The Middle Voice in Ancient Greek: A Study of Polysemy (Amsterdam Studies in Classical Philology, 11) [Illustrated] 9050633684, 9789050633680

Citation preview

ΤΗΕ

MIDDLE VOICE Α

STUDY

ΙΝ

ΙΝ

ANCIENT GREEK

POL YSEMY

AMSΊΈRDAM

STUDIES

ΙΝ

CLASSICAL PHILOLOGY

EDΠORS

ALBERT RIJKSBARON IRENE J.F. DE JONG

HARM

PINKSτER

VOLUME ELEVEN

PREVIOUS L Υ PUBLIS HED

1.

Α. RIIKSBARON,

1991.

Grammatical Observations on Euripides' Bacchae.

R. RISSELADA, lmperatives and other Directive Expressions in Latin. Α Study in the Pragmatics of α Dead Language. 1993. 3. G. W ΑΚΚΕR, Conditions and Conditionals. An Investigation of Ancient Greek. 1994. 4. C. ΚRΟΟΝ, Discourse Particles in Latin. Α Study ofnam, enim, autem, vero and at. 1995. 5. Η. DIK, Word Order in Ancient Greek. Α Pragmatic Account of Word Order Variation in Herodotus. 1995. 6. J.E. V.D. VEEN, The Significant and the Insignificant. Five Studies in Herodotus' View of History. 1996. 7. Α. RIIKSBARON (ed.), New Approaches to Greek Particles. 1997. 8. R. RISSELADA (ed.), l..atin in Use. 1998. 9. Ρ. CLAES, Concatenatio Catulliana. Α New Reading of the Carmina. 2002. 10. Α.Μ. BOLKESTEIN, a.o. (eds.), Theory and Description in Latin Linguistics. Selected Papers from the Xlth International Colloquium on Latin Linguistics, Amsterdam 2001. 2002

2.

RUTGER J. ALLAN

ΤΗΕ

MIDDLE VOICE

ANCIENT GREEK Α

STUDY ΙΝ POL YSEMY

J.C. GIEBEN, PUBLISHER AMSTERDAM 2003

ΙΝ

Ν ο part of this book may be translated or reproduced in any form, by pήnt, photopήnt, microfilm, or any other means, without wήtten permission from the publisher.

© by R.J. Allan, 2003 I

Pήnted

in The Netherlands I ISBN 90 5063 368 4

ν

CONTENTS

ACΚNOWLEDGEMENτS

CHAMΈR Ι.

ίχ

INτRODUCΠON

The Problem 1.1 νοίce: some basic notions Ι. Ι. Ι The Prototypical Transitive Clause Ι.Ι.2 Langacker' s Billiard-ball Model 1.1.3 Prototypical Transitivity and the Middle νoice l.I.4 Markedness and the Meaning of the Acti ve ν oice Ι.2. Polysemy and the Middle ν oice 1.2.Ι Langacker's Complex Network Model Ι.2.2 The Middle ν oice as a Complex Category Ι.2.3 Kemmer, The Middle Voice (1993) Ι.3 Two General lssues Ι.3.Ι The Status of the Media Tantum 1.3.2 νalency Reduction Ι.Ο

CHAPΊ"ER

2.

ΤΗΕ MIDDLE ν OICE AS Α COMPLEX NEτwORK CAΊΈGORY

Classification of Middle Usage Types 2.Ι.Ι Passive Middle 2.Ι.2 Spontaneous Process Middle 2.Ι.3 Mental Process Middle 2.Ι.4 Body Motion Middle 2.Ι.5 Collectίve Motion Middle 2.Ι.6 Reciprocal Middle 2.Ι. 7 Direct Reflexive Middle 2.Ι.8 Perception Middle 2.Ι.9 Mental Actίvity Middle 2.Ι.Ι Ο Speech Act Mίddle 2.Ι.ΙΙ lndirect Reflexive Middle 2.2 Semantic Feature Analysis 2.3 The Structure of the Network 2.3.Ι The Semantic Map of the Middle 2.3.2 The Category Prototype

Ι

6 6 8 Ι4

Ι9

30 30 39 42 49 49 53

57 58 58 60 64 76 82 84 88 95

2.Ι Α

10Ι

105 ΙΙ2 ΙΙ8 Ι20 νοίce

Ι20 Ι2Ι

vi

CΗΑΡΊΈR

3.

ΤΗΕ MIDDLE AND PASSIVE VOICES ΙΝ Π1Ε AORIST SΊΈΜ

3.0 Introduction 3.1 The Distήbution of -η- vs. -θη3.1.1 The Distribution of -η- vs. -θη- in Homer 3.1.2 The Distribution of -η- vs. -θη- in Classical Greek 3.1.3 Morphological V ariation 3.2 The Distήbution of the Sigmatic Middle vs. the Passive Aorist Form 3.2.1 Sigmatic Middle vs. Passive Aorist Form in Homer 3.2.2 Sigmatic Middle vs. Passive Aorist Form in Classical Greek 3.2.3 Morphological Variation 3.2.3.1 Between Body Motion and Direct Reflexive 3.2.3.2 Between Perception and Mental Process 3.2.3.3 Between Speech Act and Mental Process 3.2.3.4 Between Mental Activity and Mental Process 3.2.3.5 Between Reciprocality and Collective Motion 3.2.4 Root and Thematic Aoήst 3.3 The Abstract Meaning of the Passive and the Sigmatic Middle Aoήst CΗΑΡΊΈR

4.

126 126 127 127 136 141 148 148 154 157 157 158 160 166 168 170 174

ΎΗΕ MIDDLE AND PASSIVE VOICES

4.0 Introduction 4.1 Aspect Revisited: the case of φανο\>μαι vs. φανήσομαι 4.2 Other Verbs with an Aspectual Distinction 4.3 Middle is Middle and Passive is Passive

178 178 185 199 200

'SYNONYMOUS' ACΠVE AND MIDDLE VERBS 5. 5.0 Introduction 5.1 Some Active and Middle 'Synonyms' 5.1.1 όρμάω vs. όρμάομαι 5.1.2 πειράω vs. πειράομαι 5.1.3 πολιτεύω vs. πολιτεύομαι 5.1.4 &λλομαι vs. πηδάω 5.1.5 βούλομαι vs. έθέλω 5.2 Active and Middle Body Motion Verbs 5.3 Conclusion

203 203 210 211 217 225 230 236 242 246

ΙΝ ΠΙΕ F'υruRE SΊΈΜ

CΗΑΡΊΈR

νίί

CΗΑΡΊΈR

6. GENERAL CONCLUSION

248

BIBLIOGRAPHY

251

lNDEX OF NAMES AND SUBJECTS

266

lNDEX OF GREEK WORDS

269

ACKNOWLEDGEMENTS

This

is a slightly revised editiση σf my doctσral dissertatiση, writteη at the Uηiversity σf Amsterdam. Α ηumber σf peσple have, ίη σηe way σr anσther, made aη iηvaluable cσηtributiση tσ its cσmpletiση. Here, Ι wσuld like tσ express my gratitude tσ them. bσσk

