. Revue des Études Latines 72: 25–38. García-Hernández, B. 1998. Diathèse et aspect verbal dans les structures lexicales. Bulletin de la Société de Linguistique de Paris 93: 211–227.

On the development of actionality, tense, and viewpoint from Early to Late Latin  Giorgi, A. & Pianesi, F. 1997. Tense and Aspect. From semantics to morphosyntax [Oxford Studies in Comparative Syntax]. Oxford: OUP. Haverling, Gerd. 2000. On Sco-verbs, Prefixes and Semantic Functions: A study in the development of prefixed and unprefixed verbs from Early to Late Latin, (Studia Graeca et Latina Gothoburgensia 64), Göteborg: Acta universitatis Gothoburgensis. Haverling, G. 2001. Sur l’emploi du parfait et de l’imparfait dans le latin tardif. In C. Moussy (ed.), De Lingua Latina Novae Quaestiones. Actes du Xè Colloque International de Linguistique Latine, Paris–Sèvres, 19–23 avril 1999 [Bibliothèque d’ètudes Classiques 22], 355–370. Louvain: Éditions Peeters. Haverling, G. 2002. On the semantic functions of the Latin perfect. In A.M. Bolkestein†, C.H.M. Kroon, H. Pinkster, H. W. Remmelink & R. Risselada (eds.), Theory and Description in Latin Linguistics. Selected papers from the Eleventh International Colloquium on Latin Linguistics, Amsterdam June 24–29, 2001, 153–167. Amsterdam: Gieben. Haverling, G. 2003a. Sur le latin vulgaire dans la traduction “ravennate” des Aphorismes d’Hippocrate. In Latin vulgaire – latin tardif VI: Actes du VIe Colloque International sur le latin vulgaire et tardif, Helsinki 28 août–2 septembre 2000, 157–172. In H. Solin, M. Leiwo & H. Halla-aho (eds.). Hildesheim: Olms-Weidmann. Haverling, G. 2003b. On prefixes and actionality in Classical and Late Latin, Acta Linguistica Hungarica 50(1–2): 113–135. Haverling, G. 2005. On tense, viewpoint and actionality in vulgar and literary late Latin. In G. Calboli (eds.), Latina Lingua! Proceedings of the Twelfth International Colloquium on Latin linguistics, Bologna 9–14 June 2003 [Papers on Grammar IX.1], 281–291. Roma: Herder. Haverling, G. 2006. Struttura lessicale e verbale nel latino e nel greco. In: P. Cuzzolin & M. Napoli (eds.), Fonologia e tipologia lessicale nella storia della lingua greca. Atti del VI Incontro Internazionale di Linguistica Greca (Bergamo, settembre 2005), 123–142. [Materiali linguistici dell’università di Pavia], Milano: Franco Angeli. Haverling, G. 2007. Sulle categorie di tempo, aspetto e azionalità nella traduzione “ravennate” degli Aforismi d’Ippocrate. In A. Ferraces Rodríguez (ed.), Tradición griega y textos médicos latinos en el período presalernitano. Actas del VIII Coloquio Internacional Textos Médicos Latinos Antiguos (A Coruña, 2–4 septiembre 2004), A Coruña 2007, 159–176. Haverling, G. Forthcoming. ‘Actionality, Tense and Viewpoint’. In P. Cuzzolin & P. Baldi (eds.), Chapter 4 in Perspectives on Historical Syntax of Latin, Vol. II. Berlin: Mouton de Gruyter. Hedin, E. 2000. The type-referring function of the imperfective. In Dahl, 227–264. Hopper, P.J. & Traugott, E. Closs. 2003. Grammaticalization [Cambridge Textbooks in Linguistics], 2nd edn. Cambridge: CUP. Johanson, L. 2000. Viewpoint operators in European languages. In Ö. Dahl (ed.), 27–187. Josephson, F. 1977. On the function of the Gothic preverb ga-. Indogermanische Forschungen 81: 152–175. Lehmann, C. 1995. Latin predicate classes from an onomasiological point of view. In D. Longrée (ed.), DE USU: Études de syntaxe latine offertes en hommages à Marius Lavency [Publications linguistiques de Louvain], 163–173. Louvain-La-Neuve: Peeters. Leiss, E. 2000. Artikel und Aspekt: Die grammatischen Muster von Definitheit [Studia Linguistica Germanica 55]. Berlin: de Gruyter. Lloyd, A.L. 1979. Anatomy of the Verb: The Gothic verb as a model for a unified theory of aspect, actional types, and verbal velocity. Amsterdam: John Benjamins. Oldsjö, F.. 2001. Tense and Aspect in Caesar’s Narrative [Studia Latina Upsaliensia 26]. Uppsala: Acta univ. Upsaliensis.

 Gerd V. M. Haverling Pinkster, H. 1983. Tempus, aspekt and Aktionsart in Latin (Recent trends 1961–1981). In W. Haase & H. Temporini (eds.), Aufstieg und Niedergang der Römischen Welt II.29.1, 270–319. Berlin: de Gruyter. Pinkster, H. 1990. Latin Syntax and Semantics, (transl. H. Mulder). London: Routledge. Pinkster, H. 1998. Narrative tenses in Merovingian hagiographic texts. In J. Herman (ed.), La transizione dal latino alle lingue romanze. Atti della tavola rotonda di linguistica storica, Università Ca’ Foscari di Venezia 14–15 giugno 1996, 229–235. Tübingen: Niemeyer. Philippi, J. 1997. The rise of the article in the Germanic languages. In van Kemenade & Vincent (eds.), 62–93. Rijksbaron, A. 1994. The Syntax and Semantics of the Verb in Classical Greek: An introduction, (2nd edn), Amsterdam: Gieben. Rix, H. 1998. Bemerkungen zu den lateinischen Verbformen des Typs faxo faxim. In J. Jasanoff, H. Craig Melchert & L. Oliver (eds.), Mír Curad: Studies on honor of Calvert Watkins [Innsbrucker Beiträge zur Sprachwissenschaft 92], 619–634. Innsbruck: Institut für Sprachwissenschaft der Universität Innsbruck. Rosén, H.B. 1992. Die Komposita mit co(n)- in funktioneller und vergleichender Sicht. In O. Panagl & T. Krisch (eds.), Latein und Indogermanisch – Akten des Kolloquiums der Indogermanischen Gesellschaft, Salzburg, 23–26 September 1986 [Innsbrucker Beiträge zur Sprachwissenschaft 64], 357–367. Innsbruck: Institut für Sprachwissenschaft der Universität Innsbruck. Smith, C. S. 1997. The Parameter of Aspect. (2nd. edn) [Studies in Linguistics and Philosophy 43]. Dordrecht: Kluwer. Smyth, H.W. 1956. Greek Grammar, (rev. by G. Messing). Cambridge MA: Harvard University Press. Squartini, M. 1995. Tense and aspect in Italian. In R. Thieroff (ed.), Tense Systems in European Languages II, 117–134. Tübingen: Niemeyer. Szantyr (1965) = Szantyr, Anton and Johann Baptist Hofmann. 1965. Lateinische Syntax und Stilistik (= Lateinische Grammatik. Zweiter Band, by Leumann, Hofmann & Szantyr) [Handbuch der Altertumswissenschaft II.2.2]. Munich: C.H. Beck’sche Verlagsbuchandlung. Traugott, E.C. & Dasher, R.B.. 2002. Regularity in Semantic Change [Cambridge Studies in Linguistics 96]. Cambridge: CUP. Väänänen, V. 1981. Introduction au latin vulgaire [Bibliothèque Française et Romane, Serie A, Vol. 6], 3rd edn. Paris: C. Klincksiek. van Kemenade, A. & Vincent, N. (eds.). 1997. Parameters of Morphosyntactic Change, Cambridge: CUP.

Continuity and change The history of two Greek tenses* Eva-Carin Gerö & Hans Ruge The goal of this paper is to demonstrate two cases of continuity in the Greek tense system from Antiquity till today, viz. the Imperfect and the Future, both of which, in spite of partly changing morphology and other variation on a surface level, remain remarkably constant semantically and conceptually. The constant meaning of the Imperfect may be defined quite cogently, depending on the choice of theoretical framework, either as developing, intraterminal or inclusive. Also the Future, in spite of more radical surface variation diachronically, displays a remarkably high degree of continuity on a conceptual level. Already Ancient Greek had the possibility of expressing different aspects also in the future tense by means of modal expressions or by periphrasis.

1. Introduction From the oldest literary attestations of Ancient Greek to written and spoken Modern Greek many traits of the Greek language have survived more or less unchanged. Quite often the same morphology has been kept with seemingly identical meaning. (The different Aorist stems are one good example.) A particular meaning often remains the same and may be said to be an important semantic constant which, however, manifests itself in a new linguistic form when seen in a diachronical perspective. The Greek Perfect1 and the sensitivity of the Greek language for Frege’s distinction between Intensionality and Extensionality (in traditional terms largely corresponding to that between

* This article was translated from German to English by Folke Josephson, whom we sincerely thank. 1. In Gerö & Stechow (2003), a continuous “Extended-Now”-meaning is assumed for the Greek language from the Classical period till today. This approach follows the XN-analysis suggested for Modern Greek in Iatridou et al. (2001). Also in Ruge (2005 and 2006), in spite of different theoretical preferences, the Greek Perfect is treated as a semantic constant.

 Eva-Carin Gerö & Hans Ruge

Subjectivity and Objectivity) are worth mentioning.2 (In this connection one should also compare the so called atavisms of Modern Greek).3 In this paper we will discuss two specific cases of linguistic continuity in Greek, the Imperfect4 and the Future, within the framework of Temporal Semantics.

2. The Imperfect 2.1

Ancient Greek

2.1.1 Theoretical preliminaries The Ancient Greek Imperfect (as well as that of Modern Greek (cf. 2.3.3) can be understood only contrastively with the Aorist and it is indeed described in that way in the grammars of the classical philological tradition as in Gildersleeve (1900–11:103): “It (i.e. the Aorist) is one of the two great narrative tenses of the Greek language, and is best studied in connection with the other, the imperfect”.5 In descriptions of tense and aspect in Post-Classical Greek this opposition continues to be stressed, as in the case of New Testament Greek by Turner (1963: 64): “The contrast between the indicatives of the imperfect and aorist illustrates the difference between linear and punctual Aktionsart6 in its most acute form...” and Fanning (1990: 124): “The primary aspectual relationship is that which obtains between the present and Aorist aspects”. On the other hand, already the Stoic grammarians understood the Imperfect and the Present by contrast with the Pluperfect and the Perfect, which are the tenses of the Perfect system. They considered the Aorist and, as we shall see later, also the Future, as a χρόνος ἀόριστος, i.e. as a tense that is ‘indefinite’ and is therefore of a quite different nature from the Imperfect which belonged to the ‘definite’ χρόνοι ὡρισμένοι. The opposition between ‘imperfective’ (παρατατικός) and ‘perfective’ (συντέλειος) as well as that between definite and indefinite tenses both follow from the tense and aspect theories of the Stoics because the distinction between “terminated” and “unterminated” is a primary one for them (Cf. Robins [1967] 1990: 34 and Versteegh 1980). 2. For a broad spectrum of constructions, cf. Gerö (2007), as well as similia in Giannakidou (1997, 1998) and in Veloudis (1983-84 et alibi). 3. See Ruge (1985). The phenomena referred to are linguistic “items” at hand in Protohellenic and early Greek, which after having disappeared in later stages of the language reappear in Modern Greek. 4. Greek paratatikós. For Modern Greek we employ this expression. 5. Also Goodwin ([1875]1965: 16) – who in other respects seems to have had great sympathy for the Stoic system – at times emphazises the opposition between the imperfective and the Aorist paradigma, as if it were of primary importance. 6. The phenomenon referred to is aspect. Aktionsart is to us, as nowadays to almost all linguists, a distinction expressed lexically rather than grammatically.

Continuity and change 

Grammatical tradition often seems to consider both members of the oppositional pair Imperfect: Aorist as equally marked (Cf. Schwyzer & Debrunner 1950 and Bornemann-Risch 1978), the Imperfect having a specific meaning as ‘durative’ or the like and the Aorist being ‘punctual’.7 This also seems to be the case if one applies a more modern general linguistic tense theory as that of Klein (1994) to the ancient Greek tenses. According to this theory which further develops Reichenbach’s widely used description of tenses (Reichenbach 1947) the Imperfect should be understood as Inclusion and the Aorist as Identity (or Overlapping) of temporal entities. As a consequence, the distinction between ‘marked’ and ‘unmarked’ will not play any obvious role (Cf. Klein 1994). Another theory that has found some acceptance is the one according to which certain Ancient Greek tenses, especially the Aorist, express only aspect (cf. Rijksbaron 1988, McKay 1988 and (for Modern Greek) also Seiler 1952).8 A theory of this kind will have an impact on the understanding of the Imperfect. Aorist and Imperfect will no more be in clear opposition but will belong to quite different areas of verbal meaning. In the context of a discussion concerning several languages including Modern Greek and Ancient Greek, Hedin (2000) suggests that the Imperfect – and imperfective aspect generally – has a type-referring function which implies that this tense is used when the situation is seen in a static, non-temporal perspective in disregard of individual realisation. The perfective aspect, the Aorist in the case of Greek, is seen as token-referring and as the paradigm that is used for instantiations occurring in time.9 This theory seems at first sight to be radically divergent from the traditional point of view of Greek grammar. We do, however, suggest that it does not have any implications that are not reconcilable with the traditional view. 2.1.2 Data In fact, the Ancient Greek Imperfect has a series of applications that cannot easily be analysed in a uniform way.10 The classical usage is traditionally divided into the following subcategories: (1) (habitual/iterative) Lys. 1:9 ἐπειδὴ δὲ τὸ παιδίον ἐγένετο when but the child be.born.aor.3sg

ἡμῖν, we.dat

7. Other grammarians, like already the Stoics, focus on the “indefiniteness” of the Aorist with respect to properties expressed by the Imperfect (e.g. “incomplete”). 8. Gerö (1996) argues against this kind of interpretation. 9. Cf. Hedin (2000:259 sq.): “With the Imperfective, situations are considered in a static, nontemporal perspective as types (neglecting any instantiation), as opposed to the Perfective, which is used for reference to situations as tokens or instantiations occurring in time.” 10. Similarly as in the case of the Aorist, which e.g. Wackernagel (1920-24), broadly speaking, describes in terms of “Verwendungstypen”, rather than attempting a uniform analysis. Cf. also Coseriu (1992:34 f.), who explicitly dismisses the linguists’ endeavour to find a uniform Grundbedeutung in such cases as “sprachwissenschaftlichen Positivismus”.

 Eva-Carin Gerö & Hans Ruge

ἡ μήτηρ αὐτὸ ἐθήλαζεν the mother.NOM it.ACC suckle.IMPF.3SG When the child was born to us, its mother suckled it (2)

(conative) Xen. An. 7.3.7 ἄγγελοι ἔπειθον ἀποτρέπεσθαι· οἳ δὲ οὐχ ὑπήκουον messengers persuade.IMPF.3PL turn back.INF they but not listen.IMPF.3PL messengers tried to persuade them to turn back, but they would not listen to them

(3) (inchoative) Th. 7.51 ἐπειδὴ δὲ καιρὸς ἦν,.. τὰ τείχη when but time be.IMPF.3SG the walls.ACC. τῶν Ἀθηναίων προσέβαλλον the Athenians.GEN.PL attacked.IMPF.3PL when the favourable moment came, they assaulted the Athenian walls This distribution also implies a certain tendency to use the imperfective aspect within the scope of a negation as in (4): (4)

Dem. 44.17 ὁ μὲν οὐκ ἐγάμει, ὁ δὲ ἔγημεν he on the one hand NEG marry.IMPF.3SG he but marry.AOR.3SG [the one] did not marry (=did not want to marry), but {the other one} did marry

Furthermore, the imperfective aspect is preferred with verbs of movement, demand and inquiry, as commonly stated in school grammars: (5) Th. 2.6 ἄγγελον ἔπεμπον καὶ τοὺς νεκροὺς messenger.ACC send.IMPF.3PL and the dead.ACC.PL ὑποσπόνδους ἀπέδοσαν secured-by-treaty.ACC.PL delivered.AOR.3PL ...they sent a messenger...[and] gave back the dead under a truce Modern grammarians have also observed the discursive function of the Imperfect in contexts in which forms of the Imperfect paradigm supply the ‘background’ or ‘frame’ of other events as in: (6) Hdt.1.114.1 sqq. καὶ ὅτε ἦν δεκαέτης and when be.IMPF.3SG ten-year-old.NOM.MASC ὁ παῖς, the child.NOM.MASC

Continuity and change 

πρῆγμα happening.NOM.NEUT.SG

ἐς αὐτὸν τοιόνδε to he.ACC.SG such.NOM.NEUT

γενόμενον having-occurred.PART.midd.NEUT.SG

ἐξέφηνέ μιν. show.AOR.3SG he.ACC.

ἔπαιζε ἐν τῇ κώμῃ..., ἔπαιζε δὲ play.IMPF.3SG in the.DAT.SG village.DAT.SG..play.IMPF.3SG play.IMPF.3SG but μετ᾿ ἄλλων ἡλίκων ἐν ὁδῷ. καὶ with-others.GEN.PL of-same-age.GEN.PL in road.DAT.SG and οἱ παῖδες παίζοντες the boy.NOM.3PL. play.PART.PRES.NOM.PL

εἵλοντο choose.AOR.3PL

ἑωυτῶν βασιλέα εἶναι τοῦτον... own.GEN.PL king.ACC.SG be.INF he.ACC.SG When the boy was ten years old the following incident revealed his identity. He was playing in the road...with others of his age. The boys in their play chose [this boy] for their king In Koine Greek, where the Aorist forms expand at the cost of the Imperfect, the distinctions of the classical usage are, however, largely upheld. In the New Testament we therefore still find the following types of usage (cf. Turner, 1963: 64 sq., and Mandilaras, 1973: 129 sq.): (7)

(iterative/habitual) Ep. Luc. 21.37 τὰς νὐκτας.. ηὐλίζετο.. the.ACC.PL night.ACC.PL... pass-the-night.IMPF.3SG. [he] would spend the night in the open

(8)

conative/desiderative) Act. Ap. 26. 11 ἠνάγκαζον βλασφημεῖν force.IMPF.1SG blaspheme.INF.PRES I [tried] to force them to renounce their faith

(9)

(descriptive/background) Act. Ap. 21.20 ἐδόξαζον τὸν θεόν, εἶπόν τε... praise.IMPF.3PL the God.ACC.SG say.AOR.3PL and... They gave glory to God--- Then they said...

(10)

‘communicative’ verbs Act. Ap. 15.38 ἠξίου... μὴ συμπαραλαμβάνειν deem.IMPF.3SG NEG admit.INF.PRES [Paul] was not in favour of (= demanded not to) of taking along the man

In Archaic and Classical Greek there are cases where the difference in meaning between an aoristic and an imperfective form in the same context can hardly be observed. Cf. for example Hom. Il. 2.43 βάλλετο (IMPF) and 45 βάλετο (AOR). This is also

 Eva-Carin Gerö & Hans Ruge

true for Post-Classical Greek as in the New Testament 1. Ep. Cor. 10, 6–11 ταῦτα... τύποι ἡμῶν ἐγενήθησαν (AOR) as against ταῦτα...τυπικῶς συνέβαινεν (IMPF). In most cases of this kind it is, however, possible to maintain the usual difference even if it has to be interpreted in a subtle way.11 But in many cases expressions that have an original distinction of meaning seem to have frozen to mere formulae.12 As the above examples show, the Ancient Greek Imperfect had a whole spectrum of meanings for which it is not easy to find a common denominator. Traditionally, one assumes a meaning of “PAST+developing/linear verbal action” as a fundamental value which would fit, though it is not particularly explicit. Gerö & Stechow (2003) assume an ambiguity Progressive/Iterative as does Giannakidou (1998) for Modern Greek. This would cover the range of meaning rather well but it goes against the intuition of traditional grammar, which we share, that varieties of usage are subsumed by a common meaning. Though Hedin (2000) provides a comprehensive analysis of the phenomena (cf. above) it seems only to slightly widen the traditional picture.13 2.1.3 Evaluation of the material Though it is difficult to suggest a general meaning of the Greek Imperfect that will cover the different varieties of usage in their entirety, our intuition favours, and theories of tense as an analytical tool will not contradict that such a meaning may be found. Already through the tense/aspect theories of Klein (1994) whose system is built by means of a limited ontology consisting of temporal intervals (TU: time of utterance, TT: topic time, TSit: time of situation) and relations such as ‘before’, ‘after’, ‘inclusion’ and ‘overlapping’14 we obtain a picture of the Ancient Greek Imperfect that is more precise than that of traditional grammar but nevertheless confirms its intuitions. When Klein describes the imperfective aspect as “Inclusion of TT in TSit” and the perfective aspect as “Overlapping/Identity of TT and TSit”, we obtain a comprehensive definition which describes the use of aspect in (1) – (10) in a better way than “durative” or “punctual” would. The definition provided by Klein can also be made semantically more explicit so as to eliminate the ambiguity implied in Gerö & Stechow (2003). 11. See Goodwin (1889:17): “Since the same event may...be stated by the Aorist or the Imperfect according to the writer’s point of view, it is natural that it should occasionally be a matter of indifference which form is used, especially when the action is of such a nature that it is not important to distinguish its duration from its occurrence”. 12. Cf. Turner (1963:64): “indeed we can find no difference between ἔλεγεν and εἶπεν in the NT.” 13. In order to make the picture complete the periphrastic construction ἦν διδάσκων should be mentioned (Björck 1940; Rydbeck 1969). This periphrasis formed by the present participle and imperfect of εἰμί does not seem to be different from a monolectic form. Cf. Fanning 1990: 313 sq., who states that there may be a stronger emphasis on the imperfective meaning (progressive or customary). 14. “Tense relations are deictic: they locate the topic time with respect to the time of utterance, and aspect relates the time of the situation to this topic time”. (Klein 1994: 140)

Continuity and change 

The meaning suggested by Hedin (2000) (cf. 2.1.1) can be secondary at the very most and only reflect the real semantic value of the Greek Imperfect. The meaning of the Imperfect is simply too special to be understood as denoting a type. We would rather need a term like “temporal extension” in combination with a property of “divisi bility” and of “interval” in order to explain progressive and iterative meaning. We thereby obtain somewhat more explicitly, and in direct correspondence with the definitions by grammarians like Goodwin, Gildersleeve and Rijksbaron, a common meaning covering terms like background, inchoative and de conatu. This meaning will make sentences with a verb in the Imperfect seem to have an “abstract” meaning.15 Whatever framework we use in the analysis of the Greek material, whether it be Klein’s concept of Relations between intervals or an analysis like that of Davidson (2001 et passim) and understand events and activities as temporal entities with different properties, the semantic characteristics of the Imperfect will appear as something very specific, as a χρόνος ὡρισμένος. One will best understand this tense contrastively with the Pluperfect, the other χρόνος ὡρισμένος of the past. The Aorist, which is the tense of “Identity” or the tense form that describes an event without internal structure, will then remain an ἄλλο γένος allo genos and we join the Stoics again.16 We will then obtain a uniform analysis also of this tense and will not remain in the descriptivity of various manners of usage. The maybe most important test of a Tense/Aspect theory will be provided by a combination of verbal forms and temporal adverbs (cf. Stechow 1999: 2). We thereby obtain an important indication as to how to understand aspect marking as in (11) and (12) where Imperfect and Aorist of the same verb are found in the same context, in each separate case with an indefinite-durative or a non-durative temporal adverb: (11) Lys. 7.4 δημευθέντων confiscate.PART.PASS.GEN.3PL

τῶν ὄντων ἐκείνου the property.GEN.PL that-one.GEN.SG

Ἀπολλόδωρος ὁ Μεγαρεὺς δωρεὰν παρὰ τοῦ Apollodoros.NOM the Megaran.NOM gift.ACC.SG from the δήμου λαβὼν people.GEN.SG obtain.PART.AOR.NOM.SG

τὸν μὲν ἄλλον the.ACC.SG but other.ACC.SG

χρόνον ἐγεώργει time.ACC.SG cultivate.IMP.3SG ...when his property was confiscated, Apollodoros of Megara had it as a gift from the people and cultivated it for some time 15. Hedin also overlooks that the “imperfective for perfective” that she sees as an example of type referentiality can be understood as “background” or as being used parallelly to “communicative” verbs. 16. “L’ Aoriste est effectivement ce qui est dépouillé des valeurs subjectives de durée ou d’ achèvement qu’expriment présent et parfait.” (Humbert 1960: 141)

 Eva-Carin Gerö & Hans Ruge

(12) ibid. 10 ἀπεμίσθωσα let-out.AOR.1SG

Καλλιστράτῳ. Callistratus.DAT

ὃς who.NOM.SG

δύο ἔτη two years.ACC

ἐγεώργησεν cultivate.AOR.3SG I let it out (sc. the plot) to Callistratus...; he cultivated it for two years The meaning of the Imperfect, which we understand intuitively as vague or as developing or, analytically, as inclusive (Klein: TT included in TSit) or as a term for an indefinite temporal entity (Davidson), is confirmed by the combination with the adverb τὸν... ἄλλον χρόνον. The meaning of the Aorist as “distinct, punctual” etc. is confirmed by the combination with the definite adverb δὑο ἔτη.

2.2

Byzantine Greek

2.2.1 Data This period is fundamentally characterized by diglossia which from the time of the Atticist movement forms a background that will always have to be taken into account for any evaluation of the Greek language of this period. We shall therefore only cite examples from literature that are as close as possible to the living language. Already from the beginning of the last millennium preliminary stages of Modern Greek can be observed, as in the text of the epic by Digenis Akritas (11th C.), which is considered as being the first literary monument of Modern Greek, in the beggar poems ascribed to the pen name of Prodromos (12th C.) and the romances of chivalry of the 13th and 14th centuries. In one such “early Modern Greek” work Βἐλθανδρος και Χρυσἀντζα (295–321) we read: (13)

Και τὀτε βρὐσην ηὐρηκε παρἀξενον οκἀτι· 295 ψυχρὀν, χιονοκρὐσταλλον ὐδωρ εἰχεν η βρὐση. Φρἀσαι καν ὀλως απορὠ την καλλονἠν την τὀσην, ην εἰχεν υπερἀπειρον η των Ερὠτων. Ἰστατο γρὐψος λαξευτὀς, εξἠπλωνε πτερἀ του, την ρἀχην προς ισὀτηταν ἐφερνε των πτερὠν του 300 και την ουρἀν εγὐριζεν ως προς την κεφαλἠν του· τοις δε ποσἰ τοις ἐμπροσθεν εβἀσταζε λεκἀνην στρογγυλολαξευτὀμορφον απὀ πελἐκιν λἰθον. Το δε νερὀν εκ στὀματος εξἠρχετο του γρὐψου κι εις το λεκανοπἐτρινον εισἠρχετον απἐσω 305 του μη σταλἀξαι σταλαγμὀν κἀτω προς γην ολἰγον. Ὠραν εστἀθην περισσἠν ο Βἐλθανδρος προσἐχων του γρὐψου την κατασκευἠν και του νεροὐ το ξἐνον· πὠς δε το εξερχὀμενον εκ στὀματος του γρὐψου χωρεἰται εις μικρολἐκανον παντοὐ μη ἐχον πὀρον, 310

Continuity and change 

ἠ και τινἀ κατατροπἠν εις το νερὀν εκστἀζειν, και πἀλιν πὠς εκ των χειλὠν εκρἐει της λεκἀνης, του ὐδατος την εκροἠν εθαὐμαζεν εκεἰνος. Εξαἰφνης κατεπἀτησεν ο γρὐψος εκ του τὀπου· αντἰπερα του ποταμοὐ απἠγε και εστἀθην. 315 Ἐἰδε, παρεξενὠθηκεν ο Βἐλθανδρος εκεἰσε και πἀλιν το αναπὀταμον το φλογερὀν εκεἰνο. Καλὠς περιεπἀτειε κατασκοπὠν τα πἀντα και τὀπον περιἐδραμεν πανεπιτἠδειον πἀνυ. Εἰδε παλἀτια φοβερἀ, τα ου δὐναμαι αφηγεἰσθαι κατἀ λεπτὀν τας χἀριτας των παλατἰων εκεἰνων.

320

The imperfects have been underlined (εἰχεν). He then came upon a remarkable fountain whose water was as cold as snow. The beauty which the fount of the cupids possessed in boundless measure, I am wholly at a loss to describe. A carved griffin was standing there with extended wings and its back arched to a level with them. Its tail was bent down to its head. In its front paw it held a beautiful round basin carved from a precious stone. Water came from its mouth and flowed into the basin without the smallest drop falling down to the ground. For some time Velthandros stood contemplating the griffin’s construction and the strange property of the water. How was the water, which came from the griffin’s mouth, held in the small basin, which had no aperture at all? Or did the water change its direction to escape? But how could the water flow back from the basin’s lips? He marvelled at where the water went. Suddenly the griffin stamped away from where it was standing, crossed the river and stood there. Once again Velthandros saw and wondered at the blazing river. He walked around examining everything carefully and wandered over the wonderful place. He saw remarkable palaces whose charm I cannot describe in detail. The interchange between Aorist and Imperfect corresponds exactly to what we are used to in Classical Greek; Aorists (295 ηὐρηκε, 307 εστἀθην, 314 κατεπἀτησεν, 315 απἠγε, εστἀθην, 316 Ἐἰδε, παρεξενὠθηκεν, 319 περιἐδραμεν, 320 Εἰδε) carry the action forward, Imperfects present descriptions of circumstances. Thus the hero finds (295: Aorist) the fountain with the stone griffon standing there (299: Imperfect) spreading his wings (299:Imperfect). The following Imperfect forms (300, 301, 302, 304, 305) express characteristic traits of the griffon at the fountain. Then the hero remains standing for a long time (307: Aorist). One can observe the use of the Aorist when a definite time-span is mentioned (cf. above (12)). The hero stands in admiration (313: Imperfect). The griffon suddenly moves (314: Aorist) and betakes himself somewhere else (315: Aorist). The hero sees it and is astonished (316: Aorists). By moving around (318: Imperfect) he traverses the entire area (319: Aorist) and gets to see (320: Aorist) many wondrous sights.

 Eva-Carin Gerö & Hans Ruge

2.2.2 Evaluation of the material This example clearly shows that the Imperfect still played its “classical” role of the tense of background and narrative abidance during the late Byzantine period and in a stage of the language that can be called Early Modern Greek. There is no indication that the imperfect of the Byzantine period would be in need of a different analysis from that of Antiquity. According to Jannaris (1897: 436), the periphrasis by the Imperfect of ‘be’ and the Present participle which is known to occur in the New Testament was still favoured in Byzantine times. This periphrasis which is present already in the classical language seems to culminate in the New Testament, where its frequency was probably due to Semitic influence (cf. Fanning 1990: 316 sq.). Its frequency then diminished, probably because of the disappearance of the Present participle from the living language. The socalled gerund of Modern Greek (cf. Ruge 2000: 62) is not a “real” participle, as it has no part of nominality. V.I. 603A ἦσαν βαστάζοντες is an example of Byzantine Greek (cf. Rosenqvist 1986: 4). Its meaning seems to be the same as that of the monolectic form.

2.3

Modern Greek

2.3.1 Theoretical preliminaries In grammars, the Modern Greek paratatikós is generally described in the same way as the Ancient Greek Imperfect. One usually defines this tense contrastively to the Aorist. Cf. Ruge (2001: 163) who states that the aóristos refers to an event in the past. It can refer to an action that objectively seen occurs as quick as lightning but equally well to one that takes a long time to perform, or to one that forms the beginning or the end of a state or to repetition of one action, i.e. a habit. In a larger context where aóristos and paratatikós are interchangeable the paratatikós refers to the background against which the events (aóristos) stand out. Tzermias (1969: 136 sq.) also believes in a complementarity of the two tenses: “Beide zusammen decken den Verwendungsbereich des Deutschen Imperfekts”. Holton et al. (1997: 224 sq.) express the same opinion, though less explicitly. In Markedness Theory there are different opinions as to whether the paratatikós is the marked member when opposed to the aóristos. Seiler (1952) and Mirambel (1959) give a positive answer to this question as does Ruipérez in the case of the Ancient Greek imperfect, whereas Kahane & Kahane (1958) and Babiniotis (1972) surprisingly are of the opposite opinion in regard to Modern Greek. Furthermore, Seiler emphasizes the aspectual nature of the aóristos. Hedin understands the opposition aóristos/paratatikós as equipollent. Some linguists find that the Modern Greek paratatikós shows an ambiguity of meaning between iteration and habituality (cf Giannakidou 1998). As already mentioned, Hedin (2000) explains the difference between Modern Greek aóristos and paratatikós in a somewhat different manner. In her opinion the opposition is one of type/token and the paratatikós (or imperfective aspect generally) denotes situations that are seen in a static, non-temporal perspective (types) whereas the aóristos (or the perfective aspect generally) denotes such that are temporally “realised” (tokens).

Continuity and change 

2.3.2 Data The example given in (14) demonstrates the interplay between aóristos and paratatikós (the Imperfect (paratatikós) forms are underlined): (14) Ὀλη τη νὐχτα χιὀνιζε ασταμἀτητα. Whole the night rain.IMPF.3SG uniterruptedly Κρὐωνε πολὐ και ἰσως απ᾿αυτὀ να Be cold.IMPF.3SG much and perhaps from-that that εἰχε εφιἀλτες. Ἐβλεπε στον ὐπνο have.IMPF.3SG nightmare.ACC.PL See.IMPF.3SG in-the sleep του ὀτι τον στἠνουνε στον τοἰχο κι ὀτι he.GEN that he.ACC place.IMPF.3PL to-the wall and that τον αποτελεἰωνε με χαριστικἠ βολἠ he.ACC liquidate.IMPF.3SG with grace coup ο Αγκιτἀτσιας φωνἀζοντἀς του the Agitatsias cry.GER he.GEN ᾿Να πὠς τιμωρεἰ το Κὀμμα τους προδὀτες.᾿ There how honor.PRES.3SG the party.NOM the traitor.ACC.PL Ξὐπνησε με πονοκἐφαλο αλλἀ του Wake up.AOR.3SG with headache but he.GEN ἐκανε καλὀ η σκἐψη πως δεν εἰχε make.IMPF.3SG good the thought how NEG have.IMPF.3SG καμιἀ ευθὐνη και ἐμεινε ξαπλωμἐνος ως any responsibility and remain.AOR.3SG lie.PART until τις δἐκα, αφἠνοντας τη φαντασἰα του να some ten abandon.GER the fantasy he.GEN to φτερουγἰζει μὀνο γὐρω απὀ πρἀγματα πολὐ flutter.PRES.3SG only around PREP thing.ACC.PL much ευχἀριστα· Στο ποἰημα που τοὐ ᾿χε pleasant.ACC.PL In-the poem where he.GEN have.IMPF.3SG γρἀψει write.supin

ο Μπἀμπης Τσἰρος πρὠτα πρὠτα, that Babis Tsiros first first

στην κὀτα με τις χυλὀπιτες που του at-the hen with some macaroni somewhere he.GEN εἰχε φτιἀξει η παπαδιἀ, have.IMPF.3SG prepare.supin the priest’s wife στο ζεστὀ κρασἰ που ἐπινε με at-the hot wine somewhat drink.IMPF.3SG with

 Eva-Carin Gerö & Hans Ruge

τον παπἀ τις χειμωνιἀτικες νὐχτες. Ὀταν the priest.ACC some wintery.ACC.PL night.ACC.PL When τ᾿αποφἀσισε να σηκωθεἰ, δεν ἐπεφτε and-decide.AOR.3SG to rise.PRES.3SG NEG fall.IMPF.3SG πια χιὀνι any-longer snow It snowed the whole night without interruption. He was very cold and it was perhaps because of this that he had nightmares. He saw in his dream that they placed him against the wall and how Agitatsias was about to liquidate him by a coup de grace while crying: “This is how the party punishes traitors!” He woke up with a headache but the thought that he had no responsibility made him good and he remained in bed until ten letting his thought flutter around only very pleasant matters. First about the poem that Babis Tsiros had written in his honour: (then ) around a hen with macaroni that the priest’s wife had prepared for him and about the hot wine that he used to drink during the winter nights. When he decided to rise it was not snowing any more. The first verbs (paratatikós) provide a background (snowing, freezing, bad dreams). The person then wakes up (aóristos: Ξὐπνησε). Then comes a new background with pleasant thoughts (paratatikós), remaining in bed (aóristos: ἐμεινε ξαπλωμἐνος) followed by more pleasant thoughts of how he used to drink hot wine (habitual/iterative action expressed by the paratatikós). He decides to rise (aóristos). In the meantime the material background has changed (negated paratatikós): it is not snowing any more. The relevant literature does not tell us much about inchoative or conative functions of the paratatikós. Hedin (2000) shows by her example with πεἱθω that that Modern Greek is different from Ancient Greek by having no conative paratatikós. In the case of verbs meaning ‘say’ she does, however, find continuity in the fact that this tense is preferred to the aóristos. In connection with adverbials there are some preferences as regards the choice between aóristos and paratatikós.: expressions of repetition (καθημερινἀ “daily”, ταχτικἀ “regularly”) and duration (ὀλη την ημἐρα “the whole day”), are generally combined with the paratatikós, whereas limitations and particularised repetitions such as δυο φορἐς “twice” τρἰα χρὀνια “three years” etc. need the Aorist (cf. Mackridge 1987: 108sq., Ruge 2001: 165). 2.3.3 Conclusions In a diachronic perspective the Modern Greek paratatikós fundamentally has the same properties as the Imperfect during the Ancient and Byzantine periods. There are insignificant differences: Hardly any inchoative or conative employment is to be found. But one may ask how common such employment was in Ancient Greek. The main function was to provide a background to events mentioned in the Aorist. It is not sure that one can even speak of a semantic shift. There can be no doubt about semantic continuity.

Continuity and change 

From the point of view of Markedness Theory the Ancient Greek Imperfect is seen by Ruipérez (1954: 89) as the marked member of the Aorist/Imperfect opposition. Differently from Kahane & Kahane (1958), Babiniotis (1972) and Hedin (2000) we see no reason why this should not be the case also in Modern Greek. This continuity is also supported by Mackridge (1985). The type-interpretation of the paratatikós suggested by Hedin (2000) can be considered as a secondary phenomenon which can easily be combined with the traditional interpretation. Modern Greek has no periphrastic paratatikós.

3. The Future 3.1.1 Theoretical background In contrast to the Modern Greek (analytic) Future the Ancient Greek (synthetic) Future lacks a morphologically realised distinction between perfective and imperfective.17 Cf. Gildersleeve (1900–11:115): “The future denotes either continuance or attainment in the future/---/ ἄρξω “I will be ruler, I will become ruler” The asymmetry of the Ancient Greek verbal system in regard to the future and the past has caused a certain amount of interest among the grammarians. Cf. Buttman (1841: 368): The difference between what is momentary, and what implies a duration, occurs also...in the time future. The language, however, has no double form for it in the indic...

Early grammarians did not want to recognise the lack of a formal expression of aspectual differentiation in the future forms. In his efforts to find something corresponding to a coherent logical substructure in the Greek language Hermann (1801) tried to demonstrate a formal differentiation between a “präsentisch” (i.e. imperfective) and an “aoristisch” (i.e. perfective) Future. He perceived the medial Future forms in -σομαι with their passive meaning as being close to the present system and the forms in -θήσομαι as being similar to the Aorist.18 Several grammarians after Hermann held similar opinions, especially Blass (1892), who thought that the Classical Greek Future φανεῖται had a durative meaning ‘will seem’ whereas φανήσεται would have an ingressive-aoristic meaning ‘will appear’, ‘will be found to’. In cases where a verb possesses two Future forms, one of which has a morphological similarity to the Aorist, it can, however, be demonstrated that both are used with durative as well as aoristic meaning

17. Wackernagel (1920:200) asks if Ancient Greek did not already perceive this difference between Future forms. 18. Hermann finds a formal expression of the “logically necessary” paratatic Future in the active paradigm i.e. in the μέλλω periphrasis in combination with the Present infinitive. In the simple γράψω he sees an ambiguity between the marked “aoristic” and the aspectually “neutral”.

 Eva-Carin Gerö & Hans Ruge

(Wackernagel 1920–24:202–203). The grammatical handbooks also treat this problem (cf. Jannaris 1897: 441 sq. and Smyth-Messing 1956: 427 sq.).19 Most grammarians are of the opinion that non-indicative modal forms or periphrastic constructions are used to give explicit perfective or imperfective meaning to future events. Cf. Gildersleeve (1900–1911: 80): The absence of special forms for designating the kind of time in the future helps to account for many modal combinations. The subjunctive and optative often serve to express exact temporal relations in the future.