First σf all, Ι wσuld like tσ thaηk my Gymnasium teacher David Rijser. With his iηfectiσus eηthusiasm aηd stuηηiηg eruditiση, he stiπed - aηd still stirs - my lσve fσr the Classical languages. Secσηdly, Ι am grateful tσ Prσfessσr C.J. Ruijgh, my cσ­ supervisσr. Thaηks tσ σur mσηthly sessiσηs aηd his detailed cσmmeηts ση earlier versiσηs σf this bσσk, Ι was able tσ beηefit frσm his vast kησwledge σf the Aηcieηt Greek language. Thirdly, Ι wσuld like tσ take the σppσrtuηity tσ express my gratitude tσ Prσfessσr Α. Rijksbarση, my supervisσr. The stimulatiηg discussiσηs Ι had with him, aηd his critical remarks - always accσmpa­ ηied by a sσmewhat irσηic smile ('Bizar!') - have uηerriηgly laid bare the weak pσiηts σf my argumeηt. Further, Ι am grateful tσ all the peσple whσ have cσmmeηted ση earlier versiσηs σf this bσσk: Prσfessσr R.S.P Beekes, Prσfessσr S. Kemmer, Dr. F.C. vaη der Leek, Dr. R. Risselada, Prσfessσr S.R. Sliηgs, Dr. F.M.J. Waaηders, Mathieu de Bakker (whσ cσmmeηted ση aη earlier versiση σf chapter 2), and the members σf the Amsterdamse Hellenistenclub (chapter 5). Ι wσuld like tσ thank Liza Berry whσ cσπected my Eηglish aηd bσldly cσmbatted my σccasiσηal fits σf 'Duηglish' ( ... σr shσuld Ι say 'cσmbatted bσldly'?). Se Leηsseη σf the Classical Departmeηt σf the Uηiversity σf Amsterdam prσvided valuable help ίη the prσcess σf makiηg the manuscript camera-ready. Needless tσ say, Ι am sσlely respσηsible fσr any remaiηiηg eπσrs. Ι wσuld alsσ like tσ thank the Netherlands Organization for Scientific Research fσr giviηg me the σppσrtuηity tσ make a start ση my research at the Thesaurus Linguae Graecae ίη Hamburg, and the Van der ValkjΌnds which awarded me a grant, thereby eηabliηg me tσ atteηd the cσurses σf the Linguistic Summer Institute at the Uηiversity σf Califσrnia, Santa Barbara. Fiηally, my gratitude gσes σut tσ Alfruη whσ prσvided suppσrt wheηever Ι ηeeded it, aηd whσ ηever failed tσ put up with my 'ethelσ­ bσulσmai' -meditatiση sessiσηs.

Rutger Allan

Amsterdam, February 2003

CHAP'fER

1

Introduction 1.0 The Problem

The puzzling diversity of the different usage types of the middle and passive voice 1 in Greek has been the subject of debate ever since the times of the ancient grammarians. The nature of the problem becomes apparent when consideήng the brief inventory of middle uses given by Rijksbaron (cf. 2002 3 : 161-63, slightly adjusted). 2 (i)

Passive use The patient is the subject. The aorist is of the passive type (i.e. in -(θ)η- ).

(Ι)

αί νflες παρεσκ:ευάσθησαν ύπο τrον 'Αθηναίων

'The ships were prepared by the Athenians' (ii)

Direct reflexive use This use occurs with agentive transιtιve verbs, typically verbs of grooming. The subject-referent performs the action ση himself. Agent and patient are co-referential. In contrastive contexts a reflexive pronoun is added as direct object. The aorist is of the sigmatic middle type. E.g.:

(2)

a. b.

bathed (myself)' (έμαυτόν) 'Ι prepared myself'

έλουσάμην (έμαυτόν) 'Ι

παρεσκ:ευασάμην

1 With the term middle voice Ι refer to the inflectional category in Greek, i.e. in the present stem the middle voice includes rniddle-inflected verbs with passive rneaning. The passive voice (formed with the suffix -(θ)η-) is only found in the aorist and future stern. 2 This inventory is only rneant as a bήef survey of the different rniddle uses. In ch. 2, Ι will descήbe the rniddle uses rnore extensively.

2

CΗΑΡΊΈR

1

(iii)

lndirect reflexive use This use occurs with agentive transιtιve verbs. The subjectreferent perfoπns the action in his own interest. The aoήst is of the sigmatic middle type. E.g.:

(3)

παρεσκευασάμην τaς να-uς 'Ι

prepared the ships in my own interest'

(iv)

Pseudo-reflexive and pseudo-passive use The coπesponding active transitive verbs are causative. The middle verb denotes either that the subject brings about a change of state3 to himself (pseudo-reflexive; typically verbs of movement), or that the subject undergoes a change, no agent being present (pseudo-passive; typically verbs of emotion/cognition and physical processes). The aoήst is of the passive type. E.g.:

(4)

a. άπηλλάχθην 'Ι went away' b. έφοβήθην 'Ι became afraid' c. έτάκην 'Ι melted'

(ν)

Media tantum and Passiva tantum4 The middle only verbs have aπ aorist of the middle

foπn,

and

3 The notion change of state implies that an entity is in a certain state at moment that is different from the state it was in at Τ0 • This means that, for instance, a verb Iike hit does not denote a change of state, since the direct object may be in the exact same state after it was beaten as it was before. Verbs that do denote changes of state are break, melt, destroy, remoνe, frighten, persuade. Changes of state can be subdivided into physical and mental changes of state, and changes of location. 4 Media tantum (or middle-only verbs) are middle verbs that do not have active counterparts. They are sometimes called deponentia, a term boπowed from Latin grammar. This term is Iess adequate since it suggests that these verbs have 'Ίaid off' (i.e. Iost) their active forms. There is no histoήcal evidence that this is what actually happened. In the present study, the term media tantum (or middle-only verbs) will be used as a coveήng term to refer both to media tantum in the stήct sense (i.e., verbs with middle aoήst forms) and to passiνa tantum (verbs with passive aoήst forms), cf. Rijksbaron (2002 3 : 160). Ι am aware, however, that this morphological dichotomy has important implications, especially regarding the semantics of the verbs concemed. In fact, chapter 3 will be devoted almost exclusively to the relation between form and meaning in the aoήst stem.