The question of markedness is an additional problem with regard to the Future as in the case of the Aorist. When we observe the fact that the synthetic Future covers a large semantic area and can refer to events with different aspectual properties and that there are alternative expressions that are imperfective in an explicit way, should we then consider the Future is aspectually neutral or as having a marked aoristic meaning? In descriptions of the Classical Greek Future there is often an implicit preference for aspectual neutrality. In discussions about the use of tenses in the New Testament the Future forms are often considered as having only temporal meaning. Cf. Blass-Debrunner-Rehkopf (1979: 263) and Fanning (1990: 123 sq.) who seems to consider this opinion as valid also for Classical Greek:20 What is invariant about the future through all its forms is the temporal meaning of ‘future occurrence’, which can have secondary nuances of intention, potential, command and so forth as a consequence of this time reference. Thus, it contrasts with the aorist and present forms not on the basis of aspect-distinction but in regard to the time-values which are attached to their indicative forms.

Also according to those grammarians who understand the forms of the other Greek tense stems as temporally non-defined and as indicating aspect only the forms of the Future are an exception (also in regard to the optative, the infinitive and the participle) and are seen as marking an absolute temporal level, which is the future (cf. Bornemann-Risch 1978: 213). 3.1.2 Data Though the Future is often regarded as temporally absolute the Future forms express “at most a more or less reliable prediction” (Rijksbaron 1984: 33) as in: (15) Ar. Nu. 1352 χρὴ δὲ λέγειν πρὸς τὸν χόρον necessity but tell.INF.PRES to the choir.ACC.SG. 19. There is a similar discussion in regard to the New Testament. Cf. Robertson 1919:870 sq. who considers the distinction to be correct, but not applicable because of the reduced number of forms of the Future tense. 20. Turner (1963: 86) believes that the Future does not express Aktionsart but the time of the action relative to the speaker.

Continuity and change 

πάντως δὲ τοῦτο δράσεις Certainly but this do.FUT.2SG You have to tell it to the choir: for sure, you will do this It often refers to unresolved, enjoined or generic events:21 (16)

Dem. 8. 37 τί ἐροῦμεν ἢ τί φήσομεν what say:FUT.1PL or what suggest.FUT.1PL What shall we say or suggest?

(17) Xen. Cyr. 3.2.29 ὁ ἄγγελος... the messenger.NOM.SG...

ἐλθὼν arrive.PART.AOR.NOM.SG

ἐκεῖσε there

ὧδε λέξει so tell.FUT.3SG ...the messenger, … when he arrives there, shall speak on this wise (18) Men. Mon. 45 ἀνὴρ man. NOM.SG

φεὐγων escape.PART.PRES.NOM.SG

καὶ and

πάλιν μαχήσεται again fight.FUT.3SG A man who escapes shall return another time and fight The future expresses both “durative” and “punctual” events: (19) Ar. Av. 726 ἢν οὖν ἡμᾶς νομίσητε θεοὐς, if thus we.ACC.PL consider.aor.SUBJ.2PL god.ACC.PL... | ἕξετε have.FUT.IND.2PL

χρῆσθαι μάντεσι μοὐσαις |. ..κοὐκ use:INF foreboder.DAT.PL muse.DAT.PL... and-not

ἀποδράντες | run-away.PART.aor.PL

καθεδοὐμεθ᾿ ἄνω sit.FUT.1PL above

σεμνυνόμενοι παρὰ ταῖς νεφέλαις honour.PART.PRES.MID.3PL in the cloud.DAT.PL If you will consider us as gods....you will be able to use us as Muses Prophetic, and we will not run away and sit up in the clouds putting on an air

21. Cf. Humbert (1960:151): “le futur...n’est pas...un temps objectif, comme le futur du latin ou le nôtre: c’est un virtuel, qui indique qu’une chose doit se réaliser, c’est-à-dire arriver à la réalité; par conséquent il participe de la subjectivité d’un mode.”

 Eva-Carin Gerö & Hans Ruge

(20)

id. Ran. 199 οὔκουν καθεδεῖ δῆτ᾿ ἐνθαδί, γάστρων; notquest sit.FUT.2SG indeed there Potgut Now sit you there, you little Potgut

Furthermore the Future can be combined with punctual-definite as well as durativeindefinite adverbs: (21) εἰ τοῦτο ποιήσομεν, ῥᾳδίως τὰ ἐπιτήδεια if this do.FUT.1PL easily the supplies.NEUTR.ACC.PL ἕξομεν ὅσον χρόνον ἐν τῇ πολεμίᾳ ἐσόμεθα have.FUT.1PL as much time.ACC.SG in the war.DAT.SG be.FUT.1PL If we do this, we will easily provide our supplies as long as we will be at war The periphrasis μέλλω + infinitive should be specifically taken into account. Like the monolectic Future it shows an important variety of (related) meanings:22 (22) (intention) Pl. Ap. 21 b μέλλω... ὑμᾶς διδάξειν, ὅθεν μοι shall.PRES.1SG... you.ACC.PL tell.FUT.INF whence I.DAT ἡ διαβολὴ γέγονεν the accusation.NOM.SG arise.PF.3SG I shall tell you, whence the accusation against me has arisen (23)

(deontic) Eur. Ion 760 εἰρήσεταί τοι, κεἰ θανεῖν μέλλω διπλῇ tell.FUT.PASS.3SG thou.DAT.SG even-if die.INF.AOR shall.PRES.1SG twice Thou shalt be told – yea, though I die twice over

(24) (someone else’s will) Eur. El.17 ἄρσενά τ᾿ Ὀρέστην..., τὸν μὲν young.man.ACC.SG and Orestes.ACC.SG.. he.ACC indeed πατρὸς γεραιὸς ἐκκλέπτει τροφεὺς | father.GEN.SG old.man.NOM.SG steal.away.PRES.3SG fosterer.NOM.SG

μέλλοντ᾿.. χερὸς ὕπ᾿ Αἰγίσθου be destined.to.PART.PRES.ACC.. hand.GEN.SG from Aigisthus.GEN θανεῖν... die.INF.AOR The boy Orestes...doomed to die by Aigisthus’ hand, his father’s old fosterer stole away

22. Cf. Goodwin (1889:20): ”This form sometimes denotes mere futurity and sometimes intention, expectation or necessity.”

Continuity and change 

(25) (pure future [in real or possible worlds]) Pl. Resp. 412a οὐκοῦν καὶ ἐν τῇ πόλει ἡμῖν.. δεήσει τοῦ thus and in the city.DAT.SG we.DAT... lack.FUT.3SG ART.GEN τοιοῦτου τινὸς ἀεὶ ἐπιστάτου, εἰ μέλλει such.GEN someone.GEN always overseer.GEN if will.PRES.3SG ἡ πολιτεία σώζεσθαι; the constitution.NOM.SG save.PRES.INF.PASS And shall we not also need in our city...a permanent overseer of this kind, if its constitution is to be preserved? It should also be observed at this point that an aspectual difference can be expressed by an interchange between the Aorist and the Present infinitive ((23) and (24) as against (25)). In New Testament Greek there are several future periphrases, such as the one by μέλλω + infinitive (cf. Rom. 8:13: εἰ γὰρ κατὰ σάρκα ζῆτε, μέλλετε ἀποθνήσκειν “if you live according to the flesh you will die” (i.e. you are already sentenced to death). Aspectual variation (imperfective) can be accomplished in the New Testament also by means of a periphrasis with a Future form of εἰμί with the Present participle (cf. Fanning 1990: 317 sq.): (26)

1 Ep. Cor. 14.9 ἔσεσθε γὰρ εἰς ἀέρα λαλοῦντες be.FUT.2PL for in air.ACC talk.PRES.PART.NOM.PL You will be talking to the air.

3.1.3 Conclusions In agreement with Schwyzer-Debrunner (Ancient Greek generally), Fanning (NT) and other grammarians we understand the Ancient Greek (synthetic) Future as aspectually neutral. The Future is thus similar to the Aorist in this respect as was already observed by the Stoics, who are known to have classified both tenses as χρόνοι ἀόριστοι. As to its meaning, the future tense is a deictic temporal form expressing a temporal relation (‘later than’) between the verbal action and a reference point which is mostly that of the time of utterance: tı>t0 (cf. Lohnstein 1996: 207 sq.). Usages implying other meanings such as volitive, general etc., can be inferred from the fundamental temporal meaning.23 In a diachronical perspective it is important to be aware that there was a need to express durativity already in Ancient Greek and that means were available, either, as in many cases, by using a special Future form (as held by certain grammarians) or a periphrastic or modal construction such as μέλλω + infinitive or an optative with άν. It should, however, be remarked that the periphrastic expressions kept their specific meaning in the archaic and classical periods and were therefore usually not pure expressions of Future tense. The subjunctive later became a very important means of 23. Cf. Lyons 1977:816 sq. See also Fleischmann (1982) and Dahl (1985).

 Eva-Carin Gerö & Hans Ruge

expressing a future event (cf. supra), possibly through the influence of negative prohibitions with οὐ μή, as in the New Testament (cf. Ballantine 1897 and Jannaris 1897). An alternative analysis of the meaning of the future is provided by the ExtendedNow theory which is generally applied only to the Perfect (cf. McCoard 1978 and Stechow 1999, Gerö and Stechow 2003, and Anagnostopoulou et al. 1997, Iatridou et al. 2001, for Modern Greek, but recently also to the future cf. Stechow 1999, for German). According to this theory the Perfect provides an Extended Now, i.e. an interval which has the reference time as the right boundary and extends to the left into the past (cf. Stechow 1999: 39, and also Anagnostopoulou et al. 1997: 12). In a sentence like ‘I have always loved Greece’ there is an interval that ends at the utterance time and stretches into the past which can be classified contextually as ‘always’, meaning all the time since I got to know Greece. Within the interval the sentence ‘I love Greece’ is therefore valid without restriction. The same analysis can also be applied to the future. The sentence ‘I will always love Greece’ will thus be valid during an interval whose left boundary is the time of utterance and that stretches into the Future. In the case of Ancient Greek the periphrasis with μέλλω is perhaps most suitable for an analysis of this kind. By introducing an interval that could be called an Extended Now or ‘unechter Futurbereich’ (Fabricius-Hansen 1986) we believe that we are able to explain the heterogeneous use of the μέλλω construction (deontic, prospective, pure future etc.). These nuances have in common that they are oriented towards the ‘Now’ (‘the necessity to do x in the future exists now’, ‘I now intend to....in the future...’, ‘the gods have already decided now that in the future....’).24 We, however, consider the synthetic Future to be aoristic, i.e. as aspectually neutral. It can therefore be combined with definite as well as indefinite adverbs. The synthetic Future may possibly be compatible with an Extended-Now-Analysis. We would then find a parallelism with the Perfect rather than with the Aorist.

3.2

Byzantine Greek

3.2.1 Theoretical background In the Byzantine period the unstable Future category was transformed in several ways in the direction of periphrasis. But even in the Modern Greek epic of Digenis Akritas (11th C.) monolectic Future forms are often found. In the rather popular chronicles of the 6th to 12th centuries we often find the periphrasis ἔχω + infinitive (Psaltes 1913: 216 sq.). This construction provides the means of expressing aspects. Not unexpectedly, the Present infinitive rarely occurs as shown by the examples from Psaltes. Later, in the Early Modern Greek period, ἔχω + aor. infinitive becomes a Perfect (cf. Ruge 2005a and 2005b). In the periphrasis ἔχω is replaced by θέλω in combination with an infinitive, later with a subjunctive: θέλει να, which later becomes the Modern Greek Future

24. Cf. already Hermann 1801:184.

Continuity and change 

particle θα.25 In all these cases a choice of aspect is possible, but most frequently, in the unmarked “normal” use the aoristic future is chosen. Psaltes (1913: 216) cites 37 examples of Aorist infinitive in the periphrasis with ἔχω but only 3 for the present infinitive. In the Byzantine period the subjunctive can also be used with future meaning. It stands alone in the chronicles but in the early Modern Greek literature it is combined with a preceding particle να, later also sometimes with πα να (< πάω < ὑπάγω + να). 3.2.2 Data In the Modern Greek epic of Digenis Akritas the synthetic future is still present (6.769–776): (27) Συ μὀνος με ενἰκησας, συ με αποκερδἰσεις, you alone I.ACC conquer.AOR.2SG you I.ACC win.FUT.2SG ἐξεις δε με και συνεργὀν εις have.FUT.2SG but I.ACC and helpmate.ACC to τους υπεναντἰους. oυκ αποθνἠσκεις, Μαξιμοὐ, προς αυτἠν the foe.ACC.PL NEG die.PRES.2SG Maximo to her ἐφην, το δε ἐχειν σε γαμετἠν ου say.IMPF.1SG that but have.INF you.ACC wife.ACC NEG δυνατὀν μοι ἐσται· νὀμιμον ἐχω γαμετἠν ευγενἠ possible I.DAT will be lawful have.PRES.1SG wife.ACC noble και ωραἰαν, ης αγαπἠν ουδἐποτε and fair who.GEN.F.SG love.ACC never τολμἠσω αθετἠσαι. Λοιπὀν δεὐρο υπὀ σκιἀν suffer.FUT.1SG break.INF. Thus now under shadow.ACC απἐλθωμεν go away.AOR.SUBJ.1pl

του δἐνδρου και διδἀξω the tree..GEN and teach.FUT.1SG

σε ἀπαντα τα κατ᾿ εμἐ ως ἐχουν. you.ACC all that with I.ACC as have.PRES.3PL

“You alone have conquered, you shall win me all; And have me helpmate too against your foes”. “You die not, Maximo” I said to her, “But it cannot be for me to make you wife. I have a lawful wife noble and fair, Whose love I will never bear to set aside. Come let us go under the tree’s shadow, And I will teach you all that me concerns.”

25. For different and competing earlier forms of this particle cf. Joseph and Pappas 2002.

 Eva-Carin Gerö & Hans Ruge

An example from Psaltes (1913: 216 sq.) of periphrasis with ἐχω + infinitive in the chronicles: (28)

λαλἠσαι ἐχω, καλὠς ἐχομεν Speak.AOR.INF have.PRES.1.SG well have.PRES.1.PL ζἠσαι live.AOR.INF I will speak, we will live well.

Examples from Psaltes 1913: 217) of “pure” subjunctive: (29) ου δὐνηται ρὐσασθαι NEG be able.SUBJ.3 SG defend.AOR.INF.MID He will not be able to defend himself An example of the Early Modern Greek combination na + subjunctive: (30) και να σας τ᾿ αφηγοὐμαι τις and PTC you.acc.PL and tell.PRES.SUBJ.1SG who

εἰμαι be.PRES.1SG and I shall also tell you who I am (Λἰβιστρος και Ροδἀμνη 1625f)

3.2.3 Conclusions As we may expect there were several possibilities to express future events also in the Byzantine period (cf. especially Rosenqvist 1981: 68 sqq.). The above examples show only some of the possibilities and there are others, as was also the case in Antiquity. The periphrases should be seen as grammaticalization of lexical means of expression that were available in Ancient Greek. The periphrasis with ἔχω was obviously not capable of survival, as it later mutated into a periphrasis of the Perfect. It was replaced by θέλω and the particle θα that developed from it. The subjunctive is also often used as a replacement of the Future. In all these cases a possibility of differentiating aspect automatically arises, as the infinitive as well as the subjunctive have to “follow suit” as regards aspect. But in nearly all cases the aoristic (perfective) aspect is chosen. The synthetic Future continues to be used and the Modern Greek θα -construction is nowhere near to be established. Summing up, one could say that all this is characteristic of the instability of the category of Future.

3.3

Modern Greek

3.3.1 Theoretical background The Modern Greek Future is analytic/periphrastic and is formed by the particle θα followed by the subjunctive. As there is an analytic as well as a periphrastic subjunctive

Continuity and change 

there will automatically also be two Futures, Aoristic and paratatic. As a result of the orthographic reform of 1976 the paratatic subjunctive is identical with the present also in writing. The paratatic Future therefore always has the shape of θα + present. 3.3.2 Data Aoristic Future: (31) Θα θυμὠσει ο Πἐτρος PTC be angry.AOR.3SG Petros Petros will be angry (32) Θα μεἰνει στην Ελλἀδα δὐο χρὀνια PTC stay.AOR.3SG in-ART Greece two years he will stay two years in Greece (33) Θα καθἰσω στον ἠλιο PTC sit.AOR.1SG in-ART sun I will sit down in the sun (34) Σἠμερα θα βρἐξει today PTC rain.AOR.3SG It will rain today Paratatic Future: (35) Σἠμερα θα βρἐχει ὀλη την ημἐρα today PTC rain.PRES.3SG all ART day Today it will be raining all day (36) Θα υπἀρχω ὀσο υπἀρχεις εσὐ PTC exist.PRES.1SG as-long exist.PRES.2SG you I will exist as long as you exist Like the English Future construction with will, the paratatic Future can also express a supposition with regard to the present. (37) Θα κοιμἀται τὠρα PTC sleep.PRES.3SG He will be asleep now

now

3.3.3 Conclusions As we have observed, Modern Greek has a well established and grammaticalized analytic Future with obligatory expression of aspect. In comparison with the many variations and periphrases that were current in the Byzantine period a certain stability thus came about through this regularization. But what changes did actually occur after the classical period? Classical Greek had only one grammatical Future, which was morphologically akin to the Aorist but will have to be seen as aspectually neutral(ized),

 Eva-Carin Gerö & Hans Ruge

because it was not opposed to any other form. When necessary, a paratatic imperfective Future could however be formed by a periphrasis (mostly with μέλλω and a Present infinitive). In later periods, as is still the case today, there is just one grammatical Future with a choice of aspect, which is rather one-sided as the paratatic Future is used much more rarely than the aoristic in Modern Greek. We therefore observe the following continuity through the centuries: in the normal case the Future is aoristic (aspectually neutral/perfective), but there was and still is a possibility of expressing an imperfective aspect by a paratatic form.

4. Final remarks The goal of this paper was to demonstrate two cases of continuity in the Greek tense system from Antiquity to the present time. This journey through the millennia has made it clear that the Imperfect and the Future have kept their forms and functions in different ways but nevertheless in a surprisingly continuous fashion in both cases. As regards the Imperfect, there is little to say about morphology. The stem formation shows some innovation and the conjugation has been somewhat modified. The meaning has been kept unchanged. In the course of time and in the use of the Greek language in different periods and in different literary genres there were obviously minor changes in the use of the Imperfect in relation to the Aorist, but considering the multitudinous language contacts the use of aspect in the past tenses was surprisingly stable. This constant meaning of the Imperfect can be defined more or less cogently within the framework of different theoretical models as developing (traditional grammar), intraterminal (Johanson 1971) or inclusive (Klein 1994). In the case of the Imperfect all these models seem to be compatible. When defined in such ways, the specific meaning of the Imperfect can hardly be the unmarked member of the Imperfect/Aorist opposition and does certainly not have a primary type-referential function. The marked nature of the Imperfect was actually observed already by the Stoics, who categorized this tense as one of the χρόνοι ὡρισμένοι. The situation is somewhat different as regards the Future. Superficially, Modern Greek seems to show radical innovation because of the possibility to express aspectual differentiation also in the Future tense, something that – for Ancient Greek – was wished for already in the 18th and 19th centuries by grammarians interested in the philosophy of language. If we look more closely this “change” will however not seem so extraordinary: Ancient Greek already had the possibility of expressing different aspects also in the Future by means of modal expressions or by a periphrasis with μέλλω + infinitive (present as opposed to Aorist). In Byzantine times more types of periphrases were added. A more important fact is that Ancient Greek Future is aspectually neutral (the already mentioned Stoics called it χρόνος ἀόριστος μέλλων and the stem formation mostly shows an affinity to the aoristic Past) as well as the fact that the Modern Greek perfective Future of the Aorist stem is pragmatically so dominant that it

Continuity and change 

seems to be aspectually neutral. We may also mention the curious circumstance that the Modern Greek aoristic Future is different from the Ancient Greek Future only in regard to the presence of the preceding particle θα. This is valid only for the active. It should also be mentioned that the Future is a somewhat peculiar tense as it refers to something that is non-factual. It is therefore close to the non-indicative moods and ever since Antiquity it could be replaced by those moods. The historical background of the grammatical Future itself is a subjunctive, in Ancient Greek and especially transparently in Modern Greek. An analysis by Extended-Now as suggested by Stechow (1999) for the German Future does not seem to be necessary for the synthetic Future of Ancient Greek nor indeed for Modern Greek. The Ancient Greek and Byzantine periphrasis with μέλλω would seem to be a better candidate for a meaning of this kind with a focus on the contextual ‘Now’.

References Anagnostopoulou, E., Iatridou, S. & Izvorski, R. 1997. On the Morpho-Syntax of the Perfect and How it Relates to Meaning. Ms, MIT. Babiniotis = Μπαμπινιὠτης, Γ. 1972. Το ρἠμα της ελληνικἠς. Athens. Ballantine, W.G. 1897. Negative futures in the Greek New Testament. American Journal of Philology 18: 453–461. Björck, G. 1940. Ἦν διδάσκων. Die periphrastischen Konstruktionen im Griechischen. Uppsala & Leipzig. Blass, F. 1892. Demosthenische Studien,...: Das Futur. Rheinisches Museum 47. Blass, F. & Debrunner, A. 1979. Grammatik des neutestamentlichen Griechisch. (15. Aufl. herausgegeben von F. Rehkopf). Göttingen: Vandenhoek & Ruprecht. Bornemann, E. 1978. Griechische Grammatik. (Unter Mitwirkung von E. Risch). Frankfurt: Diesterweg. Buttmann, P. 1841. Greek Grammar. (Translated from the German by C. Supf, 2nd edn). London. Coseriu, E. 1992. Einführung in die Allgemeine Sprachwissenschaft. (2. Auflage). Tübingen: Francke. Dahl, Ö. 1985. Tense and Aspect Systems. Oxford: Blackwell. Dahl, Ö. (ed.). 2000. Tense and Aspect in the Languages of Europe. Berlin: de Gruyter. Davidson, D. 2001. Essays on Actions and Events. Oxford: Clarendon Press. Fabricius-Hansen, C. 1986. Tempus fugit. Düsseldorf: Schwann. Fanning, B.M. 1990. Verbal Aspect in New Testament Greek. Oxford: Clarendon Press. Fleischman, S. 1982. The Future in Thought and Language. Diachronic evidence from romance. Cambridge: CUP. Gerö, E.-C. 1996. In the footsteps of Raphael Kühner: Einige Bemerkungen zu den Theorien von Rijksbaron und McKay. Ms. Universität Tübingen (Vortrag, Promotionskolloquium 960126). Gerö, E.-C. 2007. Worlds, events and individuals. The syntax of intensionality and extensionality in Ancient Greek. Graeca Stockholmiensia. MS.

 Eva-Carin Gerö & Hans Ruge Gerö, E.-C. & von Stechow, A. 2003. Tense in time: the Greek perfect. In R. Eckard et al. (eds.), Words in Time: Diachronic semantics from different points of view, 251–293.Berlin: Mouton de Gruyter. Giannakidou, A. 1997. The Landscape of Polarity Items. PhD dissertation, University of Groningen. Giannakidou, A. 1998. Polarity Sensitivity as (Non)veridical Dependency. Amsterdam: John Benjamins. Gildersleeve, B.L. 1900–11. Syntax of Classical Greek, from Homer to Demosthenes, I-II. New York NY: Bouma´s Boekhuis. (Nachdruck Groningen 1980). Goodwin, W.W. [1887]1965. Syntax of the Moods and Tenses of the Greek Verb. London: Macmillan. Goodwin, W.W. 1889. Syntax of the Moods and Tenses of the Greek Verb. (Rewritten and enlarged). London. Hedin, E. 2000. The type-referring function of the imperfective. In Ö. Dahl (ed.), 227–264. Hermann, G. 1801. De emendanda ratione Graecae grammaticae, I. Leipzig. Holton, D., Mackridge. P. & Philippaki-Warburton, I. 1997. Greek: A comprehensive grammar of the modern language. London: Routledge. Humbert, J. 1960. Syntaxe grecque. (3e édition). Paris: C. Klincksieck. Iatridou, S., Anagnostopoulou, E. & Izworski, R. 2001. Some observations about the form and meaning of the perfect. In M. Kenstowicz (ed.), Ken Hale: A life in language, 189–238. Cambridge MA: The MIT Press. Jannaris, A. N. 1897. An Historical Greek Grammar Chiefly of the Attic Dialect. London. Johanson, L. 1971. Aspekt im Türkischen. Uppsala: University of Uppsala. Joseph, B.D. & Pappas, P. 2002. On some recent views concerning the development of the Greek future system. Byzantine and Modern Greek Studies 26: 247–273. Kahane, H. & Kahane, R. 1958. The tense system of Modern Greek. In Omagiu lui Iorgu Iordan, 453–474. Bukarest: Editura Academiei Republicii Romîne. Klein, W. 1994. Time in Language. London: Routledge. Lohnstein, H. 1996. Formale Semantik und Natürliche Sprache. Opladen: Westdt. Verlag. Lyons, J. 1977. Semantics 1–2. Cambridge: CUP. Mackridge, P. 1987. The Modern Greek Language. Oxford: OUP. McCoard, R.W. 1978. The English Perfect: Tense choice and pragmatic inferences. Amsterdam: North-Holland. McKay, K.L. 1988. Aspectual usage in timeless contexts in Ancient Greek. In Rijksbaron et al. (ed.), 193–208. Mandilaras, B. G. 1973. The Verb in the Greek Non-Literary Papyri. Athens: Hellenic Ministry of culture and sciences. Psaltes, S. 1913. Grammatik der byzantinischen Chroniken. Göttingen:Vandenhoek & Ruprecht. Reichenbach, H. 1947. Elements of Symbolic Logic. New York NY: Macmillan. Rijksbaron, A., Mulder, H.A. & Wakker, G.C. (eds.). 1988. In the footsteps of Raphael Kühner. Proceedings from the international commemoration of the 150th anniversary of the publication of Raphael Kühner’s Ausführliche Grammatik der griechischen Sprache, II. Theil: Syntaxe. Amsterdam 1986. Amsterdam: Gieben. Rijksbaron, A. 1984. The Syntax and Semantics of the Verb in Classical Greek: An introduction. Amsterdam: Gieben.

Continuity and change  Rijksbaron, A. 1988. The discourse function of the imperfect. In Rijksbaron et al. (eds.), 237–254. Robertson, A.T. 1919. A Grammar of the Greek New Testament in the Light of Historical Research. (3rd edn). New York. Robins, R.H. [1967] 1990. A Short History of Linguistics. (3rd edition). London: Longman. Rosenqvist, J.O. 1981. Studien zur Syntax und Bemerkungen zum Text der ’Vita Theodori Syceotae’. Uppsala: Almqvist & Wiksell. Rosenqvist, J.O. 1986. The Life of St Irene Abbess of Chrysobalanton. A critical edition with introduction, translation, notes and indices. Uppsala: Almqvist & Wiksell. Ruge, H. 1985. Atavismen im Neugriechischen. Glossologia 4: 31–38. Ruge, H. [1986] 2001. Grammatik des Neugriechischen. (3. Auflage). Köln: Romiosini. Ruge, H. 2005a. Morphologisches Recycling in byzantinischer Zeit: Vom Perfekt Aktiv zum Aorist Nichtaktiv. In L. M. Hoffmann & A. Monchizadeh (eds.), Zwischen Polis, Provinz und Peripherie. Beiträge zur byzantinischen Geschichte und Kultur, 639–648. Wiesbaden: Harrassowitz. Ruge, H. 2005b. Wozu braucht man ein Perfekt? Aussterben und Wiedergeburt einer grammatischen Kategorie in byzantinischer Zeit. In S. Kolditz & R.C. Müller (eds.), Geschehenes und Geschriebenes. Studien zu Ehren von Günther S. Henrich und Klaus-Peter Matschke, 255–262. Leipzig: Eudora-Verlag. Ruipérez, M. S. 1954. Estructura del sistema de aspectos y tiempos del verbo griego antiguo: Análisis functional sincrónico. Salamanca: Colegio trilingüe de la Universidad. Rydbeck, L. 1969. Bemerkungen zu Periphrasen mit εἴναι + Präsens Partizip bei Herodot und in der Koine. Glotta 47: 186–200. Schwyzer & Debrunner = Schwyzer, E., Griechische Grammatik, I. München 1939, II. vervollst. u. hrsg. von A. Debrunner, München 1950 (Handbuch der Altertumswissenschaft, II.1.1–2). Seiler, H. 1952. L’ aspect et le temps dans le verbe néo-grec. Paris: Société d’Éditions ‘Les Belles Lettres’. Smyth, H.W. [1920] 1956. Greek Grammar. (Revised by G. M. Messing). Cambridge MA. Stechow, A. v. 1999. Eine erweiterte Extended Now-Theorie für Perfekt und Futur. Zeitschrift für Literaturwissenschaft und Linguistik 113: 86–118. Turner 1963 = Moulton, J. H. A Grammar of New Testament Greek, Vol III: Syntax. Edinburgh: Clark. Tzermias, P. 1969. Neugriechische Grammatik. Bern & München: Francke. Veloudis, I. 1983–84. Η υποτακτική στις αναφορικές προτάσεις. Glossologia 2(3): 111–135. Versteegh, H. 1980. The Stoic verbal system. Hermes 108: 338–355. Wackernagel, J. 1920–24. Vorlesungen über Syntax, mit besonderer Berücksichtigung von Griechisch, Lateinisch und Deutsch. Basel.

Actionality and aspect in Hittite Folke Josephson The Hittite reflexive construction has perfective-like functions like Spanish and Italian reflexives. We will consider if they express ‘inner’ aspect. Most present tense forms in narrative tense-switching belong to verbs with a -ske/a-suffix. These verbs can be seen as +INTRA items with actional and aspectual features. -kan and -san are frequently found in the last slot of Wackernagel clitics. -san is allative-directional, -kan locational. -kan indicates initial and final limitation and is similar to Latin con- and Gothic ga-. Lexical (inner) aspect is an important element in modern analysis of Russian perfectivity. An aspectual ‘pair’ system of the Russian kind with an interplay between telicity and quantification as expressed by preverbs and imperfectivity expressed by suffixation did not arise in Hittite though -kan and -ske/a- which could have allowed a similar development were present as were con- and -sco in Latin. Hittite preverbs did not play the same role for aspectuality as preverbs did in many other IE languages but -kan and -san were involved in the VP and affected verb semantics and aspectuality. The discussion will be focused on the function of clitics and verbal suffixation.

1. The reflexive construction The function of the Hittite reflexive particle -za, which belongs to the second to last slot in the long chain of enclitic particles in the Wackernagel position at the beginning of the sentence, which is a typical trait of the Anatolian languages, was discussed in a general typological perspective in Josephson (2003). The synchronic as well as the diachronic relation of the Hittite reflexive construction to the inherited medio-passive verbal forms which indicate Middle Voice was discussed in that article. It also contains a discussion of the complicated valency-changing functions of the reflexive. A typological parallel with the subjective use of the Spanish reflexive was seen to exist. This function is not found in most other Romance languages but exists in Italian. There is a valency-changing function of the Hittite reflexive which works two ways and, quite differently from the Middle form of the verb, may sometimes increase transitivity and the agentive function of the subject.

 Folke Josephson

In Hittite, the continued existence of many functions of the medio-passive seems to have prevented a full-scale development of objective and deagentive reflexive construction. A spreading use of reflexives with middle function did however occur (cf. Boley 1993: 212.). The Hittite reflexive construction with -za can have a telicizing function. This was observed by Garrett (1996: 95) with respect to -za esari ‘sits down’, a middle verb in reflexive construction. The following examples clearly show the telic function: (1) kissann-a le tesi arha-wa parkunummi so-and not say.PRES.2SG off.PTC clean.PRES.1SG And do not say: I will thoroughly clean up which is followed by: (2) parkunusi-ma-za UL kuit Clean.PRES.2SG-but-REFL not anything but you (actually) clean nothing up The promise refers to a cumulative action. It is subsequently “quantized” by -za in the way described by Nishida (1994) who uses that term for the telic function of Spanish se, a function that is not shared by reflexive markers of other Romance languages except Italian. Maldonado (2000: 167) discusses the “focusing function” of se which focalizes some portion of the event, which may be spontaneous or unexpected. The whole event is bounded and no gradual adverb is compatible with the construction. Another variant is the completive import of the reflexive combined with transitivity increase (Maldonado 2000: 167). This completivity is understood by Jacquesson (2001: 101) as aspectual. Verbs of consumption with full subject involvement (beber, comer etc.) were discussed by Maldonado in this context and by Nishida (1994) in the context of telic and what he calls “quantized” situations. A similar use of the reflexive is found in Hittite. Maldonado (2000: 180–183) emphasizes that what looks as an aspectual use in Spanish is a consequence of the focusing function and the location of the object in the subject’s dominion which are proper to the reflexive. Completiveness derives from the internal properties of the verb as they are profiled by the clitic. This is “construction meaning” in the sense of Goldberg (1995). Sanz & Laka (2002: 326) understand the internal properties as Aktionsart. A possible aspectual function of the reflexive in Spanish was suggested by several Romance scholars (cf. Nishida 1994, Zagona 1996, and Jacquesson 2001: 101). This function has, however, been considered by other scholars as being aspect-like but not properly aspectual (cf. Maldonado 2000). Sanz and Laka, in spite of understanding the internal properties as Aktionsart, consider themselves as following the aspectual view of Nishida and Zagona who use “transitions” to cover the use of se with verbs designating consumption and motion. Sanz & Laka (2002: 326) prefer the term realización. D’Alessandro (forthcoming) understands the Italian phenomenon as inner aspectual projection that encloses telicity.

Actionality and aspect in Hittite 

The situation in Hittite is similar. The use of the reflexive in order to stress the initiation of the action by the subject (or as a spontaneous event) in this language may be seen as aspect-like but is probably not aspectual in the sense of outer aspect. The meaning of the reflexive clitic in (2) seems to be an example of the kind of inner aspectual projection described by D’Alessandro. The enclitic reflexive particle in second to last position in a chain of enclitic Wackernagel particles can be followed in the last position by the particle -kan which indicates immediate effectuation of the action or by -san which indicates direction towards a term and ultimate arrival at the term (cf. Josephson 1972, 1995, 2003). Subjective reflexive function and different kinds of focusing function including completive or telic and inner aspectual properties are important traits of the reflexive construction in the Anatolian branch of Indo-European (cf. Josephson 2003: 228).

2. The -ske/a- suffix The Hittite verbal suffix -ske/a- was inherited from Indo-European *-sk’e/o-. It has been considered as a mark of “verbale Pluralität” (Dressler 1968). According to Harðarson (1998: 339) the inherited function of the suffix was iterativity. Atelicity is a predominant characteristic of the Hittite -ske/a- verbs. They certainly have a frequentative as well as a progressive meaning. Grønn mentions “subinterval property (atelicity)” for the progressive and habitual-iterative meanings of atelicity in his contribution to this volume. The suffix -ske/o- is of frequent occurrence, with similar functions, in Ionian Greek. The important functional similarities of the Hittite -ske/a-verbs and the -ske/overbs of the Greek epic were noted by several scholars (cf. Puhvel 1991, Watkins 2001). The productive Latin -sco-suffix was thoroughly treated by Haverling in a major monograph (Haverling 2000): The action of the Latin sco- verbs is dynamic and non-terminative and may be gradual (pp. 155, 159). This may be true also for a prefixed verb like permanasco ‘seep through gradually’ (p. 135). In a verb like albesco it is added to a fientive stem albe- and indicates gradual entrance into a state (cf. Harðarson 1998: 339). sco-verbs prefixed by actional con- may refer to the completion of an action without emphasising its final stage (Haverling 2000: 261). In early Latin some common scoverbs are always prefixed by con- as consuesco, conticesco, concupisco, or by some other prefix as in the case of obdormisco (p. 138). The dynamic suffix and a prefix are necessary to indicate a change from one state of another in verbs like condormisco /obdormisco ‘fall asleep’ or conticesco ‘stop talking’. There is an interesting relationship between ad- and con- with ad- indicating gradual change as in adsuesco (p.280) as against consuesco, where con- indicates terminativity (p. 273). The IE *-sk’e/o- suffix has other functions in certain other IE languages. It has an aoristic function in Armenian and a causative function in Tocharian where it forms the regular causative in Tocharian A and the present causative Class IX in Tocharian B.

 Folke Josephson

That the Hittite -ske/a- form has an imperfective function was first suggested by Bechtel (1936). This was more recently advocated by Melchert (1997, 1998) and by Hoffner/Melchert (2002). In a paper presented at the meeting of the Indogermanische Gesellschaft in Cracow 2004 Alexandra Daues (handout) notes several similarities as well as some differences in regard the functions of the iterative-distributive and durative *-ske/o- forms in Hittite and Greek. Syntactically, Hittite -ske/a- and Greek -ske/o- verbs are not infrequently found in relative clauses and several -sk-forms often occur in a row. They also give background information or additional information. In explaining one of her Hittite examples Daues suggests that “backgrounding” could be one of the functions of the -ske/a- form. The examples given for Homeric -ske/o- are all in the past tense and formed either on the imperfect or on the aorist. There is only one example of the present tense in the Hittite material presented by Daues. A backgrounding function of the present tense of Hittite -ske/a- verbs and of the present of some other verbs was suggested by Josephson (2006) at the Dichtersprache (langue poétique) conference of the Indogermanische Gesellschaft in Paris 2003 in connection with the question of a specific narrative style in Middle Hittite epic texts translated from Hurrian. A similar function was seen to exist in ritual protocols and other texts and was found to belong to a narrative style proper to Hittite. The use in narratives of the present tense of -ske/a- verbs and of the present tense of some unsuffixed verbs in obvious alternation with preterits of unsuffixed verbs was seen to imply an aspectual function of the present tense form as opposed to the preterit. This kind of opposition between the tense-forms in tense switching was examined in a typological perspective. Tense switching in Middle Hittite narrative texts was found to be similar to that found in the Old French Song of Roland, in certain types of Russian narrative texts as well as in narrative texts of other related and unrelated languages. The Song of Roland does not show the later French opposition of passé simple and imperfect. The present is used for stative and durative background situations. This has been called narrative subordination or grounding (cf. Fleischmann 1985: 851). The narrative present may also be used for “High foreground” or saliency which is the case in the French medieval text Aucassin et Nicolette. This is similar to the alternation that is found in a Russian folktale: vyšel starik na bereg i govorit ‘the old man came out on the shore and says’, where the present is obligatorily imperfective and has the foregrounding effect “only when contrasted with a perfective past” (cf. Chvany (1990: 223). An interesting case of tense switching between a durative present (which is not marked morphologically by any suffix) in its normal use for custom and repetiton and the aorist was observed in Greek by Peppler (1933) citing two passages from Aristophanes (Nub. 348–352 and Plutos 234–244).

Actionality and aspect in Hittite 

This type of tense switching is also found in Hittite as in the following passage of Ullikummi epic: (3) nu-kan and.PTC

Istanus nepisaz Istanu.NOM.SG heaven.ABL

[katta] sakuwait down look.PRET.3SG

nu-za Ullikummin sakuiskizzi and-REFL Ullikummi.ACC see.PRES.3SG The sun(god) looked down from heaven and he sees Ullikummi (KUB1 XVII+ IV 33–34) In the Hittite passage the perfective past is expressed by the simple preterit and the present tense of the suffixed -ske/a- verb is used for saliency or narrative focus. It should be observed that the reflexive verb -za...sakuiskizzi refers to a discovery and is neither durative (continuative) nor iterative (repetitive) in this passage. Hittite possesses a non-augmented preterit but does not have the IE aorist and the IE perfect. (It has a periphrastic perfect and pluperfect.) There are no augmented verbal forms and the question of the augment, which is relevant for Greek (the -ske/oforms are never augmented) is of no importance for Hittite. The fact that the interchange of present and preterit in Middle Hittite narrative style seems to belong to the aspectual domain does not lead us to suggest that aspect should be considered as a main function of the tenses in the Hittite language. But when we find an aspectual function of a preterit which is opposed to the imperfective present tense in Hittite narrative style, the conclusion will be that the -ske/a- morpheme in most of the present forms occurring in tense switching in the Middle Hittite narrative texts is not a fundamental marker of grammatical aspect, but rather a marker of aktionsart. Aktionsart is what is normally morphologically marked by suffixes in IE languages. One of those suffixes is the continuative or repetitive *-sk’e/o- suffix. There are other suffixes of similar function in Hittite. One of those is -anne/a- which has been said to be intensive or iterative-durative. Several suffixes may occur simultaneously with one verb as when -ske/a- is added to the causative suffix -nu- or in the case of the interesting combination -anneske/a-. In this combination, -anne/a- should be understood as indicating +INTRA progressive action of a kind in which the agent is located in the midst of the activity: “be in the course of ” (cf. Johanson 2000: 95) and -ske/a- as referring to continuative action. That the -anne/a-suffix has a +INTRA progressive function of the kind discussed by Johanson is indicated by the following examples: (with -anne/a-): (4) n-at-kan kasa pahhunaz huittiyannahhi and-it-PTC lo fire.ABL draw.PRES.1SG lo, I pull it from the fire (KUB II 9 IV 21)

1.

KUB = Keilschriftsurkunden aus Boghazköi. Berlin 1921-

 Folke Josephson

(with -anne/a- and -ske/a-): (5) kinun-a-tta kasa humandaza EGIR-pa now-and-you.ACC lo all.ABL back huittiyanneskimi pull.PRES.1SG but now, lo, I am drawing you back altogether! (KUB X 72 II 13–15) (with only -ske/a-): (6) GI2ummiyaz-iya-wa kuez KAŠ-essar huittiyeskimi Straw.ABL-and-PTC which.ABL beer.ACC draw.PRES.1SG the straw with which I draw beer (KBo3 XIII 64 obv.12) Example (4) refers to the subject’s involvement in the removal of the object from its initial position. In (5) the added -ske/a- suffix indicates progressivity. In (6) it has frequentative (habitual) meaning. It should be observed that the great majority of occurrences of huet-, huittiya- in Hittite texts are unsuffixed and that they are very often present forms. The -ske/a- forms are progressive-like but are not +INTRA high focal items. Progressive-like forms of this kind preserve their original iterative or continuative meaning (cf. Johanson 2000: 95). They are preaspectual and may occur in habitual contexts, i.e. be compatible with pluri-occasionality (Johanson 2000: 97). The -anne/a- form, on the other hand, should be seen as a truly progressive +INTRA form, which is high focal. If we follow the argument of Johanson this kind of form would be more likely than the -ske/a- form to develop into a grammaticalised aspectual form. This did not happen to the -anne/a- form in Hittite. We mentioned that what may be seen an aspectual function of the tense opposition present/preterit occurs in Middle Hittite narrative texts. The reason for the aspectual opposition of the tense forms may in many cases be described as backgrounding and foregrounding but the subtleties of the use, which is connected with narrative technique will have to be studied thoroughly in a typological perspective, as I have attempted to show (Josephson 2006). That the progressive -ske/a- suffix which naturally indicates an action that is (completed) was frequently combined with the present tense in narrative passages of Middle Hittite may be easily understood, but the occurrence of a continuative form in combination with an imperfectively used present tense in narrative style should not lead to an a priori confusion between the aktionsart expressed by -ske/a- and the temporal/aspectual sphere.

2. Sumerian determinative. 3.

KBo = Keilschrifttexten aus Boghazköi. Leipzig 1916-23/Berlin 1924-

Actionality and aspect in Hittite 

Verbs of energeia (according to the categories of Dowty 1977) are found in tense switching without the use of the -ske/a- suffix as in the case of ars- ‘flow’ in KUB XXXIII 113+ I 29’-30’: (Ullikummi epic): (7) [nu-za] dU-as dagan esat and-REFL Stormgod.NOM ground.LOC sit.PRET.3SG nu-ssi- kan ishahru [par]a PA5.HI.A-us man arsanzi and-he.DAT-PTC tear-NOM.PL forth stream.NOM.PL as flow.PRES.3PL The stormgod sat down on the ground and his tears flow forth like streams. Verbs of kinesis use the continuative -ske/a- suffix together with the present tense in tense switching when the present of these verbs creates an interval in the chain of events (cf. the discussion in Dowty op.cit.). The passage of the Hittite Laws I paragraph 43: (8) takku LÚ-as GU4 ÍD-an zenuskizzi If man.NOM.SG ox river.ACC.SG cross.CAUS.PRES.3SG tamais-an Another.NOM.SG-he.ACC.SG

suwaizzi nu KUN GU4 push.PRES.3SG and tail ox

epzi ta ÍD-an zai take.PRES.3SG and river.ACC.SG cross.PRES.3SG if a man crosses a river with his ox (i.e. makes his ox cross a river: causative -nu- form of zai-with -ske/a-) and another man pushes him off and grasps the tail of the ox and crosses the river... has been interpreted by scholars in different ways. Obviously, the -ske/a- form here provides a background to the following action. The meaning of the verb is not iterative in this passage. It indicates low focal progressivity. The Hittite -ske/a- verbs are similar to Bulgarian verbs with the suffix -va- which were analysed by Dimitrova-Vulchanova (1999: 13–14) as expressing frequency, iterativity or (significantly!) interval. The Russian verbs suffixed by -ivaj- in the birch bark letters of Novgorod express iterativity, distributivity and activity (as opposed to action!) according to Nørgaard-Sørensen (1997). Sørensen believes that the category of aspect did not yet exist in the language of the Novgorod texts. In Serbian, which possesses the classical Slavic aspectual system with perfectivizing prefixes, the suffix -iva- checks the features of aspectual prefixes and adds the meaning of repeated action (cf. Progovac 2002: 204–205). It may on occasions be paraphrased in terms of universal quantification: “on every (relevant) occasion”. The suffix also has a frequent use as durative. In modern Russian the imperfective suffix can “take telic predicates and map them into predicates that closely mimic the behaviour of atelic ones” (Kratzer 2004: 405). This is a kind of Higher imperfective operator that takes scope over the culmination implication of a predicate with a telic prefix. Such derivational morphemes are exponents of inner aspect and modifiers of eventuality types whereas the exponent of

 Folke Josephson

imperfective aspect belongs to the domain of outer aspect (cf. Filip 2000, Filip and Rothstein 2006). The Hittite -ske/a- verbs are dynamic and non-terminative like the Latin -scoverbs. Melchert and Hoffner claim that they are imperfective. For the origin and development of this kind of use I cite Johanson (2000: 95): “When a preaspectual progressive-like item develops into an early aspectual +INTRAHF item, there are stages where it exhibits both actional and aspectual features.” Johanson refers to his 1995 study of this kind of ambiguity in Turkish. This stage had been reached in Hittite. The aktionsart meanings of iterativity and universal quantification are obviously present and are very frequent meanings of the -ske/a- forms. It is however not obvious that there was a fullscale grammaticalisation of the -ske/a- form.

3. The sentence particles The enclitic so called “sentence particles” -san, -kan and -(a)sta are present in Hittite from an early period of the language. At the early period there were two more particles of this kind: -(a)pa and -an. These “sentence particles” are normally placed in the final slot in the chain of particles that follow the initial word of the sentence, which is often a sentence introductory particle, usually nu or ta. -san and -kan are also found in other positions in the sentence. The “sentence particles” are a typical trait of Hittite, but the phenomenon is Common Anatolian as similar particles are found in Luvian and in other related Anatolian languages. In Hittite the order of particles in the Wackernagel chain is fixed as follows: 1. citation particle-wa(r) 2. dative pronoun 1.2.3. person plural: -nnas, -smas, -smas 3. nominative pronoun: common gender sing. -as, or accusative pronoun sing. -an. The plural forms are subject -e and -at, DO -us/-as. Neuter sing. -at, plur. -e/-at in both functions. (The nominative pronoun is never used with a transitive verb!) 4. dative pronoun 1.2.3. person sing.: -mu, -tta/-ddu, -sse/-ssi 5. the reflexive particle -za 6. the “sentence particles” -a and -ma ‘but’ are placed before other particles. This is also the case of the potential and optative particle -man. These particles do not belong to the set of particles that most frequently follow a sentence introductory nu-(n-), ta- and (in the oldest language) su- (s-). Legendre (2000) discusses Romanian verbal clitics. She cites several studies by S.R. Anderson concerning second position Wackernagel clitics, where it is claimed that there is a need to resort to non-syntactic mechanisms in order to characterize their distribution. She particularly refers to Klavans (1985) who maintains that clitics are phrasal affixes. According to Legendre (2000: 219–220) verbal clitics are best

Actionality and aspect in Hittite 

analyzed as phrasal affixes. The verbal clitics of Romanian show clustering and a rigid order which “results from the same violable alignment constraints that operate in South Slavic Wackernagel languages”. The approach of Legendre, who favors Optimality Theory as do several scholars in recent literature concerning clitics in Romance and Slavic languages, seems to be appropriate for explaining the behaviour of Hittite clitics that belong to verbs, have a rigid order and are moved into a well-ordered chain in Wackernagel position. Besides being important for questions of the diachronic development of syntactical and morphological categories in various Indo-European languages the category of Anatolian clitics which includes the “sentence particles” is obviously of relevance to the general typology of clitics. The clitic -(a)sta, which has functional similarities to -kan, but alternates with it for reasons connected with deixis and directivity, went out of use at a certain stage of Hittite language history. The fundamental opposition between -san and -kan survived for a longer period but -san was gradually replaced by -kan. -san had lost much territory in the late phase of the language but apparently never disappeared completely. The situation in Hittite became similar to that of the related Hieroglyphic Luvian, which has only one particle of this kind: -ta, with the same function as that of late Hittite -kan. Cuneiform Luvian, as being more conservative, shows two particles: -tar and -ta. The function of these particles, as local indicators or as modifiers of actionality or possibly of aspect has been subject to a long and at times lively debate. The idea of -kan as a marker of perfectivity was originally suggested by Laroche (1961: 35). In Josephson (1972) and in later papers I put the main emphasis on the function of these clitics in the domain of actionality or aspect while not neglecting their origin and continued function as deictic elements or local markers. The function of -kan in relation to the verb can in many respects be compared to that of the Gothic preverb ga- (cf. Josephson 1976), to which it is etymologically related (Josephson 1976: 171–175) and to the Latin prefix con-, which is of the same origin and of similar function. It is an interesting fact that a prefixed ka-, which is probably a shorter ablaut form of -kan (<*kom) as is Gothic ga-, is found in Hittite as in the verb kappuwai-, which corresponds exactly in structure and meaning to Latin computare. A prefixed ka- is also found in Lydian, another Anatolian language. Gothic ga- was suggested by Kuryłowicz (1964) to possess an aspectual function. This Gothic particle may indicate a terminative manner of action. The action qualified by ga- may be seen as directed towards a term and be of cursive nature, but the particle may also indicate resultative manner of action and a complete event. When ga- is used in this way the stress is placed on the reaching of a term or the passing of a limit. In the latter case the limit may be the initial limit of the action, which is also true in the case of Hittite -kan (cf. Josephson 1976: 170 sqq.). The function of the Latin preverb con- is similar to that of Gothic ga- as the action of the con-verb may refer to the term without being seen as completed. In Hittite, however, different particles are used for direction towards a term and for a completed event. -kan is the mark of punctual occurrence and completed event.

 Folke Josephson

The function of the preverb ge- in Middle High German was treated by Solms (1991) and more recently by Herbers (2002), who uses a diachronic approach. It has ingressive and effective functions. In one of its uses (ge2 in the classification of Solms) it indicates (like ver- and er-) “vollständige Durchführung”. ge3, according to the same classification, indicates establishment of a contact. ge4 indicates that the goal has been reached and the effect on the accusative object is then also stressed. ge5 stresses the beginning of the action, the change into another state or the momentaneous occurrence of an action: MHG gefangen can have the sense of later German anfangen. Furthermore, there is a syntactic use of ge- with verbs in the preterit which indicates that the action had been concluded before another action occurring in the context. With verbs in the present tense ge- may refer to a future action. The moment of completion is stressed. In Gothic, ga- is used in a similar way. It may be terminative without being resultative just as MHG ge2 but it is also used in the way of MHG ge4 and ge5. MHG ge2 and ge3 are functionally similar to Hittite -san and ge4 and ge5 to Hittite -kan. The syntactic function of ge- can also be found in respect to Hittite -kan. The future meaning of ge- with the present tense is reminiscent of the Slavic present perfective (Herbers 2002: 218 sqq.). This function is not a characteristic of Hittite -kan. The most important meaning of OHG ge- is that of “vollständige Durchführung”: “daran, zusammen, hinzu, durcheinander, gemeinsam”(Schwartz 1986). As for the function of this particle it has been considered difficult to distinguish aspect and aktionsart (cf. Herbers 2002: 225 with reference to Obouzar (1984): “Phasenopposition”). If there was in fact an aspectual system in MHG it did in any case not exist later than 1400 A.D. My present view of the function of the sentence particles was stated in Josephson (1995). One important function of -san and -kan is modification of the actionality of the verb. The transformative function of -kan as indicating effectuated completion may be seen as coming close to being aspectual. The function of -san as signalling the crucial limit of an action also belongs, like that of -kan, to the dimension of transformativity but -san does not refer to an achieved term but rather to direction towards and ultimate arrival at the term. In addition to these functions with the verb the “sentence particles” -san and -kan serve as links between the actionality of the verb and the allative, ablative, dative or locative case or a local adverb. This use of the particles represents the formal expression of a kind of link that has been construed as a theoretically motivated category of case grammar (cf. Josephson 1995: 173). The notions of abstract directional clause, locative and directional relationship and non-directional and directional interpretation of verbal action as found in John M. Anderson (1971) are of relevance to the question (cf. Josephson 1995: 173 n. 16 ). Directional -san frequently occurs together with the allative case (as long as the allative case subsisted in Hittite) as I mentioned in Josephson (1981(82)), and with the locative that replaced it. It also occurs quite commonly with the dative. It is a significant fact that -kan is never found with the allative. It often occurs with a locative or an ablative case form. It is found in a regular manner with a local adverb/

Actionality and aspect in Hittite 

preverb/postposition such as arha ‘off, away’. It occurs with arha and an ablative as in the expression kez-kan INIM-za arha (from this (ABL)-kan reason (ABL) away) ‘except for this reason’ (cf. Melchert 1977: 185). When arha serves as a preverb with such verbs as arha ak- ‘die off ’ or arha wemiya- ‘find out’, where it has a directive function and indicates the final limit in accordance with its etymological sense of ‘to the boundary’, -kan is not used. On the other hand, para-kan pawar ‘going forward’ (pawar is a verbal noun) translates Akk. SĪTU ‘exit’ in a bilingual lexical entry. anda-kan pawar ‘going in’, ‘entering’ occurs in a lexical context and may interestingly be compared to Gothic innga-leiÞan (Josephson 1976: 172). -san is also found in lexical entries such as katta-san arnumar which means ‘bringing to a close’, or ‘termination’. The literal sense is ‘bringing down (katta) to a close (-san)’. Cf. also anda-san tiyawar ‘stepping inside’. luli-kan anda ‘in the pond’ is an example of the use of -kan with a locative and a following local adverb specifying the position. There is no specific connection with a verb. The combination katta-kan means ‘in a low position’ as in katta-kan kuit harzi ‘what she has below’. The opposition between -san and -kan with verbs and with local adverbs is obviously founded on the opposition between -san which may be combined with allative and dative case as against -kan which is appropriate with locative and ablative (cf. Josephson 1995: 167, 1997). -san can refer to GOAL, -kan is used with SOURCE. -kan, in its initially or finally limiting and transformative function, indicates absence of direction whereas -san, when used with non-stative verbs, indicates development directed towards an ultimate goal. With directional verbs as well as local adverbs -san points to the far end and ultimate fulfilment (accomplishment) of the action. -kan indicates initial as well as final limitation. When -kan indicates final limitation the action qualified by it is seen as achieved, the limit as reached and extension as annihilated. -san is however also used with verbs that do not imply any motion as with es- ‘to be’. This kind of use should be understood as the consequence of an original lative and perlative function of -san which is in keeping with the relation of the clitic with the allative case. (An original perlative function of the allative itself was discussed in Josephson and Carling [1998] 1999.) The fundamental function of -san seems to be adessive and its meaning of cohesiveness, adherence, togetherness and conjunction is plainly that of its IE etymon *sem/som and of Vedic sam-. -kan, on the other hand, has an inessive and illative dimension. An initially limiting function of -kan is obvious with the stative verb as(s)- ‘remain’ in the sense of ‘be left over’, with which it is almost always present (unless it is replaced by -(a)sta when the meaning is that of ‘comes to (or: will) remain’). Latin consto ‘remain’ is a formation showing con- with the same kind of function. In a few cases when the verb refers to remaining as being in one’s place or staying on at a location the focus is changed, there is less emphasis on the state as established and -san is used instead of -kan. A finally limiting completive function of -kan similar to that of con- in Latin conficio is obvious with the verb kuen- ‘kill’, which shows -kan in a regular manner, also when the verb has the -ske/a-suffix: kuwask- ‘kill regularly ’ or the -nne/a- suffix: kuenna-.

 Folke Josephson

In Russian studies, the notion of lexical aspect has come to the fore in recent years as I mentioned (1995: 167) with reference to studies by Klimonow (1996) and Mehlig (conference paper 1996)4. Johanson (2000) and Bertinetto and Delfitto (2000) emphasized the aktionsart character of Russian aspect. Quite recently an aktionsart character of Russian aspect was stressed by Tatevosov (2003: 881) in a study of the Russian delimitative po-. po- fails to quantize the predicate, which is also true for the prefix na(Tatevosov 2003: 885 with reference to a study by Filip 2000). It is interesting to note that Old Hittite possesses a clitic -(a)pa, which sometimes has the totalizing function of an exhaustive transformative (consumative) character of the kind described by Johanson (2000: 69) as in s-an-ap atanzi ‘they eat him (up)’ (III 60 II 5) or s-an-za-pa assu suw[a.... ‘fill him up with good’ (KBo XII 18, 6). Cf. the use of Russian po-: “Verbs containing completive po- just refer to situations that attain a resultant state without any further qualificational or quantificational refinements” (Tatevosov 2003: 876). po-verbs may also be gradual completion verbs such as pokrasnet’ ‘redden’ (Tatevosov 2003: 877). The clitic -san in its aktionsart function is similar to the Russian preverb na-, which is of adessive and allative meaning (Josephson 1995: 168). I referred to Russell (1985) who studied the aspectual properties of this prefix as well as its semantics and the lexical aspect of verbs prefixed by na-. I also mentioned in that article that there are some common properties of Hittite -(a)sta and Russian za-, which was analyzed by Janda (1985) as a trajector which may indicate deflection and fixation in the extradomain. The latter of these meanings leads to a meaning of change of state. An interesting approach is the cognitive model of Rosch, who sees time as a subcategory of cognitive space. It has important consequences for the problem of actionality in Anatolian languages and shows that an investigation of local and deictic-directional categories will have to precede final pronouncements about aspect. We have seen that the dimension of terminality and the expression thereof by a system of + / – directional enclitic particles, sometimes in conjunction with directional adverbs/preverbs, plays a great role in Hittite. The most important of the particles involved are -san which has an adessive and allative-directional meaning, which can be compared with that of the Russian prefix na- and the punctual and limiting -kan (originally inessive and illative) which may be initially limiting and refer to the initial boundary of the action. From Middle Hittite on they are frequently combined with directional adverbs. With -kan an event is seen as +completed. With -san this is not the case. Both particles may be called aspectual markers only if aspect is defined in a way to cover both these functions, as is common in Slavic studies. Recent trends in the study of Slavic aspect tend to cast doubt on the exclusive appropriateness of this kind of terminology. Weighty reasons have been adduced for reintroducing lexical aspect or aktionsart into the discussion of the complex Slavic perfectivity. Gehrke (2005) argues that Slavic verb prefixes are resultative and that quantification plays an important role. 4. The paper presented by Mehlig in 1995 was entitled “Verbalaspekt und Phasenaktionsarten im Russischen”. Important parts of its content were presented in Mehlig (1996).

Actionality and aspect in Hittite 

Cf. recent publications by Arsenijevic (2007a, 2007b). Arsenijevic (2007b) makes a unified analysis of internal and external Slavic verb-prefixes which explains similarities between the two classes of prefixes “which are expected if they are generated in the same template: (perfectivity is then just a particular realization of resultativity and telicity)”. The imperfective suffix in Slavic languages corresponds to the English progressive (cf. Borer 2005). An application of progressive structure covers the phenomenon of imperfectivity in the Slavic languages (Arsenijevic 2007a). According to Arsenijevic the meaning of the Slavic imperfective suffix is slightly more general than that of the progressive. An establishment of a Slavic type of aspectual system which was effectuated through the interplay between preverbal terminal force with +INTRAHF actionality expressed by suffixation did not take place in Hittite though structures which could have allowed a similar development were present in the language as in Latin (cf. Haverling 2000). In Hittite, the enclitic “sentence particles” which were able to serve as a link between actionality and case became increasingly involved in the VP and in verb semantics. During the same period in the history of the language adverbs started to develop into preverbs. (Cf. Boley 1993: Summary.) This fact created a complicated situation which had serious ulterior consequences for the stability of the system of directional clitics. -kan remained in late Hittite, largely as a syntactical marker.

4. Conclusions In my study of the functions of the reflexive particle in Hittite (Josephson 2003), which was carried out within the theoretical framework of Nedjalkov’s classification of reflexives, interesting valency-changing functions of the Hittite reflexives were discovered. A completive import of the reflexive particle in Hittite was seen to have a typological parallel in Spanish and Italian, in regard to which several linguists have studied the situation. Telicity and inner aspect are involved as in Italian. An aspect-like use of the tenses is present in Hittite narrativity and can be compared typologically with similar use in a variety of languages. There seems to be arguments for connecting tense-switching with aspect. A similar use of tenses is found in Russian narrative texts, in old French and in many other languages. Studies such as those carried out by Johanson, to which I referred when discussing the atelic -ske/a-verbs show a complicated theoretical situation. The categories proposed by Johanson provided us with a tool for examining the possibility of understanding the -ske/a- verb as an aspectual item. This verb is a +INTRA item. We stayed at discussing aktionsart for reasons connected with the functions of the suffixed verb as compared with the+INTRAHF function of the suffix -anne/a- and because it was not possible to establish that there was a full-scale grammaticalisation. Recent discoveries in Russian aspectual studies do not make it obvious that outer aspect is involved. In any case, a clear binary system of the Russian kind is lacking in Hittite.

 Folke Josephson

We observed in passing that the use of the -ske/o- form in Ionic Greek has something to do with narrative technique and that there are similarities with the Hittite use of the similar iterative-durative form. The study of the Latin sco-verbs and the semantic role of prefixes as carried out by Haverling (2000) and the treatment by Brachet (2000) of adverbial particles such as ex- and the directive particle de-, which became preverbs and finitransformative phase structure markers of the kind described by Johanson (2000: 68–70) reveal two systems in a certain interaction which never reached the stage of grammaticalised imperfective/perfective aspect. The function of Hittite -kan is important for understanding traits of actionality in Indo-European languages such as Gothic, MHG and Latin, which have related preverbs. The function of Hittite -san, which is similar to that of Old Indic sam- and Greek sun-, is also relevant in that context as are other directional particles of Hittite. Establishment of the Russian type of aspectuality which arose from the interplay of preverbial force with +INTRA actionality did not take place in Hittite though fundamental elements which could have allowed a similar development were present in that language: suffixation of a +INTRA character, local adverbs in the process of developing into preverbs and, more significantly, in a rather early phase of Anatolian, a system of directional clitics which became indicative of verbal actionality. There is much about the Anatolian “sentence particles” that may serve as comparative material for the establishment and the synchronic function of Slavic aspectuality. Modern trends in Slavic aspectology cast some doubt on the appropriateness of the traditional terminology and reasons have been adduced for introducing lexical (inner) aspect as an important element in the analysis of Russian perfectivity. Establishment of a Russian type of aspectual “pair” system which arouse by an interplay between telicity and quantification as expressed by preverbs, and +INTRAHF actionality expressed by suffixation did not take place in Hittite though structures that could have allowed a development of this kind were present as they were in Latin with its preverb con- and suffix -sco. Hittite -kan was involved in the VP and was important for verb semantics and aspectuality. There was a certain tendency for Hittite local adverbs to develop into preverbs but they did not play the entire role that preverbs played in other IE languages such as Latin and Slavic. The roles of -kan and other clitics are therefore of fundamental importance for a general understanding of Hittite aspect. More research will have to be done on the means of expressing telicity and quantification before a final judgement on the extent of such a “slightly more general” meaning of the -ske/a-suffix that Arsenijevic considers possible for the Slavic imperfective suffix can be pronounced.

Actionality and aspect in Hittite 

References Anderson, J. M. 1971. The grammar of case. Towards a localistic theory. Cambridge: CUP. Arsenijevic, B. 2007a. Slavic verb prefixes are resultative. Cahiers Chronos 17: 197–213 Arsenijevic, B. 2007b. A unified analysis of Slavic verb-prefixes. In S. Blaho. L. Vicente & E. Schoorlemmer (eds.), Proceedings of ConSOLE XIV, 21–36. Leiden: University of Leiden. Bechtel, G. 1936. Hittite verbs in -sk-. Ann Arbor MI: Edwards Brothers. Bertinetto, P. M. & Delfitto, D. 2000. Aspect vs. actionality: Why they should be kept apart. In Ö. Dahl, (ed.) Tense and Aspect in the Languages of Europe, 189–225. Berlin: Mouton de Gruyter. Boley, J. 1993. The Hittite Particle -z/-za [Innsbrucker Beiträge zur Sprachwissenschaft]. Innsbruck:University of Innsbruck. Borer, H. 2005. Structuring sense. Vol. I: In name only. Oxford: OUP. Brachet, J.-P. 2000. Recherches sur les préverbes de- et ex- du latin. Bruxelles: Collection Latomus. Chvany, C. V. 1990. Verbal aspect, discourse saliency, and the so-called “Perfect of result” in modern Russian. In N.B. Thelin (ed.), Verbal Aspect in Discourse, 213–235. Amsterdam: John Benjamins. D’Alessandro, R. Forthcoming. Inner aspect and verb-object agreement in impersonal si constructions. In: S. Manninen S., K. Hietaam, E. Keiser & V. Vihman (eds.), Passives and Impersonals in European Languages. Amsterdam: John Benjamins. Daues, A. 2004. Funktionen der Verbalaffixe im Hethitischen. Anmerkungen zu den -ske-Formen im Hethitischen und Homerischen. Handout at the XII. Fachtagung der Indogermanischen Gesellschaft. Cracow. Dimitrova-Vulchanova, M. 1999. Verb Semantics, Diathesis and Aspect [Lincom Studies in Theoretical Linguistics]. München: Lincom. Dowty, D.R. 1977. Towards a semantic analysis of verb aspect and the English “imperfective” progressive. Linguistics and Philosophy 1: 45–77. Dressler, W. 1968. Studien zur verbalen Pluralität. Wien: Österreichische Akademie der Wissenschaften. Filip, H. 2000. The quantization puzzle. In J. Pustejovsky & C. Tenny (eds.), Events as Grammatical Objects. From the combined perspectives of lexical semantics, logical semantics and syntax, 3–60. Stanford CA: CSLI. Filip, H. & Rothstein, S. 2006. Telicity as a semantic parameter. Annual Workshop on Formal Approaches to Slavic Linguistics. The Princeton Meeting 2005, 139–156. Ann Arbor MI: Michigan Slavic Publications. Fleischman, S. 1985. Discourse functions of tense-aspect oppositions in narratives: Toward a theory of grounding. Linguistics 23: 851–882. Garrett, A. 1996. Wackernagel’s law and unaccusativity in Hittite. In A.L. Halpern & A.M. Zwicki (eds.), Approaching Second, 85–133. Stanford CA: CSLI. Gehrke, B. 2005, The prepositional aspect of Slavic prefixes and the goal-source asymmetry. In J. Zwarts & H. de Hoop (eds.), Proceedings of the ESSLLI Workshop on Formal Semantics and Cross-linguistic Data, 47–56. Edinburgh: Heriot Watts University. Goldberg, A. 1995.Constructions. A construction grammar approach to argument structure. Chicago IL: Chicago University Press. Harðarson, J. A. 1998. Mit dem Suffix *-eh1 bzw. *-eh1-ie/o-gebildete Verbalstämme im Indogermanischen. In W. Meid (ed.), Sprache und Kultur der Indogermanen. Akten der X.

 Folke Josephson Fachtagung der Indogermanischen Gesellschaft Innsbruck, 22–28. September 1996, 323–339. Innsbruck: Institut für Sprachwissenschaft. Haverling, G. 2000. On Sco-verbs, Prefixes and Semantic Functions. A study in the development of prefixed and unprefixed verbs from early to Late Latin [Studia Graeca et Latina Gothoburgiensia LXIV]. Göteborg: Acta Universitatis Gothoburgensis. Herbers, B. 2002. Verbale Präfigierung im Mittelhochdeutschen. Eine semantisch-funktionale Kor pusanalyse. Tübingen: Niemeyer. Hoffner, H. & Melchert, C. 2002. A practical approach to verbal aspect in Hittite. In S. di Martino & F. Pecchioli Daddi (eds.), Anatolia Antica, Studi in memoria di Fiorella Imparati, 377–391. Firenze: LoGisma. Janda, L. A. 1985. The meanings of Russian verbal prefixes. Semantics and grammar. In M.S. Flier & A. Timberlake (eds.), The Scope of Slavic Aspect, 26–40. Columbus OH: Slavica Publishers. Jacquesson, F. 2001. Review of Z. Frajzyngier & T.C. Curl (eds.), Reflexives, Forms and Functions. BSL XCVI(2): 83–108. Johanson, L. 1995. Mehrdeutigkeit in der türkischen Verbalkomposition. In M. Erdahl & S. Tezan (eds.), Beläk Bitig. Sprachstudien für Gerhard Doerfer zum 75. Geburtstag [Turcologica 23], 81–101.Wiesbaden: Harrassowitz. Johanson, L. 2000. Viewpoint operators in European Languages. In Ö. Dahl (ed.), Tense and Aspect in the Languages of Europe, 27–187. Berlin: Mouton de Gruyter. Josephson, F. 1972. The Function of the Sentence Particles in Old and Middle Hittite [Acta Universitatis Upsaliensis. Studia Indoeuropaea Upsaliensia 2]. Uppsala: Skriv Service. Josephson, F. 1976. On the function of the Gothic preverb ga-. IF 81: 152–75. Josephson, F. 1981(82). Review of Frank Starke: Die Funktionen der dimensionalen Kasus und Adverbien im Althethitischen. Kratylos 36: 94–105. Josephson, F. 1995. Directionality in Hittite. In W. Smoczyński (ed.), Kuryłowicz Memorial Volume. Part One, 165–176. Cracow: Universitas. Josephson, F. 1997. Review of J. Boley: The Sentence Particles and the Place Words in Old and Middle Hittite. Bibliotheca Orientalis LIV 1(2): 155–160. Josephson, F. & Carling, G. [1998]1999. PERLATIVE and INSTRUMENTAL in Sanskrit and Indo-European. In: Josephson, F. (ed.), Categorisation and Interpretation. Indological and comparative studies from an international Indological meeting at the Department of Comparative Philology, Göteborg University [Meijerbergs arkiv för svensk ordforskning 24], 77–105. Göteborg: Meijerbergs institut för svensk etymologisk forskning. Josephson, F. 2003. The Hittite reflexive construction in a typological perspective. In B.L.M. Bauer & G.-J. Pinault (eds.), Language in Time and Space. A festschrift for Werner Winter on the occasion of his 80th Birthday, 221–232. Berlin: Mouton de Gruyter. Josephson, F. 2006. Début d’un emploi créatif de la langue Hittite. In G.-J. Pinault & D. Petit (eds.), La Langue poétique indo-européenne. Actes du colloque de travail de la Société des Études Indo-européennes (Indogermanische Gesellschaft/Society for Indo-European Studies), Paris, 22–24 octobre 2003 [Collection linguistique publiée par la Société de linguistique de Paris 91], 141–156. Leuven: Peeters. Klavans, J. L. 1985. The independence of syntax and phonology in cliticization. Language 61: 5–120. Klimonow, W. 1996. Zur Interaktion von grammatischen und Lexikalischen Aspekten. Folia Linguistica XXX(1–2): 73–86.

Actionality and aspect in Hittite  Kratzer, A. 2004. Telicity and the meaning of objective case. In J. Guéron & J. Lecarme (eds.), The Syntax of Time, 389–423. Cambridge MA: The MIT Press. Kuryłowicz, J. 1964. The Inflectional Categories of Indo-European. Heidelberg: Winter. Laroche, E. 1961. Notes de linguistique Anatolienne. Revue hittite et asianique XIX(68): 25–37. Legendre, G. 2000. Positioning Romanian verbal clitics at PF. An optimality-theoretic analysis. In B. Gerlach & J. Grijzenhout (eds.), Clitics in Phonology, Morphology and Syntax, 219–254. Amsterdam: John Benjamins. Maldonado, R. 2000. Conceptual distance and transitivity increase in Spanish reflexives. In Z. Frajzyngier & T.C. Curl (eds.), Reflexives, Forms and Functions [Typological Studies in Language 40], 153–185. Amsterdam: John Benjamins. Mehlig, H.R. 1996. Some analogies between the morphology of nouns and the morphology of aspect in Russian. Folia Linguistica XXX(1–2): 87–109. Melchert, H.C. 1977. Ablative and Instrumental in Hittite. PhD dissertation, Harvard University. Melchert, H.C. 1997. Traces of a PIE aspectual contrast in Anatolian. Incontri Linguistici 20: 83–92. Melchert, H.C. 1998. Aspects of verbal aspect in Hittite. In S Alp & A. Süel (eds.), Acts of the Third International Congress of Hittitology, Çorum, September 16–22,1996. Ankara: Uyum Ajans. Nishida, C. 1994. The Spanish reflexive clitic se as an aspectual class marker, 413–418. Linguistics 32: 425–58. Nørgård-Sørensen, J. 1997. Tense, aspect and verbal derivation in the language of the Novgorod Birch Bark Letters. Russian Linguistics 21: 1–21. Peppler, C.W. 1933. Durative and aoristic. American Journal of Philology 54: 47–53. Progovac, L. 2002. Aspect phrases, perfective prefixes and congruent prepositional phrases in Serbian. Formal Approaches to Slavic Linguistics 10: 201–221. Ann Arbor MI: Michigan Slavic Publications. Puhvel, J. 1991. Homer and Hittite. Part 2: The East Ionic and Hittite iteratives [Innsbrucker Beiträge zur Sprachwissenschaft, Vorträge und Kleinere Schriften 47], 13–20. Innsbruck: University of Innsbruck. Russell, P. 1985. Aspectual properties of the Russian verbal prefix na-. In M.S. Flier & A. Timberlake (eds.), The Scope of Slavic Aspect, Columbus OH: Slavica Publishers: 59–75. Sanz, M. & Laka, I. 2002. Oraciones transitivas con se: El modo de acción en la sintaxis. In: Sánchez-López, C. (ed.), Las construcciones con se. Madrid: Visor Libros: 309–338. Schwartz, H. 1986. Präfixbildungen im deutschen Abrogans. Analyse und Systematik. Göppingen: Kümmerle. Solms, H.-J. 1991. Zur Wortbildung der Verben in Hartmann von Aues >Ivein< (Hs.B) und >Gregorius< (Hs.A): Das Präfix ge- im System der verbalen Präfigierung. Zugleich ein Beitrag zur Diskussion historischer Wortbildung. Zeitschrift für deutsche Philologie 110. Sonderheft: 110–140. Tatevosov, S. 2003. A theory of Slavic aspect and the Russian delimitative. In P. Kosta et al. (eds.), Investigations into Formal Slavic Linguistics II [Contributions of the Fourth European Conference on Formal Description of Slavic Languages], 873–891. Frankfurt: Peter Lang. Watkins, C. 2001. An Indo-European linguistic area and its characteristics: Ancient Anatolia. In A.Y. Aikhenwald & R.M.W. Dixon (eds.), Areal Diffusion as a Challenge to the Comparative Method? Areal diffusion and genetic inheritance, 44–63. Oxford: OUP. Zagona, K. 1996. Compositionality of aspect: Evidence from Spanish aspectual se. In C. Parodi C. et al. (eds.), Aspects of Romance Linguistics. Selected Papers from the Linguistic Symposium on Romance Languages XXIV. March 10–13, 1994, 475–488.Washington DC: George town University Press.

Imperfectivity and complete events Atle Grønn Imperfectivity is cross-linguistically associated with the subinterval property and a modal component induced by the famous ‘imperfective paradox’. These properties arguably hold for both the progressive and habitual-iterative readings. However, both in Romance and Slavic, the imperfective may also refer to complete events instantiated in the world of evaluation: the so-called Imparfait narratif in French and the Factual Imperfective in Russian. I propose an analysis of viewpoint aspect in terms of temporal inclusion relations between the event time and the assertion time. Importantly, however, the source of the two complete event readings in question are quite different inasmuch as the Russian imperfective is unmarked and is used whenever the marked perfective aspect is inappropriate, while the French Imparfait is marked. This means that the French Imparfait retains its meaning of contemporaneity even when it has a complete event interpretation.

1. Introduction1 The category of viewpoint aspect (Smith 1997) – that is, the opposition between perfective (PF) and imperfective (IPF) verb forms – plays an important role in language families such as Slavic and Romance. Imperfectivity is cross-linguistically associated with the subinterval property (atelicity), which is characteristic of both the progressive and habitual-iterative readings. If a sentence, whose main verbal predicate is marked with IPF, is true of a certain temporal interval, then the same sentence is also true of subparts of this interval. However, both in Slavic and Romance, IPF may also refer to complete events in which case this inference does not go through. If a complete event of the type of the VP holds at a time t, it is not necessarily true that a complete event of the same type also holds at an arbitrary subinterval t’ of t. What are the implications of these puzzling complete event readings for the universal notion of imperfectivity? 1. I would like to thank Philippe Schlenker, Nikolaj Vazov and the audiences of Gothenburg 2004 and Kjell Johan Sæbø’s circle of theoretical linguistics (Oslo 2005) for valuable comments at various stages of this work.

 Atle Grønn

In this paper I propose an analysis couched in event semantics of the Factual Imperfective (obshchefakticheskoe znachenie) in Russian and Imparfait narratif in French which shows that the two complete event readings in question are quite idiosyncratic. A more fine-grained analysis reveals their different semantic properties, explaining why these two complete event readings are not attested in the same range of contexts. The point is that these specific interpretations of the imperfective viewpoint arise in competition with other tense-aspect grams in the verbal system, and the role of IPF in Russian is very different from IPF in French, inasmuch as only the latter is marked and therefore expected to get a uniform interpretation.

2. The data The most important subsets of the complete event readings of IPF in Russian are labelled ‘existential IPF’ (ex. 1) and ‘presuppositional IPF’ (ex. 2) in (Grønn 2004). (1) Vanja chital Vojnu i mir. Vanja read.IPF.PAST War and peace Vanja has read War and Peace. (2) A: Krasivo ukrasili elku. beautifully decorate.PF.PAST.PL spruce.ACC A: They decorated the Christmas tree beautifully. B: Kto ukrashal? (from Rassudova) Who.NOM decorate.IPF.PAST B: Who decorated it? A semantic representation of example (1) should encode the speaker’s assertion that there exists a complete past event of Vanja’s reading War and Peace. The imperfective verb in speaker B’s utterance in the discourse of (2) also conveys the fact that there exists a complete past event of the type denoted by the VP. However, in this case the existence of the event is not asserted, but presupposed since it is already entailed in the input context through speaker A’s utterance containing the perfective verb ukrasili (‘decorated’). In the dynamic two-dimensional representational format of Discourse Representation Theory (Kamp 2001), the difference between (1) and (2) is primarily captured by declaring the verb’s event argument in the assertoric and presuppositional component, respectively. The complete event interpretation of IPF in French is often referred to as Imparfait narratif, reflecting the fact that it typically occurs in contexts of narrative progression, as witnessed by examples such as (3) below:

Imperfectivity and complete events 

(3) A huit heures, les voleurs entraientimparfait dans la banque, ils discutaientimparfait avec un employé, puis se dirigeaientimparfait vers le guichet principal. (from Jayez 1999) At eight, the robbers entered the bank. They discussed with a clerk and then they moved towards the main desk. This usage of telic VPs with the imperfective viewpoint (entraient – ‘entered’) is completely ruled out in Russian, where the factual IPF can never move the narration forward. In Russian, PF occurs happily in contexts of narrative progression and the factual IPF is effectively blocked in this environment. Indeed, the complete event reading of the factual IPF is compatible with event anaphora, as we saw in example (2) above, but IPF in Russian cannot be used to assert the existence of a complete event at a punctual or contextually given interval, as in the French case above. Hence, phenomena such as temporal anchoring and temporal anaphora interact quite differently with the imperfective viewpoint in French compared to Russian. Furthermore, we note that Imparfait narratif combines with Vendlerian achievements, such as entrer, while the factual IPF applies to accomplishments like chitat’ Vojnu i mir or ukrashat’ elku. This distinction in the telic properties of the VPs being input to the imperfective operators in the two languages under discussion is not merely a coincidence, as we will see in section 5.