Τ1

INrRODUCΠON

3

are agentive. The passive only verbs have aπ aoήst of the passive fonn, and are mostly non-volitional. If so, they are similar to pseudo-passives. (5)

Media tantum: ~τιασάμην αύτόν 'Ι accused him', jumped' Passiva tantum: ησθηv 'Ι enjoyed myself'

ήλάμην 'Ι

Sirnilar enumerations of middle uses are found in all grammars of Ancient Greek, e.g. Goodwin (1895: 265-8), Kϋhner-Gerth (1: 100-29), Gildersleeve (1900-11: 64-79), Stahl (1907: 42-74), Smyth-Messing (1956: 389-398), Schwyzer-Debrunner (11: 228-41), Garcίa Gual (1970), and Martίnez Vazquez, Ruiz Yamuza & Femandez Garήdo (1999: 229-253). 5 However, such lists almost always leave the reader with the following questions: 6

(I)

ls there a semantic element common to these usage types? If so, how should it be defined?

(11)

Assuming that the various middle uses constitute a polysemous structure: in what way are the middle uses related to one another?

(ffi)

What is the explanation of the differences in tion?

aoήst

forma-

Until recently, the middle voice has been the exclusive domain 5 Other contributions to the study of the rniddle (and passive) voice ίη Ancient Greek are: Carlsson (1912), Chantraine (1927), Margulies (1929-30), Wistrand (1941), Vaillant (1942-5), Schwyzer (1943), Vendryes (1948), Blass-DebrunnerRehkopf (1979 15 : 255-65), Hartrnann (1954), Hurnbert (19542: 103-9), Boeder (1961), Bechert (1964), Zsilka (1964), Zsilka (1965), Benveniste (1966), Jahnkuhn (1969), Bader (1972), Plenat (1974), Barber (1975), Michelini (1976), Cock (1981), Ruiperez (1986), Risselada (1987), Delaunois (1988: 79-93), Anderson (1993), Bakker (1994), Perel'rnuter (1995), Sicking & Stork (1996: 131-7), Basile (1998: 320-32), Kurzova (1999). 6 Gonda ( 1960: 31) expressed his discontent about the existing treatrnents of the rniddle voice in the following rnanner: ''It is true that the old-fashioned enurnerations of the functions of this category ίη class-books (... ) created the irnpression of aπ incoherent rnixture of functions, and that atternpts to give a general definition laboured under delusions and vagueness: (... )".

4

CΗΑΡΊΈR

σf classίcal

hσwever,

the

1 aηd Iηdσ-Eurσpean

mίddle vσίce

has

liηguίsts.

becσme

the

Duήηg

the last decades,

σbject σf ίηcreased ίηterest

research. Οηe σf the factσrs that brσught abσut thίs the strίkίηg sίmίlarίty that mίddle vσίce systems dίsplay acrσss languages, eveη ίη thσse whίch are geηetίcally ηση­ related. Of great ίmpσrtaηce tσ the receηt fruίtful research ση mίddle vσίce pheησmeηa ίs the ίηsίght that reflexive systems - as thσse fσuηd ίη maηy σf the mσdern Eurσpeaη laηguages - are fuηctίσηally hίghly simίlar tσ (iηflectίσηal) mίddle vσίce systems as we fίηd them ίη Aηcieηt aηd Mσdern Greek aηd Sanskrίt. Οη accσuηt σf thίs remarkable fuηctίσηal sίmίlaήty betweeη reflexίve aηd mίddle vσίce systems (here used ίη the strίct seηse), Kemmer (1993) subsumes bσth grammatίcal systems uηder a sίηgle, semaηtίcally defίηed ησtίση middle voice. Mίddle vσίce systems and reflexίve systems are ησt σηly fσuηd ίη ancίeηt laηguages lίke Greek, Latίη and Sanskrίt, but alsσ ίη mσdern Eurσpeaη laηguages lίke, fσr example, Rσmaηce, Slavίc, Germanίc (Eηglίsh beίηg a ησtable exceptίση), aηd ίη ηση-Ιηdσ­ Εurσpeaη laηguages, as ίη Tamίl (Dravίdίaη) aηd Fula (Nίger-Cσηgσ ), see Klaίmaη (1991). ίη typσlσgίcal

ίηcreased ίηterest ίs

Majσr cσηtrίbutίσηs tσ σur kησwledge σf mίddle vσίce pheησ­

are frσm a crσss-lίηguistίc staηdpσίηt, are: Faltz (1977), Shίbataηί (1985, 1988), Lίchteηberk (1985), Geηίu~ίeηe (1987), Crσft, Shyldkrσt & Kemmer (1987), Klaίmaη (1988, 1991, 1992), Kemmer (1993, 1994), Fσχ & Hσpper (1994), aηd Gίνόη (1994). Besίdes these geηeral wσrks ση mίddle vσίce pheησmeηa, several ίmpσrta.ηt language-specίfίc studίes have appeared durίηg the last decades. These ίηclude the wσrks σf Geπίtseη ( 1990, 1992) [Russίan], Maldσηadσ (1999) [Spanίsh], aηd Manηey (2000) [Mσdern Greek]. 7 Apart frσm studίes that are cσηcerned exclusίvely wίth mίddle­ reflexίve systems, there have beeη a ηumber σf ίmpσrtaηt develσp­ meηts ίη liηguίstίc theσry that may ίηcrease σur ίηsίght ίηtσ the mίddle vσίce as a grammatίcal categσry. These develσpmeηts wίll be the tσpίc σf the fσllσwίηg sectίσηs. Ιη sectίση 1.1, Ι wίll dίscuss the ησtίση σf Prσtσtypίcal Traηsitίvίty (Hσpper & Thσmpsση 1980) aηd ίts crucίal ίmpσrtaηce tσ vσίce dίstiηctίσηs. Ιη sectίση 1.2, the ηature σf cσmplex meηa, mσst σf whίch

Comprehensive studies ση the semantics of other ancient Indo-European Ianguages are: Neu (1968a, 1968b) [Hittite], Flobert (1975) [Latin], Gonda (1979) [Vedic Sanskήt], Schmidt 1969 [Tocharian].

INrRODUCΠON

5

polysemous categoήes (Lakoff 1987, Langacker 1987, 1991 a, 1991 b, 2000) and the notion of category prototype will be discussed.

6 CHAP'fER 1

1.1 Voice: some basic notions 1.1.1 The Prototypical Transitive Clause

Voice altemations can be characteήzed as encodings of different choices of clausal subject (cf. Langacker 1991a: 335). For exarnple, the farniliar passive construction can be descήbed as a marked expression of the special configuration in which the expected subject is bypassed in favour of a less typical subject. It has often been observed that voice altemations can be fruitfully descήbed as markings of departures from the prototypical transitiνe eνent (Hopper & Thompson 1980; Givόn 1984: 157, 2001a: 126-8; Langacker 1991a: 335). 8 The importance of the notion of prototypical transitivity for grammatical organization was first fully recognized by Hopper and Thompson in their seminal article published in 1980. On the basis of vast typological evidence they claimed that the grammatical category transitiνe is structured around a prototype that can be defined by means of a cluster of semantic properties. These properties are listed as follows (Hopper & Thompson 1980: 252; the terminology used is theirs):

8 In this study, the tenn 'event' is used in a broad sense subsuming several, more specific, event types such as states (static; non-volitional), processes (dynamic; non-volitional), and actions (volitional). Thus it is roughly equivalent to Dik's tenn State of Affairs (Dik 1997a).