3. Diachronic pressure towards a binary opposition The tense-aspect systems of Russian and French have developed quite differently from a common Indo-European source. While French partly retains the original inflectional category Imperfect:Aorist, Russian, like other Slavic languages, is now equipped with a more derivational like aspectual category Perfective:Imperfective (See Bermel (1997) for a diachronic analysis of the emergence Russian aspect). Interestingly, both contemporary French and Russian have the formal means to express a complete/incomplete event opposition, and, furthermore, both French and Russian will eventually end up with a pure binary opposition between two members of the aspectual category in past tense, cf. table 1 and 2. Table 1. First round. Inventory of past tense-aspect forms of the verb ‘to read’ (1sg.) Written/earlier French

Old Church Slavonic

Imparfait (lisais) Passé simple (lus) Passé composé (ai lu)

Imperfect (chitaakh’’) Aorist (chitakh’’) Perfect (jesm’ chital’’)

 Atle Grønn

Table 2. Second round. Inventory of past tense-aspect forms of the verb ‘to read’ (1sg.) Spoken/modern French

Russian

Imparfait (lisais) Passé composé (ai lu)

Imperfective (chital) Perfective (prochital)

The emerging simplified picture of a binary opposition in table 2 is not witnessed in some other Slavic and Romance languages, e.g. Bulgarian and Portuguese, which display a greater variety of tense-aspect grams, and therefore also a somewhat different set of usages of the imperfective viewpoint. The hypothesis to be explored here is that the more or less idiosyncratic complete event readings of IPF in Russian and French should be understood in light of the competition between two forms in a closed system. I will claim that the ways in which the aspectual opposition is morpho-syntactically expressed in table 2 largely determine the use of imperfectivity with complete event readings in the languages under consideration. This implies that the analysis presented here does not straightforwardly extend to cover imperfective complete event interpretations in, say, Bulgarian, which does not reduce to a clear-cut aspectual opposition, but where a derivational (Slavic) system coexists with an inflectional (Romance) system and furthermore interacts with overt manifestations of perfect tenses.

4. Key ingredients of the analysis 4.1

Markedness

The analysis to be presented restores some traditional insights, notably concerning the rather problematic notion of markedness. The Jakobsonian idea of applying markedness theory to grammatical categories such as aspect has a long history in Slavic linguistics, but it has lately lost in popularity. However, with the fast-growing framework of Bidirectional Optimality Theory, that is OT from both the hearer (interpretation) and speaker (production) perspective (Blutner 2000), we finally have the tools to implement markedness theory in a formal analysis. Importantly, Bidirectional OT is primarily a diachronic theory, which can explain why certain form-meaning pairs are optimal in an evolutionary setting both for the hearer and speaker. It is often unreasonable to expect the discourse participants to calculate the optimal form-meaning pairs in online communication (synchronically), since this mental process can be utterly complex (the speaker has to consider a set of forms; the hearer has to consider a set of interpretations for each form; and both have to consider a possibly large number of form-meaning pairs in order to arrive at the optimal outcome). However, speakers and hearers learn from (mis-)communication

Imperfectivity and complete events 

and eventually, hence diachronically, agree on the optimal form-meaning combinations.2 Furthermore, markedness theory can serve as a guide for the semanticist: Since it is well-established that PF is marked in Russian – viewpoint aspect in Russian is primarily a grammaticalisation of telicity expressed through prefixation – the analysis should start with an invariant definition of perfectivity, and the range of usages of the unmarked IPF should follow from competition between the marked and the unmarked forms. In French, the picture is the opposite inasmuch as the perfective like aorist (Passé simple) is historically the unmarked form, cf. (Breu 1998). According to the traditional view of French aspectologists like Guillaume, Passé simple stands outside the complete/incomplete event opposition (Jayez 1999). I will therefore propose an explicit semantic interpretation of Imparfait, which is expected to have an invariant meaning due to its markedness. Note that in contemporary spoken French the marked member of the original aspectual opposition has proven more viable than the unmarked. The presence of IPF and the morphologically orthogonal composite perfect has made the aorist (Passé simple) superfluous in the following way: Whenever Passé simple was used historically with atelic predicates it can easily be replaced by IPF, as in (4) below: (4) Jean travaillapassé simple beaucoup. ⇒ Jean travaillaitimparfait beaucoup. John worked a lot. In many cases of Passé simple being applied to telic predicates with a complete event interpretation, Passé composé enters the scene, since perfect tenses by default have a complete event interpretation with telic predicates, cf. (Bohnemeyer and Swift 2004). However, the possibility of Imparfait narratif in certain aoristic contexts blurs the straightforward correlation between Passé simple and Passé composé, since it shows that also Imparfait may, in the right context, function on the traditional domain of Passé simple.3 In Russian, like in French, we have a neat aspectual opposition in past tense. However, in Russian the unmarked form is in no danger of disappearing. The imperfective

2. See Grønn (2008) for a bidirectional OT approach to Russian aspect, and Grønn (2007) for a purely diachronic view on the emergence of Russian aspect. 3. Breu (1998), like the present paper, compares the tense-aspect systems of Slavic and Romance from the perspective of markedness theory. However, in view of Imparfait narratif, Breu is wrong when he claims that in a translation of viewpoint aspects from Romance to Slavic (or the other way) the marked aspect of the source language, say French, always corresponds to the same unmarked form in the target language, say Russian. Breu thus predicts that the marked Imparfait translates into the unmarked IPF in Russian. This cannot be correct since the marked Imparfait narratif in French always corresponds to the marked PF in Russian. On the other hand, Breu is right when he points out that the unmarked form of the source language does not tell us which form to choose in the target language (Breu 1998: 95).

 Atle Grønn

viewpoint is simply necessary to maintain the aspectual opposition, since no other aspectually loaded grams are available in the grammar. The fact that morphologically composite perfect tenses disappeared in Russian was related to a more general loss of the copula, a process which was orthogonal to the emergence of the new PF:IPF dichotomy. As shown in (Grønn 2004), the functions of perfect tenses may still be rendered in contemporary Russian by simple past forms of perfective and imperfective verbs. In the case of present perfect readings, this is rather trivial since the grams present perfect and simple past are truth-conditionally equivalent (with telic predicates). The difference is merely that a simple past makes use of the utterance time as a default evaluation time, while a present perfect explicitly designates the utterance time as the time for which the resultant state of the event holds.4

4.2

Viewpoint aspect in Russian as temporal relations

Following contemporary approaches to viewpoint aspect – in the spirit of (Klein 1995) – I treat aspect as a temporal phenomenon which interacts with tense through Reichenbachian temporal parameters. The complete event reading of PF in Russian can be captured by the inclusion relation e ⊂ t, saying that the run time of the event e is properly included in the assertion time t. The latter notion refers to the time interval on which the speaker focuses, an interval which is typically denoted by a temporal adverbial, as in (5): (5) Chasov v shest’ vechera poobedali. Hours.GEN at six evening.GEN dine.PF.PAST.PLUR We had dinner around six p.m.

(Internet)

A semantic representation for this sentence should contain a condition like e ⊂ t, where e represents the event of eating dinner, while t equals the frame time of approximately one hour denoted by the adverbial. In absence of temporal adverbials or context times, the assertion time parameter is merely restricted by past tense, and then equals ‘the whole past preceding the current evaluation time’.5 Note that the Reichenbachian reference time parameter is split into the assertion time and the evaluation time, and it is thereby given a more empirical foundation inasmuch as the assertion time gets its value from overtly expressed temporal adverbials and tenses.

4. The case of covert past perfect readings in Russian, however, is much more complicated. See Grønn (2004) for a thorough discussion of relative past tense (for instance in complements of verba dicendi). 5. The evaluation time parameter will here be represented by the distinguished variable t0, which equals the utterance time whenever it occurs free in the semantic derivation, cf. von Stechow (1995).

Imperfectivity and complete events 

The progressive reading of IPF can be rendered through the opposite inclusion relation, that is, t ⊂ e. Hence, we have the tools to capture the fundamental aspectual opposition between (5) and (6): (6) My obedali we-NOM dine.IPF.PAST kogda u moego druga proizoshel pristup. (Internet) when with my friend.GEN happen.PF.PAST attack.NOM We were having dinner, when my friend had a heart attack. In example (6), the assertion time t, which corresponds to the time denoted by the temporal adverbial (the kogda/when-clause), is properly included in the time of the event of having dinner. In other words, the speaker merely focuses on a proper subpart of the eating event, and the aspectual configuration should only license an inference to an incomplete event of having dinner. For all we know, the dinner event may have been interrupted in the real world (the imperfective paradox). In virtue of being unmarked, IPF in Russian may also refer to complete events, as in example (7): (7) Ty segodnja obedal v restorane! (Internet) You.NOM today dine.IPF.PAST in restaurant.LOC You had dinner in a restaurant today! In (7), we get an ‘existential IPF’ interpretation, that is, the speaker asserts the existence of a complete event of having dinner within the proper subinterval of segodnja (‘today’) which precedes the utterance time. Thus, the unmarked imperfective is compatible with totally opposite inclusion relations, and it seems impossible to find an invariant meaning of this viewpoint. This is in fact also what markedness theory suggests: There is no such thing as the meaning of IPF in Russian since the imperfective viewpoint merely amounts to the non-use of PF. An imperfective sentence such as (8) below can therefore be considered as underspecified without further context, but will in practice (i.e. given a specific context) get one out of three interpretations, as indicated in the English translations: (8)

Ja obedal. I.NOM dine.IPF.PAST I was having dinner. (at some particular time) I used to have dinner. (regularly) I have had dinner. (today)

 Atle Grønn

In the semantics proper, what we should retain from the discussion above is the invariant analysis accorded to PF, which can be spelled out as follows in a compositional version of Discourse Representation Theory: (9) Perfective aspect in Russian ⇒ λPλt [e| P(e), e ⊂ t] The perfective operator takes a set of events P (a tense- and aspectless VP) as input and returns a set of assertion times t as output. The aspectual operator contributes to the declaration (existential quantification) of the event variable e of type P and further asserts that e is properly included in the assertion time t. Later in the derivation, the value of t will be provided by the ‘tense branch’. Hence, tense is assumed to have scope over viewpoint aspect at the syntax-semantics interface. In section 6, I will return to the issue of how the Russian IPF, given the right context, can receive a complete event interpretation in competition with PF.

5. A semantics for IPF in French I will now turn to the French aspectual system, which requires an explicit semantics for the marked form Imparfait.

5.1

An anaphoric/presuppositional past tense

As argued in the previous section, aspect is a temporal phenomenon, which interacts closely with tense. However, the syntax-semantics interface appears to be somewhat different in Russian and French. In Russian, viewpoint aspect (e.g. the perfective prefix po-) is assumed to be located below tense (e.g. the past tense morpheme -l) in the syntactic input to the semantic interpretation. If we respect the surface morpho-syntax of French, we have to treat Imparfait (the suffix -ai- in ‘lisait’) as a marker of both aspect and past tense. This implies that the logical type of Imparfait in French must be different from that of aspectual operators in Russian: While an aspectual operator in Russian is a function from predicates of events to predicates of times, the French Imparfait is a function from predicates/sets of events to (dynamic) truth values. Thus, in the compositional semantics, when Imparfait applies to the VP (an aspect- and tenseless sentence radical), we end up with an expression which contains information pertaining both to tense and aspect. Nevertheless, a sentence with the French Imparfait is still in need of further temporal anchoring in the context: (10) ?? Marie ouvraitimparfait la porte. (Jayez 1999) Mary was opening/opened the door.

Imperfectivity and complete events 

I will argue that examples like (10) are infelicitous out of context because of a temporal presupposition of IPF in French. The Imparfait is not a self-sufficient past tense, but requires anchoring to a salient past time. This past time interval can be provided anaphorically (intersententially) or through the presence of temporal adverbials (intrasententially). This observation points to a significant difference between Imparfait in French and a past IPF in Russian. In this respect, there is a connection between narrative progression, characteristic of Imparfait narratif but unavailable for the Russian Factual IPF, and anaphoricity: The French IPF picks up a contextually salient time provided by the preceding event and moves the narration forward. At the same time, given our uniform treatment of Imparfait, this anaphoricity effect is not restricted to Imparfait narrative. As noted by Smith, there is a link between anaphoricity and imperfectivity tout court: More generally, sentences with imperfective viewpoints are often dependent on other information in the manner typical of anaphora [...] The partial information given by imperfectives, and the fact that they are frequently used for backgrounding, conspire to produce this effect. (Smith 1997: 90).

Smith certainly has the incomplete event reading (progressivity) in mind, but what is the exact semantic contribution of the French IPF? As mentioned above, the answer to this question must relate to different parameters of temporality since Imparfait is both a past tense and an aspect. Concerning the tense part, I propose a presuppositional treatment, such that the contribution of past tense amounts to presupposing that the assertion time is located prior to the current evaluation time: t < t0, cf. the approach to tense semantics in (Heim 1994). In our dynamic DRT-framework, the semantic representation in (11) below is split into an assertoric part and a presuppositional part (Kamp 2001). Accordingly, the condition relating to past tense belongs to the latter, which is marked with the label ‘Presupposition’.

5.2

The inclusion relation

More challenging is the question of which aspectual configuration should be accorded to Imparfait. Importantly, the imperfective viewpoint in French, as a marked verb form, should get a uniform interpretation, like PF in Russian. Given the characteristic atelic properties of imperfectivity, the configuration t ⊂ e (as in the Russian sentence (6) above) is a natural candidate, but, obviously, this proper inclusion relation cannot capture the complete event interpretation of Imparfait narratif. This simple empirical fact represents a serious problem for most previous uniform treatments in formal semantics of the French IPF. For instance, in the classical DRT-tradition, following the important work of (Kamp and Rohrer 1983), Imparfait is treated as a kind of stativiser, which only produces atelic interpretations. At the same time, the perfective inclusion relation e ⊂ t clearly runs afoul of the Hauptbedeutung of IPF, that is, the progressive. This leaves us with the configuration

 Atle Grønn

t ⊆ e, where t = e is the extreme case of Imparfait narratif, cf. similar considerations in (Jayez 1998). I therefore propose, in a compositional framework, the following uniform analysis of Imparfait, which solves the paradox of coexisting complete and incomplete event interpretations: (11) Imparfait in French (preliminary version) ⇒ Assertion: λP [e| Pw(e), t ⊆ e] Presupposition: [t| t < t0] The progressive reading follows straightforwardly from (11) since the configuration t ⊆ e allows for the assertion time to only focus on a proper subpart of the event. Note that the modality of the progressive and habitual-iterative readings – the notorious imperfective paradox – requires manipulation of the world index w, an issue which will not be dealt with here.6 Of course, the imperfective viewpoint in French is not a pure progressive marker (unlike the periphrastic construction être en train de), but it still encodes contemporaneity. Indeed, concerning Imparfait narratif, the representation in (11) captures an important intuition among French semanticists: The [assertion time] is viewed as spanning the whole event [...] but not beyond [...], whence the peculiar flavour of the narrative Imparfait: one follows completely some eventuality, then shifts to the next one, etc. (Jayez 1999)

This is exactly what we get if we assume that the complete event interpretation of IPF in French amounts to the aspectual configuration t = e.

5.3

The assertion time

This neo-Reichenbachian analysis puts considerable restrictions on the size of t in the representation of IPF, and it can therefore explain why Imparfait narratif is ruled out in the main clause of contexts such as (12), where Passé composé has to be used: (12) Après que Marie est arrivéepassé (??partaitimparfait). After Mary came, Paul left.

composé,

Paul est partipassé

composé

The temporal clause après que Marie est arrivée denotes an interval which spans from Mary’s arrival to an infinite future. Without any further restrictions, this interval is not a candidate for the value of t in the condition t ⊆ e. Such a big interval cannot possibly equal (or be included in) the temporal trace of the event argument. On the contrary, this temporal frame is ideally suited for entering the inclusion relation e ⊂ t of Passé composé. Perfect tenses, like Passé composé in French, prototypically combine with big, indefinite assertion times, which easily include a complete event of the type denoted by

6. See Ferreira (2005) for a recent treatment of imperfectivity and modality.

Imperfectivity and complete events 

the VP.7 Of course, Passé composé is also restricted to past tense contexts, but this should naturally follow from a proper treatment of the morpho-syntax- semantics interface, where the present tense auxiliary est further delimits the interval denoted by the “après-clause” by establishing a right boundary (= the utterance time) of the assertion time interval. Contrary to (10) and (12), an appropriate context for the narrative IPF is the following, where the value of t is clearly identifiable and of the right size: (13) Quelques minutes plus tard, Pierre et Bruno apparaissaientimparfait dans un panache de poussière, slalomant entre les croissants de dunes à plus de 100 km/h devant le vieillard hébété. (Internet) A couple of minutes later, Pierre and Bruno could be seen to emerge (were emerging) out of a cloud of dust, slaloming among the dunes at more than 100 km/h before a dazed old man. The overt temporal adverbial in (13) supersedes the need for a context time and invites the presupposition of past tense to be verified intrasententially, i.e. by the sentence itself. Thus, the variable t in the presupposition is unified with the interval denoted by the adverbial quelques minutes plus tard – ‘a couple of minutes later’. There is naturally a complementary distribution between the two sources of presupposition resolution for past tense. In the presence of a temporal adverbial which explicitly provides the value of the assertion time, there is in general no need to invoke covert context times.8 This being said, the adverbial quelques minutes plus tard in (13) displays an interesting mixture of intersentential and intrasentential presupposition verification with the anaphoric element plus tard (‘later’).

5.4

The divisibility constraint

There is still one piece missing: the question of why complete event readings of IPF in Romance predominantly occur with Vendlerian achievements, that is, predicates 7. This feature of the perfect is reminiscent of the kind of complete event interpretation of IPF in Russian which I refer to as the ‘existential Ipf ’. The Russian IPF thereby represents the opposite picture of what we observe with Imparfait narratif in French, inasmuch as the former typically combines with a big, indefinite assertion time within which the event in question is properly included, cf. the ‘life time effect’ of Vanja’s having read ‘War and Peace’ in example (1). This feature of the existential Ipf in Russian follows from competition with PF, which is characterised by the property of ‘definite temporal anchoring’. 8. Perfect tenses, like Passé composé, prototypically occur in contexts where no context times or overtly expressed frame times are present. In the temporal calculus developed in (Grønn 2004), the default ‘all time’ (represented by the distinguished variable t∝) is invoked for these cases. This interval is further restricted by past tense to produce the interval ‘all time before the current evaluation time’. Even if we allow for this default interpretation in the temporal system, it is not relevant for Imparfait, since conditions such as t∝ ⊆ e or t∝ < t0 do not make sense.

 Atle Grønn

which only consist of a culmination. For French, this correlation between punctual Aktionsarten and the narrative reading of IPF has recently been confirmed in the survey of (Hobæk Haff 2005). One could argue that in order to obtain the configuration t = e, the intervals of both t and e should both be conceptualised as punctual. But this is apparently not quite what is going on: French informants (and aspectologists) rather emphasise the opposite idea, viz. IPF zooms in on the whole event and produces a stretching effect which in a certain sense cancels the inherent punctuality of achievements. In order to capture the data and these judgments, I will argue that the imperfective operator in French comes with a sortal condition requiring the input VP to satisfy the property of divisibility (i.e. if an event e of type P holds at t, there exists a proper subpart t’ ⊂ t, such that an event e’ also of type P holds at t’). The final version of the French IPF then looks like this: (11’)

Imparfait in French (final version) ⇒ Assertion: λP [e| Pw(e), t ⊆ e] Presupposition: [t| t < t0] i. if P satisfies divisibility ii. undefined otherwise

This additional condition explains, for instance, why Imparfait gets only a coerced group interpretation with semelfactives: (14) Jean frappaitimparfait à la porte. John knocked on the door. Due to the subinterval property (here: the constraint on divisibility), IPF in examples like (14) cannot refer to a single complete event of knocking on the door. The predicate must be interpreted as referring to a group of knocking events. What, then, happens with punctual achievement predicates like entrer in (3) and apparaître in (13)? Like the semelfactive in (14), achievement predicates must be coerced due to the sortal condition on the imperfective. For achievements this coercion typically results in a complex event predicate which includes the preparatory process preceding the culmination. There may be many such complex event predicates (e.g. all the possible ways in which Pierre and Bruno could emerge out of a cloud of dust in (13)). For some of these event predicates P1, P2, P3 etc. the events satisfying, say, P1 stand in a proper subpart relation with the events of type P2. In this way, the divisibility constraint of Imparfait is satisfied through coercion. It thus seems that achievement predicates become acceptable with a complete event interpretation of IPF when they are conceived of as more complex “accomplishments”. Why, then, are true accomplishments ruled out (or marginal) with a narrative complete event reading of Imparfait? Indeed, it is not clear whether this restriction is absolute. For the Italian Imperfetto, Bonomi (2004) claims that although achievements are clearly preferred in the complete event interpretation, accomplishments may be possible, as well. Concerning the French IPF, in example (3) above – from Jayez (1999)

Imperfectivity and complete events 

– the achievement predicate entraient is followed by the activity/accomplishment discutaient avec and the accomplishment se dirigeaient vers. Interestingly, however, my informants seem to prefer Passé simple or Passé compose for the two last VPs. The suggestion I want to make here is that accomplishments (and activities) already contain a preparatory process, which satisfies the divisibility constraint, and therefore favour an ‘incomplete event interpretation’. In other words, when the imperfective viewpoint combines with a predicate of the accomplishment type, it typically zooms directly in on the preparatory phase leaving the culmination part out of focus. This option is conceptually simpler than the alternative, which would be to coerce the accomplishment into a complex event predicate which includes the end point and still satisfies the divisibility constraint.

6. Markedness, competition and pragmatic strengthening Given the semantics accorded above to the marked PF in Russian and the marked IPF in French, I briefly return in this section to the issue of markedness and the interaction of tense-aspect grams in the verbal system.

6.1

Competition in Russian

The Factual IPF, that is, the form-meaning pair , is expected to be blocked both from the point of view of production and interpretation. There is a better form for expressing complete events (PF), and there is a better interpretation of IPF (the progressive/processual reading). Still, the data provided above suggests that blocking does not leave this problematic form-meaning pair unemployed in Russian. In optimality theoretic terms, we get what is known as partial blocking (and also deblocking), cf. Grønn (2008). What the complete event readings of the Factual IPF have in common, is that they occur in contexts where the hearer is not misguided by the use of IPF. In the case of existential IPF (cf. ex. 1), the assertion time is big and indefinite (all time preceding the utterance time), which effectively excludes a progressive interpretation. Hence, if there is no particular focus on the result of the event, the speaker sees no reason not to use the unmarked and more economic IPF. A similar kind of reasoning applies to the presuppositional IPF in examples like (2) above, where IPF is used with event anaphora. Since a complete event of the right kind is already given in (entailed by) the input context, no misunderstanding can arise from the use of the unmarked IPF. The unmarked status of the imperfective viewpoint in Russian is arguably the crucial factor which allows the Factual IPF to reappear on the scene and become the winner in a second round optimisation.

 Atle Grønn

The pervasive competition between PF and IPF leads to pragmatic strengthening and in some cases to rather surprising implicatures: [T]he unemployed form may soon find a new job, generally expressing something closely related to but subtly different from the canonical interpretation that one might have expected. (Beaver and Lee, 2003: 140)

These phenomena are well-known from text-books on Russian aspect. Here, I restrict myself to one example: the convention of annulled result (dvunapravlennoe znachenie) associated with IPF in the following minimal pair: (15) Vanja priekhal v Moskvu na proshloj nedele Vanja.NOM come.PF.PAST to Moscow.ACC in last week.LOC i ostanetsja do zavtra. and stay.PF.FUT till tomorrow Vanja came to Moscow last week and will stay until tomorrow. (15’) Vanja priezzhalIPF-past v Moskvu na proshloj nedele (*i ostanetsja do zavtra). Vanja came to Moscow last week (*and will stay until tomorrow). When the lexically encoded result state following the event holds at the utterance time, PF must be used as in (15). IPF, as in (15’), gets the opposite interpretation: The result state is cancelled, hence the agent must have left Moscow at some point prior to the utterance time. Such implicatures arise because of the tight competition between two members of a grammatical category. At different levels of optimisation, we get a polarisation of the interpretations accorded to the two aspectual competitors.

6.2

Competition in French

The French case is perhaps more tricky to capture in terms of a binary competition since the category of viewpoint aspect in contemporary French belongs to an intermediate phase, for which it would be an idealisation to say that Passé simple has disappeared, although the form is clearly declining. The two stages of the evolution of the tense-aspect system in French referred to in section 3 are therefore coexisting in some narrative texts, where we find both the “old” Passé simple and the “new” Imparfait narratif with complete event interpretations in narrative progression, cf. example (16): (16) Maigret repritpassé simple la petite auto noire et, quelques minutes plus tard, il pénétraitimparfait (narratif) dans la brasserie de la place de la République. (from Hobæk Haff) Maigret once more took the little black car, and a few minutes later he entered the café at the Republic square.

Imperfectivity and complete events 

Still, the French system is clearly developing into a pure aspectual opposition in past tense between Imparfait and Passé composé, cf. the two different historical stages in table 3 and 4. Table 3. Tense-aspect grams in early/written French Form/meaning

t⊆e

e⊂t

explicit reference to t0

Imparfait Passé simple

OK (OK)

!* OK

!* !*

!*

(OK)

OK

Passé composé

Table 4. Tense-aspect grams in modern/spoken French Form/meaning

t⊆e

e⊂t

explicit reference to t0

Imparfait Passé composé

OK (OK)

!* OK

!* OK

In the first round (cf. table 3), Imparfait, as the marked aspect, is accorded the semantics familiar from section 5. Passé simple is the aspectually unmarked past tense, and is in principle neutral with respect to the complete/incomplete event distinction.9 On the other hand, as a pure simple past, Passé simple does not explicitly designate the utterance time t0, hence the barred third cell (marked ‘!*’) in table 3. The composite perfect is naturally the winner in case of temporal relevance of the utterance time parameter. As the unmarked viewpoint, facing competition from both Imparfait and Passé composé, Passé simple is quite fragile. The competition only increases, when Imparfait narratif challenges Passé simple in cases of narrative progression. The narrative complete event reading of the French IPF appeared quite late, in the beginning of the 20th century, but, as shown in section 5, it is compatible with the invariant meaning of IPF in French. In narrative texts, a division of labour naturally arises between Imparfait narratif and Passé composé, such that the former is used when contemporaneity holds (typically with a coerced achievement predicate and a small assertion time), while the latter occurs with accomplishments and/or big assertion times. In table 4, I have suggested that Passé composé has developed into an unmarked tense-aspect gram, compatible with any temporal configuration. This is an interesting point, suggested to me by Philippe Schlenker (p.c.). The idea and the relevant data definitely need further investigations. If Passé composé is a semantic default in contemporary French, we expect this gram to occur whenever the marked Imparfait is 9. Whether Passé simple exhibits incomplete event interpretations is not obvious (due to the strong competition from Imparfait), but it clearly combines with atelic predicates.

 Atle Grønn

inappropriate. Schlenker provides some highly intriguing data with a progressive, incomplete event context where Imparfait turns out to be unavailable for purely morphological reasons. The crucial observation is that Imparfait seems to be replaced by a composite perfect in such environments. The relevant contrast is between the standard example (17) and (17’/17’’): (17) Lorsque le directeur est entrépassé composé, Marie étaitimparfait (*a étépassé composé) en train de danser. When the director entered, Mary was dancing. (17’) Bien que Marie ait étésubjunctive passé composé (*étaitimparfait) en train de danser lorsque le directeur était entrécomposite past perfect, elle ne sera pas licenciée. (no imperfective form in the subjunctive, and the past imperfective subjunctive is obsolete) Even though Mary was dancing when the director entered, she will not be fired. (17’’) Avoir étécomposite perfect (*étaitimparfait) en train de danser lorsque le directeur est entré – quelle malchance! (no imperfective form in the infinitive) Having been dancing when the director entered – what bad luck!

7. Conclusion In this paper, I have argued that some apparent similarities between imperfective viewpoints in Russian and French are quite accidental. Insights from traditional markedness theory, notably the idea that only the marked form should be accorded an invariant meaning, have here been explored in a framework of formal semantics, with some allusions also to optimality theory. Given the basic assumption that the Russian IPF is unmarked, while the French IPF is marked, I have shown how the different imperfective complete event readings arise in competition with other tense-aspect grams in a closed grammatical system. Crucially, while the Russian IPF is fully compatible, in the right context, with the perfective-like inclusion relation e ⊂ t, the marked Imparfait retains its meaning of contemporaneity even when it has a complete event interpretation in contexts of narrative progression. The so-called Imparfait narratif therefore encodes the condition t = e, and Imparfait tout court amounts to the improper inclusion relation t ⊆ e.

Imperfectivity and complete events 

References Beaver, D. & Lee, H. 2003. Form-meaning asymmetries and bidirectional optimization. In J. Spenader, A. Eriksson, & Ö. Dahl (eds.), Proceedings of the Stockholm Workshop on Variation within Optimality Theory, 138–148. Stockholm: Stockholm University. Bermel, N. 1997. Context and the Lexicon in the Development of Russian Aspect [Linguistics Vol. 129]. Berkeley CA: University of California Press. Blutner, R. 2000. Some aspects of optimality in natural language interpretation. Journal of Semantics 17: 189–216. Bohnemeyer, J. & Swift, M. 2004. Event realization and default aspect. Linguistics and Philosophy 27: 263–296. Bonomi, A. 2004. Semantical remarks on the progressive reading of the imperfective. Ms. Breu, W. 1998. Sopostavlenie slavjanskogo glagol’nogo vida i vida romanskogo tipa (aorist:imperfekt:perfekt) na osnove vzaimodejstvija s leksikoj. In M. Chertkova, (ed.), Tipologija vida, 88–99. Moscow: Jazyki russkoj kul’tury. Ferreira, M. 2005. Event Quantification and Plurality. PhD dissertation, MIT. Grønn, A. 2004. The Semantics and Pragmatics of the Russian Factual Imperfective. Dr. art. dissertation. Acta Humaniora Vol. 199. Oslo. Grønn, A. 2007. Horn strategies and optimization in Russian aspect. In A. Benz, C. Ebert, & R. van Rooij, (eds.), Proceedings of Language, Games, and Evolution, 17–24. Dublin: ESSLLI. Grønn, A. 2008. Russian aspect as bidirectional optimization. In F. Marusic & R. Zaucer (eds.), Studies in Formal Slavic Linguistics [Linguistik International Vol. 19], 121–137. Frankfurt: Peter Lang. Heim, I. 1994. Comments on Abusch’s theory of tense. In H. Kamp (ed.), Ellipsis, Tense and Questions, 143–170. Amsterdam: University of Amsterdam. Hobæk Haff, M. 2005. L’Imparfait narratif – l’enfant terrible de l’univers aspectuo-temporel français. Révue Romane 40(1): 137–152. Jayez, J. 1998. DRT et Imparfait. Un exemple de traitement formel du temps. In J. Moeschler (ed.), Le temps des événements, 123–154. Paris: Kimé. Jayez, J. 1999. Imperfectivity and progressivity: The French imparfait. In T. Matthews & D. Strolovitch (eds.), Proceedings of Semantics and Linguistic Theory IX, 145–162. Ithaka NY: Cornell University Press. Kamp, H. 2001. The importance of presupposition. In C. Rohrer, A. Rossdeutscher & H. Kamp (eds.), Linguistic Form and its Computation, 207–254. Stanford CA: CSU Publications. Kamp, H. & Rohrer, C. 1983. Tense in texts. In R. Bäuerle, C. Schwartze & A. von Stechow (eds.), Meaning, Use, and Interpretation of Language, 250–269. Berlin: de Gruyter. Klein, W. 1995. A time-relational analysis of Russian aspect. Language 71: 669–695. Smith, C.S. 1997. The Parameter of Aspect. (2nd edn). Dordrecht: Kluwer. von Stechow, A. 1995. On the proper treatment of tense. In T. Galloway & M. Simons (eds.), Proceedings of Semantics and Linguistic Theory V, 362–386. Ithaka NY: Cornell University Press.

Predicative expressions of transition in Portuguese and Spanish A cognitive approach to aspect, aktionsart and tense* Ingmar Söhrman and Kåre Nilsson Contrary to the Germanic languages, the Romance languages have a whole range of verbs instead of one (or two) specific, polysemic copula verb which can be used to express all kinds of transition from one state or situation into another, i.e. serving as a dynamic counterpart of a corresponding static verb used to describe existing states of affairs related to animate or inanimate subjects in a predicative context. There also exist more synthetic ways of expressing change in the state of affairs through the so-called ingressive verbs. These verbs reflect the typical situation in Latin, where transition, i.e. change in the state of affairs, was normally represented by lexical and/or synthetic means of expression. Cf. excandesco and irascor ‘to get angry’, where the rise of a new situation is made explicit by the inchoative infix -sc-. Thus the nowadays current analytical expressions of transition by means of a dynamic copula or a factitive verb followed by a predicative complement of the subject or object, respectively, represent a typological shift from the ancient, synthetic ways of expressing situational change.

1. Introduction A typological difference with didactical consequences between Romance and Germanic languages is the existence of very few polysemantic lexemes to mark transition in Germanic languages whereas Romance languages need a whole range of verbs to express this function. Verbs used to indicate a transition imply a change of positions of the subject, i.e. A ⇨ B. This transition can be either a transformation that changes A into B (1) and (2) or a mere going from nothing to something (3). This description is a * This article is a comparative approach to a problem that has only been dealt with separately and the diachronic dimension has been added here. The basic data used in this article come from two earlier separate investigations: Nilsson 1999 and Nilsson & Söhrman 2007.

 Ingmar Söhrman & Kåre Nilsson

cognitive elaboration of the expresiones fientivas presented by Wesch (2004: 218–219 & 222–229).

(1) She got angry.

(2) He became a judge.

(3) The job turned out to be more difficult than expected.

In example (1) the subject obviously changes her mood or is made to change it, and in example (2) the subject became a professional after being a student, while in example (3) the predicative (difficult) stands out as the final result of a previous development from a zero-point of departure. As will be shown further down this, in many respects, is a problem that touches heavily upon the notions aspect and Aktionsarten (dos Santos, 1996) and to a lesser degree on tense, but it is by no means a clean-cut aspectual problem. On the contrary it is anything but clean-cut. Evasiveness and lexical transgression is more the rule than the exception. However, our intention is to bring some more order into this complex and vague contrastive problem. In Scandinavian languages the two verbs in question are blive (Danish)/bli (Norwegian and Swedish) and gøre (Danish)/gjøre (Norwegian)/göra (Swedish). These correspond to German werden/bleiben and machen, which are like their Dutch parallels worden/blijven and maken. The old Scandinavian verb was (Norwegian) verða/ (Swedish) varða that got replaced by bliva (later bli) at the end of the Middle Age due to Low German influence blîven (Faarlund 1994: 71). English differs, since there are more verbs to be considered. We have just seen two examples – get and become. However, there are more lexical possibilities: turn, grow, go, be made and turn out to be etc. Let us just look at the first two examples once more. In Scandinavian both examples will take bli as in (4), (5), (7) and (8) and in German we find werden, (6) and (9). In written German the imperfect dominates, but in the spoken language perfect (ist geworden) is often used in these cases.

(4) Hun ble sint. (Norwegian)

(5) Hon blev arg. (Swedish)

(6) Sie wurde böse. (German)

She got angry

(7) Han ble dommer. (Norwegian)

(8) Han blev domare. (Swedish)

(9) Er wurde Richter. (German) He became (a) judge There also exists another possibility. The ingressive verbs already contain the semantic value of transition. They seem to be slightly more frequent in Scandinavian languages

Predicative expressions of transition in Portuguese and Spanish 

than in English. Be that as it may. In Scandinavian languages these verbs have a -na/-ne suffix: mörkna/mørkne (get dark, darken). Originally this originates from *-nan, a suffix with inchoative value that derived verbs from adjectives like Old Scandinavian blána ‘get blue’ < blá ‘blue’ + -nan (Faarlund 1994: 70). Among Romance languages this ingressive value is very frequent; cf. Sp. anochecer (‘get dark’), enojarse (‘get angry’), Port. anoitecer, aborrecer-se, French s’ obscurcir, se fâcher and Italian oscurarsi, arrabbiare. As English often differs from the other Germanic languages due to strong French influence, French differs from the other Romance languages as a consequence of Frankish (Germanic) influence, and French actually has a corresponding verb devenir and the reflexive se rendre, with an ingressive possibility in se fâcher, cf. (10), (11) and (12).1 It seems possible that in Vulgar Latin a prefixed variant of the verb venire gave birth to these verbs in French and Italian (cf. Rey 1992, quoted below), although this does not seem to have occurred in Old Spanish with a few exceptions (Pountain 1984: 101–102). However, venire had the possibility of an ingressive (or, in our terms, transformative) reading in Old Spanish as well, as can be found in the earliest manifestations of Spanish literature, cf. El Cid: 1853 “ricos son venidos todos sos vasallos” (Pountain 1984: 102). DEVENIR v. intr. Est un emprunt ancient (v. 980) au latin devenire, de de et venire (→ venir), proprement “venir de, arriver à” d’ où au figuré “aboutir à, recourir à” et, à basse époque, “être engagé dans un processus qui aboutit à un changement d’ état”. Devenir, emprunté dans ce dernier sens, ne s’ est employé à la forme impersonnelle il deviant qu’ en français classique (mil. XVIIIe s. Prevost) [...] (A. Rey, 1992: 594).

(10) Il est devenu juge. (Fr.) He be.PRES become.PERF.PART.MASC judge He became a judge (11) Elle s’ est rendue furieuse. (Fr.) She herself be.PRES get.PERF.PART.FEM angry or its ingressive equivalent: (12) Elle s' est fâchée. (Fr.) She herself be.PRES become angry.PERF.PART.FEM (both 11 and 12) She got angry The French verb devenir has been adopted as a learned borrowing into Spanish, but it is not used outside the specialised world of linguists (Pountain 1984: 101; Wesch 2004: 220).2

1. This is also the case in Italian with the verbs diventare/divenire, but to a much lesser degree than in French. 2. A fuller view of the development of this ingressive verb in Romance languages is given in Pountain 1992.

 Ingmar Söhrman & Kåre Nilsson

In Spanish and Portuguese the analytic variants of the last three examples would be: (13) Se hizo juez. (Sp.) himself make.PRET.3SG judge He became a judge (14) Se quedó enfadada. (Sp.) herself become.PRET.3SG angry PERF.PART.FEM (15) Tornou-se juiz. (Port.) turn.PRET.3SG-himself judge He became a judge (16) Ficou zangada. (Port.) become.PRET.3SG angry.PERF.PART.FEM. She got angry (ficou = Sp. se quedó, also without a direct parallel in English or other Germanic languages) In Classical Latin these verbs basically correspond to fieri (with the present fio etc.), the passive counterpart of facere. In this case we are dealing with a suppletive perfect, factus sum, since fio did not have a perfect, (Rubenbauer &Hofmann 1975 :111) and the reflexive form se facere (Pountain 1984 :101; García Hernández: 2000: 59–60.). Cf. 16: (17) Iudex factus est. (Lat.) Judge make.PERF.PART.MASC be.PRES.3SG He became a judge This is the Latin equivalent to examples (13) and (15). Nevertheless, sometimes both transition and result have to be expressed by an inceptive verb form (Haverling 2000) as in (18), excanduit from the verb excandescere. (18)

Excanduit Lichas hoc sermone turbatus. (Lat.) Anger.PERF.3SG Lichias this speech distort.PERF.PART Lychas got angry when he heard this speech (Petronius, Satyricon, 105.1)

So far we can just see that there is a typological difference between Romance and Germanic languages, as well as between Latin and modern Ibero-Romance ways of expression. If we go back to what was indicated at the beginning we notice that there are two cognitive categories: one resultative and one transformative (cf. Fabricius-Hansen 1975; Nilsson 1999.) The first one focuses on the result of a development from “nothing” to “something” (0 ⇨ +) as in the examples (3), (19) and (20), while the transformative change presupposes a transition from a contrary or at least a different earlier stage (- ⇨ +) as in (1), (2), (21) and (22). Accordingly, the resultative category does not necessarily imply a previous situation, being merely a way of announcing how something finally turned out, while the second focuses on the transitional process itself.