INrRODUCΠON

7

Table 1: Semantic properties of the prototypical transitive clause

Α.

Participants

Β. Κinesis

C. Aspect D. Punctuality Ε. Volitionality F. Affirmation G. Mode Η. Agency Ι. Affectedness of Ο J. Individuation of Ο

HIGH Transitivity

LOW Transitivity

2 or more participants action telic punctual volitional affirmative realis Α high in potency9 Ο totally affected Ο highly individuated

1 participant non-action atelic non-punctual non-volitional negative iπealis Α Ο Ο

low in potency not affected non-individuated

Clauses can be ranked ση a scale of transitivity according to whether they have properties in common with the Prototypical Transitive Clause. In other words, clauses can be more or less transitive. Hopper & Thompson (1980: 253) give the following examples (6)

a. Jerry likes beer b. Jerry knocked Sam down

Example (b) is much higher in transitivity than (a) because it displays the following properties: • • • • •

Kinesis: action Aspect: telic Punctuality: punctual Affectedness of 0: total Individuation of 0: high; rejerential, animate, and proper.

For each of the semantic properties of the prototypical transitive clause (τable 1), Hopper and Thompson have cited languages in which clauses that possess this particular property display transitive behaviour. Conversely, clauses that differ from the Transitive Prototype with regard to only one of these properties are found to display intransitive behaviour, even if they possess other prototypical transiti-

9 Following Dixon's example (Dixon 1979), Α and Ο denote Agent and Object. These notions roughly coincide with transitive agent and patient.

8

CΗΑΡΊΈR

1

ve properties. Of the features enumerated by Hopper and Thompson, Givόn ( 1984: 20) considers two of pήmary importance: agency and affectedness. In this way, the prototypica1 transitive is pήmaή1y conditioned by: a. the presence of a visible, volitional, and controlling cause/agent; and b. the presence of a clearly visible result-registeήng effect/patient.

(7)

Furthermore, according to Givόn, agency and affectedness (and, therefore transitivity) are strongly conditioned by the clausal property of perfectivity. The connection between these notions can be descήbed as follows (Givόn 1984: 157; italics are his): a. Affectedness of patient: ''τhe more completed an event is, the more likely it is that the patient in fact registers to the full the effect of the action'' b. Effectiveness of agent: ''The more successfully completed the event is, the more likely it is that the agent was in fact the deliberate, direct, effective cause of that successful completion"

(8)

Examples of prototypical transitives given by Givόn (1984: 20) are: 10

a. Mary cut the meat b. John destroyed the house

(9)

1.1.2 Langacker's Billiard-ball Model The notion of prototypical transitivity also plays a major role in Langacker's Billiard-ball model. The Billiard-ball Model is a cognitive model. In Cognitive Grammar, it is claimed that meanings are to be analyzed in relation to cognitive domains. The notion of cognitive domain can be defined as "[a] context for the characterization of a

10

More examp\es can be found in Givόn (200\a: 127).

INrRODUCΠON

9

semantίc uoίt" (Laogacker 1987: 147). For example, the meanίog of the word glass evokes the cogoίtίve domaίos of space [ίο whίch all coocrete objects exίst], of shape [typίcally cylίodήcal], material [glass], size [easίly held ίο the hand], but also more complex domaίos such as ίts typίcal function. The typίcal fuoctίoo of a glass could be characterίzed as an ίostrumeot used ίο the process of dήokίog: ίt ίs fίlled wίth a lίquίd that ίs suίtable for human coosumptίoo, ίt ίs grasped and lifted wίth the arm, etc. Cognίtίve domaίos can, therefore, be vίewed as knowledge bases relatίve to whίch the meanίog of ao expressίoo cao be assessed. 11 The οοtίοο of cogoίtίve domaίo ίs very sίmίlar to Lakoffs οοtίοο of Idealίzed Cognίtίve Model (Lakoff 1987), and to other ootίoos such as frame (Fίllmore), scene, schema, or script. The character of a cognίtίve domaίo cao be of a more ίdealίzed or archetypal oature. Such an ίdealίzed cogoίtίve domaίo can be refeπed to as a cogoίtίve model, whίch bήogs us back to Langacker's Bίllίard-ball Model. The Bίllίard-ball model ίs an archetypal cognίtίve model that structures our cooceptίoo of eveots. The model cooceίves the world as cootaίoίog dίscrete objects whίch are coostaotly movίog arouod, makίog cootact wίth ooe aoother, aod partίcίpatίog ίο eoergetίc ίoteractίoos (Laogacker 1991a: 283). As a result of the ίoteractίoos of dίffereot objects an action chain aήses: Fίg.

1: Action Chain

Ooe eotίty (represeoted by a cίrcle) ίs charged wίth eoergy, and makes forceful cootact wίth a secood eotίty. As a result, eoergy ίs traosmίtted from the first source-eotίty to the secood (thίs ίs showo as a double aπow). Thίs eotίty ίs thereby dήveo ίοtο cootact wίth a thίrd, whίch 11 Typical1y, as we have seen in the case of the word glass, a 1inguistic unit invokes rnu1tip1e cognitive dornains, and the nurnber of dornains evoked cannot be sharp1y delirnited. This view can be characteήzed as the encyclopedic view of sernantics (as opposed to the dictionary view). In the encyc1opedic view, it is c1airned that there is no specific boundary between 1inguistic and non-linguistic know1edge (cf. Langacker 1987: 154-166).

10

CΗΑΡΊΈR

1

again brings about a transmission of energy. This process can continue indefinitely, until, finally, an entity absorbs the energy and thereby undergoes a change of state. The simplest instance of an action chain is one in which the initial energy-source (which Langacker calls the head of the action chain) interacts directly with the final energy-sink (the tail of the action chain). This configuration is depicted in Figure 2:

Fig. 2: The Prototypical Transitive Event

Agent

Patient

Here, the action chain head is the agent, and its tail the patient. In the Prototypical Transitive Event, the agent and the patient figure in their most prototypical role. Langacker characteήzes agent and patient as follows: ( 1Ο)

''The archetypal agent is a person who volitionally initiates activity resulting, through physical contact, in the transfer of energy to an external object. lts polar opposite is an archetypal patient, an inanimate object that absorbs the energy transmitted via externally initiated physical contact and thereby undergoes an internal change of state.'' (Langacker 1991 a: 285) phy~ical

The patient's change of state is depicted by a squiggly arrow in Fig. 2. lt should be noted that Langacker's charactization of agent and patient concerns archetypes. Obviously, not every clause conforms to these archetypes since human experience is too ήch and varied. For instance the archetypal patient, as descήbed by Langacker, concerns an object that is affected by the event. However, an important class of patients that are similar to this archetype concerns effected objects. These are objects that have no prior existence but rather are created by the event itself, as in She built α house. Effected objects are generally not

lNrRODUCΠON

11

distinguished from affected objects by any special marking (see Langacker 1991a: 362). The concept of the prototypical transitive event is also highly relevant to the grammatical relations of subject and object. The unmarked coding of the prototypical event is that the subject is the agent, and the object is the patient (see Langacker 2000: 25, Givόn 2001a: 126, inter al.). Other, less-prominent participants in the event, such as instruments or expeήencers, are typically coded by oblique cases or prepositional phrases. This brings us to the issue of semantic roles. 12 There is much controversy on the definitive number of semantic roles and their exact definition. Every linguistic theory, or perhaps even every individual linguist will posit a different inventory of roles that are, furthermore, defined differently. However, these inventoήes generally display a considerable similarity. In this study, Langacker's description of a number of archetypal semantic roles is adopted. The list of semantic roles enumerated by Langacker ( 1991 a: 285-7) is as follows: (11)

• agent and patient (see definition above) • instrument: a physical object manipulated by the agent to affect the patient, thereby serving as an intermediary ίη the transmission of energy • experiencer an animate entity engaged in a mental event • zero: an entity that merely occupies a location or exhibits a static property.

Langacker' s role archetypes listed above are of a highly abstract character. 13 This is a reflection of their cognitive fundamentality and their ubiquitous manifestation in the languages of the world. However, Langacker (1991a: 285) admits that these roles do not form an exclusive club, and that finer distinctions can be made. Α number of additional roles are useful in an adequate descήption of the semantics 12 Semantic roles are also known as case roles, thematic relations, theta-roles, notional roles, and semantic functions. 13 Langacker also distinguishes the semantic role of moνer. The mover, according to Langacker, is aπ entity that undergoes a change of location, as Rocky in Rocky droνe to the beach and the door in Ι opened the door. In this study, however, moving participants will be viewed either as agents (if the motion is volitional) or as patients.

12 CHAP'fER 1 of the middle voice in Greek. These roles, controversial: (12)

Ι

suspect, will not be very

• beneficiary: an animate entity receiving benefit as a result of the event. • recipient: an animate entity into whose possession something is transfeπed. • cause: an inanimate entity that causes a physical or mental change in another entity. • source: the location from which an entity moves, or, metaphorically, the stimulus of a mental event. • goal: the location towards which an entity moves.

lt should be noted that many accounts of semantic roles do not distin-

guish the cause-role from the instrument role (cf. Givόn 1984: 126, 2001a: 161-2; Palmer 1994: 5). Α cause is similar to Dik's semantic function Force (Dik 1997a: 118; cf. also Van Valin & LaPolla 1997: 85). The cause-role differs from an instrument in that it does not imply the presence of an initiating agent: the cause itself is conceptualized as an autonomous initiator. In Greek the cause is typically expressed by the dative case (dative of cause), but a cause can also have the status of subject in a transitive clause. Since causes are autonomous initiators (i.e. only differing from agents in volition), this is not surpήsing. For instance, (13)

a. ό σίδηρος τψ πυρ1αusε τήκεται 'The iron is being melted by the fire' b. το ΠUPcAUSE τήκ:ει τον σίδηρον 'The fire is melting the iron'

The cause-participant may appear in mental events, e.g., (14)

ilσθην άπειλαι~Αusε (Ar. Eq. 696) Ι

was delighted at your threats

Α source is basically a spatial role, but it can also appear in an extended metaphorical sense. Examples of mental sources (or stimuli) are the genitival complements of verbs of perception, emotion, and cognition:

(15)

a. Perception: άκ:ούσαντες τf\ς σάλπιyyοςsοuRcε (Xen. An. 4.2.8)

INτR.ODUCΠON

13

Having heard the trurnpet ( ... ).

b. Emotίon: ό άνηρ ( ... ) κήδεται τ&ν ΘηβαίωνsοuRcε (Xen. HG 6.4.5) The rnan (... ) cares about the Thebans.

c. Cognίtion 14 : μάθε δέ μoυsouRcε κ:αi τάδε (Xen. Cyr. 1.6.44) Learn frorn rne this too.

In the above, the notion of prototypical transitive event, which involves a physical transmission of energy from an agent to a patient, was set out. lt ίs, however, a widely-occurring phenomenon that the transitive clause structure is extended to code other .situation types. An example of this phenomenon is the way mental events (perception, cognitίon, and emotion) are treated. Examples of these are 'see', 'know', 'understand', 'want', and Ίove' (see Langacker 1991a: 303-4, 310, 2000: 26). These events are coded ίη English as a transitίve with a subject and a direct object, although they obviously do not ίnvolve a physical transmίssion of energy. (16)

Ι

{see/know/understand/wantllove} it (Langacker 1991a: 303)

This extension, from the prototypical transitive event to the mental event, has a metaphorical character. Its motivation can be found ίη the abstract commonality that is inherent ίη both types of events. On the one hand, we have the transmission of energy from an active initiator (the agent) to a passive endpoint (the patient), and on the other hand, we have the concept of a metaphorical mental path leading from a more active, conscious participant (an experiencer) to a more passive object-participant. In other words, mental phenomena such as gazes and direct attention can be conceived of as paths, analogical to a physical path like that of an energy flow. 15 Examples ίη Greek of

14 With verbs of cognition and perception, the genitival cornplernent can also denote the entity about which one perceives or leams sornething, e.g. πατρος άιcούσας (δ ι 14) ('having heard about father'). This type of genitive is often viewed as a partitive genitive. 15 Often, verbs of perception and cognition are based on a rneιaphor, e.g. English perceive frorn Latin -cipiό 'grasp'; English grasp 'understand'; Dutch be-grijpen frorn grijpen 'grasp' On the irnportance of rneιaphor to the verbs of perception and cognition, see Sweetser (1990).

14

CHAPτER

1

coded as transitive coπstructioπs are πumerous: γιγνώ­ σκω τι 'Ι realize sth.', οtδά τι 'Ι kπow sth.', όράω τι 'Ι see sth.', τ ρέω τι 'Ι fear sth.' . 16 meπtal eveπts

1.1.3 Prototypical Transitivity and the Middle Voice the previous sectioπ it was observed that iπ the uπmarked case, the of the prototypical traπsitive is coded by the subject, aπd the patieπt by the direct object. Coπsider ποw the followiπg clauses (from Laπgacker 1991a: 335): Ιπ

ageπt

( 17)

a. He opened the door. b. The door opened very easily. c. The door suddenly opened. d. The door was opened.