Predicative expressions of transition in Portuguese and Spanish 

(19) ...al fin resultó una empresa ardua (Sp.) at-DEF.ART end turn.PRET.3SG a undertaking troublesome finally this turned out to be a troublesome undertaking, (JMM, 329) (20) A quase tragédia acabou reduzida DEF.ART quasi-tragedy end.PRET.3SG reduce.PERF.PART.FEM a uma perna fraturada. (Port.) to a leg break.PERF.PART.FEM. What could have become a tragedy was reduced to a broken leg’3, (Veja, 86) (21) ¿Por qué se hizo comunista? (Sp.) Why him-/herself make.PRET.3SG communist Why did (s)he become a communist?’ (VM 227) (22) Com o tempo, tornou-se o maior As the time turn.PRET.3SG-itself the most important fabricante de vinhos da comarca. (Port.) producer of wines from-DEF.ART region. As time went by he turned into the most important wine producer in the region. However Portuguese and Spanish (and Catalan as well) get more complicated, as there are two or even three verbs for to be. In Spanish and Portuguese there are ser and estar, and in some instances the Spanish quedar(se) or the Portuguese ficar may even be considered as more or less an alternative to the former two verbs of being. The main difference between these verbs seems to be a mainly aspectual one, insofar as ser generally describes a norm that does not change so easily (23) and (24) while estar (like quedar(se)/ ficar) in combination with an adjective or participle (and in a few cases with a noun) characterizes a transitory or non-norm situation (25) and (26) (cf. Falk 1979b). (23) La naranja es amarilla. (Sp.) DEF.ART orange be.PRES.3SG yellow (24) A laranja é amarela.(Port.) DEF.ART orange be.PRES.3SG yellow. The orange is yellow (23 and 24) (25) El agua está caliente. (Sp.) DEF.ART water be.PRES.3.SG hot. (26) A água está quente. (Port.) DEF.ART water be.PRES.3SG hot The water is hot (25 and 26)

3. The translations of the Portuguese and Spanish examples are not intended to be idiomatic but to facilitate the understanding of the given examples.

 Ingmar Söhrman & Kåre Nilsson

From a general point of view ser could be seen as representing a durative Aktionsart and estar as a limitative, since it expresses what seems to be a transitional state but not necessarily reversible. This goes for the examples just quoted, but the semantic values are more complex than what we have seen. In the examples (25) and (26) the water can be warm, hot, cold and varies continuously. The main thing, however, is that there is no norm for the temperature as opposed to a case like la nieve es fría (‘the snow is cold’, Sp.) or a neve é fria (Port.) inasmuch as the coldness of the snow is an immanent (and unchangeable) property (or norm). In (25) and (26) there does not exist a prototype or prototypical situation, and, furthermore, the different states of things do not have to be reversible. This is evident especially when it comes to participles, as in the following examples: (27) El vaso está roto. (Sp.) DEF.ART glass be.PRES.3SG break.PERF.PART.MASC (28) O copo está partido. (Port.) DEF.ART glass be.PRES.3SG break.PERF.PART.MASC The glass is broken (27 and 28) (29) La carta está escrita. (Sp.) DEF.ART letter be.PRES.3SG write.PERF.PART.FEM. (30) A carta está escrita. (Port.) DEF.ART letter be.PRES.3SG write.PERF.PART.FEM. The letter is written (29 and 30) As will be seen further down many of the copulative verbs of transition in Portuguese and Spanish predicative expressions of change have a basic semantic value as verbs of movement ‘go’, ‘put’ etc. However, this could not explain the semantic change, even if a transition could be regarded a figurative movement. We agree with Coseriu when he suggests that these verbs function as aspectual forms of the copula ser, and that this might have its origin in an early Greek influence on spoken Latin inasmuch as in ancient Greek we find verbs with aspectual functions of be like γίγνομαι ‘become, get’ and πέλω/ πέλομαι ‘exist, be, become’ etc. (Coseriu 1977: 75–77). Without developing this complex syntactic and semantic problem we can only conclude that it is not entirely an aspectual problem. It has just got some aspectual links that should be taken into regard. How does this affect the usage of these copulative verbs of transition? First of all there has to be an existing or presupposed situation that describes how things are or were, although this might only be in the speaker’s own mind. In any case a transition requires a (static) situation to react to or interact with, i.e. a dynamic process as the transition must become stable or leave a stable state. The yellow colour of an orange is a sign that it is now ripe, and it has thus become ripe. In its first (unripe) stage it is green, and the process of becoming what is described in (23) and (24) would be as in (31) and (32) where we use verbs of transition, if we want to stress just that and not the final state.

Predicative expressions of transition in Portuguese and Spanish 

(31) La naranja se volvió amarilla. (Sp.) DEF.ART orange itself turn.PRET.3SG yellow The orange turned yellow. (32) A laranja tornou-se amarela. (Port) DEF.ART orange turn.PRET.3SG-itself yellow The orange turned yellow. Before getting into the actual verbal usages in Portuguese and Spanish there is one more important dichotomy that should be emphasised: intransitive and transitive constructions. The temporal and aspectual sides will become clearer further down. The intransitive (or formally reflexive) constructions are paired with a predicative complement of the subject. (33) He went mad. The transitive constructions have objects which, in their turn, go with a predicative complement. Compare the intransitive construction in (33) and its transitive counterpart (34). (34) He’s driving me mad. In order to obtain a quantitative basis we have studied thoroughly some novels, short stories and other texts that are given in the bibliography and all percentages shown here refer to this corpus, which is entirely modern. The few historical Spanish examples are taken from Hernán Núñez (late 17th century).4

2. Portuguese In Portuguese the main intransitive verbs indicating a change substantiated by a predicative complement are ficar, tornar-se, pôr-se, fazer-se, acabar, chegar a (ser), passar a ser, vir a ser, ser (alone) and, in Brazil, virar, while the main transitive, which all have a causative value, are deixar, tornar and pôr. Three of these verbs are, as could be foreseen, intransitive when they go with a reflexive pronoun and transitive when not. The intransitive verbs and verbal idioms are the most frequent ones, and the verb that stands out as the most frequent is ficar (de Carvalho 1984 and Lohse 1981). In 46 % of the cases of our corpus this verb is followed by an adjective. In another 25 % ficar goes with participles that function as adjectives as well.

4. Hernán Núñez’ texts consist of 90 letters from 1670/1671 that are being edited by Ingmar Söhrman.

 Ingmar Söhrman & Kåre Nilsson

(35)

Ficou perplexo, become.PRET.3SG puzzle.ADJ.MASC, irresoluto. (Port.) indecisive.ADJ.MASC (Andrade, 84)

(36) A cidade ficou arrasada. (Port.) DEF.ART city become.PRET.3SG tear.PERF.PART.FEM. The city was reduced to ruins. Ficar can also be used with an adverbial of duration that makes it virtually incompatible with the ingressive/inchoative Aktionsart, as in (37), where ficar corresponds more or less to estar, as can be seen in (38), although only the first of these examples implies, as Hundertmark Santos-Martins puts it, “eine Zustandsänderung als Folge” (Hundertmark-Santos Martins 1982: 331). (37) Ficaram algum tempo calados. (Port.) become.PRET.3PL some time silent.PERF.PART.MASC.PL We became silent for some time. (Andrade, 94) (38) Estiveram algum tempo calados. (Port.) be.PRET.1PL some time silent PERF.PART.MASC.PL We kept quiet for some time. Furthermore, it seems worth noticing that in (36) ficar also assumes an intransformative (non-change) component in addition to that of the change that gives birth to the subsequent situation. It also seems that this verb lends itself to both animate and non-animate subjects, without any marked preference for one or the other. Tornar-se also seems to prefer by far adjectives as a complement, as in 73 % of the examples of this verb. (39) Samuel tornou-se íntimo da Samuel turn.PRET.3SG-himself intimate of-DEF.ART legação da China. (Port.) embassy staff of-DEF.ART. China Samuel got close to the Chinese embassy staff. (Andrade, 80) This verb, however, mainly goes with static situations (like ser) and most of the subjects are animate, while the inanimate subjects represent only half that frequency. This, however, does not exclude the possibility of combining the verb with an infinitive (40) or a whole phrase (41). (40) Para atender ao serviço de estradas [...], To attend to-DEF.ART service of streets torna-se imperioso lancer novos impostos.(Port.) turn.PRES.3SG-itself necessary launch new taxes In order to attend the road service it became necessary to launch new taxes. (Andrade, 41)

Predicative expressions of transition in Portuguese and Spanish 

(41) Se o objeto indireto e o sujeito não forem If DEF.ART object indirect and DEF.ART subject not be.PRET.3PL correferenciais, torna-se possível coreferential turn.PRES.3SG-itself possible que o objeto indique o LOP5. (Port.) that DEF.ART object indicate.3SG DEF.ART LOP. If the indirect object or the subject are not be coreferential, it gets possible for the object to indicate a LOP. (Moreira, LdA, 47) The verbs fazer-se and pôr-se turned out to be very infrequent, which was a bit surprising. They tend to carry the same semantic value as tornar-se (41) and ficar (42). Besides, fazer-se with a transitional value is used in the impersonal idiomatic phrase, fazer-se tarde (‘to be getting late’). (42) ...frequentemente não se faz necessária …often not itself make.PRES.3SG necessary tal especificação. (Port.) such specification ...often such a specification does not become necessary. (Moreira, LdA, 40) (43) As prodigiosas mãos de alumínio puseram-se DEF.ART.PL prodigious hands of aluminium put.PRET.3PL-themselves trémulas. (Port.) trembling The prodigious aluminum hands started to tremble. (de Melo, IP, 86) The verb ser itself is normally used to describe existing facts. Referring to the future, however, it takes on a perfective aspect, implying the rise of a new situation: (44) Quando for grande, quero ser When be.FUT.SUBJ.1SG big, want.PRET.1SG be.INF espada. (Port.) sword When I grow up I want to become a bull-fighter. (Costa, IP, 227) When ser is extended to passar a ser the predicate represents explicitly a change of quality in the subject: (45) … viver passa a ser pior …live turn.PRES.3SG to be.INF worse do que morrer. (Port.) of-DEF.ART than die.INF …to live becomes worse than to die. (Port., Veja, 127) 5.

LOP = Local de Origem do pacente (place of origin of the pacient).

 Ingmar Söhrman & Kåre Nilsson

With the construction chegar a (ser) the final is intensified somehow, meaning as Hundertmark puts it “es kommt dahin, das etwas geschieht” (Hundertmark-Santos Martins 1982: 355). And, with an animate subject (the typical case), implies endeavous to achieve the result indicated by the predicative (adjective or noun). (46) Viste já a luminosidade deste rio? see.PRET.2SG already the light of-this river?

Chega a ser agressiva [...] (Port.) come.PRES.3SG to be.INF aggressive Have you seen the light of this river. It is almost getting aggressive. (Carvalho, IP, 172)

(47) Viera do Norte, morava come.PLUPERF.3SG from-DEF.ART. North, live.IMPF.3SG em Laranjeiras, chegar a gerente de banco. (Port.) in Laranjeiras come.PLUPERF.3SG to director of bank He had come from the north, lived in Laranjeiras, [and] had finally become a bank director (Andrade, 74) As to the transitive varieties in Portuguese, deixar corresponds to ficar and, for obvious reasons, tornar goes like tornar-se, which means that the first one is “linked” to the estar-field and the second one to ser, as in (48) and (49). (48) A senhora me deixa aflita! (Port.) DEF.ART mistress me leave.PRES.3SG distress.PERF.PART.FEM You make me distressed (Trevisan, SE, 45) (49) ...o mercado planetário dos filmes e a ...DEF.ART market planetary of-DEF.ART. filmes and DEF.ART Internet tornam tudo obsessivamente presente. (Port.) Internet turn.PRES.3PL everything obsessively present …the world film market and internet turn everything obsessively present. (Revista, 50)

3. Spanish In Spanish, the intransitive verbs used in the transitional and resultative contexts in question are quedar(se), hacerse, llegar a (ser), pasar a ser, volverse, ponerse, resultar, salir and ser (alone), whose transitive counterparts are dejar, hacer, volver and poner. The verb quedar(se) turns out to be, by far, the most frequent one, although this is not the impression given in most grammar books. It is used both with and without a reflexive pronoun, which stands out in comparison with the normally non-reflexive Portuguese ficar that largely corresponds to quedar(se). All the cases belong to the estar-

Predicative expressions of transition in Portuguese and Spanish 

field. It has a long tradition in Spanish and has been established as a verb of transition for a very long time. Note that in (51) we have the same doubleness as in ex. (37). (50) Éste, efectivmente, se quedó mudo This actually himself remain.PRET.3SG mute unos instantes. (Sp.) some moments He actually shut up for a couple of moments. (M, 13) (51) ... que Su Mgd. no querra que yo ... that Her Majesty not want.FUT.3SG that I quede priuado remain.PRES.SUBJ.1SG deprive.PERF.PART.MASC.

en tal dia del sacrifiçio dela missa. (Sp.) in such day of-DEF.ART sacrifice of-DEF.ART mass... …that Her Majesty does not want me to be deprived of the mass on such a day…(HN, letter 28)

In most cases the subject of quedar is inanimate, while that of quedarse is animate, but it has to be pointed out that quedar often takes a perfect participle and then the reflexive form is mostly avoided, even if the subject is animate: (52) …la primera Soledad había DEF.ART first Soledad have.IMPF.3SG quedado atrapada remain.PERF.PART.MASC6 trapped.PERF.PART.FEM. entre los dolores de un parto. (Sp.) between DEF.ART.PL pains of a birth. … the first Soledad had been trapped between the labour pains… (JMM 397) It can also be used with a gerund. In that case it seems that the subject is almost always animate, and the verb appears in its reflexive form. (53) Los tres se quedaron mirando DEF.ART.PL three themselves remain.PRET.3PL look.GER a Carvalho...(Sp.) at Carvalho The three [of them] stopped and looked at Carvalho…(VM, 67) Hacerse always corresponds to ser, which means that it has a long stable (durative) result. It normally implies an active role for the subject, when this is an animate one, 6. The VP había quedado is analyzed word by word, but it constitutes the two parts of an analytic pluperfect, i.e. = had remained.

 Ingmar Söhrman & Kåre Nilsson

which is most often the case (54). There are much fewer examples with an inanimate subject (55). (54) …llama ahora y hazte socio …call.IMP.2SG now and make.IMP.2SG-yourself member de Médicos sin Fronteras. (Sp.) of Doctors without borders …call now and become a member of Doctors Without Borders. (El P, 36) (55) Vivir en Euskadi se le hizo insoportable. (Sp.) live.INF in Euskadi itself him make.PRET.3SG unbearable To live in the Basque Country became unbearable to him. (El P, 38) Llegar a (ser) (normally with an animate subject) indicates a long (and hard) process to achieve the resulting situation (like chegar a (ser) in Portuguese) or what has been achieved in the end (56); whilst pasar a ser explicitly focuses the transformative process (57). (56) En el instituto llegó a ser In DEF.ART academy go.PRET.3SG to be.INF un músico experto. (Sp.) a musician expert At the academy he became an expert musician. (El P, 41) (57) De amenazadora has pasado From threaten have.PRES.3SG pass.PERF.PART.MASC7 a amenazada. to threaten.PERF.PART.FEM (Sp.) After being threatening you have become threatened. (VF, 197) Both volverse (58) and ponerse (59) normally take an animate subject, and while the first one points to a long-lasting result (thus corresponding to ser), the second seems to reflect resultative situations of a shorter ‘life-span’ (corresponding to estar). (58) ... al volverse sordo, pasó ..to-DEF.ART turn.INF-himself deaf pass.PRET.3SG a desempeñar el cargo de guardabosques. (Sp.) to fullfill.INF DEF.ART work of forest warden. …when he became deaf he took on the work of a forest warden (M, 297) (59) …el niño se puso enfermo …DEF.ART child himself put.PRET.3SG ill

7. The VP has pasado is here analyzed word by word, but it constitutes the two parts of an analytic perfect, i.e. = have become.

Predicative expressions of transition in Portuguese and Spanish 

de la gripe terrible … (Sp.) of DEF.ART flu terrible …the child fell ill to the horrible flu…( MM, 113) It is also interesting to note that this last verb is fairly infrequent, although most grammars uphold the contrary. Resultar often substitutes ser putting emphasis on the resultative value and the lasting effect. (60) En los países árabes, eso resultaba patente. (Sp.) In DEF.ART countries Arab this result.IMPF.3SG obvious In the Arab countries this became obvious. (VF, 99) It is also a verb that has been used in sentences of this kind for a very long time. (61) ... por que desto les resulta tan grande ...since from+this them result.PRES.3SG so great Vtil [=utilidad] que me pareçe seria

usefulness that me seem.PRES.3SG be.COND.3SG buena seguridad... (Sp.) good security …therefore this becomes so useful for them that it seems to me that it would be a good security measure… (HN, letter 47)

Furthermore it is of frequent usage, although it can not compete with quedar. The verb resultar might sometimes also be combined with ser to make the resultative value clearer. (62) is a typical example of this value, representing a ‘subjective change’ only felt by the speaker – i.e. when things turn out to be different from what was expected (a surprise) this is new information to that person, albeit it might not be so from a more objective point of view. Thus the surprise is only reflected in the eyes of the beholder. (62) Un día rechazó el original de un dibujante One day reject.PRET.3SG DEF.ART original of a drawer que resultó ser el hijo de Smith... (Sp.) who result.PRET.3SG be.INF DEF.ART son of Smith One day he rejected the original [drawing] of a drawer who turned out to be Smith’s son… (RM, 82) Salir, on the other hand, is often considered equivalent to resultar, but it is far less frequent. (63) Tuve esas seis entrevistas [...], have.PRET.1SG these six interviews

 Ingmar Söhrman & Kåre Nilsson

y no salí contenta de ninguna... (Sp.) and not leave.PRET.1SG happy of none I had these six interviews […] and did not get happy with any [of them]. (JMM, 164) Here the lexical meaning is kept intact – and this is also the case in (64) although in a metaphorical way. (64) ¿Han salido las fotos bien? (Sp.) have.PRES.3PL leave.PERF.PART.MASC8 DEF.ART.PL photos well? Have the photos turned out well?’ (VM, 152) It might be that resultar tends to avoid animate subjects and that salir has taken its role in these phrases, but that has not been sufficiently proven by our corpus. As in Portuguese, ser is interesting inasmuch as it implies more of a judgement of how things turned out or are/were expected to turn out than a real transition. It often goes from nil to something: (65) Aquel fue un verano horrible. (Sp.) That be.PRET.3SG a summer horrible That turned out to be a horrible summer. (RM, 97) It is also clear that tense plays an important role here, as the verb, when representing some sort of change or development, is normally used in the future and rarely in the present tense (66) with the exception of the expression ser cada vez más that evidently is a process-orientated lexicalised idiom (67). (66) También será inevitable un encuentro con Fonseca. Also be.FUT.3SG inevitable a meeting with Fonseca A meeting with Fonseca will be inevitable. (VM, 57) (67) Los casos son cada vez más DEF.ART.PL cases be.PRES.3PL every time more sofisticados. (Sp.) sophisticated.PERF.PART.MASC.PL The cases are getting more and more sophisticated. (El P Negocios, 12) As in Portuguese the transitive constructions are few, and while dejar (68) and poner (69) ‘belong’ to the estar part, hacer (70) and volver (71) are linked with ser (as copula

8. The VP han salido is here analyzed word by word, but it constitutes the two parts of an analytic perfect, i.e. = have left.

Predicative expressions of transition in Portuguese and Spanish 

of the resulting situation)9. None of these is very frequent, which makes it difficult to give any coherent and general idea of their usages. (68) Por fin, cayó Finally fall.PRET.3SG

en la cuenta de in DEF.ART count of

que lo habían dejado solo. (Sp.) that him have.PRES.3PL leave.PERF.PART10 alone Finally, he understood that they had left him alone. (M, 30) (69) No hay que ser sectario, Not have.PRES to be.INF sectario. pero a veces lo ponen muy difícil. (Sp.) but at times it make.PRES.3PL very difficult You do not have to be a sectarian, but sometimes they make it very difficult. (VM, 232) (70) Una te hará rico, otra te One.FEM you make.FUT.3SG rich another you encumbrará, otra te hará feliz. (Sp.) exalt.FUT.3SG another you make.FUT.3SG happy One will make you rich, another will exalt you, [and] a third one will make you happy. (M, 102) (71) La tipa es una golfa, pero le volvió DEF.ART type be.PRES.3SG a tart but him turn.PRET.3SG loco. (Sp.) crazy The bitch is a tart, but she drove him crazy. (VF, 146)

4. Conclusions Generally speaking, Spanish seems to have developed a more complex system than Portuguese as regards different kinds of change expressed by means of a copulative ‘auxiliary’ (followed by a complement representing the resulting situation). In Portuguese this field is mostly covered by the intransitive ficar (linked to estar) and tornar-se (linked to ser) and their transitive counterparts deixar and tornar. Their Spanish equivalents, on the other hand, are divided between the intransitive quedar(se) and ponerse 9. With the possible exception volver loco a alg (‘drive someone mad’), since this construction may imply both estar and ser loco as a result, depending on how durative the situation is seen, i.e. if it can be regarded as a property (norm) or not. 10. The VP habían dejado is here analyzed word by word, but it constitutes the two parts of an analytic pluperfect, i.e. = had left.

 Ingmar Söhrman & Kåre Nilsson

(linked to estar) to translate ficar, as well as hacerse and volverse (linked to ser) to translate tornar-se, with their transitive counterparts hacer and volver to translate tornar. Furthermore, two more Spanish verbs – resultar and salir – enter into the picture in this context, without obvious equivalents in Portuguese. As for the rest of the constructions treated in this study, there are no noticeable differences between Portuguese and Spanish. Ser as a means of expressing change or resulting (permanent) situations functions in the same way in both languages, and so do the periphrastic expressions chegar/llegar a (ser) as well as passar/pasar a ser. Finally, we would like to point out that what we have presented here is just a brief outline of a most intriguing and complex verb category, which we hope to give a theoretical structure in due time. So far we will just limit the conclusions to the following reflections: Our analysis of the complexity of this verbal category has hopefully illustrated that the combination of the three underlying structures – i.e. resultative vs. transformative value, normative vs. non-normative be-verb and transitivity vs. intransitivity – provides a good basis for understanding the intriguing semantic game and also serves as a starting point for undertaking deeper research work into this matter. Furthermore, we hope that we have been able to show how these verbal functions and categories are partly linked with the ideas of aspect, Aktionsart and tense, and how the animate or inanimate character of the subjects (and objects) also influences the choice between one or another verb or expression.

Corpus Portuguese Andrade = Carlos Drummond de Andrade. 1984. Contos de Aprendiz, 22.a ed. Rio de Janeiro: Ed. Record. IP = Imaginários Portugueses – Antologia de Autores Portugueses Contemporâneos. 1992. Coimbra: Fora do Texto. LdA = Linhas d'Água, 12. Dec. 1997. São Paulo. = Semesteremne i portugisisk – Noveller og dikt, (an anthology of Brazilian SE and Portuguese short stories). 1997. KRI, Universitetet i Oslo. Revista = Revista Expresso No 1328, 10.4.1998. Veja = Veja No 15, 16.4.97

Spanish AMM = Antonio Muñoz Molina. 1992. Los misterios de Madrid. Madrid: Alfaguara. El P = El País 27/06/04.

Predicative expressions of transition in Portuguese and Spanish 

= Hernán Nuñez. Cartas. (These letters from 1670–1671 are presently being edited by Ingmar Söhrman). JMM = José María Merino. 2003. El heredero. Madrid: Alfaguara. M = Eduardo Mendoza. 1990. La ciudad de los prodigios. Barcelona: Seix Barral. RM = Rosa Montero. 1988. Amado amo. Barcelona: Madrid: Debate. VF = Alberto Vázquez-Figueroa. 1996. Marea negra. Barcelona: Plaza & Janes VM = Manuel Vázquez Montalbán. 1981. Asesinato en el Comité Central. Barcelona: Planeta HN

References de Carvalho, J. G. H. 1984: Ficar em casa / Ficar pálido: Gramaticalização e valores aspectuais. In J.G. Herculano de Carvalho & J. Schmidt-Radefeldt (eds.), Colecção Linguística, 1”, do vol. Estudos de Linguística Portuguesa. Coimbra: Coimbra Editora. Coseriu, E. 1977. Estudios de lingüística románica. Madrid: Gredos. Faarlund, J. T. 1994. Old and Middle Scandinavian. In E. König & J. van der Auwera (eds.), Germanic Languages, 38–71.London: Routledge. Fabricius-Hansen, C. 1975. Transformative, intransformative und kursive Verben [Linguistische Arbeiten 26]. Tübingen: Niemeyer. Falk, J. 1979a. SER y ESTAR con atributos adjetivales. anotaciones sobre el empleo de la cópula en catalán y en castellano I [Studia Romanica Upsaliensia 29]. Uppsala: Acta Universitatis Upsaliensis. Falk, J. 1979b. Visión de ‘norma general’ versus visión de ’norma individual’. Studia Neophilologica 51: 275–293. García Hernández, B. 2000. Complementariedad intersubjetiva y secuencia intrasubjetiva. Desplazamientos históricos. In M. Martínez Hernández et al. (eds.), Cien años de investigación semántica: de Michel Bréal a la actualidad. Actas del congreso internacional de semántica. Universidad de La Laguna 27–31 de octubre de 1997, Vol. 1: 45–64. Madrid: Ediciones clásicas. Haverling, G. 2000. On Sco-verbs, Prefixes and Semantic Functions. A study in the development of prefixed and unprefixed verbs from Early to Late Latin [Studia Graeca et Latina Gothoburgiensia LXIV], Göteborg: Acta Universitatis Gothoburgensis. Hundertmark-Santos Martins, M.T. 1982. Portugiesische Grammatik, Tübingen: Niemeyer. Lohse, B. 1981. Det portugisiske verbum, ‘ficar’ – en syntaktisk og semantisk undersøgelse (hoved fagsspeciale). København: Romansk Institut, Københavns Universitet. Nilsson, K. 1999. Expressões predicativas de mudança em português. In J. Nystedt et al. (eds.), Acta del XV Congreso escandinavo de romanistas, Estocolmo 10 a 15 de agosto de 1999, 1434–1445. Stockholm: Stockholms universitet. (CD-rom). Nilsson, K. & Söhrman, I. 2007. Expresiones predicativas de cambio en castellano. In K. Benson, J.L. Girón Alconchel & T. Riiho (eds.), Actas del I congreso de hispanistas y lusitanistas escandinavos, 298–311. Madrid: Instituto Ibero-Americano de Finlandia. Pountain, C. 1984. How ‘become’ became in Castilian. In R.A. Cardwell (ed.), Essays in Honour of Robert Brian Tate from his Colleagues and Pupils, 101–111. Nottingham: University of Nottingham.

 Ingmar Söhrman & Kåre Nilsson Pountain, C. 1992. La notion de devenir en roman. Revue de linguistique romane 56: 427–437. Rey, A. (ed.). 1992. Dictionnaire historique de la langue française. Paris: Dictionnaires Le Robert. Rubenbauer, H. & Hofmann, J. B. [1975] 1995. Lateinische Grammatik. 12. Anflage. C. C. Buchner Verlag, J. Lindauer Verlag, R. Oldenburg Verlag. München. dos Santos, D. M. de S.M.P. 1996. Tense and Aspect in English and Portugese: A contrastive semantical study (tese de doutoramento em Engenharia Informática e de Computadores). Lisboa: Universidade Técnica de Lisboa. Wesch, A. 2004. La expresión de la noción ‘devenir’ en español. In J. Lüdke & C. Schmitt (eds.), Historia del léxico español. Enfoques y aplicaciones, 217–232. Madrid-Frankfurt-am-Main: Lingüística iberoamericana.

The Old Nordic Middle Voice in the pre-literary period Questions of grammaticalisation and cliticization* Kjartan Ottosson In this paper, data from sources predating the earliest preserved manuscripts, runic inscriptions as well as poetry, is used to shed light on the earliest history of the morphologically defined Nordic Middle Voice, ending in -sk in Old West Nordic and in -s in Old East Nordic. This category has not only the reflexive meaning inherited from the reflexive clitic –sk, but also i.a. reciprocal and anticausative meaning. The semantic development must be independent from the formal development including cliticisation, as shown by comparison with the German Middle Voice, which lacks cliticisation. The formal development includes generalisation of the clitic -sk from the 3rd person and infinitive as well as from the accusative to the dative (and genitive). The cliticisation is not restricted to arguments to the verbal hosts, but can involve elements very loosely connected to the verb. Morphological and syntactic arguments show that the clitic must have become an affix as early as the 9th and 10th centuries.

1. Introduction 1.1

General background

The topic of this paper is the morphologically defined category of Middle Voice (MV) in Old Nordic, characterised by the ending -sk (-mk in 1st sing.) in Old West Nordic, -s in Old East Nordic. Uncontroversially, this morphological marker is derived from an enclitic version of the reflexive pronoun, which is sik in the accusative, while the variant -mk in Old West Nordic derives from the 1st sing. personal pronoun mik. The * This paper was written while I was a fellow of the research project “Language change and grammatical theory” led by Jan Terje Faarlund at the Centre for Advanced Study in Oslo in the academic year 2004–2005, although it incorporates material presented at ICHL 14 in Vancouver 1999. It has benefitted from comments from the participants in the research project, as well as from the audiences in Vancouver and in Göteborg 2004.

 Kjartan Ottosson

ultimate origin of this category must be Proto-Nordic, although no relevant examples can be found in the very limited corpus of runic inscriptions from that period. The development of the Nordic Middle Voice raises some interesting questions concerning grammaticalisation and cliticization, some of which I will try to answer in this paper. On the basis of the oldest Nordic manuscripts, we can construct paradigms of the type exemplified in Table 1, for Old Icelandic and Old Swedish. Note the contrast in the 1st sing. between Old Icelandic kϞllumk and Old Swedish kallas. Table 1. The West and East Nordic Paradigms West Nordic -sk kallask ‘be called’

sing. 1st 2nd 3rd

plur. kϞllumk kallask kallask

kϞllumsk kallizk kallask

1st 2nd 3rd

kallas kallas kallas

kalloms kallins kallas

East Nordic -s kallas

A strong indication of the origin of the Middle Voice can be seen in the alternation found between some MV forms and active verbs with free reflexives, as in the Old Norse examples in (1) and (2), adapted from Nygaard (1905: 156–160). (1) a. Ek vil fǿrask undan þrælkan hans – ‘I want to remove-myself from his slavery’ b. þeir fǿrðu sik (ACC) undan sϞkum – ‘They removed themselves from charges’

(2) a. steyptisk konungsliðit ofan á þá – ‘the king’s army threw-themselves upon them’ b. hann steypir sér (DAT) út af þekjunni – ‘he throws himself out from the roof ’

The individual middle verbs can in their bare form govern either accusative or dative case on their objects, as illustrated in examples (1) and (2). A few middle verbs even govern genitive. It is clear that the -sk variant is phonologically derived from an enclitic form of the accusative sik of the reflexive pronoun. The phonological origin of the -s variant, on the other hand, has been disputed. In this paper, I will be concentrating on the evidence on the development of the Middle Voice provided by linguistic sources from the period before the earliest manuscripts, which for West Nordic – in Norway and its colonies – is the mid-12th century, for East Nordic – in Sweden and Denmark – the mid-13th century. The period in focus

The Old Nordic Middle Voice in the pre-literary period 

is roughly speaking the Viking Age, with the bulk of the West Nordic evidence coming from the 10th century, whereas most of the East Nordic evidence is from the 11th century. The main emphasis will be on West Nordic, and comparison with classical Old Norse will help to see lines of development, providing a partial basis for reconstructing the Proto-Nordic Middle Voice. That reconstruction, which in addition must draw on evidence from other Old Germanic languages, will hopefully be undertaken in a future publication. In addition to mapping and interpreting the development of this category in general, some specific questions of interest to grammaticalisation theory will be addressed in the paper.

1.2

Questions of grammaticalisation of the Old Nordic Middle Voice

Some aspects of the grammaticalisation that has been assumed to be involved in the development of the Old Nordic Middle Voice are rather well known from the literature, especially the Cambridge textbook by Hopper and Traugott (1993: 150–153). These authors (1993: 151f) take the development of the reflexive pronoun into a suffix in Old Norse to be a classic example of grammaticalisation. The main points in their account of the Old Nordic Middle Voice can be summarised as follows: Cliticization Paradigmatization (morphologisation): a. Syncretism of person and number: spreading of the MV marker from 3rd person and infinite forms to 2nd person and 1st plural. b. Case syncretism: spreading of an originally accusative form to dative (and genitive). Semantic generalisation: a shift from reflexives to reciprocals, middle voice, even passive, typical for the development of reflexives (as shown by Kemmer 1993). Phonological assimilation: The pronoun merges phonologically with the host. Hopper & Traugott (1993: 103–113) further discuss the syntactic aspects of what they call “decategorialisation”, which is a tendency for a form to lose the syntactic properties that would identify it as a full member of a major grammatical category, e.g. for a noun to lose the ability to appear with articles or adjectives, to serve as a verbal argument, or appear in certain syntactic positions. Although Hopper & Traugott do not apply their concept of decategorialisation to the Old Norse middle voice, it can be shown to be relevant there also. Hopper & Traugott’s account seems to be based entirely on Heusler’s 1921 Old Icelandic textbook (Altisländisches Elementarbuch), without any primary research. It is reconstructive in nature and concentrates on the cumulated effect of the components it assumes. Their account is more or less atemporal, as they do not attempt to date the change, although they do to some extent claim overlapping in time of some of its components. Subsequent to Hopper and Traugott’s book, the grammaticalisation properties

 Kjartan Ottosson

of the Old Nordic Middle Voice have received some attention. Thus, the unpublished Ottosson (1999) concentrated on some empirical issues relevant to grammaticalisation, and the important aspects of that paper are incorporated here. It was i.a. pointed out that the element had little semantic content to begin with, and that the result is largely derivational. This last point is one of Enger’s (2003) major concerns insofar as this represents a non-standard case of grammaticalisation, as that process is commonly assumed to create inflectional categories. Evidence bearing on the derivational properties of the MV should be sought from the much more abundantly attested Classical Old Norse (and East Nordic), and will only be given limited attention in this paper. On a more basic level, I would like to address the question whether the individual components which Hopper & Traugott see as part and parcel of any grammaticalisation process are that closely connected to each other in the case of the Old Nordic Middle Voice. Such a decomposition approach has been advocated by Andersen (2005), who sees the semantic aspect of grammaticalisation, which he calls grammation, a change from lexical to grammatical content, as the pivot of such processes, with the formal aspects more contingent. With respect to the Middle Voice, it turns out that the situation in the earliest sources (9th century) is strikingly similar to that of the oldest preserved manuscripts (ca. 1200). The decisive semantic development resulting in the Middle voice, more properly called “regrammation”, being a transition from one grammatical content to another, must have taken place before the period under investigation. Its logical independence of the morphological development is shown by the fact that Old High German uses the collocation of an active verb and a free reflexive for most of the same semantic variants as we find in Old Nordic, but shows no signs of cliticization of the reflexive pronoun or extension of the reflexive pronoun into the 1st and 2nd persons. Thus, the semantic development is prior to, and independent of, the formal development. The case syncretism part of paradigmatization in Old Nordic seems to be essentially independent of other formal aspects. On the other hand, “normalisation” of the phonological shape of the Middle Voice exponent from -sk to -st comes surprisingly late in West Nordic.

1.3

Questions of cliticization

In Hopper & Traugott’s scenario, the development of the Old Nordic Middle Voice involves morphologisation along the whole cline from lexical item through clitic to morphological affix. However, it seems reasonably clear that Hopper and Traugott (1993: 150–153) assume that -sk is still a clitic at the Classical Old Norse stage. One of the main issues for Enger (2003) is that -sk is an affix rather than a clitic in Old Norse, in agreement with the brief argument in Ottosson (1992: 68f). Against this, Faarlund (2005) has maintained that the forms involve cliticization at this stage, at least in “the original reflexive function”. A problem with such a split analysis is that the morphological properties of the middle voice are the same irrespective of which semantic and syntactic properties it shows in individual cases. More concretely, the 1st sing. forms

The Old Nordic Middle Voice in the pre-literary period 

which by Faarlund’s admission “strongly suggest” that the ending is a suffix rather than a clitic, as will be discussed presently, are just as current in reflexive use as in other semantic variants. The reflexive element –sk/-mk originated as a clitic, but not a simple clitic. It must have been a so-called “special clitic” (e.g. Zwicky & Pullum 1983: 510, Halpern 1998: 101), as it did not have the same syntactic distribution as the corresponding full forms. Further, -sk was what Halpern (1998: 104) calls a verbal clitic, i.e. one that always appears adjacent to a verb. As shown by examples such as (3), from Faarlund (2005), the reflexive element may appear in positions where the free reflexive is excluded in Classical Old Norse, and the same must be assumed for pre-literary Old Norse. (3) a. bjósk hann þá til holmgϞngu – ‘then he prepared (himself) for a duel’ b. *bjó sik hann þá til holmgϞngu When determining whether the reflexive element was a clitic or an affix in Classical Old Norse or pre-literary Old Nordic, it is surprisingly hard to find any hard-and-fast criteria for distinguishing between clitics and affixes in the theoretical literature (Zwicky 1977, Zwicky & Pullum 1983, Klavans 1985, Carstairs 1987, Halpern 1998), but in preparation for the discussion of individual aspects of the pre-literary examples it is necessary to mention briefly some of the main criteria that have been applied to Classical Old Norse. Faarlund’s (2005) main arguments for clitic status of the reflexive element in Old Norse are two. First, he argues that it meets the criterion that clitics alternate with independent words, and second that it is an independent element of syntax, in that it functions as an argument. In examples involving alternation between active verbs plus reflexive and middle verbs, as in (1)-(2) above, one might take the presence of the reflexive pronoun in the active alternant to indicate the presence of a “hidden” argument in the middle alternant. This alternation is reasonably regular, but not perfect, in Classical Old Norse (Nygaard 1905: 155–160, cf. Ottosson 1992: 66). Against this argument one can point to a case like the middle voice in Classical Greek, which on the basis of alternation with free reflexive pronouns could be argued to incorporate a reflexive argument semantically. However, the Greek middle is expressed by a set of inflectional endings which cannot possibly be analysed morphologically as containing an enclitic reflexive element. More generally, there exist certain morphological categories that reduce the dyadicity of a verb. At least in many cases, these categories are uncontroversially expressed by means of an affix rather than a clitic. One such category is the passive, and another and more pertinent one is that of antipassive, which has an intransitivising function. In view of such phenomena, an independent element of syntax might be more persuasively posited in cases where the reflexive element cannot be analysed as an argument of the verb it attaches to, but is instead the subject of a clause which is an argument of that verb (ECM or AcI subjects), or where it has syntactic dependents of whatever kind. Pursuing that line of argumentation, the reflexive element in (4) would be an instance of Exceptional Case Marking, or raising to object.

 Kjartan Ottosson

(4) Austmaðr kvezk á þat hætta mundu the Norwegian says-himself on that venture would ‘The Norwegian says he will venture that’ An alternative analysis of cases like this, however, is that the subject of the argument clause is not expressed by the reflexive element, but by a phonologically empty category PRO. The availability of that analysis may be taken to neutralise the argument in question. On closer look, constructions like (4) actually provide a strong argument against the clitic analysis for Classical Old Norse and for the affix status of the element. This can be most clearly seen by looking at the alternation often found between the Middle Voice and active plus free reflexive, as illustrated in (5). In (5a) we see agreement in case between the subject sik of the object clause and its syntactic dependent, the predicative Noun Phrase, both of these constituents showing accusative case. In (5b), on the other hand, where one might hypothesize that the reflexive has been cliticised to the finite verb form, the predicative NP actually appears in the nominative, agreeing with the subject of the finite verb kváðusk. Thus, -sk is not visible to agreement. In the terminology of Hopper and Traugott (1993), the reflexive has been decategorised. (5) a. Svasi kvað sikACC vera þann FinninnACC – ‘S. said he was that Lapp’ b. þeir er biskuparNOM kváðusk vera – ‘who said they were bishops’ It should be added that clitics are not by nature excluded from triggering agreement. Clitics may even trigger agreement when a free argument doesn’t, as in the well-known case of French object clitics, which trigger agreement on the perfect participle, unlike full NPs. Another argument from Classical Old Norse for the suffix status of the reflexive element is morphological, namely that -mk regularly attaches to a base that is different from the active form (cf. Zwicky & Pullum’s 1983: 504 Criterion C). The most striking cases are found in the strong verbs, as illustrated in (6). In the present indicative of the verb bjóða ‘offer’, the active form has i-umlaut which is absent in the middle form of the 1st sing., although the 2nd and 3rd sing. has umlaut in both active and middle. In the past indicative of the same verb, the middle has a different ablaut grade than the active, the grade otherwise restricted to the plural forms in both active and passive.