Here we can observe that ίπ the uπmarked prototypical transitive clause (a), the ageπt is subject, the patieπt is object, aπd the verb is ίπ the active voice. The other clauses depart from this prototype ίπ οπe crucial respect: the patient is coded as the subject. This departure resides ίπ the fact that the participaπt that is expected to be the subject (the ageπt) is bypassed ίπ favour of a less qualified candidate: the patieπt. Clauses (b) aπd (c) are formally ίπ the active voice, but they desigπate oπly the patieπt's participatioπ. Ιπ (b) the adverb very easily implies the efforts of aπ uπspecified ageπt, which are facilitated by the iπhereπt characteristics of the patieπt-subject (the door). Ιπ (c), the implicit refereπce to aπ ageπt is πoπ-salieπt aπd may be abseπt altogether. Clauses like (b) aπd (c) are sometimes called cases of middle diathesis, siπce they are formally ίπ the active voice (at least ίπ Eπglish), but semantically beloπg to the middle domaiπ. Ιπ clause (d), though the ageπt is left uπspecified (as is the case ίπ most passive clauses), the efforts of aπ ageπt are defiπitely implied. Ιπ this respect the passive clause (d) differs from clauses (b) aπd (c). Ιπ the case of the Eπglish passive the deviatioπ from the prototypical traπsitive aπaπgemeπt is marked by the alternatioπ of voice: iπstead of the

The non-prototypicality of these transitive verbs is manifested by the fact that they do not have a passive (οίδα, τρέω), or acquire a passive only after Homer's time (γιγνώσκω, όράω). 16

INΓRODUCΠON

15

unmarked actίve voίce the verb has the marked passίve voίce. 17 In the example of the famίlίar English passίve, ίt can be observed that the notίon of prototypίcal transίtίvίty ίs hίghly relevant to voίce altematίons. Below, Ι wίll argue that the prototypίcal transίtίve event ίs also essentίal to the characterίzation of the middle voice ίη Ancient Greek (cf. Questίon (I) in (1.0)). There have been many attempts to capture the essence of the semantics of the Greek middle voice. Thίs ίs not aπ easy task ίf one consί­ ders the dίversίty of mίddle usages such as passίve, ίntransίtίve 18 , dίrect reflexίve and ίndίrect reflexίve. lt is clear that a core-meanίng, if there ίs any, could only be of a hίghly abstract nature. lt ίs, therefore, useful to consίder some of the most ίmportant defίnitίons of the middle voίce. The natural startίng-poίnt of thίs overvίew ίs, of course, the unrivalled grammar of Kϋhner and Gerth (K-G): ( 18)

''Dίe Medίalform bezeίchnet eίne Thatigkeίtsausserung, welche von dem Subjekte ausgeht und auf dasselbe wίeder zurϋckgeht. Dίese von dem Subjekte ausgehende und auf dasselbe wieder zurϋckgehende Thatίgkeίtsausserung kann entweder bloss auf das Subjekt beschrankt seίn, als: βουλεύομαι, ίch berate mίch, λούομαι, ίch wasche mich, oder auf eίn Objekt seίner Sphare, (... ), als έκοψάμην τftv κεφαλήν ίch schlug mίr das Haupt, κατεστρεψάμην τitv -yίlv, ich unterwarf mir das Land (... )." (K-G, 1: 100)

K-G's

definitίon

is of a remarkable oήginalίty since, as we shall see, strongly from later definitions. However, ίt ίs interesting to note that K-G's definition seems to convey aπ ίmagery that ίs related to modem vίews ίη whίch the middle voice is claimed to express that the subject is the Startingpointllnitiator (cf. "ausgeht") as well as the Endpoint ("zurϋckgeht") of the actίon chain (cf. Croft, Shyldkrot & Kemmer 1987; for sίmίlar vίews: Κlaiman 1988: 26-27, Κlaiman ίt dίffers

17 The main function of the passive construction is often taken to be to defocus the agent (see Shibatani 1985), or, as Gίνόη puts it: "The agent is exιremely nontopical ('suppressed', 'demoted'), so that the patient is the surviving topical argument in the clause" (Gίνόη 200lb: 94). ΙR Here used in the semantic sense. The intransitive middle covers Rijksbaron's pseudo-reflexive and pseudo-passive.

16

CΗΑΡΊΈR

1

1991, Geπitsen 1990: 11, Croft 1994: 103). Ι will elaborate on this matter later. However, as is the case with these more recent views, KG's characterization only concems the rniddle in the strict sense, i.e. it does not include the passive middle. 19 More farniliar formulations of the meaning of the middle voice are:

(19)

"The Middle Voice denotes that the subject is in some especial manner involved or interested in the action of the verb.'' (Gildersleeve 1900: 64)

(20)

"Verba (... ), die ihren Schauplatz in der Sphare des Subjekts haben, bei denen das ganze Subjekt als beteiligt scheint.'' (Brugmann 1903: 104)

(21)

"Dans l'actif les verbes denotent un proces qui s'accomplit a partir du sujet et hors de lui; dans le moyen, qui est la diathese a definir par opposition, le verbe indique un proces dont le sujet est le siege; le sujet est interieur au proces.'' (Benveniste 1966: 172)

(22)

indo-iranien et en grec les desinences moyennes indiquent que le sujet est interesse d'une maniere personelle au proces." (Meillet 1937: 244)

(23)

''The implications of the middle (when it is in opposition with the active) are that the 'action' or 'state' affects the subject of the verb or his interests." (Lyons 1969: 373)

'Έη

19 In modern linguistic literature there is much confusion about the terms νoice and diathesis. Many authors seem to use the terms interchangeably. Others seem to use diathesis as pertaining to the formal properties of the verb (e.g. Klaiman 1988: 27; Duhoux 2000: l 17), and νoice as pertaining to the semantic properties of the verb. Again, other authors use the terms the other way around (e.g. Geniu§iene 1987). In the Latin grammatical tradition νδχ (the ancestor of νoice) seems to refer to the form of the verb, i.e. the way a word sounds. lt is, indeed, a translation of the Greek term φωνή which means 'sound' as well as 'voice' (in the physical sense). However, it could also be used by the grammarians in a more technical fashion as in the sequence of sounds that form a word. The meaning of the Greek grammatical term διάθεσις is very hard to pin down. For valuable discussions of the ancient grarnmarians on διάθεσις and μεσότης (middle), Ι refer to Collinge (1963), Garcfa Gual (1970), Rijksbaron (1987), Anderson (1989, 1993), and Lallot (1989, 1997).

INτRODUCΠON

17

With regard to these definitions one can object either that they seem to focus on the indirect reflexive meaning of the middle (Gildersleeve and Meillet: the subject's interest), or, conversely, that they seem to be more suitable for the passive and intransitive uses of the middle (Brugmann and Benveniste: "the whole subject participates" and "the subject is intemal to the process", respectively). Only Lyons' characterization pertains clearly to both the 'passive' meanings ("affects the subject") and the indirect reflexive meaning ('Όr his interests"). Lyons' recognition of these two extremities of the meaning of the middle, and the transparency of his formulation, make his definition appealing. Lyons uses the verb affect in a sense that is broad enough to be applicable to all middle meanings. lt is this broad sense of affect and affectedness that we need in aπ adequate definition of middle meaning. 20 If we are tempted to interpret affect and affectedness in a naπower sense, that is, as aπ equivalent of the ancient term πάθος, we inevitably run into trouble, since the notion πάθος pertains to passίvίty21 , as opposed to ένέργεια, which pertains to actίvίty. The ancient grammarians, however, are clear in that they consider indirect 20 The teπn oήginated from the narrower, emotional meaning 'affect', which appears especially in the Geπnan tradition. For instance, Schwyzer (11: 228) emphasizes that "das affektische Moment" is crucial to the semantics of middle verbs. The verba affectuum (νerbs of affection, e.g. fίδομαι, φέβομαι) are a special and productive group, of which the active transitive counterparts are either rare or nonexistent. Therefore, Schwyzer concludes, the 'affective moment' must also be the distinctive feature in opposition to the active in the other verb classes. This conception of 'affectedness' is obviously too narrow. The difficulties are especially apparent in the case of an inanimate subject (e.g. ό σίδηρος τήκεται 'the iron is melting'). 21 The teπns πάθος and παθητική are applied by the ancient grammaήans not only to the passive in the stήct sense (e.g. τύπτομαι 'Ι am being beaten'), but also to direct reflexives (e.g. έλουσάμην 'Ι washed myself (Apollonius, Macrobius and Choeroboscus; see Rijksbaron 1987: 434n8, Lallot 1997, 1: 216), and even to active verbs like νοσ& 'be ill' and όφθαλμι& 'suffer from ophthalmia' (Macrobius; see Rijksbaron 1987: 438). Furtheπnore, Apol\onius Dyscolus (150 Uhlig) has a subclass of παθητική called αύτοπάθεια 'auto-passivity', which includes active verbs like θν~σκω, όφθαλμι&, πάσχω (see Lallot 1997, 1: 255). In sum, it appears that πάθος and παθητική are semantic teπns, designating events in which the subject undergoes the action. This means that the teπn πάθος only partly relates to the meaning of the middle voice. Incidentally, according to Lallot (1997, Π: 246n359), the element αύτοίn αύτοπάθεια expresses that the element of passivity is designated by the verbal lexeme itself (i.e. not by means of passive morphology).

18

CΗΑΡΊΈR

1

reflexive middle verbs such as

έποιησάμην, έγραψάμην

as having an

ένέργεια-meaηίηg. 22

Lyons' definition was taken up by Barber (1975). She presents the semantic implications of the active and the middle ίη Greek ίη the following dίagram (Barber 1975: 21): Fig. 3: Active vs. Middle (Barber 1975)

ACΠVE Plaίn

(no incoming aπows)

Active

ss ---+

MIDDLE (incoming aπows) Fώl

Middle

ss~

Reflexive

ss~

Reciprocal

ss .....

Passive

ss.-

The letters 'SS' stand for Sentence Subject. The arrows indίcate the direction of the action resulting ίη affectedness. The term 'Full Middle' coπesponds roughly to the indίrect reflexive middle. Barber adds that "( ... ) the active represents the one case ίη which the subject is not specified as being affected by the action; whereas the middle subsumes all cases ίη which the subject is affected by the action ( ... )" (Barber 1975: 21). In sum, Barber sees the middle voice as a marker of subject-affectedness, taken ίη the broad sense. This compή­ ses, on the one hand, affectedness ίη which the subject is very much like a patient (as ίη the passive, reflexive and reciprocal middle), and, on the other hand, affectedness ίη which the subject is similar to an 22 This seems to be implicated by Apollonius III, 30 (see Lallot 1997, 1: 216; ΙΙ: 172). In the ancient grammarians, the term μέση/ μεσότης (middle) seems to refer to a formal category, consisting of (i) 'passive' forms with active meaning (e.g. έποιησάμην), and (ii) active forms with passive meaning (e.g. γf:yονα). The term could also apply to those forms that can both have an active meaning and a passive meaning, e.g., βιάζομαι Ί assault s.o.' and Ί am being assaulted', διέφθορα Ί have destroyed' and Ί am destroyed' (cf. Dionysius 13.8; Lallot 1989: 55, 166-7).

IN'rRODUCΓION

19

indirect object (as in the indirect rniddle). As a working-hypothesis, Ι will subscribe to the notion of subject-affectedness as the abstract rneaning of the rniddle voice. The advantages of the notion of subjectaffectedness are that it is not too vague, and that it potentially subsurnes the different rniddle uses in an adequate fashion. The validity of this working-hypothesis will be tested in chapter 2, in which each of the rniddle uses is discussed in rnore detail. Ι would now like to discuss the question as to how the notion of prototypical transitivity is relevant to the sernantics of the rniddle voice. As Ι have expounded in section 1.1.1, the prototypical transitive clause can be defined in the following rnanner: an agent-subject volitionally initiates physical activity resulting in a transfer of energy to a patient-object that absorbs the energy and thereby undergoes an intemal change of state. As a rule, the verb in a prototypical transitive clause has the active voice. Now the rniddle voice can be defined as a rnarked coding of a departure frorn the prototypical transitive. Contrary to the prototypical transitive, the subject, in sorne way or other, undergoes an effect of the event. This effect can be of a physical or a rnental nature, and it can be direct or indirect (in that it involves an extemal object). In chapter 3, it will be argued that the notion of prototypical transitivity is also relevant to the sernantics of the activerniddle-passive tήchotorny in the aorist stern.

1.1.4 Markedness and the Meaning of the Active Voice In the previous section, Ι have been concemed with the abstract rneaning of the rniddle voice. The rneaning of the rniddle voice was characterized as a rnarked coding of a departure frorn the prototypical transitive event. Now the question arises as to the rneaning of the active voice. Can we sirnply conclude that the active voice designates the absence of subject-affectedness in opposition to the rniddle voice? Ι will argue here that this conclusion cannot be drawn. Instead, the active voice rnust be taken as the unrnarked rnernber of a privative opposition. In other words, the active voice is neutral as to the sernantic feature of subject-affectedness. 23

The markedness of the midd1e voice in Greek has been advocated earlier by Gua1 (1970: 11-12, 29-32), and Ruiperez (1986). Gonda (1979: 39) reaches the same conc1usion with respect to the middle voice in Rigvedic Sanskrit, which is 23

Garcίa

20 CΗΑΡΊΈR 1

useful approach to markedness-phenomena can be found in Croft (1990). Croft argues that the many criteria that have been proposed (notably by Greenberg (1966)), can be reduced to three general ones (cf. 1990: 64ff.): Α

1 Structural:

number of morphemes used to express marked and unmarked values.