(6) a. ek bjóðumk ~ ek býð mik, pres. ind. – ‘I offer myself ’

b. ek buðumk ~ ek bauð mik, past ind. – ‘I offered myself ’ (cf. plural vér buðum(sk)) c. ek kϞllumk ~ ek kalla mik, pres. ind. – ‘I call myself ’ A related argument against clitic status has to do with u-umlaut as in (6c) (cf. Enger 2003: 91). u-umlaut is a quite regular phonological process in Old Nordic, as it affects every a followed by an unstressed u, of whatever origin, within the same word. However, it is not to be expected that a clitic should trigger such a change in its host. The

The Old Nordic Middle Voice in the pre-literary period 

putative clitic must be analysed as -umk rather than -mk synchronically, because a host ending in -u would be otherwise unattested and thus replicate the situation in the strong verbs. The suffixed definite article provides an interesting contrast, as it never triggers u-umlaut in its host, e.g. in stað+num, dat. sing. of staðr ‘place’. Furthermore, in hiatus the final vowel of the host repels the initial vowel of the article, as in staða + inna > staðanna, gen. plur. of staðr. By contrast, the final vowel of the active stem is systematically repelled by the u of -umk: kalla+umk > kϞllumk, 1st sing. pres. ind. of kalla ‘call’. In this respect, -umk acts like any other inflectional suffix, e.g. kalla + ið > kallið, 2nd plur. pres. ind. of kalla. If the Old Nordic reflexive element indeed is an affix, the question is at what point in the development it became one. Therefore it is necessary to look closely at the possible affix properties of -s(k) and -mk at various stages of the development. Another question concerning cliticization which has been neglected in the literature is that regarding the nature of the cliticization which must have taken place at some stage. As it turns out, attention to this question gives interesting results which throw light on the ultimate origin of the category.

1.4

Organisation of the present paper

This paper is organised as follows. After a brief discussion of the sources, I will go through the individual components with the just discussed questions of grammaticalisation and cliticization in mind. The components of grammaticalisation assumed by Hopper & Traugott will provide the organising principle, as they apply in the areas of morphology, semantics, syntax, and phonology, with a discussion of the lessons to be learned from a related construction, the non-middle -mk construction, intercalated after the discussion of the syntactic aspects. Questions concerning cliticization will be discussed in their natural places, especially in sections 3.3 and 5.

2. The sources of pre-literary Old Nordic For the period before the earliest Old Nordic manuscripts, we have mainly two types of sources. With regard to East Nordic, there is a large number of runic inscriptions, most of them very short. The majority of these inscriptions are Swedish and most of them are from the 11th century. I have (in Table 2 in 3.1) collected the examples from the standard editions in Sveriges runinskrifter and Danmarks runeindskrifter, supplemented and corroborated by information from Peterson’s (1994) Runic Swedish concordance, as well as from the data base Rundata. For West Nordic, there also exist runic inscriptions from before 1100, but the few examples of Middle Voice found there don’t add much to our knowledge: N 235 Bore V antaþisk ‘died’, N 252 (b)ar(i)þis(k) ‘fought’. Thus, almost the only sources for West

 Kjartan Ottosson

Nordic are orally transmitted poetry, preserved in manuscripts from the 13th century on. This poetry is of two kinds, Skaldic poems and Eddic poems. Most of the Skaldic poems are attributed to poets about whom we know when they were active. The majority of the early Skaldic poems are in a strict metrical form (mostly in the metre “dróttkvætt”), with alliteration, internal rhyme and a rather welldefined number of syllables in each line. This rigid form has been assumed to have preserved the poems from corruption, although some striking examples of the contrary are known (Helgason 1953: 107f). Their complicated style often makes them hard to interpret (cf. Frank 1985: 165–169). One type of skaldic poems is court poetry, the authenticity of which is generally not called into question (Helgason 1953: 117, Frank 1985: 173, cf. Fidjestøl 1982: 81f). Only in a few cases, the primary sources attribute the same poem to more than one poet of the same period. However, some old court poems in metres less strict than “dróttkvætt” may well have been subject to some linguistic updating. Helgason (1953: 119) claims this explicitly for Haraldskvæði by ÞorbjϞrn hornklofi, and other important poems where this is a distinct possibility include Ynglingatal by Þjóðolfr of Hvin and Hákonarmál by Eyvindr skaldaspillir (cf. Helgason 1962). Concerning single stanzas (lausavísur), attributed especially to protagonists of the Sagas of Icelanders, there are serious questions of authenticity which radically affect dating (e.g. Frank 1985: 172–175), in addition to occasional disagreement in the primary sources concerning author (Frank 1985: 164, Helgason 1953: 143). Thus, Helgason (1953: 145) suspects various chronological layers in the poetry of Egill Skallagrímsson, and some of Gísli Súrsson’s poems show suspect Christian influence (Helgason 1953: 152). More radical proposals in recent decades argue that all the poetry attributed to a Viking Age love poet like Kormákr are actually from the 13th century. Studies of the language of skaldic poems, however, generally show clear linguistic development according roughly with the traditional dating. My corpus of skaldic poems is a selection from the standard edition by Finnur Jónsson, Den norsk-islandske Skjaldedigtning 1912–15, consisting of well preserved and uncontroversially interpreted examples from both court poetry and single stanzas. I have included all poems dated by Jónsson to the 9th and 10th centuries, and supplemented them by all the poems, from ca. 1010–1040, of the prolific poet Sigvatr Þórðarson.1 I have (in the Eddic poems as well as the skaldic poems) excluded examples only preserved in young and poor manuscripts and such where the manuscripts show serious disagreement, as well as forms that exist in no manuscript, but are reconstructed by editors. The interpretation of all skaldic examples passing this test was then compared with the interpretation of Jónsson’s most systematic critic, Ernst A. Kock (1923–1944, Den norsk-isländska skaldediktningen 1946–1950). Those examples were excluded where there is disagreement between the two, unless it seemed clear to me that one of them had the best interpretation. This selection process had to involve considerable 1. I have omitted Sendibítr by Jórunn skaldmær, as Fidjestøl (1982: 180f) argues that it might be from after the Viking Age.

The Old Nordic Middle Voice in the pre-literary period 

subjective evaluation, but this is not the place to discuss the details concerning the application of the criteria or enumerate the approximately two hundred examples. The Eddic poems are anonymous and hard to date. On the other hand, they are generally much easier to interpret than the Skaldic poems, due to their simpler poetic devices. As their metrical form is less rigid than that of the Skaldic poems, they ought to be more liable to modification in oral tradition. It seems to be a realistic assumption, supported by some concrete textual evidence (Helgason 1953: 30f), that the Eddic poems were subject to false memory, omissions, poetic revisions and mutual influence at the oral stage. Until recent decades, when some scholars have taken an oral formulaic view of the Eddic poems, considerable agreement on the relative chronology of most of the poems had been reached (cf. Harris 1985: 93). As in the case of skaldic poetry, the Eddic poetry preserves some old linguistic features which are hard to attribute to archaization. In view of the more fluid nature of the text of the Eddic poems (as well as some of the skaldic poems), however, it is wise to be careful in dating innovations based on a limited number of occurrences in these poems. In selecting Eddic poems for excerption, I have chosen to follow Sveinsson (1962), limiting myself to those that he assumes cannot be much younger than ca. 1000. A further requirement was that the manuscript preservation should be as reliable as possible. Thus, my Eddic corpus includes the following mythological poems: VϞluspá, Hávamál, Vafþrúðnismál, and Grímnismál, as well as the more controversially dated Skírnismál, Lokasenna, and Þrymskviða, whereas Hárbarðsljóð has been excluded because of its unusually fluid text (Sveinsson 1962: 292f). The following heroic poems have been excerpted: VϞlundarkviða, Atlakviða, Hamðismál, Helgakviða hundingsbana II, Brot af Sigurðarkviðu, GrottasϞngr, Helgakviða HjϞrvarðssonar, Reginsmál, and Sigrdrífumál, but the parts of the last three poems that might be younger (Sveinsson 1962: 446, 460f, 466f) have only been given limited consideration. The edition in Norrøn fornkvæði 1867 was excerpted, and compared to Helgason’s edition (Eddadigte 1951–1952) where available, otherwise to the edition by Neckel and Kuhn (Edda 1983). All suggestions for textual emendation listed in these editions were noted, and examples where the emendations indicated corrupted text were excluded. In interpreting the forms used, I have taken into critical consideration Nygaard’s 1865 and Wisén’s 1865 description of the syntax of the Eddic poems.

3. The morphological development 3.1

Paradigmatization 1: Case syncretism

One aspect of the morphologisation of the middle voice under Hopper & Traugott’s scenario is the generalisation of one morphological element for accusative and dative (and even genitive in a few cases). The view that has been standard, by virtue of being espoused by Noreen in his 1923 handbook (Noreen 1923: 367), is that the -s variant is

 Kjartan Ottosson

phonologically derived from *-sR, which is an enclitic form of dative séR of the reflexive. This form supposedly underwent assimilation to -ss and subsequent shortening. In view of the fact that -sR(-) was always assimilated to -ss(-) in the Nordic languages, the phonological aspect is unproblematic. The claim is that the -sk variant was originally selected by verbs governing accusative, while those governing dative had -s or -ss. This dichotomy was, under this view, obliterated later, by generalisation of one of the variants. Other scholars, most notably the syntactician Nygaard (1905: 155, cf. 1865: 50f), assumed that the middle had universal -sk from the time when the category was formed, and that -s is derived from -sk. Although one cannot assume a priori that the distinction facts fall out in the same way in the 1st sing., this has been commonly assumed in the literature, and the evidence for non-distinction seems about as strong there.2 Looking at the sources, we find that even in the oldest preserved manuscripts of West Nordic, there are no traces of a difference according to original case selection by the verb (cf. Ottosson (1992: 90–92, 101). The way in which the Viking Age poetry is preserved in the manuscripts does not in general allow any conclusions on the original form of the element in individual instances, as any differences of this kind, mostly variation between -sk and -s, are so minor that they would have been levelled out to accord with current usage. For what it is worth, the preserved poems even in the oldest and best preserved manuscripts show the -s variant just as sporadically as the prose texts in the oldest manuscripts. One might think that rhymes in Skaldic poems, especially its characteristic metre “dróttkvætt”, could provide secure examples, but the metrical rules are such that certain evidence is excluded. As often as not, only the first post-vocalic consonant counts (e.g. Sveinsson 1962: 124f), and the desinences do not participate in the rhyme (e.g. Helgason 1953: 17). One might think that the reflexive element is better preserved when a negation (-a, -at) is cliticised to its right, so that a k is lacking when the verb governs the dative. There is reasonably good evidence from around 1000 for the 2nd sing. past tense þóttiska ‘you didn’t think of yourself ’ (Hárbarðsljóð 26, Lokasenna 60), where -a is the negation clitic. It is an open question, however, what the case of the reflexive element should be with this verb.3 Indirect, but highly suggestive evidence comes from the nonmiddle -mk forms discussed in section 6, forms like erumka ‘it is not for me’. As the East Nordic runic inscriptions are preserved in original, they should be more instructive than the West Nordic poetry in this respect, see Table 2. Determining the case properties of the individual verbs involved is rather problematic, though. Case assignment is quite idiosyncratic and unpredictable in classical Old Norse. Which case is assigned to a particular argument is to a large extent dependent on individual verbs rather than determined by general rules. Therefore, one can only feel reasonably confident of the original case of the reflexive element if one knows which case a syntactically 2. For a reasonably full review of the history of research into the preliterary history and prehistory of the middle forms, see Ottosson (1981: 4-20). 3. In the carefully reconstructed form *letskaðu (from letja) ‘you aren’t discouraged’ (Lokasenna 47), the verb governs accusative, so that -sk- would be just as expected.

The Old Nordic Middle Voice in the pre-literary period 

Table 2. Middle Voice in the East Nordic runic inscriptions (-sk to the left, -s to the right) DENMARK Jylland kuask ‘said of himself ’ 131 Aars, barþusk ‘fought’ 66 Aarhus SWEDEN Småland itaþisk ‘died’ 46 Erikstad

[:h]uilis ‘rests’ 6 Slesvig (carved by a Swede)

etaþis ‘died’ 29 Ingelstad, ...aþis (for etaþis) ‘died’ 27 Berga cemetery

Västergötland

bþiþus (for barþus) ‘fought’ 40 Råda

Östergötland

eatatis ‘died’ 81A Högby, aitaþis ‘died’ 81A Högby, etaþis ‘died’ 155 Sylten, furs ‘perished’ 145 Dagsborg, -ublas (for kumblas) ‘be buried’ 174 Skärkind

Gotland

entaþis ‘died’ 207 Stenkumbla

Södermanland

entaþis ‘died’ 9 Lifsinge, antaþis ‘died’ 33 Skåäng, entaþus ‘died’ 34 Tjuvstigen, anlaþis (for antaþis) ‘died’ 65 Djulefoss entaþis ‘died’ 85 Västerby, entaþis ‘died’ 216 Aska, furs ‘perished’ 335 Ärja

Uppland

etaþis ‘died’ 518 V. Ledinge, antaþis ‘died’ 136 Broby, itaþis ‘died’ 358 Skepptuna, entaþis ‘died’ 140 Broby, antaþus ‘died’ 153 Lissby, furs ‘perished’ 201 Angarn, furs ‘perished’ 1016 Fjuckby, furs ‘perished’ 349 Odenslunda, fors ‘perished’ 363 Gådersta, kas at ‘liked’ 226 Bällsta

Hälsingland hafsk... rit furiR ‘has right for himself ’ Forsa ring

 Kjartan Ottosson

Table 3.4 Expected case selection in East Nordic runic inscription and the form of the

suffix

berjask

acc (?)

1 sk

kveðask endask(/andask) farask hvílask kumblask hafask fyrir getask at

acc acc (?) dat acc acc (?) dat ?

1 sk 1 sk

1s 17 s 6s 1s

1s 1 sk 4 sk

1s 27 s

and semantically comparable free object of the verb in question has. Because the runic corpus is so small, however, we have little direct evidence about the case selection of verbs at this language stage. We thus have to rely heavily on comparative evidence from later stages and from related languages, with all its inherent uncertainties. It turns out that even when using all possible sources of evidence it is often unclear what the case on the reflexive element was at the time when it attached to the verb. In Table 3 (above), I show the results of an analysis of the case selection of the middle verbs in the East. We must conclude from Table 3 that no correlation between case selection and affix form can be seen at this language stage. In spite of all the inherent uncertainty, one may be reasonably certain that case selection did not play any significant role in determining the form of the affix in the oldest Nordic sources. Notice for instance that berjask ‘fight’ has both variants, as does endask ‘die’, with a predominance of the unexpected -s. Further, both hvílask ‘rest’ and (probably) kumblask ‘be buried’, show a suffix form which goes against the robustly established case selection. It should also be kept in mind that accusative is much more common than dative for verbal objects in general in Old Nordic. 4. The verbs are given in normalised Old Icelandic. Although neither East nor West Nordic sources have active berja ‘hit’ with an independent object in the specialised meaning seen in berjask, it is customary (e.g. Nygaard 1865: 54) and reasonable to assume that the -s(k) element in berjask is historically derived from the regular accusative object with berja. Swedish runologists (e.g. Peterson 1994) take andask to be just a phonological variant of endask, a verb which has an active counterpart with an accusative object. It might be suggested that andask actually is an independent verb identical to West Nordic andask, a verb where the original case of the reflexive element cannot be determined with any degree of certainty. However, andask would in that case only account for four or five examples (all with -s) of the total of eighteen. The collocation hafask rétt fyrir ‘have right for oneself ’, has a parallel in Old Norse hafask fyrir with concrete objects (cf. 5.1), where the original case can be assumed to have been dative, and in Modern Icelandic hafa rétt fyrir sér ‘be right’. The origin of the collocation einhverjum gezk at einhverju is so unclear that it is futile to speculate.

The Old Nordic Middle Voice in the pre-literary period 

Unlike the case distribution evidence, the geographical evidence for the distribution of the -sk and -s variants at the earliest visible stage gives a quite clear picture. A quick glance at Table 2 suffices to show that at this early stage, the -s variant is restricted to Sweden, whereas Denmark and the southern and northern periphery of Sweden (Småland in part, Hälsingland) show -sk. The natural conclusion is that the -s variant is secondary and spread from central Sweden. The Viking Age Nordic loanwords in Old English (cf. Björkman 1900–1902), more specifically verbs which incorporate the element -sk, tend to confirm what we learn from the runic inscriptions. The verb busk ‘prepare (oneself)’, in particular, is universally seen as derived from the Old Nordic verb búask, and the majority opinion seems to be that bask is also derived from Old Nordic baðask ‘bathe (oneself)’ (Onions et al. 1966, Simpson & Weiner 1989). Although the Danes were most numerous among the settlers in England during the Viking Age, there were also many Norwegians. As we have just seen, Danish had -sk in the Middle Voice in the Viking Age, and the same goes for Norwegian until the 13th century. Swedes were not well represented among the Nordic settlers in England.5 The natural conclusion from all of this must be that -sk is universally the oldest form of the Nordic suffix. It is reasonable to assume that the -s variant represents a later phonological simplification, spreading from central Sweden. The evidence does not lend support to the assumption that there was ever an enclitic form corresponding to the dative variant, -sR (>-s). The case against a dative variant in the 1st sing. is considerably stronger, based in part on the non-middle -mk construction discussed in section 6. The dative variant may have been barred from surfacing as a clitic. It is also quite possible that the dative and the accusative had the same enclitic form already at a stage even earlier than the one seen in the runic inscriptions, and even before the oldest skaldic and Eddic poems (9th century). That non-distinction of accusative and dative clitics may have held at the Proto-Nordic stage. If that is the case, this may have been part of the process of cliticization. But whether that cliticization process had anything to do with the “grammaticalisation” (or regrammation) is another story.

3.2

Paradigmatization 2: Syncretism of person/number

One component of Hopper & Traugott’s grammaticalisation account is the generalisation of -sk from the 3rd person and infinitive to all person-number forms except 1st sing. in Old West Nordic, which had -mk from the 1st person pronoun in reflexive use. In East Nordic, there is no evidence that a form based on the personal pronoun ever existed in the 1st sing. Reflexive pronouns are otherwise only used with 3rd person and infinite forms in the Nordic languages, with personal pronouns standing in for 5. There are other cases, however, where the reflexive element is not incorporated. The best known example is thrive, which must derive from þrífask, but there is also addle from the West Nordic deponent Ϟðlask (Simpson & Weiner 1989, Onions et al. 1966).

 Kjartan Ottosson

reflexives in 1st and 2nd person. It is reasonable to assume that a clitic variant was only available in 3rd person (from sik) and in 1st sing. (from mik), and that -sk was subsequently generalised to fill the gaps in the paradigm. This is the view advocated by Nygaard (1865: 50f) and Wisén (1883: 378f), who postulate a stage where only the free personal pronouns were used in the 2nd person and in the 1st plural while cliticised -sk had gained common currency in other forms. Both Nygaard and Wisén think that the extension of -sk into the other categories only started after non-reflexive semantic variants had emerged, that is, in modern terms, that the grammaticalisation, or more precisely, the regrammation process had started. Other scholars such as Specht (1891: 44–55) and Kock (1919) had pressed the case for phonological development, but in spite of considerable imagination regarding sound development and interpretation of individual scribal forms, they had to resort to considerable analogical extension, not least for the 2nd plural. The paradigmatization view adopted here is supported by the fact that no cliticised object forms of other personal pronouns than the 1st sing. are attested in Nordic or Proto-Nordic. One may speculate that the sound shape of the other relevant personal pronouns was not as suitable for cliticization.6 In investigating the emergence of the full paradigm, the Early Nordic runic inscriptions are of no use, as only 3rd person forms and infinitives are attested in them. The old poems are our only source of evidence on this count. Starting with the 2nd person, singular and plural, it must be conceded that there are not too many examples found in the oldest poems, and none of them are of the most reliable kind. The most reliable ones for the 2nd sing. are given in (7) below, and for the plural in (8).7 On balance, it does not appear to be reasonable to call into question the existence of such forms in pre-literary Old Nordic. Nygaard (1865: 51) claims that there are “still” relatively few examples of the 2nd person in the Eddic poems. To the extent that this generalisation is true for the oldest Eddic poems, one might think that the scarcity of such forms might be due to the subject-matter. What is interesting in this connection is whether there is a higher proportion of active with free pronoun in reflexive use in the 2nd person than in the 3rd in the poems. Further scrutiny of the available data might yield some indication whether this is the case.

6. An enclitic version of the 2nd sing. pronoun þik should thus have been *-þk, the 1st plur. pronoun oss could only cliticise as the ambiguous -s, while it is not clear to me what to expect from the cliticization of 2nd plur. yðr. As for the dual forms, both 1st person ok(k)r and 2nd person yk(k)r would probably have yielded -k, homophonous to the 1st sing. subject clitic. 7. The example from Haraldskvæði is the only one found in Skaldic poems before 1000. Additional examples from Eddic poems in my data base are: setzku niðr Skírnismál 29, létztu eigi mundu Lokasenna 9, þóttiska þú Lokasenna 60, ef afli treystisk Helgakviða HjϞrvarðssonar 22, sakask eigi þú Helgakviða HjϞrvarðssonar 33, Huggastu, Sigrún! Helgakviða hundingsbana II 29, þú sjá þykkisk Helgakviða hundingsbana II 41, kømstu Sigrdrífumál 10. The only additional example for the plural is it reiðir vegizk Lokasenna 18.

The Old Nordic Middle Voice in the pre-literary period 

(7) a. þykkisk skil vita – ‘you think you have certain knowledge’ (ÞórbjϞrn hornklofi Haraldskvæði 18)

b. Atla þóttisk þú stríða – ‘you thought you caused Atli sorrow’ (Hamðismál 8) c. Hví þú þá, Gagnráðr, mælisk af golfi fyr? – ‘Why do you then speak for yourself from the floor, G.?’ (Vafþrúðnismál 9) d. nálgastu mik – ‘approach me’ (Grímnismál 53)

(8) a. it Gymir finnizk – ‘you and Gymir meet’ (Skírnismál 24)

b. þrøngvizk ér of ungan... gram – ‘you throng around the young king’ (Sigvatr Lausavísur 13) Going on to the first plural, there are rather few examples in the oldest poems, some with dual, others with plural reference, and they tend to be less than optimally preserved. Some of the most reliable ones are given in (9).8

(9) a. mælumk í sessi saman – ‘let us [two] converse in the seat’ (Vafþrúðnismál 19)

b. BϞrðumk vér – ‘we fought’ (Egill Lausavísur 22) We may conclude from the preceding that a full paradigm for the Old West Nordic Middle Voice existed as early as the 9th or 10th century. One cannot, however, exclude the possibility that the forms of the 2nd person and the 1st plural were not as much used as the original 3rd person and 1st sing. forms, and that the independent pronouns were relatively more used than in these latter cases. Furthermore, it is clear, unlike what we concluded on the case syncretism in 3.1., that there was a change involving the expansion of -sk to the 2nd person and the 1st person sing., as discussed by Hopper & Traugott (1993). However, it seems reasonable to assume that the extension of the reflexive pronoun occurred later than the original change in content, the regrammation.

3.3

Morphological idiosyncracies in the 1st singular

For the 1st sing. of the Middle Voice, the poems are our only source, as the Viking age runic inscriptions contain no examples of this category. As already mentioned (1.3) the 1st sing. in the oldest preserved West Nordic manuscripts (mid-12th century on) shows forms which deviate markedly from common patterns of the inflectional system of the verb. This is most striking in the indicative of the strong verb, where the 1st sing. shows the same vocalism as the plural. In the examples in (10), we see plural vocalism in the

8. Other examples, with plural reference: finnumk vér Egill Lausavísur 35, with dual reference gørðumsk (possibly 1st sing.) Hamðismál 28, báðir vit komumk Skírnismál 10, hittumk Helgakviða HjϞrvarðssonar 22.

 Kjartan Ottosson

strong present indicative, in the past indicative in (11). The examples in (12) illustrate similar deviations in 1st singular forms of weak verbs.9 (10) a. búumk til vígs – ‘I prepare myself for a fight’ – contrast active bý (mik) (Egill Lausavísur 29) b. ráðumk þér – ‘I advise you’ – contrast active ræð (Hávamál 112) (11) Eigi brugðumk – ‘I did not fail’ – contrast active brá (Sigvatr BersϞglisvísur 7) (12) a. BϞrðumk einn – ‘I fought alone’ – contrast active barða (Egill Lausavísur 40) b. ek sjá þykkjumk – ‘I think I see’ – contrast active þykki (Helgakviða hundingsbana II 40) There is also indirect evidence for plural ablaut stage in the past tense of strong verbs from the 9th century, in that the non-middle -mk construction, to be discussed in 6, presupposes the existence of strong past forms like buðumk. The conclusion from all of this must be that the classical paradigm (with -sk/-mk as an affix rather than a clitic) existed already in the 9th century. Reconstruction of the earliest history and prehistory of the 1st sing. forms provides interesting indirect evidence on the question of cliticization in particular. Most categories of 1st sing. forms in most inflectional classes can be derived by regular sound change, supported by parallel development in other words, as argued already by Wisén (1883), cf. Noreen (1923: 367f, cf. 361, 355f), on the the subjunctive Benediktsson (2002a, b). However, we can only be certain to get the attested forms “for free” if the -mk element became a suffix before the Old Nordic stage. The first set of arguments for that claim concerns cases where active forms have -a while the corresponding middle forms have -u-. Although the majority of the active 1st sing. forms ends in -a in Old West Nordic, these same forms ended in long ō in Proto-Nordic: the subjunctive present and past, past indicative of weak verbs in general, and pres. ind. of weak ō-verbs (cf. Krause 1971: 121–123, 125–128, Antonsen 1975: 22). The standard assumption is that ō changed into -a in final position, but changed into u before m by regular sound change. If the element -mk became a suffix before the ō changed into -a, influence of 9. Other examples of the 1st sing. in my database are the following. From Skaldic poems up to ca. 1000: minnumk Goþþormr Hákonardrápa 7, þóttumk Egill Lv 43, undrumk Egill Lv 43, minnumk Kormákr Lv 10, hugðumk Kormákr Lv 26, forðumk Kormákr Lv 62, reiðumk Gísli Súrsson Lv 19, þóttumk Þórarinn Máhlíðingavísur 8, þóttumk Víga-Glúmr Lv 5, tegumk BjϞrn Breiðvíkingakappi Lv 1, hræðumk Hallfrøðr Lv 28; from Sigvatr Þórðarson: óðumk Austrfararvísur 2, spurðumk Aust 4, hræðumk Aust 5, séumk Aust 6, þóttumk Vestrfararvísur 7, minntumk Lv 21, minnumk Lv 24. From the Eddic poems: þóttumk Hávamál 47, 99, létumk Háv 106, ráðumk 18 other places in Hávamál, óumk Grímnismál 20, sjámk Grímn 20, hétumk twice Grímn 46, one each Grímn 48, 54, vilnumk VϞlundarkviða 31, þǿttumk Hamðismál 21, komumk Skírnismál 11, óumk Skírn 16, þykkjumk Lokasenna 54, sjámk Helgakviða hundingsbana II 16, þykkjumk Reginsmál 8, hræðumk Reginsmál 9, hugðumk Fáfnismál 16.

The Old Nordic Middle Voice in the pre-literary period 

the following m could safely be invoked to explain the emergence of u in the 1st sing. middle forms in the categories in question. If, on the other hand, -mk remained a clitic after ō changed to -a and -u-, which presumably happened in the transition from Proto-Nordic to Old Nordic, some additional explanation of the development would be needed. The problem is essentially that as long as -mk is a clitic, we would not expect its initial m to affect the quality of the final vowel of its host. Even if the unexpected had happened, resulting in a string of active form and a reflexive clitic such as bjóðu + mk (pres. subj. of the strong verb bjóða ‘offer’), such strings would probably have been unstable and tended to replace the deviant active component with the regular form used in the absence of an attached clitic, e.g. bjóða + mk. If -mk changed from clitic to affix in such a remodelled string in Early Old West Nordic, it is hard to tell whether the m would have been able to round the a to u. This intricate problem cannot be pursued further here, but I think it is safe to say that the -umk ending corresponding to active -a favors an early date for the transition from clitic to suffix. Note that the numerical superiority of these cases makes analogical remodelling based on other forms where -umk arose by regular sound change unlikely. The second set of arguments for early transition of -mk to a suffix concerns cases where the middle form retains Proto-Nordic -u, which has disappeared by analogy in Old Nordic. This is the case in the pres. sing., where strong verbs and weak (i)ja-verbs ended in -u in the active in Proto-Nordic (cf. Krause 1971: 121f, Antonsen 1975: 22), so that attaching -mk yielded -umk without any complications. Moreover, forms without i-umlaut in the strong verbs, e.g. bjóðumk, are derived automatically, as there is no i present in the ending at any stage. The Old Nordic active forms with umlaut, e.g. býð, are based on analogy with 2nd and 3rd sing. forms. It is unclear when the -u of the active was lost, but as there are no signs of it in any early Old Nordic texts, this should have happened in the transition from Proto-Nordic to Old Nordic. In this case, the attested middle forms must be remnants from the Proto-Nordic stage, which were unlikely to survive in a string of active form and enclitic reflexive and could not arise by any regular sound change in Early Old Nordic.10 This leaves the past indicative of strong verbs, forms like buðumk, which can only have emerged by analogy with other forms which by regular sound development had come to look like they contained the plural stem. It is reasonable to assume that a form like buðumk, with u instead of au as in the other persons of the singular, got its vocalism 10. Someone might want to allow for the possibility that forms like bjóðumk were constructed on the analogy of the pres. ind. of weak ō-verbs, where 1st sing. forms like kϞllumk had the same vocalism as the 1st plur. kϞllumsk or kϞllumk. I find it quite unlikely that these weak verbs, although they were more numerous than any other class, could have affected the strong verbs in this way, because strong verbs tend to be quite common individually, and also because u-umlaut (unlike i-umlaut and ablaut) is phonotactic in nature, occurring as it does before any unstressed u. The only pres. ind. forms not yet discussed are those of ē-verbs, but these verbs are so rarely found in the middle voice that analogical pressure can be assumed, as traditionally done (e.g. Noreen 1923:367).

 Kjartan Ottosson

from the plural buðumsk, on the model of the present indicative with bjóðumk, plur. bjóðumsk. It is possible that before this analogical remodelling, the u in the ending of buðumk had been introduced from other forms of the 1st sing., yielding the unattested type *bauðumk. In any event, strong verb forms like buðumk, which are found from the earliest Skaldic poems on, thus presuppose a rather complex prehistory of developments within an inflectional paradigm. If we want to allow some time for that development to happen, the original change to affix status must be assumed to have taken place in Proto-Nordic. The 3rd sing. forms in the present indicative, ending in -sk, present an interesting contrast to the 1st sing.. The active ending in -þ found in Proto-Nordic runic inscriptions starts being replaced with one in -R, later -r, in the 7th century (Krause 1971: 54, 122, 140f, 164f; Antonsen 1975: 22, 85–88). The attested middle ending most probably derives from the younger variant, judging by the 2nd plur., which retains a dental stop in the ending (-dsk, -tsk, mostly normalised to -zk in editions) until the early 13th century in Old West Nordic. The evidence from the 3rd sing. thus suggests that the reflexive element became an affix in late Proto-Nordic, not earlier than the 7th century.

4. Semantics: regrammation Turning now to the semantic development of the middle voice, it must be stressed that the element subject to grammaticalisation – that is, sik, and mik in 1st sing. in West Nordic – had very little semantic content from the beginning. Its original function was to mark coreference with another argument. Thus, we have here a rather clear instance of regrammation, in Andersen’s (2005) terms, a change from one grammatical meaning to another. As for the semantic types of the middle in the earliest sources, one is not surprised to find the reflexive content retained. We also see semantic generalisation in a certain sense, less properly described as bleaching than as the development of new semantic variants. As a matter of fact, all the main semantic variants of the manuscript period are already found occurring in the oldest sources. Contrary to claims by Jónsson (1921: 273–276), no clear semantic development can be seen in the oldest poems. However, the relationship between the verb and the syntactic position where the reflexive element would be, if independent, can be looser than in classical Old Norse, as we will see. In this section, then, we shall mostly be concerned with cases restricted to direct arguments of the verb, i.e. arguments receiving their case from the verb rather than from a preposition or in other ways. Cases involving more indirect relationships will mostly be discussed in the next section. Before proceeding, it must be noted that the semantic development of the middle voice involves a large measure of lexicalisation. In other words, one cannot confidently predict whether a specific bare active verb has a corresponding middle variant, and what the meaning of such a counterpart is. One aspect of this lexicalisation is the

The Old Nordic Middle Voice in the pre-literary period 

existence of middle verbs without any bare counterpart, e.g. Ϟðlask ‘acquire’ without any hypothetical verb *Ϟðla. Thus, many middle verbs have to be entered into the mental lexicon as independent lexical items. Although it seems reasonable to assume that this lexicalisation increased after the reflexive element became an affix, it must have existed to some degree before that stage. Even in Modern German, which retains the Old High German non-cliticised middle, there are cases where a reflexive collocation has no active counterpart. One may start by stating that all the main semantic types found in Classical Old Norse are already solidly established in the Viking Age. As is to expected in view of the lexicalisation involved, often leading to lexical idiosyncracies, a number of verbs do not fall neatly into any major category. Some of those have been put into the heterogeneous category 3 here, while others had to be ignored. Most examples, however, can be put into one of the five categories discussed here below. 1. Reflexive. The reflexive category, where it is reasonably clear that an agent does something to himself, is at least as well represented in Viking Age Norse as in Classical Old Norse. At the classical stage, most middle verbs with reflexive meaning alternate with a collocation of an active verb and a reflexive pronoun, like in (13), and we may assume that this was at least as true at the pre-literary stage. Most often, the meaning of the middle verb in such pairs is the same as that of the active-reflexive counterpart, as seems to have been the case in (13a-e) and in hvílask in (13f), but sometimes the middle verb has a more specialised (lexicalised) meaning, as in setjask in (13f-g). (13) a. gerðut... vægjask – ‘they did not spare themselves’ (Hallfrøðr vand ræðaskald Óláfsdrápa, erfidrápa 13) b. nú vill dyljask dϞglingr fyr mér – ‘now the prince wants to hide [himself] from me’ (Helgakviða hundingsbana II 13) c. Svá skal frǿkn fjándum verjask – ‘Thus a brave one should defend himself against his enemies’ (Atlakviða 19) d. Fyllisk fjϞrvi feigra manna – ‘fills himself with the life of dead men’ (VϞluspá 41) e. Hreyfðisk enn hϞsfjaðri – ‘the one with grey feathers moved [himself]’ (ÞórbjϞrn hornklofi Haraldskvæði 4) f. þær á sævarstrϞnd settusk at hvílask – ‘they sat down (“set themselves”) to rest [themselves] on a beach’ (VϞlundarkviða 1) g. settisk Ϟrn – ‘the eagle landed (“set itself ”)’ (Þjóðolfr ór Hvini HaustlϞng 2) 2. Reciprocal. The reciprocal semantic type, common in classical Old Norse, is already abundantly attested in the Viking Age sources, as illustrated in (14). This variant seems like an innovation linked to the emergence of the Middle, as there is no evidence that the independent reflexive pronoun was used in a reciprocal meaning in Old Nordic or Proto-Nordic. Thus, classical Old Icelandic uses hverr annan ‘each other’ and not sik

 Kjartan Ottosson

for reciprocals, and the free reflexive sik is never reciprocal in the pre-literary poems any more than in Classical Old Norse.11 (14) a. of drepask kváðu – ‘said they had killed each other’ (Þjóðolfr ór Hvini Ynglingatal 11) b. kvϞddusk – ‘greeted each other’ (ÞórbjϞrn hornklofi Glymdrápa 4) c. hjϞggusk hvárirtveggja heggir mækis eggja – ‘the warriors on both sides exchanged blows’ (Glúmr Geirason Gráfeldardrápa 10) d. Vápnum ok váðum skulu vinir gleðjask – ‘Friends shall make each other glad with weapons and clothes’ (Hávamál 41) e. finnask vígi at Surtr ok in svásu goð – ‘Surtr and the dear gods meet each other in a battle’ (Vafþrúðnismál 17) f. hvar ýtar túnom í hϞggvask hverjan dag – ‘where men slay each other every day’ (Vafþrúðnismál 40) g. ef við einir skulum sáryrðum sakask – ‘if we alone shall reproach each other with insults’ (Lokasenna 5) h. at it reiðir vegizk – ‘that you, angry, fight each other’ (Lokasenna 18) i. þá er þau lint skyldu optar um faðmask – ‘when they should embrace each other tenderly again’ (Atlakviða 40) 3. Other verbs with agent-type subject. Many verbs appearing in the middle voice do not fall naturally into either the reflexive or the reciprocal group, although their subject is to varying degrees agentive in nature, in clear contrast to the anticausative and passive classes. One case concerns berjask ‘fight’, which was most probably reciprocal originally, with a specialised sense of the basic verb berja ‘beat’. Already in the oldest sources, however, the verb is no longer necessarily reciprocal, as it can take a subject in the sing., as illustrated in (15a). Another example where the subject is much farther from the proto-agent is given in (15b), just to give an idea of the range of these sundry agent-type subject verbs. (15) a. barðisk konungr – ‘the king fought’ (ÞórbjϞrn hornklofi Haraldskvæði 7) b. Vitr þóttisk valkyrja – ‘the valkyrie thought she was wise’ (‘thought-ofherself wise’) (ÞórbjϞrn hornklofi Haraldskvæði 2)

11. Nygaard (1905:164f) mentions a couple of examples where the reflexive is used in a reciprocal sense: síðan tϞluðust þeir við konungr ok Finnr ok rǿddu sín á milli Hkr 580.12, þeir rǿddu með sér, hvárt jarl mundi fallinn Hkr 605.4. I think, actually, that this is an instance of something close to the reciprocal, but rightly distinguished from it i.a. by Dal (1966: 74), namely individuation within a group. The prepositions á milli and með probably don’t allow hverr annan, the regular expression of reciprocity, in this context.

The Old Nordic Middle Voice in the pre-literary period 

4. Anticausative / spontaneous event. In Classical Old Norse, the middle voice very commonly has an anticausative meaning (Ottosson 1992: 67f).12 In anticausatives strictly speaking, a transitive verb has its agentive argument not only suppressed as in the passive, but actually conceptually eliminated. Thus, it is possible to modify the event by adverbials meaning “by coincidence”, “of its own accord”, but not “intentionally”. Anticausatives are thus semantically similar to unaccusatives (or “ergatives” in some generative literature), but differ in being morphologically derived. Admittedly, the difference between anticausative, passive and reflexive can be subtle, and in determining which examples are anticausative I have partly been guided by my native speaker intuition concerning that category in Modern Icelandic (Ottosson 1986). In pre-literary Old Norse, an anticausative meaning is abundantly attested already in the earliest poems. In (16a), for instance, we have an agentive transitive verb hrista ‘shake’, but in the middle, only the shaking effects are expressed, and it is understood that the event is brought about by forces of nature rather than an agent in the full sense. The same focussing on the effects is seen in (16b-f), although in some of these cases, participants that are potential agents may contribute to bringing these effects about. In (16g), the warriors contribute to the slaughter by fighting, but they cannot quite control the opportunities for striking blows and whether they prove fatal. In (16h), the infant being born has no control over the situation, but the mother has limited control as well. Similarly, in cases like (16i-k), the lexical meaning of the verb is such that the subject participant does not have control over the outcome but can only try to bring it about. Some verbs allow an agentive reading although the non-agentive one is more common, and supported by the context in a case like (16l). The typical anticausative verb is dynamic rather than stative, and this can be the case also when referring to events where a psychological element is central, as in (16m-n). In some cases, such as the ones in (17a-c) (and 18b), there exists no formally active counterpart to a verb marked with -sk, or the semantics of the active counterpart is quite different. Still, the semantics of the middle verb is basically the same as in examples like those in (16). In such cases it may be more fitting to speak of spontaneous events (cf. Enger 2003: 87f). Finally, it should be noted that psychological verbs often have a less dynamic semantics than verbs referring to more concrete and publicly observable situations. They can be ambiguous between a dynamic and a stative reading, as in (18a), or purely stative (18b). (16) a. hristusk bjϞrg – ‘the cliffs shook/got shaken’ (Þjóðolfr ór Hvini HaustlϞng 16) b. í veltanda vatni lýsask valbaugar – ‘French rings get brighter in rolling water’ (Atlakviða 27)

12. Nygaard (1905: 165-171), who apparently did not have a special term for this, calls this “bleached reflexive meaning” (fordunklet reflexivbeydning). Some of the examples which Ny gaard (1905: 171-174) subsumes under transition to passive meaning might also be categorised as anticausative.

 Kjartan Ottosson

c. eyðisk land ok láð – ‘land and people perish’ (Eyvindr Finnsson skal daspillir Hákonarmál 21) d. fǿrask fjϞll en stóru... í djúpan ægi – ‘the big mountains get displaced into the deep see’ (Kormákr Lausavísur 42) e. á sér hon ausask aurgum fossi – ‘she sees the river get poured on with a muddy waterfall’ (VϞluspá 27) f. á því hvéli er snýsk – ‘on the wheel that turns’ (Sigrdrífumál 15) g. gerðisk mest at morði mannfall – ‘very big slaughter arose in the battle’ (Einarr Helgason skálaglamm Vellekla 24) h. áðr an... gǿðingr myni fǿðask – ‘before a chieftain will be born’ (Hallfrøðr vandræðaskald Óláfsdrápa, erfidrápa) i. komsk Urðr ór brunni – ‘Urd got out of the well’ (Kormákr Sigurðardrápa 4) j. Hlakkar móts... rauðbríkar fremsk rǿkir ríkr... slíku – ‘the mighty warrior gets advanced by such things’ (Einarr skálaglamm Vellekla 16) k. Ráð lukusk – ‘that marriage came about’ (Hallfrøðr vandræðaskald Hákonardrápa 5) l. vanðisk hann... á þat snimma – ‘he got accustomed to that early’ (Hallfrøðr vandræðaskald Óláfsdrápa (1996) 4) m. þat verðr mϞrgum manni, es of myrkvan staf villisk – ‘it happens to many a man, that he errs in an obscure rune’ (Egill Lausavísur 38) n. ǿstusk austkylfur ok... hljópu – ‘the eastern clubs became mad and jumped’ (ÞórbjϞrn hornklofi Haraldskvæði 11) (17) a. niðkvísl... þróttar Þrós of þróazk hafði – ‘Odin’s family had grown (strong)’ (Þjóðolfr ór Hvini Ynglingatal 35) b. jϞrð bifask – ‘the earth shakes’ (Skírnismál 14) c. er hrørask tekr – ‘when he becomes worn out by age’ (Fáfnismál 6) d. þreifsk reiddra øxa rymr – ‘the roaring of brandished axes thrived’ (Þór bjϞrn hornklofi Glymdrápa 7) e. endisk rauðra randa rϞdd – ‘the sound of the red shields was sufficient’ (ÞórbjϞrn hornklofi Glymdrápa 4) (18) a. ek þess minnumk – ‘I remember that’ (Goþþormr sindri Hákonardrápa 7) b. óðusk malmþings meiðar – ‘the warriors were afraid’ (Hallfrøðr vand ræðaskald Óláfsdrápa, erfidrápa 12) 5. Passive. A clearly passive as opposed to an anticausative meaning is found in Classical Old Norse, rarely but solidly attested in popular style (Ottosson 1992: 68, cf. Ny gaard 1905: 171–173). The situation is strikingly similar in the oldest poems, with few but robust examples of clearly passive meaning. It should be noted that a clearly passive

The Old Nordic Middle Voice in the pre-literary period 

meaning is also seen in the East Nordic runic kumblas, see section 3. It must be conceded that it is often hard to distinguish passive meaning from an anticausative one on the basis of context alone. However, it seems reasonably clear from the context in the examples in (19) below that an agent is crucially involved and thus we have a passive rather than an anticausative meaning. (19) a. rakðisk JϞrmungandr at sandi – ‘J. was dragged along the sand little by little’ (Bragi gamli Ragnarsdrápa 16) b. barsk at borði – ‘[the king] was carried aboard’ (ÞórbjϞrn hornklofi Glymdrápa 4) c. hruðusk riðmarar Róða rastar – ‘the ships were stripped’ (Tindr Hallkelsson Drape om Hakon jarl 1) d. Þaðan verða fϞt fyrða... SϞrla rjóðask bjϞrt í blóði – ‘Therefore the shining armour must be reddened with the blood of men’ (Hallfrøðr vandræða skald Hákonardrápa 9) Before leaving the semantic properties of the pre-literary Middle Voice, two points must be briefly mentioned. One aspect of grammaticalisation that is often discussed is divergence, that is, the retention of the targeted element in original form alongside its grammaticalised incarnation. In this particular case, the emergence of the anticausative meaning made a contrast with active verb form plus reflexive possible. Thus, the East Nordic runic fors ‘perished’, Old Norse fórsk from the verb farask, contrasts with fara sér ‘commit suicide’. Further, as we shall see better in the next section, the semantic development follows only in part what we would expect from Kemmer’s (1993) account of the cross-linguistic clustering of semantic subtypes of the middle. In particular, various indirect semantic relationships are quite common, and can be explained from the largely semantically blind cliticization involved in the earliest history of this category in Old Nordic.

5. Syntax 5.1

The syntactic connection between verb and reflexive

In the pre-literary West Nordic sources one can see remnants of a hypothesised earlier stage where the reflexive cliticised rather freely to the verb, without too much regard for the semantic relationship between the two. In other words, not only the objects of the verb can cliticise to it, but rather, the position which reflexive -sk is semantically related to can be structurally more distant from verb. Although this is possible to some degree in classical Old Norse, it seems more wide-spread at this early stage. One aspect of this freedom is that reflexive objects of prepositions can cliticise to the verb, as illustrated in (20). In many such cases, the preposition seems clearly

 Kjartan Ottosson

subcategorised by the verb, forming some kind of a complex verb semantically speaking. The only two prepositions commonly found in such instances with more or less clear reflexive meaning are fyr(ir) and um(b). In many cases, the standard dictionaries (Cleasby & Vigfusson 1874, Fritzner 1886–96, Jónsson 1931) record an alternant with more or less the same meaning where the preposition actually takes an object. Even when such alternants are not known, prepositional origin is generally the most likely, although both fyr and um(b) can be used as verb particles already at this early stage (e.g. ræðr umb við sik ‘deliberates with himself ’ Fáfnismál 33). This is because the middle verb is either not attested without the putative particle or is not known to have a meaning that would give the appropriate contextual interpretation when combined with the particle. (20) a. Hví þú þá, Gagnráðr, mælisk af golfi fyr? (=mælir fyr þér) – ‘Why do you then, Gagnráðr, speak for yourself from the floor?’ (Vafþrúðnismál 9) b. Hyggsk vætr hvatr fyrir – ‘the bold one takes no precautions (thinks naught for himself)’ (=hyggr fyrir sér) (Lokasenna 15) c. Svá nýsisk fróðra hverr fyr – ‘Thus each of the wise ones enquires before him[self]’ (=nýsir fyr sér(?)) (Hávamál 7) d. um at þreifask (=þreifa um sik(?)) – ‘grope around’ (Þrymskviða 1) e. Útan varðk... húsdyrr fyrir spyrjask – ‘I had to enquire outside the house doors’ (Sigvatr Vestrfararvísur 2) f. um sásk – ‘looked around them’ (VϞlundarkviða 4) g. þylsk hann um – ‘he speaks uncessantly’ (Hávamál 17) (21) a. þeir hϞggvask orðum á – ‘they fight each other with words (Reginsmál 3) b. á gengusk eiðar – ‘oaths were broken (on both sides)’ (‘oaths broke each other’) (VϞluspá 26) One can also find reciprocal meaning combined with a preposition, as in (21a) and possibly (21b). Such examples are rare at this early stage, however, possibly more so than in later language, where the use of the preposition á ‘on’ seems to have been generalised in reciprocal use (cf. Nygaard 1905: 164). Further evidence for such cliticization across a preposition is provided by cases where the preposition cannot have been subcategorised by the verb to form a complex verb in any sense. One example of this type is shown in (22). This sort of long-distance cliticization does not seem to be possible in later language, as hafask fyrir as a collocation in this use does not seem to be attested there (although a specialised use is attested, cf. footnote 4). (22) a. hϞfðusk hlífar fyrir (=hϞfðu hlífar fyrir sér) – ‘had (held) shields before themselves’ (Eyvindr skaldaspillir Hákonarmál 11)

The Old Nordic Middle Voice in the pre-literary period 

b. hefisk lind fyrir (=hefir lind fyrir sér) – ‘has a shield before himself’ (VϞluspá 50) Apart from the prepositional cases, there are various other indirect or adverbial relationships between the clitic and the verb. Thus, in (23) we see a kind of possessor raising. (23) hamri fórsk í hǿgri hϞnd (=sér í hǿgri hϞnd) – ‘took a hammer in his right hand’ (Bragi gamli Ragnarsdrápa 15) In some of the examples that follow in (24), involving the verbs for ‘be’ and ‘become’, the reflexive element -sk may seem to function as a genitive to some non-subject element, but the correct generalisation (esp. in view of 24c) seems to involve a sort of ethical dative, or dative of advantage/disadvantage. (24) a. brǿðr tveir at bϞnum urðusk (=urðu hvárr Ϟðrum at bϞnum) – ‘two brothers became each other’s slayers’ (Þjóðolfr ór Hvini Ynglingatal 13) b. Brǿðr munu berjask ok at bϞnum verðask – ‘Brothers will fight and become each other’s slayers’ (VϞluspá 45) c. Gumnar margir erusk gagnhollir (=eru hverr Ϟðrum gagnhollir) – ‘many men are devoted to each other’ (Hávamál 32) d. viðrgefendr ok endrgefendr erusk lengst vinir (=eru hverr Ϟðrum vinir) – ‘those who give and give back are each other’s friends the longest’ (Hávamál 41) Cleasby & Vigfusson (1874) give some examples of verask from Classical Old Norse that show that this form must have been in active use until at least the early 13th century, and it survives in copying into the second half of the 13th century at least.13 This form is also attested in a Norwegian runic inscription after 1100, N 462 from Trondheim: liut--(r): ouk: iohan: erusk: uinir, in normalised spelling Ljót[gei]rr ok Jóhann erusk vinir ‘L. and J. are (each other’s) friends’. On the other hand, verðask does not seem to be attested after 1000. All the “long-distance” relationships we have discussed so far stay within the clause. However, one may find binding relationship across clause boundaries (other than logophoric), as in (25). (25) a. lét... ofan sígask (=lét [sik síga]) – ‘let himself slide down’ (Þjóðolfr ór Hvini HaustlϞng 4)

13. The examples are from three sources. Jómsvíkinga saga in AM 291 4to c1275-1300, originally composed shortly after 1200, has several examples. The older Christian law in AM 334 fol., c1260-70, originally formulated in the years 1122-32, has one example, and the fragment of a homiliary in AM 655 XXI 4to from the first half of the 13th century has one; as the latter is little investigated, it is unclear when the Old Norse version was first committed to writing.

 Kjartan Ottosson

b. ef ek bindask læt brúðar líni (=ek-i læt [PRO binda mik-i]) – ‘if I let myself be adorned with a bride’s linen’ (Þrymskviða 17) c. hví hvetjask létzt (=létzt [þú-i] [PRO hvetja þik-i]) – ‘why did you let yourself be encouraged’ (Fáfnismál 5) d. rétt er, at sá maðr... láti kaupast verk at (=sá-i láti [PRO kaupa verk at séri]) – ‘it is right that that man lets work be bought of him, i.e. works for wages’ (Grágás, Fritzner 1886–96: 2: 422) In (25b-c), the object of the infinitival small clause has cliticised to its governing infinitive, while it is co-referential to, and bound by, the subject of the matrix clause. In (25a), on the other hand, it is the subject of the infinitive which has cliticised “downwards” to the infinitive. One might expect cliticization “upwards” as in the logophoric use to be discussed presently. It is conceivable that it was a factor in the development that cliticising to the embedded infinitive kept the use of the verb as ‘let’ apart from the verb-of-saying use of the same verb láta (e.g. 26a-c). Reflexes of cliticization of both the infinitival object and the infinitival subject are also found in classical Old Norse (Nygaard 1905: 158, Fritzner 1886–96 under láta 6, Cleasby & Vigfusson 1874 under láta IV.2). Most remarkably, in the archaic language of the law code Grágás, one can find a reflexive element bound by a matrix subject cliticising to the infinitive across a preposition, as in (25d). Pre-literary West Nordic also has logophoric use of the reflexive element, as in Classical Old Norse. As noted by Kemmer (1993), logophoric use of the middle voice seems rare cross-linguistically. One may assume that its presence in Old Norse has something to do with its origin as a verbal clitic which was not restricted to argument relations. (26) a. Búinn lézk... at halda... velli – ‘declared himself to be ready to keep the ground (on the battlefield’ (Einarr Helgason skálaglamm Vellekla 21) b. Eigi látask ýtar... fremra mann of finna – ‘the men don’t say they will find a more able man’ (Hallfrøðr vandræðaskáld Óláfsdrápa, erfidrápa 10 c. VϞll lézk ykkr ok mundu gefa – ‘he said he would also give you a field’ (Atlakviða 5) d. Eitt lýti kvazk... á mér þekkja – ‘she said she knew one fault in me’ (Kor mákr Lausavísur 5) e. hafa kveðask lϞg búendr verri – ‘the peasants say they have worse laws’ (Sigvatr BersϞglisvísur 8) f. svá kvazk veita mundu fullting... – ‘he said he would give assistance in such a way’ (Hamðismál 13) g. hafa kvazk hon Helga hylli skyldu – ‘she said she would have the favour of Helgi’ (Helgakviða hundingsbana II 17)

The Old Nordic Middle Voice in the pre-literary period 

h. lýgr hinn, es telsk betri (NOM) – ‘the one lies, who considers himself better’ (contrast telr sik betra (ACC)) (Sigvatr Þórðarson Et kvad om Erlingr Skjalgsson c. 1026 i. Ósnjallr maðr hyggsk munu ey lifa – ‘An unwise man thinks he will live forever’ (Hávamál 16) j. hyggsk aldauðra arfi að ráða – ‘he intends to manage an inheritance without an heir’ (Helgakviða HjϞrvarðssonar 11) It should especially be noted that when the argument which -sk corresponds to has an adjectival element predicated of it, as in (26h), the predicative element does not appear in the accusative as one would expect if -sk kept the accusative case, but in the nominative, just like the matrix subject. This is a point that we will return to presently. The conclusion from this must be that we are not just dealing with the (further) development of a “Middle” category of the kind discussed by Kemmer (1993). Rather, one can see the effects of “mechanic” cliticization, in the style of “special clitics”. Logophoric dependencies are typologically rare in the Middle Voice according to Kemmer. As a matter of fact, Old Norse is the only example she adduces of this phenomenon.

5.2

The relationship with syntactic dependents

One change which Hopper & Traugott (1993) assume is a part of the grammaticalisation process is what they call “decategorisation”, as discussed in 1.2. By this is meant that the clitic lost some of its categorial features, i.a. making it unable to act as a visible syntactic constituent providing inflectional features, or phi-features, to other syntactic constituents. We have already illustrated the inability of -sk / -mk to trigger agreement in Classical Old Norse with verbs of saying in Exceptional Case Marking or AcI constructions (1.3.). This same inability can also be seen in other cases, like (27a) (from Fritzner 1886–96 under nefna), where the suffix -sk has no features to share, so that the argument Skrýmir can only agree with the nominative subject. (27) a. sá nefndisk Skrýmir (NOM) – ‘this [man] called himself Skrýmir’ b. hann nefndi sik (ACC) Óla (ACC) – ‘he called himself Óli’ In pre-literary Old Nordic, we see this same kind of inability of the reflexive element to trigger agreement. We saw this already in 5.1 in logophoric contexts, in (26h), and (28) provides further illustrations. In (29) we see non-agreement in other kinds of contexts. (28) a. einn rammari (NOM) hugðumk Ϟllum vera – ‘I thought myself alone stronger than all’ Fáfnismál 5 b. Fúss (NOM) læzk maðr... at deyja – ‘that man says he is willing to die’ (Sigvatr Lausavísur 22)

 Kjartan Ottosson

(29) a. Hétumk Grímr (NOM), hétumk Gangleri (NOM) (Grímnismál 46, cf. 54) – ‘[I] called myself Grímr, [I] called myself Gangleri’ b. Gerðisk (subjunctive) hilmis HϞrða húskarlar þá jarli... ofvægir (NOM) – ‘then the man-servants of the HϞrðar king would become too lenient towards the earl’ (Sigvatr Lausavísur 18)

6. Excursus on the non-middle -mk construction: 3rd person subject, 1st singular object In addition to the evidence presented in 3.3 for the old age of the idiosyncracies seen in 1st sing. forms of the middle, indirect evidence may be seen in a superficially similar construction involving (as a first approximation) cliticization of -mk to a verb with a 3rd person subject. Such forms are attested from the earliest Skaldic poems on and are also found in Eddic poems, but seem to be rare after the mid-11th century. What is striking about examples of this kind is that instead of the expected third person form with -mk added to it, what we get is a form identical to the 1st sing. of the middle voice. (30) buðumk hilmir lϞð (for bauð mér hilmir lϞð) – ‘offered-to.me the.king hospitality’ (Egill Skallagrímsson HϞfuðlausn 2) (31) Erumka leitt (for er mér eigi [-a(t)] leitt (Egill Lausavísur 25)) – ‘[it] is-to.menot unpleasant’ By virtue of the universal morphological agreement with a subject in the nominative, such as hilmir ‘king’ in (30), one would expect bauð-mk. It should be noted that the form of the suffix seems to have been -mk originally, also when representing dative, as it does in most of the attested cases. According to the standard neogrammarian account (e.g. Noreen 1923: 202), dative -mR would have yielded -m by regular sound development (although the alternative development -mR > -mr cannot be excluded as far as I can see). The manuscripts vacillate between -mk and -m, but they are so late that they cannot be expected to faithfully reflect the original form, long obsolete. More telling is the fact that when the negative clitic -a is added, the only attested ending is -mka irrespective of case government, as seen in (31). This is not the place for a full discussion of the philological problems connected to this construction, but after giving a very brief general overview of the morphological properties of this construction, I will discuss its origin and the light it can shed on the history of the middle voice. I will conclude the discussion with a brief note on the syntactic properties of the construction. There are no more than some dozens of reliable examples of this construction preserved. Almost all of the examples were collected and explained by Kock (1919), but a number of his examples are discounted here because the manuscript tradition is unreliable or because they only exist as grammarians’ reconstructions. A representative

The Old Nordic Middle Voice in the pre-literary period 

sample of the more reliably preserved examples is given below, normalised with -mk or -m according to the weight of the manuscript evidence. There are far more examples of vera, esp. the present erum(k), than of any other single verb, e.g. (32). Especially before ca. 1000, most of the examples involve strong verbs in the present or the past, as in (33), but the evidence still indicates that this construction was not restricted to strong verbs in the oldest period, cf. the examples of weak verbs in (34). (32) a. Þat erum(k) sýnt (=er mér) – ‘It is shown to me’ (Bragi Ragnarsdrápa 14) b. norn erum grimm (=er mér) – ‘the witch is cruel to me’ (Kveldulfr Lausavísa) c. várum sjón sϞgu slíks ríkari (=var mér) – ‘the sight of that was stronger to me than hearsay’ (Þórarinn loftunga Tøgdrápa (1028) 3) (33) a. Ræs gáfumk reiðar mána Ragnarr (=gaf mér) – ‘Ragnar gave me the shield’ (Bragi Ragnarsdrápa 7) b. Urðumk leið in ljóta landbeiðaðar reiði (=varð mér) – ‘The wrath of the ugly land-claimer became unpleasant to me’ (Egill Lausavísur 27) c. Fellumk hϞlf... auðván – ‘half the hope of riches failed me’ (=fell mér) (Glúmr Geirason Gráfeldardrápa 11) d. brennumk feldr fyrir (=brennr fyrir mér) – ‘the fur burns before me’ (Grímnismál 1) e. stϞndumk til hjarta hjϞrr (=stendr mér) – ‘the sword touches me to the heart’ (Fáfnismál 1) (34) a. hugr tjóðum mjϞk mága (=tjóði mér) – ‘the good-will of my in-laws helped me much’ (Egill Lausavísur 26) b. hϞfum gramr... framðan (=hefr mik framðan) – ‘the king has honored me’ (Hallfrøðr vandræðaskald Lausavísur 5) c. þeirar er lϞgðumk arm yfir (=lagði yfir mik) – ‘of the one who laid her arm over me’ (Hávamál 108) d. Ϟll of røk fira vϞrumk, dvergr, at vitir (=varir mik) – ‘I expect that you know about all the history of men, dwarf ’ (Alvíssmál 9) (35) a. hnekðumk heiðnir rekkar (=hnekðu mik) – ‘the pagan men turned me away’ (Sigvatr Austrfararvísur 4) b. yfir ok undir stóðumk jϞtna vegir (=stóðu yfir/undir mér) – ‘the giantways (rocks) stood above and below me’ (Hávamál 106) In the great majority of cases, the subject is in the singular, as in (32)-(34) (or the verb is in default singular form as in (34d)), but there are a few examples with a subject in

 Kjartan Ottosson

the plural, as in (35). Practically all the reliable examples are in the indicative.14 In view of the relative scarcity of this construction, the absence of subjunctive forms may be a coincidence. On the other hand, it seems safe to assume that strong verbs were well represented from the beginning. With the nature of the attested forms in mind, we turn to their origin. It seems impossible to reconstruct a development following the general sound laws from a stage where -mk cliticises to a 3rd sing. form. In the present indicative, the endings had a final -ð, which in the 7th century changed analogically into -R and later -r, as mentioned in 3.3. To derive the attested forms, we would first of all have to assume that the final -ð and later -R/-r was unstable in contact with the enclitic -mk. In that way, -mk would be preceded by a vowel in all the major verb classes at the Proto-Nordic stage. However, the quality of that vowel would only be of the right kind in the weak ō-verbs for a form like *kϞllu-mk to emerge. The other weak verbs and the strong verbs had a front unrounded vowel, i in most cases, which should yield -imk, if not syncopated, cf. the pres. subjunctive 1st plur. minnim(s)k. Furthermore, one may wonder to what extent a pure clitic would have been able to modify the final vowel of its host at the ProtoNordic stage, and one would expect the host to follow the sound development of the form without a clitic, leading to, e.g., kalla(r)-mk. Thus, the syncope regularly seen in the strong verbs and weak ja-verbs should lead to monosyllabic forms like *stend(r)mk, and irregular u-epenthesis would be required to yield a form like stϞndumk in (33e) (with subsequent u-umlaut). In the past indicative 3rd sing., the weak verbs uniformly ended in -e (originally long) in Proto-Nordic, and in -i in Early Old Nordic, expected to result in forms like *lagðimk. The strong verbs had no ending in the past indicative at any stage, e.g. varð. This situation is strikingly different from that in the 1st sing. of the MV, where it can be argued that regular sound development gives the ending -umk everywhere but in the past indicative of strong verbs as discussed in 3.3. Thus, it is only in a limited set of forms, namely the pres. ind. of weak ō-verbs, that forms of the kind attested had any chance of arising by regular sound development, and this set of forms turns out to be absent in the texts. Assuming any analogical generalisation from such a tenuous base to yield the attested forms would be entirely implausible. Thus, another source for the forms must be found. We are forced to assume not only generalisation of the ending -umk of the middle to such non-middle forms, but wholesale importation or copying of the entire middle forms, including the deviant plural vocalism, as one could conceive of hybrid forms like *bauð-umk for the non-middle. In view of the chronological distribution of the different types of forms, generalisation from plural forms as in (35) is also implausible, even if one were to allow for the possibility of forms of the marked number extending into the unmarked one for phonotactic convenience or other special reasons. The middle voice with all of its pecularities in the 1st sing. must have been well entrenched at the time when the 14. A single example is on record where -(u)mk is assumed to have attached to a 2nd sing. form in the imperative: gϞngumk firr, funi! (=gakk mér) ‘keep farther away from me, fire’ (Grímnismál 1).

The Old Nordic Middle Voice in the pre-literary period 

non-middle construction emerged in the earliest Old West Nordic. These forms, in turn, are an argument that the reflexive element had already become an affix at this early stage, in the 9th century. Going over to the syntactic properties of this non-middle construction, it should be noted that the syntactic connections between the attached element and the verb are typically loose or distant, in that -m(k) only in the minority of cases represents a direct argument of the verb. In the examples given above, direct argument status can be argued only in (33a), (34a), (34d) and (35a), whereas in (33c) -mk might represent an adjunct beneficiary connected to the verb. The indirect connections seen in other cases are familiar from the discussion of the Middle in 5.1. With the verbs vera ‘be’ and verða ‘become’, -m(k) represents an element in some way dependent on a predicate embedded under these verbs, and in (34b) (as well as in (36) below), it represents the object of the verb governed by hafa ‘have’ (thematically, that is, although the object may be syntactically governed by hafa, with the participle as a secondary predicate). In (33d), (34c) and (probably) (35b), -m(k) is the object of a preposition. If we assume, as I think we must, that this non-middle construction reflects cliticization that took place before the time of the oldest sources, it is most natural to see in this long-distance option a property that was common to reflexive as well as non-reflexive -mk. Another interesting syntactic property of this non-middle construction is that the clitic is visible for agreement. This is different from the middle voice, where the reflexive element is syntactically inert as far back as we can see, as discussed in 5.2. The examples of agreement with the non-middle construction concern participles governed by hafa ‘have’ as in (36), where an argument in the accusative has been incorporated into that verb as -m(k), without affecting the agreement between this argument and the accusative participle. No examples have been found where the -m(k) element could trigger agreement but fails to do so. Thus, this non-middle construction retains the expected ability of a clitic to trigger agreement, although it exhibits unexpected radical modification of its host, which I take as the consequence of it having become parasitic on the middle voice in its morphological expression. (36) a. hϞfum gramr... framðan (ACC) (=hefr mik framðan) (Hallfrøðr vandræðaskald Lv 5) – ‘has-me [the]king honoured’ b. hϞfumk orkn of skemðan (ACC) (=hefr mik skemðan) – ‘a seal has bitten me’ (Þórðr Kolbeinsson Lv 5) c. sjá hϞfumk veltistoð stiltan (ACC) straumtungls (=hefr mik stiltan) – ‘this woman has caused me trouble’ (Steinarr Lv, 11th century)

7. Phonological concomitants of morphologisation in the middle Hopper & Traugott (1993) mention certain phonological adjustments which they present as expected concomitants of morphologisation. Such are adjustments at the

 Kjartan Ottosson

seams between the active base form and -sk, as exemplified in (37). Referring back to the runic inscriptions, these changes seem quite old in the 3rd sing. pres. ind., as witnessed by [:h]uilis ‘rests’ in Slesvig, Denmark, and hafsk, Forsa, Sweden, but relevant examples are not attested for the 2nd plur. In the poems, forms deviating from current usage in such small details are not expected to survive through oral and manuscript transmission, and the few forms where metre would show whether r is preserved are not found secured by rhyme (vocal stems like fæsk, býsk from bú). Although readjustments such as these go against the nature of cliticization strictly speaking, as noted by Enger (2003), they seem to be rather common with elements that by other criteria must be seen as clitics. When a clitic has become an affix, such changes are commonplace. I think one should be careful not to put too much store in these facts as an argument in determining whether the middle is enclitic or affixal. Another story is adjustments which affect the stem, as in (37b). In Classical Old Norse such adjustments do not follow general sound rules but are lexeme-specific, see Ottosson (1992). Such forms are found in the runic inscriptions, kuask (for Kvað-sk) ‘said of himself ’ in Aars, Denmark, and kas (at) (for gat-sk) ‘liked’ in Bällsta, Sweden. More radical are cases in Classical Old Norse where the middle voice has a form that is not predictable by any sound rule, such as vizk, bizk (i.e. vitsk, bitsk) in (38), see Ottosson (1992: 80). I am not aware of relevant forms of that type in the pre-literary sources. (37) Sandhi readjustments, cf. Table 1 (cf. Hopper & Traugott 1993: 151) a. affecting the endings kallar + sk > kallask 3rd sing. pres. ind. (contrast e.g. superlative verst ‘worst’, gen. hamars from hamarr ‘hammer’) kallið + sk > kallizk (=kallitsk) 2nd plur. pres. ind. b. affecting the stem kvað + sk > kvazk (=kvatsk) 3rd sing. past ind. from kveðask ‘say of oneself ’ (contrast e.g. gleðsk 3rd sing. pres. ind. of gleðjask ‘rejoice’) (38) vizk cf. active vindr bizk cf. active bindr (39) Readjustments of the shape of the middle exponent: East Nordic -sk > -ss > -s or directly -sk > -s (10th – 11th century) West Nordic -sk > -st (13th century, see Ottosson 1992) I would also like to claim that another phonological change that has been generally overlooked is relevant in this connection. I have argued (Ottosson 1992: 135–137) that the shape of the middle exponent is adapted to its new status as a suffix rather than a clitic. The form sk is unnatural for a suffix in the Old Nordic languages, at least syllablefinally. East Nordic and West Nordic naturalise the phonological shape of -sk at different times and with somewhat different results, cf. (39). In East Nordic, -sk changed to -s rather early, by simple drop of the k or by total assimilation of -sk with subsequent simplification. It must be admitted, however, that the causal connections seem rather

The Old Nordic Middle Voice in the pre-literary period 

weak. Most importantly, there is nothing preventing a clitic changing from -sk to -st or -s(s) without any change in its morphological status. On the other hand, marked features can be retained for quite some time, as is well known. It is not surprising that it is the generally more conservative West Nordic that retains -sk longer than East Nordic. That resistance may have been facilitated by the existence of clitics, some of them similar in form to the -sk affix, e.g. -k for the 1st sing. pronoun ek. Finally, however, as the clitics had mostly disappeared, West Nordic -sk underwent normalisation in the form of partial assimilation to -st in the 13th century (see Ottosson (1992: 107–152) for detailed documentation).

8. Conclusions In this paper, I have looked at the ingredients of what has been called grammaticalisation of the Old Nordic middle voice, as seen in texts from the pre-literary period. I have shown that the semantic change involved in the Old Nordic middle voice, which is most properly called regrammation, as it changes the grammatical content of coreference for a multi-faceted middle voice content, must be very early, well within the the Proto-Nordic period. This can be concluded from the fact that from the earliest visible stage, the middle has developed new uses, a reciprocal one, an anticausative one and to some extent a passive one. Furthermore, this semantic development is logically independent of the paradigmatization and cliticization that occurs, as Old High German has developed reciprocal and anticausative meanings from the reflexive, while retaining personal pronouns in reflexive use in the 1st and 2nd persons and not showing cliticization. The regrammation seen in Old Nordic may thus well precede these formal developments by quite some time. The case syncretism between accusative and dative can be shown to be early, from the time of the earliest Old Nordic records, and this syncretism might be independent of the regrammation, having to do with cliticization as such. The other aspect of paradigmatization, the extension of the reflexive pronoun proper, sik, to the 2nd person and the 1st person plural seems to have been underway also at that early stage, but it is conceivable that the extension was still optional at that stage. Decategorisation, or the loss of the syntactic independence or syntactic visibility of the reflexive element, is seen in the earliest sources, but not in the non-middle –mk construction, which can be taken to show that decategorisation was not connected to cliticization as such. The phonological adjustment to the regrammation happened partly quite late. As for the cliticization process and the transition to a suffix status, a number of arguments have been adduced to show that the reflexive element was already an affix before the Old Nordic stage (approximately 700 A.D.), but apparently not too long before that. Further, the earliest sources show reflexes of free, semantically blind cliticization of sik to the verb. This helps to explain the divergences from the typical middle voice semantics reported by Kemmer (1993).

 Kjartan Ottosson

References Andersen, H. 2005. Review of B.D. Joseph & R. D. Janda (eds.), The Handbook of Historical Linguistics. Diachronica 22: 207–228. Antonsen, E. H. 1975. A Concise Grammar of the Older Runic Inscriptions. Tübingen: Niemeyer. Benediktsson, H. 2002a. The Germanic subjunctive: A morphological review. In G. Þórhallsdóttir et al. (eds.), Linguistic Studies, Historical and Comparative by Hreinn Benediktsson, 275–286. Reykjavík: Institute of Linguistics, University of Iceland. (Originally published in 1983). Benediktsson, H. 2002b. The 1st Singular Preterit Subjunctive in Germanic. In G. Þórhallsdóttir et al. (eds.), Linguistic Studies, Historical and Comparative by Hreinn Benediktsson, 287–297. Reykjavík: Institute of Linguistics, University of Iceland. (Originally published in 1987). Björkman, E. 1900–1902. Scandinavian Loan-words in Middle English. Halle: Niemeyer. Carstairs, A. 1987. Diachronic evidence and the affix-clitic distinction. In A. Giacalone Ramat, O. Carruba & G. Bernini (eds.), Papers from the 7th International Conference on Historical Linguistics [Amsterdam Studies in the Theory and History of Linguistic Science 48], 151–162. Amsterdam: John Benjamins. Cleasby, R. & Vigfusson, G. 1874. An Icelandic-English Dictionary. Oxford: Clarendon Press. Dal, I. 1966. Kurze deutsche Syntax. Tübingen: Niemeyer. Edda. Die Lieder des Codex Regius nebst verwandten Denkmälern. (Ed. by G. Neckel, 5th revised edn. by H. Kuhn. 1983). Heidelberg: Winter. Eddadigte 1–3. Nordisk filologi A. (Tekster 4, 7, 8. Ed. by J. Helgason. 1951–1952). København: Munksgaard. Enger, H.-O. 2003. The story of Scandinavian -s(t) retold: Grammaticalising a clitic to a derivational affix. Folia Linguistica Historica 23: 79–105. Faarlund, J. T. 2005. From clitic to affix: On the history of Scandinavian reflexive verbs. NOWELE 2005: 53–72. Fidjestøl, B. 1982. Det norrøne fyrstediktet [Nordisk institutts skriftserie 11]. Universitet i Bergen. Øvre Ervik. Frank, R. 1985. Skaldic Poetry. In Old Norse-Icelandic Literature. A Critical Guide. (Ed. by C. J. Clover & J. Lindow). [Islandica 45], 157–196. Ithaca NY: Cornell University Press Fritzner, J. 1883–1896. Ordbog over Det gamle norske Sprog 1–3. Christiania. Halpern, A. L. 1998. Clitics. In A. Spencer & A.M. Zwicky (eds.), The Handbook of Morphology, 101–122. Oxford: Blackwell. Harris, J. 1985. Eddic Poetry. In Old Norse-Icelandic Literature. A Critical Guide. (Ed. by C. J. Clover & J. Lindow). [Islandica 45], 68–156. Ithaca NY: Cornell University Press. Helgason, J. 1953. Norges og Islands digtning. Litteraturhistorie: Norge og Island. In S. Nordal (ed.), Nordisk kultur 8 B, 3–179. Stockholm: Bonnier. Helgason, J. 1962. Skjaldevers. [Nordisk filologi A 12]. København. Hopper, P. J. & Traugott, E.C. 1993. Grammaticalisation [Cambridge Textbooks in Linguistics]. Cambridge: CUP. Jacobsen, L. & Moltke, E. (eds.). 1941–1942. Danmarks runeindskrifter 1–3. København. Jónsson, F. 1901. Det norsk-islandske skjaldesprog omtr. 800–1300. [Samfund til udgivelse af gammel nordisk litteratur 28]. København: STUAGNL. Jónsson, F. 1921. Norsk-islandske kultur- og sprogforhold i 9. og 10. årh. København.

The Old Nordic Middle Voice in the pre-literary period  Jónsson, F. 1931. Lexicon poeticum antiquæ linguæ septentrionalis. Ordbog over det norsk-islandske skjaldesprog. (2nd edn). København: Møller. Kemmer, S. 1993. The Middle Voice [Typological Studies in Language 23]. Amsterdam: John Benjamins. Klavans, J. L. 1985. The independence of syntax and phonology in cliticization. Language 61: 95–120. Kock, A. 1919. Fornnordiska böjningsformer. Arkiv för nordisk filologi 35: 55–99. Kock, E. A. 1923–1944. Notationes Norroenæ. Lund: Gleerup. Krause, W. 1971. Die Sprache der urnordischen Runeninschriften. Heidelberg: Winter. Noreen, A. 1923. Altnordische Grammatik I. Altisländische und altnorwegische Grammatik (4th edn). Halle/Saale: Niemeyer. Den norsk-islandske Skjaldedigtning A1–2, B1–2. (Ed. by F. Jónsson). Copenhagen: Rosenkilde & Bagger. 1912–1915. Den norsk-isländska skaldediktningen 1–2. (Ed. by E. A. Kock. 1946–50). Lund: Gleerup. Nygaard, M. 1865. Eddasprogets Syntax I. Bergen. Nygaard, M. 1905. Norrøn Syntax. Kristiania. Onions, C.T., Friedrichsen, G.W.S. & Burchfield, R. 1966. The Oxford Dictionary of English Etymology. Oxford: Clarendon. Ottósson, K. G. 1981. Þróun miðmyndarendinga í íslensku. Cand. mag. thesis, University of Iceland, Reykjavík. Ottósson, K. G. 1986. Mörk orðmyndunar og beygingar: Miðmynd í nútímaíslensku. Íslenskt mál 8: 63–119. Ottósson, K. G. 1992. The Icelandic Middle Voice. The morphological and phonological development. Lund. Ottosson, K. 1999. From cliticisation to inflection. The emergence of the Scandinavian mediopassive. Paper presented at ICHL 14 in Vancouver. Peterson, L. 1994. Svenskt runordsregister. (2nd rev. edn). [Runrön 2]. Uppsala. Simpson, J. & Weiner, E.S.C. 1989. The Oxford English Dictionary 1–20. Oxford: Clarendon. Specht, F. 1891. Das Verbum reflexivum und die Superlative im Westnordischen. (Offprint from Acta Germanica III,1). Berlin. Sveinsson, E. Ó. 1962. Íslenzkar bókmenntir í fornöld 1. Reykjavík. Sveriges runinskrifter 1–9, 11–15. 1900–1981. Stockholm. Wisén, T. 1865. Om Ordfogningen i den äldre Eddan. Lund. Wisén, T. 1883. Om norröna medialformer på -umk i första personen singularis. Arkiv för nordisk filologi 1: 370–384. Zwicky, A. M. & Pullum, G.K.. 1983. Cliticization vs. inflection: English n’t. Language 59: 502–513. Rundata [data base]

The relevance of tense and aspect in Semitic Languages The case of Hebrew and Arabic Sven-Olof Dahlgren This article presents an introduction to the verbal system of the Semitic languages and an historic account of the research on the verbal system of Hebrew as a background to the main subject of the paper: the verbal system of Hebrew and Arabic. At variance with the common view that these systems are based on aspect, we propose that Biblical Hebrew and – to a large extent – Arabic have a relative tense system. Although this is not a new notion it introduces the foreground/ background distinction in this context, as well as the concept of marked and unmarked verbal forms as an explanation of the mysterious verbal system of Biblical Hebrew. The notion of aspect is, however, not discarded. Its use in modern written and spoken Arabic is dealt with at the end of this article.

1. Introduction1 The nature of the Semitic verbal system of such languages as Arabic and Hebrew has been debated for a long time. The editor of a recent work on Semitic languages, Robert Hetzron (1997), discusses the use of a uniform terminology: The most delicate case is the tense versus aspect controversy: with which one of these categories do Semitic languages operate their verbal system; tenses like a) past, (b) present-future or aspects like (a) completed ~ perfective and (b) non-completed ~ imperfect(ive) (where the two (a)’s and (b)’s are respectively equivalent)? I myself seem to be in a minority in being fully convinced that the relevant category is tense. (Hetzron 1997: xv) 1. Parts of this paper were written in spring 2003, during my stay at SIL, England. I am thankful for the inspiring teaching and helpful comments by Ivan Loew and David Crozier. I am also grateful to Professor Jan Retsö at Göteborg University for many helpful hints and remarks.

 Sven-Olof Dahlgren

Another recent author of a similar work, Edward Lipinski, claims: “While the classical verbal system of Semitic languages is based on aspect, modern speech tends to found the verb inflection on the notion of time and to express it by means of tenses” (Lipinski 1997: 346). The famous Arabic grammarian Sibawayhi emphasised the prominence of tense already at the outset of his monumental grammar al-Kita6b ‘The Book’. The meaning of the Arabic name for the suffix conjugation the perfect is the past (al-ma6d2i& ), and that of the prefix conjugation the imperfect is al-mud2a6ri< ‘the similar’ i. e. similar to nouns, since indicative and nominative has the same -u ending and subjunctive and accusative the same -a ending. The medieval Jewish grammarians followed this division; the verb form qatal (explained below) was past tense, qotel (active participle) was present and yiqtol future tense (McFall 1982; Waltke & O’Connor 1990: 459).2 However, a well known Semitist, Bergsträsser, in his Introduction to Semitic languages (1927: 55) asserted that Hebrew has a complicated tense syntax which has no parallel in any other Semitic literary language, which has given rise to many and diverse proposals regarding its nature. McFall (1982: 7) presents three different main theories in 1827. Then McFall proceeds to survey the opinions put forward between 1827 and 1954. He found ten main different theories based under six basic headings. At the same time he stated that no fundamentally new solution has been put forward after 1964. Before reviewing these theories we present the different elements that need to be accounted for in a comprehensive theory. First of all Arabic and Hebrew are Semitic languages. These are generally characterized by word formation built on three consonants called radicals in the bulk of the vocabulary. The three consonants convey a general meaning e. g., q t l the act of killing in Hebrew, which is the verb used in Hebrew verb paradigms since it is one of the verbs that undergo least changes – such as assimilations, or dropping of elements – in the paradigm. From these three generally arbitrary radicals thousands of words and word forms can be coined through doubling of any of these radicals, which is usually the second; adding, dropping, exchanging or lengthening of vowels; use of different prefixes or suffixes for marking gender, number and person, or adding of a preformative that gives a modification of the general meaning. The verbs we deal with here belong to the category of three radicals, which is overwhelmingly the largest one and the relevant one for discussing the verbal system. The table below gives an overview of different forms in some Semitic languages. It is common in Semitistics to use a certain verb in a Semitic language as a model for describing different verb forms, for instance iprus ‘to cut’ (literally ‘he cut’) in Akkadian and fa

The relevance of tense and aspect in Semitic Languages 

Table 1. Language

Qtl

Yqtl

Jussive

Hebrew Aramaic Arabic

qatál qetál qátala

yiqtól yiqtúl yáqtulu

yiqtól yiqtúl yáqtul

Ethiopic Akkadian Egyptian

qátil qátil sádm

yáqtul yáqtul sédmet

yáqtul yáqtul sedmét

Yaqáttal

Imperative qetól qetól

yeqáttel yaqáttal sedémmet

>uqtul qétel qútul sdém

As for Arabic there are two main conjugations: the suffix conjugation, which is often referred to as perfect (past tense or completed action according to interpretation). In the word faaf
2. The Hebrew verbal system Niccacci lists five basic verb forms. Corresponding to fa
 Sven-Olof Dahlgren

occurrences 10 830 were translated with past tense, 2454 with present, 255 with future, 56 with non-past modal, 115 with past modal, 17 with imperative, 38 jussive/cohortative, and 109 non-verbally translated. The 14299 yiqtol occurrences are translated 774 times with the past tense, 3376 times with the present tense, 5451 times with the future tense, the non-past modal 1200 times, the past modal 423 times, the imperative 2133 times, jussive/cohortative 789 times, and non-verbally 153 times. This means for indicative sentences where genuine time reference is made, 91% non-past expressions. For the wayyiqtol he reports that out of 14972 instances the RSV translates 14202 instances (95%) with the past tense, while 308 were present tense, 30 future, 12 non-past modal, 35 jussive/cohortative, and 367 non-verbal. Of the 6378 we-qatal 2932 were translated with future tense, 484 with past, 545 with present, 405 with non-past modal, 63 with past modal, 1647 with imperative, 70 with jussive/cohortative, and 232 non-verbal. 1335 we-yiqtol forms exhibited 438 occurrences with future tense, 195 with present tense, 71 with past tense, 285 non-past modal, 51 past modal, 22 imperative, 239 jussive/cohortative, 34 non-verbal (McFall1982: 186–188).

3. Theories before 1827 The mysterious waw of future was known by the Karaites already at the beginning of the 10th century, and possibly among orthodox exegetes in Palestine at the same time. Menahem ben Saruq (ca 910-ca 970) thought that the waw possessed a converting influence on the verb to which it was prefixed, i.e., changing a past verb form to a future one, or a present to a past, a view that was denounced by other Hebrew scholars in his day and afterwards. Levita (1525) and Abraham de Balmes were the first to call the waw that is connected with the qatal with future reference for waw hippuh5 “the conversive waw”, since, in his opinion, it converts a past action into a future one. There is also another waw, they noted, the copulative waw, which is a mere conjunction and hence was called waw hibbur. In the first and second person singular of qatal they can be distinguished by the position of the accent, which is indicated in the Hebrew script (1525, from McFall 1982 :10). With Schroeder one notices a real departure from the Jewish view as he, as it appears, introduces the notion of a relative tense: “…the Future form has another usage which is clearly unique and peculiar to the Hebrews, in that it receives the force of our past, and designates a matter as really past, not, however by itself and absolutely, but in relation to a past action which has preceded it. For when a number of events are to be narrated which follow one another in some kind of continuous series, the Hebrew writers consider the first action as past; however the others, which follow it, are considered as future on account of what has gone before, since something is described

The relevance of tense and aspect in Semitic Languages 

which in relation to another past action is itself later and future, so it may be called the Relative Future” (in Institutiones, 1785). In a rather modern way of thinking Gell (1818) suggested that every genre or category of literature was comprised of major and minor systems. A major system would be an historical narrative for example. It starts with a governing verb that establishes the general time of the narrative. The following subordinate verb forms retain their own individual sense through the waw of the main verb, which induces or passes on the temporal power to the subordinate verbs. This waw he calls the waw inductive. This view, he explained, would do away with the ‘barbarism of the waw conversive’ (McFall 1982: 2–16; Gell 1818: 45). McFall reduces these different view-points to three different theories: the oldest Conversive Theory, followed by the Relative Tense Theory proposed by Schroeder. This latter is at variance with the Conversive Theory in that it has a constantly changing standpoint according to the time of the narrator in relation to a fixed time-scale. The Conversive Theory reverses these two parameters. The third theory is the Waw Inductive Theory suggested by Gell (McFall 1982: 17–26).

4. Later theories (1827–1892) Lee (1827) regards the qatal as a compound of a concrete noun and a suffixed pronoun. Bearing no tense value in itself it has received its past notion by the priority (by being prior to) of the verbal action over the agent. The yiqtol he considers to be an infinitive supplied with pronominal affixes (Lee 1827: 109–135, 190, 193, 341). Both yiqtol and wayyiqtol are present tense in every context. Both tenses can be used absolutely or relatively, i.e., the time reference is other than that of the speaker as in Isaiah 9:6, “For unto us a child is born” (relative past), and “His name is called present” (relative present) (Lee 1827: 342–351). Weir (1850) highlighted the fact that many intransitive verbs do not have past time value which led him to conclude that qatal is the Hebrew present (in the past) and yiqtol as well as wayyiqtol is future (in the past). He consequently reads Genesis 1:1: “In the beginning God creates the heavens and the earth”, asserting that ‘creates’ is present in relation to ‘in the beginning’. Murphy (1850) also laborated with a relative tense system in his Anterior-CentralPosterior (=Qtl – Participle – Yqtl) theory. This theory states that a speaker/writer takes up a position in the neighbourhood of the action and describes the events in relation to his position in regard to them, so that in a connected series of past events the writer establishes a fixed standpoint in the past, from which events are contemplated as either anterior (qtl), central (participle), or posterior (yqtl). The two basic verb forms indicate Anterior (=an action ended) and Posterior (=an action to begin), whereas the participle indicates a mid-way position.

 Sven-Olof Dahlgren

Ewald (1879) asserted an aspect view of the Hebrew verbal system using the terms Perfect and Imperfect, which signified complete and incomplete respectively. Concerning wayyiqtol, he regarded the prefix to be a combination of waw and *az, which together forms a temporal particle referring to the past corresponding to the augment ein the Greek Imperfect tense. This augment in Hebrew throws an action into the past and the tense is used to describe past events as coming into being. The form, however, always follows a qatal already mentioned and is regarded as a consecutive form like the weqatal. In the weqatal form the prefix is a simple conjunction. weqatal is characterized by a strong tone on the final syllable in first and second person. The meaning of this form is the same as simple Imperfect. These two consecutive verb forms with waw are regarded as two different new constructions, and each is to be considered as a single whole in analogy with a prepositional phrase. There are in his system also two absolute tenses; the Perfect + waw, and the Imperfect + waw; but he has little to say about them. The former is distinguished from weqatal by the position of the tone. The ideas of Ewald became very popular and were taken up by authorities on Ethiopic, Syriac, and Arabic in their grammatical works (Ewald 1879: 3, 18). Nordheimer presented his Relative Tense theory in 1841, actualizing the same ideas of Schroeder eighty years earlier. He finds only two tenses in Hebrew: Past and Future. Yiqtol denotes an “action yet to take place” and qatal “an action that is past”. These tenses are called the Absolute Future and the Absolute Past. The consecutive forms are regarded as secondary tenses; they are called Relative Future (weqatal), and the Relative Past (wayyiqtol) (Nordheimer 1841: 159, 183). Driver put forward a theory (first time in 1874) that was very close to Ewald’s. It had an aspectual outlook but instead of Imperfect he used the term Nascent. The yqtl describes a nascent (beginning to exist or become manifest) or incipient (beginning to exist or become manifest) action. He also uses the terms “inchoative” (that which begins or expresses beginning), “inceptive” (noting the beginning, commencing), and “egressive” (tending to issue forth). The yiqtol is also used for frequentative actions. Besides these two basic meanings it may be used for subjunctive, potential, concessive, conditional etc. It is used in present and future time, and past time, when the writer changes his real standpoint for an ideal one which places him in the time he is describing (Driver 1874: 3–5, 27–30). The wayyiqtol predominantly connects together a series of past events into a consecutive narrative but may also be used for present and future events (Driver 1874: 73). As for the two different qatal he states that the qatal consecutive, i. e. weqatal (with ultima stress in first and second person) is used relatively; it receives its meaning from the preceding dominant verb, so that if yiqtol is future so is weqatal. It means a single action in the present or the future or a frequentative action in the past. The conjunctive qatal indicates completed action in an absolute way irrespective of preceding or following verbs (Driver 1874: 115, 143–145). Regarding the Hebrew verbal system Driver has probably influenced the English-speaking scholarly world more than anybody else by his writing, which to a large extent was a popularization of Ewald’s Perfect-Imperfect theory (McFall 1982: 76).

The relevance of tense and aspect in Semitic Languages 

Turner (1876) asserted that the qatal expresses the action or state as the attribute of the person or thing spoken of whereas the yiqtol form expresses or represents the verbal action. Following Lee he sees the noun as prior to the verb and even regards verbs as primarily nouns. In the similar endings in verbs and nouns he finds support for this view. Wayyiqtol is not different from yiqtol in his system. Agreeing with Lee as to the origin of the qatal and the yiqtol forms he departs from him by not viewing the resultant forms as Past and Present, but as pointing to an objective (qatal) and a subjective (yiqtol) view on the part of a writer (Turner 1876: 365, 378–9, 384).

5. The Comparative-Historical approach The Assyrian/Akkadian language was discovered in 1841. Before long some scholars came to view it as the Sanskrit of the Semitic languages, whereas others objected to this classification, pointing out that it, for instance, lacked a qatal form. Others noticed the astonishing similarity between two languages distant from each other: Ethiopic and Assyrian. The Assyrian Present iqatal corresponded to the Ethiopic Imperfect yeqatel and a shorter form iprus in Assyrian resembled the yeqtel in Ethiopic. These findings led to a new approach in Hebraistic studies, the historical-comparative approach. Before that, Hebrew had been studied in relative isolation although Arabic, Ethiopic, and Aramaic were known. All of them lacked consecutive tenses but had qtl and yqtl forms. The first one to embark on this approach was Knudtzon, who presented his views in 1892. However, since his description of the reflection of Akkadian verb forms in Hebrew is not clear, one had to wait for Bauer (1910) to present a more cogent theory based on this material (McFall 1982: 89–92). Bauer claimed that the Semites originally only had one verb form, the prefix conjugation (yaqtul). He called it “aorist” since it was originally timeless. Then the suffix conjugation developed from a participle. The meaning of this new verb form became either perfect (‘he is a murderer’ > ‘he murdered’, or present (‘he is skilful’ > ‘he knows’). The perfect meaning prevailed in the West Semitic languages to which Hebrew belongs with the result that the prefix conjugation became the present and future. In Akkadian, on the other hand, the suffix conjugation was transformed into a prefix form (supposedly iparas) that was used for the present tense, with the result that the older prefix form (yaqtul, iprus) became a past form. This form he saw reflected in wayyiqtol and explained the presence of both West Semitic and East Semitic prefix conjugations in Hebrew as a result of their wanderings from the East to the West. Later research showed however that the Akkadian present iparas actually had a double middle radical (iparras) and therefore could not be linked to the West Semitic perfect but rather to the Ethiopic indicative imperfect (yeparres). Nevertheless, Bauer’s work is estimated as the crucial outline of a modern view of the Hebrew tenses (Bauer 1910: 7, 19, 26, 45; McFall 1982: 93–114; Waltke & O’Connor 1990: 466–470).

 Sven-Olof Dahlgren

Driver (1936) regards qatil as the first and universal Semitic verb form. Among the reasons for it is the fact that it is the simplest form only consisting of a composite of a noun or an adjective followed by a suffix in contrast with yaqtul that also has a prefix, and compound forms cannot be older than their constituent parts. And when the ambiguities concerning tense were felt with qatil the first requirement for the Semites was a present-future tense, which became (y)qát(t)i/al. For man is more likely to be occupied with the needs of daily life than with history. Further, the use of qatil is much wider than forms like yaqatal and yaqtul which is an indication of the priority of qatil. In a second stage (y)qát(t)i/al developed for reasons given. Subsequently, a preterit yaqtol evolved, followed by the imperative qutul in the fourth stage, and the jussive and the cohortative in a fifth stage. In the sixth stage the West Semitic yaqtúl became a tense of “every kind of incomplete action”. This new imperfect was only developed in languages in which qatil and especially qatal served as purely perfect tenses for complete actions in past time and in languages where yaqattal was absent. Even Driver regards Hebrew as a composite language with elements from both East and West Semitic (Driver 1936: 26, 29–30, 107; McFall 1982: 116–150). Thacker (1954), whose main contribution was to link the Egyptian and proto-Semitic verbal systems, accepts the view of Hebrew as a conglomerate language. He also found remarkable similarities between Egyptian and Hebrew. Like Driver he regards Akkadian as having preserved almost intact the proto-Semitic verbal system. The West Semitic system on the other hand is late. He follows Driver to a large extent but diverges from him concerning the origin of the yaqattal form and the order of the verb forms to develop. The yeqattal, he asserts, is not derived from qatil but from a nominal base with a doubled root consonant. And concerning the basic forms, he accepts qatil to be the first but considers yáqtul to have developed after it, instead of yaqattal (Thacker 1954: 118, 189, 228, 322). The outlook of Bauer on the Semitic verbal system is clearly tense based, whereas G. R. Driver is famous for having thought of the Hebrew forms as not expressing tense. He is not altogether clear though particularly since he appears to suggest that yáqtul was a preterit tense from its inception. Thacker, on the other hand, leaves us in no doubt that the proto-Semitic verbal system was not a tense but an aspectual system. Tense categories came very late in the history of the Semitic languages according to Driver. Rundgren was a prominent supporter of the aspectual view of the Semitic languages. Inspired by Indo-European scholarship, especially the grammar of the Slavonic languages he suggested that the “perfect” and “imperfect” of the classical Semitic languages do not express time at all, but the subjective way of the speaker to relate certain events. He regarded the Hebrew verbal as having the basic dichotomy qatal – yiqtol/ wayyiqtol, where in a logical opposition A – not A qatal corresponds to ‘A’ and yiqtol and wayyiqtol corresponds to “not A’. Qatal is to be regarded as the neutral value, indicated by its capacity to express both stative (as in ‘he is a murderer’) and constative fiens (‘he has murdered’), as well as being used for the future w-qatal. Yiqtol expresses the

The relevance of tense and aspect in Semitic Languages 

cursive aspect, i.e., what is actually occurring in the present or past and wayyiqtol the constative aspect in his system (Isaksson 2005; Rundgren 1961: 37–38, 62–63, 84–97). All the preceding theories about the Hebrew verbal system have been based on either a tense- or aspect-oriented view. Zuber (1986) proposed a verbal system based on mood. In this system qatal and wayyiqtol give indicative propositions, and yiqtol and weqatal modal and future propositions respectively (Zuber 1986: 25–27).

6. Evaluation McFall, who reviewed all theories until 1954, was able to criticize all of them for lack of consistency in some respects. Given the state of the art of Semitic studies during most of the period as well as the lack of necessary tools from general linguistics this is no wonder. Erudite and brilliant Hebraists have tried to climb this mountain without the necessary equipment and it has proved an impossible enterprise. There are also potential sources of error in the material itself. Some of the crucial points are discussed from the aspect of vocalism of different verb forms. The text is, however, basically a consonant text and the vowels and prosodic signs were added by the Massorets 700–1500 years after the writing. They are not unlikely to be mistaken in several cases. For example, McFall, in his critique of the Conversive theory, lists six occurrences of wayyiqtol with future meaning. In two of these the editors of the Biblia Hebraica Stuttgartensia suggest emendations to a more expected yiqtol form, another has a textual variant with yiqtol, and a fourth occurrence has a variant with a possible yiqtol; it was 1 sg. where there is no difference between wayyiqtol and yiqtol (McFall 1982: 18–19, 133). There are also new advances within linguistics which can be employed to shed light over the verbal systems. The notion of markedness and prototypical concepts are two such important tools to bring into the discussion. If, for instance, a form like wayyiqtol exhibits 95% occurrences with past meaning, there is good reason to conclude that past tense is its basic meaning and the remaining occurrences are marked constructions. In text linguistics it has been shown that in the peak of a story the normal use of tenses ceases to function, which would provide one such example of a marked construction (Dahlgren 1998: 82–83).3 Even more important is the concept of foreground and background. It was shown in a pioneering study by Hopper (1979) that languages may use word order, morphology, or tense variation to mark foreground, i.e., the consecutive order of the main points in a narrative, and background. The difference between passé simple (foreground) and l’imparfait (background) in French is one example (Hopper 1979: 213–228). Surprisingly enough this important perspective is rejected in the voluminous and excellent An Introduction to Biblical Hebrew Syntax, by Waltke and O’Connor (1990: 54). 3.

Text example from the book of Esther.

 Sven-Olof Dahlgren

7. Text linguistic oriented studies The concept of foreground and background is highlighted in a study on Biblical Hebrew by Eskhult (1990). He sees the binary opposition activity–stativity as corresponding to foreground and background respectively. Qatal represents basically ‘stativity’,since it originated from a verbal noun or adjective linked to a person suffix. Later it developed to be used even for verbal actions. For instance in the common chiastic pattern wayyiqtol-obj: obj-qatal it has the same value as wayyiqtol. The foreground, however, with consecutive events was indicated by wayyiqtol. Yiqtol is used for cursive aspect, i.e. ongoing actions during which another event occurs (Eskhult 1990: 17, 19, 20, 22, 34). Eskhult finds a fundamental difference between foreground and background in that the subj-qatal order signalizes a break in the sequential narrative chain to give a statement about the subject; it may be a description or a circumstance with a bearing either on the nearby clause or on the whole episode. In dialogue, on the other hand, this function is practically absent (Eskhult 1990: 34). Niccacci (1990) underlines the importance of diachronic studies, but aims primarily at making a synchronic text linguistic description of Classical Hebrew, which is principally Biblical Hebrew. He makes a basic distinction between narrative and discourse. The latter he defines as a genre where the speaker addresses the listener directly (dialogue, sermon, prayer).4 Narrative discourse employs wayyiqtol and waw-xqatal, whereas in discourse one meets with volitive forms, simple nominal clauses, indicative x-yiqtol, weqatal and (x-)qatal (x refers to a nominal or adverbial phrase). Within the parameter foreground/background wayyiqtol is used in foreground in narrative, and in background (x-)qatal clauses and weqatal clauses are met with. As for discourse, foreground (not defined) is characterized by volitive forms, (x-)qatal, x-indicative yiqtol, and simple noun and background by waw-noun clauses and waw-(x)qatal clauses. Another parameter employed by Niccacci is that of linguistic perspective, which refers to i) retrieved information (flashback, “antecedent” to the ensuing account), ii) degree zero (the level of the story itself), iii) anticipated information (“disclosure”, reveals the end of the story). In narrative recovered information corresponds to past perfect, which is given by waw-x-qatal in background, degree zero is given by wayyiqtol and anticipated information, corresponding to conditional, by yiqtol.

4. This division is not altogether felicitous, in my opinion. The organizing of monologue discourse by Longacre (in Grammar of Discourse, 49) along the parameters “contingent succession”, “agent orientation” and “projection” would separate at least dialogue and sermon/prayer. Dialogue (p. 51-75), with its basic utterances question – answer, proposal – response and remark – evaluation, is to be seen as a separate entity, and it is also treated as such by Longacre. The crucial distinction consecutive (=contingent succession)-non-consecutive clauses is, however, kept in Niccacci’s system.

The relevance of tense and aspect in Semitic Languages 

8. A new proposal The following model was developed during my work on Word Order in Arabic (1998), by observing the diverse uses of verb forms in foregrounding and backgrounding function. It describes the Hebrew verbal system as a relative tense system, structured around the foreground/background distinction.5 I believe it is an improvement on both earlier relative tense systems and Niccacci’s work, who although he applies foreground and background to his theory, does not speak clearly of a relative tense system. To the present writer these concepts need to be brought together for a better understanding of the Hebrew verbal system, as well as that of markedness, which is generally absent in writings about the Hebrew verbal system. We suggest for the Hebrew verbal system that wayyiqtol is the form par excellence for expressing the foreground in narrative prose. Being connected to the Akkadian iprus/yáqtul it is a past tense. The form qatal is also a past tense but with a wider application; it may be used for pluperfect, preterit and perfect. However, this does not apply to stative verbs; although of qatal form they are often to be translated by present tense. Qatal may also be used in some cases in foregrounding, when the verb is preceded by an adverbial clause and when the subject is topicalized (see below). For nonstative verbs it is a general past tense that can be used for the pluperfect and other tenses in the past. There is no specific pluperfect construction in Hebrew which in our opinion is due to the relative tense system; since the backgrounded pluperfect already has a reference point in the past there is no need for a separate form. Yiqtol and weyiqtol may be used for past tense when they describe events that are co-occurrent with the foreground. According to Waltke/O’Connor (1990: 54), we-yiqtol is not used consecutively but appears in clauses that are logically or temporally subordinate to the preceding clause. Their tense value is non-past i.e., present and future tense. The reason why yiqtol may be used for the past is that different deictic points of time reference are used: the main story, the foreground, is told with the narrator’s time as the deictic point of time reference; backgrounded sentences on the other hand have the narrative itself as the time reference, which makes it possible to use forms with present tense value for the past. The same phenomenon applies to pluperfect in Hebrew. It is constructed with a single qatal, since the reference point is already in the past in that it is a backgrounded sentence.6 We present below a text example from Early Biblical Hebrew with three backgrounded yiqtol and one pluperfect qatal. For qatal we write SC (suffix conjugation), and for yiqtol we write PC1 (prefix conjugation1), and for wayyiqtol PC2. For the 5. I am much indebted to Professor Retsö who suggested to me to regard Arabic as tense based. 6. Our Hebrew notation is a pure transliterary one, where plene (full) writing (of the vowels) is indicated by “^” for yo6d§ (y) and wa6w (w), and “6”for he6. The s9wa6 is not marked. Following Retsö I do not reckon with long vowels in Hebrew.

 Sven-Olof Dahlgren

idiomatic translation we employ the New American Standard Bible (NASB), which is supposed to closely follow the original text. (1) 2 Samuel 12:1–4. 1. wayyis9lah 2 YHWH >ät-nåt§ån >äl-dåvid§ wayyåvo> send.3M.SG.PC1 LORD DEF.OBJ-Natan to-David come.3M.SG.PC2

>elåyw wayyo>mär lô to him say.3M.SG.PC2 to him The Lord sent Nathan to David. When he came to him, he said to him:

s9ne& >anås9ım håyû b-<ır >äh2åt >äh2åd§ <ås9ır two men be.3M.SG.SC in-town one, one rich

w->äh2åd§ rå>s9: and-one poor There were two men in one city, the one rich, and the other poor.

2. l-n u-b§åqår harbeh m>od§ to-rich be.3M.SG.SC sheep and-cattle much very The rich man had a great many flocks and herds, 3. w-lå-rås9 >e6n-kol ki >im-kib§såh >ah2at§ q6tanna and-to-poor nothing-all but if – ewe lamb one small

>as9är qåna6 which buy.3M.SG.SC but the poor man had nothing except one little ewe lamb which he had bought.

w-y6h2ayyä-hå wattig7dal

yah2dåw mip-pitt-ô together from-bread-his He raised it, and it grew up with him and his children. It shared his food,

t§o>k§al-u-mik-kos-ô t§is9täh u-b§-h2e6q-ô eat.3F.SG.PC1-and-from-cup-his drink.3F.SG.PC1 and-in-bosom-his

t§is9kåb§ watthı-lô k-b§at§ : sleep.3FSG.PC1 become.3FSG.PC2-to him as-daughter. drank from his cup and even slept in his arms It was like a daughter to him.

4. wayyåb§o> heläk l->ıs9 hä-<ås9ır wayyahmol come3M.SG.PC2 traveler to-man the-rich abstain.3M.SG.PC2

The relevance of tense and aspect in Semitic Languages 

lå-qah2at§ mis2-s2o>n-ô u-mib-bqår-ô to-taking from-sheep-his and-from-cattle-his Now a traveler came to the rich man, but the rich man refrained from taking one of his own sheep or cattle.

In verse 3, >as9är qåna6 ‘which he had bought’, the verb is in simple perfect; with the reference point in the foreground, i.e., in the past, there is no need of establishing a new reference point in the past through a pluperfect form. In verse 3 we also encounter three backgrounded verses with present tense value: to>k§ al ‘it eats’, tis9täh ‘it drinks’, tis9kåb§ ‘it sleeps’. They co-occur with the foreground, and therefore have past meaning. With this system probably most of the mysterious 774 instances of yiqtol in the past are explained and accounted for, which abrogates one of the main reasons for an aspect oriented view of the Hebrew verbal system. The active participle (PART) qotel, which has become the unmarked form for present tense in modern Hebrew, may also be used for the background of an actual present event or situation as in the following example. (2) Genesis 18:1 (NASB) wayyerå> >elåyw YHWH b->elonê mamre> we-hû> yos9eb§ appear.3M.SG.PC2 to him LORD by-oaks/of Mamre and-he sitting pät§ah2 hå->ohäl k-h2om hay-yôm door/of the-tent as-heat/of the day. Now the LORD appeared to him by the oaks of Mamre, while he was sitting at the tent door in the heat of the day. There are also numerous backgrounded noun clauses without the past copula håya6, i.e. in the present tense. The following example from Genesis exhibits three such instances in main backgrounded clauses as well as one instance with yiqtol. In two of the noun clauses the active participle qotel is the predicate. Although they have verbal power, such clauses are construed as noun clauses. (3) Genesis 2:9–11 (NASB) 9. wayyis2mah2 YHWH > älohîm min hå->åd§åma6 plant3M.SG.PC2 LORD God from the-ground

kål-<es2 all-tree

näh2måd§ pleasing And out of the ground the LORD God caused to grow every tree that is pleasing

l-mar>äh w-t2ôb§ l-ma> ak§ål w-<es2 ha-h2ayyîm for-sight and-good for-eating and-tree/of the-life

 Sven-Olof Dahlgren

b-tôk§ hag-gån in-middle/of the-garden to the sight and good for food; the tree of life also in the midst of the garden

w-<es2 had-da and-tree/of the-knowing good and-bad and-river going out

me-<ed§än from-Eden and the tree of the knowledge of good and evil. Now a river flowed out Eden

l-has9qôt§ >ät§-hag-gan u-mis9s9 for-watering DEFOBJ-the-garden and-from-there

yippåred§ w-håya6 is divided3M.SG.PC1 and-become.3M.SG.SC to water the garden; and from there it divided and became

l->arbås9îm s9em hå->äh2åd§ pîs9ôn hû< to-four rivers name/of the-one Pishon he

has-sob-eb >et§ the-surround.PART DEFOBJ four rivers. The name of the first is Pishon; it flows around

kål->äräs2 ha-ha2 wîla6 all land/of the-Havilah the whole land of Havila.

The transition in the end from past tense to present in the translation of New American Standard Bible is not very well motivated by the text itself; it should be “The name of the first was Pishon, it flowed around the whole land of Havila”. The new Swedish translation “Bibel 2000” starts with the present tense in verse ten: “A river flows up in Eden.” We regard this as groundless and stemming from a lack of awareness of the foreground/background distinction and its bearing upon the tenses. However, the Swedish version is supported by the early Greek translation Septuaginta from about 200 BC. Thus we provide another example where three common English translations as well as Bibel 2000 translate similar constructions by past tense. (4) 2 Samuel 1:12–13 (NASB) 12. w-håya6 ki hirbt§a6 l-hit§pallel li-fnê Happen.3M.SG.SC as make-much.3F.SG.SC to-praying to-face/of

YHWH w-<elî s9ome LORD and Eli guarding

The relevance of tense and aspect in Semitic Languages 

Now it came about, as she continued praying before the LORD, that Eli was (‘is’) watching

>ät-pî-hå DEFOBJ-mouth-her her mouth.

13. w-h2anna6 hî> md§ abbärät§
s'efåtê-hå nå<ôt§ lips-her moving As for Hanna she was (‘is’) speaking in her heart, only her lips were (‘are’) moving,

w-qôl-åh lo> yis9s9åmea< wayyah2se9 b§ä-hå <elî and-voice-her not is heard3M.SG.PC1 reckoned-it3M.SG.PC2 Eli

l-s9ikkora6 to-drunkenness but her voice was (‘is’) not heard.

So Eli thought she was drunk.

The New International Version, the King James Version and the Bibel 2000 all follow the translation of NASB in rendering by past tense the four instances marked with present tense in the parentheses. However, they should have been translated by present tense with the same way of translating as in the previous examples of Genesis 2:11 (and even 2:10 in Bibel 2000). That would have been correct without a reference point in the past. Here they are all correct, according to our system, since they are backgrounded there is such a reference point in all four instances. In the beginning of a story, however, before the foreground has been established, the past copula (which may also mean ‘to become’) is likely to occur, as in the following example. (5) Genesis 3:1 (NASB) we-han-nah2as9 håya6 <årûm mik-kol h2ayyat§ has-såd§äh and-the-snake be.3M.SG.SC crafty from-all beast-of the-field >as9är <åsåh YHWH >älohi6m. which make.3M.SG.SC LORD God. Now the snake was more crafty than any beast which the LORD God had made. The following model describes vital parts of the Hebrew verbal system in narrative prose (Dahlgren 1998: 76–77).7

7. Here is also an example of how Egyptian Arabic fits into such a diagram.

 Sven-Olof Dahlgren

Diagram. Foreground/background in relation to time B A C K G R O U N D

Y (events)

a

b t(0)

t(0)

A

c

B

t(0)

C

FOREGROUND

N X(time)

Events in foreground: A, B, C Events in background: a, b, c N T(0): narrator’s time

– Event a is pluperfect in English, but perfect in this system, since its reference point is in A. – Event b (located at t(0)) is past tense in English, but present time in this system, since it’s reference point is in B. – Event c is past future in English, but simple future here, since it’s reference point is in C. We can now summarize the different verb forms in narrative prose in Biblical Hebrew. Table 2. Hebrew verb forms in narrative prose

Foreground Background

unmarked

marked

wayyiqtol qatal, yiqtol, we-yiqtol, qotel,we-qatal

qatal

Since background is itself a marked category in relation to foreground marked categories are not noted here. Qatal in foreground is not common but it appears to be more commonly employed in Late Biblical Hebrew (LBH). Qatal with SV ( subject-verb) order may be used in foreground for topicalization or focusing. In the following example from Early Biblical Hebrew we have a topicalized subject.

The relevance of tense and aspect in Semitic Languages 

(6) 2 Samuel 13:35–36 (NASB) wayyo>mär yônåd§åb§ >äl-ham-mäläk§ hinneh b§nê-ham-mäläk§ say.3M.SG.PC2 Jonadab to-the-king look sons/of – the-king bå>û come.3M.PL.SC And Jonadab said to the king: “Behold, the king’s sons have come; ki-d§b§ar û and-look sons/of the-king come.3M.PL.SC and behold, the king’s sons came.” The demonstrative particle hinneh ‘lo!’, ‘behold!’ corresponds to the Arabic particle >inna, which is regarded as a topicalizing particle by some linguists (Dahlgren 1998: 214–216). In Hebrew, however, its basic use is together with a qatal verb to introduce a surprising or sudden referent on the scene. As such it may also introduce a focused referent in the foreground as in the following example from LBH. (7)

Daniel 10:10 (NASB) w-hinneh yad§ någ<åh bî and-look hand touch.3F.SG.SC Then behold, a hand touched me

A topicalized constituent w-k§ol-yhûd§a without hinneh but with qatal heb§î>û (SC) appears in the following example from the book of Nehemiah (LBH). (8) Nehemiah 13:11–12 (NASB) wå>årîb§a& >ät§ -has-sgånīm wå>åmrå& maddûa< reprimand.1M.SG.PC2 DEF.OBJ-the officials say.1M.SG.PC2 why nä< äzab§ is forsaken.3M.SG.SC So I reprimanded the officials and said: ‘why is bêt§ hå-> älohîm wå>äqbs2-em wå>a
 Sven-Olof Dahlgren

w-k§ol-yhûd§a and-all-Juda

heb§ î>û ma<s´ar had-dåg7ån w-hat-tîrôs9 bring.3M.PL.SC tenth/of the-wheat and-the-wine

w-hay-yis2hår lå->ôs2årôt§ and-the-oil to-storehouses All Juda then brought the tithe of the grain, wine, and oil into the storehouses. In the following we present the tense value of each verbal form, regardless of discourse type. We disregard modal values here as we regard them as secondary in that they have developed from the tenses. A division is made into marked and unmarked forms. Unmarked is defined as the normal or default form which, according to Givón, is characterized by less structural complexity, higher frequency, and less cognitive complexity than marked forms (Givón 1991: 336–337). Table 3. Hebrew verbal forms and the tenses

qatal yiqtol we-yiqtol wayyiqtol we-qatal qotel

unmarked

marked

past, present (stative verbs) present, future present, future consecutive past future present

present, future past past present, future present, past past

Below we proceed to display the different verbal forms that are used for past, present and future. Table 4. The tenses and Hebrew verbal forms

Past Present Future

unmarked

marked

qatal, wayyiqtol, yiqtol, we-yiqtol, qatal (stative) yiqtol, we-yiqtol,

yiqtol, we-yiqtol, qotel qotel, we-qatal

The we-qatal is reckoned as marked for two reasons: it has a more restricted distribution than yiqtol, and it is less common for future than yiqtol (2932 instances to be compared with 5451 instances for yiqtol, as was mentioned above).

The relevance of tense and aspect in Semitic Languages 

Arabic The closest correspondence to the Hebrew relative tense system to be found in Arabic is probably the Bedouin speech of today. One indication of its adherence to this system is the absence of pluperfect verb forms different from the simple past forms (Dahlgren 1998: 75–81). The structure of the Arabic verbal systems also points to tense prominence; the suffix conjugation, called perfect, only has one type, the fa
Pers. pr.

nr 2011 24

% 99 89 1

nr 2 3

% 0 11 (60)

nr 23 0

% 1 0 0

nr 2036 27

% 100 100 1

EMMA Perfect Imperfect % Imperfect

1043 105

65 68 9

191 25

12 16 12

368 25

23 16 6

1602 155

100 100 9

Bedouins Perfect Imperfect % Imperfect

831 133

83 35 14

76 229

8 60 75

94 19

9 5 17

1001 381

100 100 28

Classical Arabic Perfect Imperfect % Imperfect

Dem. pr.

Tot.

 Sven-Olof Dahlgren

The sparse use of personal and demonstrative pronouns in Early Arabic is due to its written character; the corpus consisted of the 800 pages of prose in the biography of Muhammad by Ibn Ishaq. Since there is no requirement in Arabic for an independent noun phrase to mark the subject, such a phrase is mainly used when there is a switch of topic, and in this case a definite noun is needed to refer to the new topic. The number of imperfect, personal and demonstrative pronouns rises in the EMMA dialects, which reflect the spoken type of language. The ample use of demonstrative pronouns is due to their function as discourse pronouns, i.e. indicating the main participant(s) in a narrative. This occurs especially in Syria (Dahlgren 1998: 171). The Bedouin dialects deviate significantly from the preceding types with a significantly larger employment of imperfect (yiqtol), although the perfect (qatal) still is the unmarked form in foreground. Here the use of the personal pronouns together with imperfect is a common device to mark the main participant(s) in the narrative. It is also worthwhile noticing that of the 368 demonstratives with perfect in EMMA 243 were of VS order, whereas 225 of the 229 instances with personal pronouns in Bedouin speech had SV order (Dahlgren 1998: 232, 234).

9. Relative tense in Early and Classical Arabic Early and Classical Arabic also have relative tense, but more as an option compared to Biblical Hebrew and modern Bedouin speech of Arabic. The following example from the Biography of the Prophet shows how both absolute and relative tenses may occur together (Guillaume 1859: 438; Wüstenfeldt 1961: 653).8 (N=nominative, A=accusative, G=genitive) (9) >inna-hu h2uddit§a >anna-hum istaqallu6 bi-n-nisa6>-i TOP-it is told.3M.SG.SC that-they carry.3M.PL.SC with-the-women-G wa-l->abna6>-i and-the-sons-G It is told that they carried [their camels] with their women, children and property, wa-l->amwāi- ma
8. The Arabic text.

The relevance of tense and aspect in Semitic Languages 

yainna fı-him yawma>id§in play3F.PL.PC behind-them and-TOP in-them that day la->umm-a

The synchronic and diachronic phonology of ejectives

Language and Rhythm in Plautus: Synchronic and Diachronic Studies (Sozomena Studies in the Recovery of Ancient Texts)

Language and Rhythm in Plautus: Synchronic and Diachronic Studies (Sozomena Studies in the Recovery of Ancient Texts - Vol. 3)

Langue and Parole in Synchronic and Diachronic Perspective

Preaspiration in the Nordic Languages: Synchronic and diachronic aspects

Modality and Ellipsis: Diachronic and Synchronic Evidence (Trends in Linguistics. Studies and Monographs)

Quotative Indexes in African Languages. A Synchronic and Diachronic Survey (Empirical Approaches to Language Topology)

Studies on German-Language Islands (Studies in Language Companion Series)

Linguistic Diversity And Language Theories (Studies in Language Companion Series)

Linguistic Diversity And Language Theories (Studies in Language Companion Series)

Diachronic Change in the English Passive (Palgrave Studies in Language History and Language Change)

Diachronic Clues to Synchronic Grammar (Linguistik Aktuell Linguistics Today)

Diachronic Clues to Synchronic Grammar (Linguistik Aktuell Linguistics Today)

Synchronic and Diachronic Perspectives on Contact Languages (Creole Language Library, Volume 32)

Synchronic and Diachronic Perspectives on Contact Languages (Creole Language Library, Volume 32)

Space in Languages of China: Cross-linguistic, Synchronic and Diachronic Perspectives

Pathways of Change: Grammaticalization in English (Studies in Language Companion)

Language Complexity: Typology, contact, change (Studies in Language Companion Series)

Quantity and Prosodic Asymmetries in Alemannic: Synchronic and Diachronic Perspectives (Phonology and Phonetics, 5)