2 Behavioural: Inflectional: number of cross-cutting distinctions the (a) marked and unmarked values contain. Distributional: number of syntactic environments in (b) which the marked and unmarked values occur; Cross-Iinguistic: number of language types in which the (c) marked and unmarked values occur. 3 Frequency: (a) Textual: number of occuπences of the marked and unmarked values in text; Cross-Iinguistic: number of languages in which the (b) marked and unmarked values are found. will argue now that these markedness criteria support the claim that the active voice is the unmarked category, whilst the middle voice is the marked category. Note that two criteria relate to cross-linguistic comparison (2c, 3b). These criteria Ι will leave out of consideration.z4 The criterion of structural markedness (1) involves counting the morphemes that signal the category in question. Croft (1990: 73): Ι

(24)

Structure: The marked value of a grammatical category will be expressed by at least as many morphemes as the unmarked value of that category.

highly similar to that of Ancient Greek: "( ... ) the active is not essentially the exact opposite of, or contrary to, the medium, but it is characteήzed by not expressing that which is conveyed by the medium, viz. some special reference to the subject. That means that it can be used when the author [of a Sanskήt text, RJA] does not deem it necessary to express the medial modification explicitly". 24 That the 'medio-passive voice' was typologically marked compared to the active voice has already been observed by Greenberg (1966: 45-6).

INτRODUCΠON

2ι

Compare the regular thematic endings (plus thematic vowel, since they are not always clearly separable) of the present stem indicative in Attic Greek: Table 2:

Actίve

and

mίddle:

present and imperfect

Present

ι sg 2 3 2 du 3 ι pl 2 3

ίndίcatίve endίngs

Imperfect

Active

Middle

Active

Middle

-ω

-ο-μαι

-ο-ν

-ο-μην

-εις

-η ι

-ε-ς

-ου

-ει

-ε-ται

-ε-0

-ε-το

-ε-τον

-ε-σθον

-ε-την

-ε-σθην

-ε-τε

-ε-σθον -ε-σθον -ο-μεθα -ε-σθε

-ου σι

-ο-ντα ι

-ε-τον -ε-τον -ο-μεν

-ε-τε

-ο-μεθα -ε-σθε

-ο-ν

-ο-ντο

-ο-μεν

Greater morphological complexity of the middle compared to the active is found in the ι sι person sg. present (-ω vs. -ο-μαι), 3'd person sg. (-ει vs. -ε-ται), and 3'd person singular of the imperfect (-ε-0 vs. -ε-το). ln the ι sι and 3'd sg. present, the active form contains a unanalyzable portmanteau-morpheme (-ω and -ει) that simultaneously expresses aspect stem, tense, voίce, and person. Conversely, the middle forms contain two morphemes; one expresses aspect stem (the thematic vowel, -ο- or -ε-), the other expressing tense, voice, and person. The fact that a number of middle endings display a greater complexity is evidence that the middle voice is marked as compared to the active. 25

25 lt is interesting to note that the middle endings in Proto-Indo-European (PIE) are a composite of active (secondary indicative) endings and sιative endings. E.g., the I sg. middle ending *-m-h 2 consists of the active ending *-m and the stative ending *-h2 ; the 3 sg. middle ending *-t-o is composed of the active ending *-t and the stative ending *-ο (see Beekes 1995: 240-2). Thus, in contrast to the active, the middle endings in PIE are clearly morphologically marked, through the addition of the stative morpheme. This sιative morpheme was quite possibly a clitic personal pronoun in oήgin. In the course of time the oήginal compositionality of the middle endings in PIE had become bluπed. For the different theoήes about the oήgin of the PIE middle endings and the semantics of the middle voice in PIE, see Narten (1968),

22

CΗΑΡΊΈR

1

Another way of establishing structural markedness involves comparing the phonological heaviness of the respective morphemes. Croft ( 1990: 76) expresses reservations about the validity of this cήteήon 26 • Although there is a tendency for morphemes expressing marked values to be physically longer, there is frequently a mismatch between physical length and markedness. The example given by Croft is from Spanish: the 1sι person plural of the verb hablar 'speak' is hablamos 'we speak' vs. 2nd person hablάίs 'you speak'. Here, the fact that the ending of the first-person plural contains more phonemes, conflicts with the typological evidence for person markedness. There is a cross-linguistic tendency for third persons (indicative) to be least marked, followed by first persons, and for second persons to be most marked (see Croft 1990: 93, citing Greenberg). For a clear example of this kind of 'markedness mismatch' in Greek, consider the second-person singular endings of the middle, -ηι (present) and -ου (imperfect). Although there is a tendency for 2nd persons to be marked as compared to the 1sι and 3n1 persons, the Greek 2nd person present and imperfect endings are shorter than the other persons. Generally, these mismatches will be the result of a histoήcal­ phonological change. One example is the disappearance of a phoneme as a result of a sound-law. In the case of the second person middle in Greek, an intervocalic -σ- has disappeared (*-ε-σαι > -ε-αι > -ηι and *-ε-σο > -ε-ο > -ου). In the Spanish example, an intervocalic -t- was deleted (cf. its Latin predecessor fabulatίs). 21 It is clear that we must, indeed, exercise some caution when we attempt to determine markedness on the cήteήon of phonological heaviness. However, when establishing a markedness pattem in the Neu (1968a, 1968b, 1976, 1985, 1989), Oettinger (1976, 1993), Jasanoff (1978), Strunk (1980), Kort\andt (1981), Georgiev (1985), Rix (1988), Stempel (1996), Kurzova (1999), and Pooth (2000). 26 Cf. a\so Comήe (1976: 111): "( ... ); in general, morphological cήteήa are the least telling, since the morphology often reflects systematic coπespondences of an earlier peήod of a Ianguage." 27 Interestingly, the iπegular 2nd person middle ending -ηι is Iater replaced by the transparent ending -ε-σαι. This restoration of the 2nd person ending first occuπed in the athematic verbs (e.g. ϊστα-σαι in Homer), analogically after forms like ιcάθημαι κάθησαι (where the -σ- resulted from -σσ-). In the Koine, the thematic verbs are affected, e.g., -ποιε'i-σαι (see Schwyzer, 1: 668-9). In this way, the expected markedness pattem is restored. The new middle ending -ε-σαι is morphologically more complex, and phonologically more heavy than the active ending -εις.

INτRODUCΠON

23

active and middle voices in Greek, we can take a number of fonns (namely two times eight personal fonns, see the table above) into consideration, instead of just one pair (as in the examples above). In this fashion, we may be able to detect a tendency that is more robust. Thus, when we contrast the active endings with the coπespon­ ding middle endings (-ω vs. -ο-μαι, etc.), we can observe that the middle endings contain more phonemes in 14 of the 16 fonns of the paradigm. Only in two cases, namely the two second-person singular fonns, is the pattem different. The primary endings (-εις vs. -η ι: both syllables have three morae28 ) and the secondary middle endings (-ες vs. -ου: both have two morae) are equivalent as to phonological heaviness. Here again, the exceptional case is due to a sound-change: the fonn *-ε-σο was originally longer than the coπespondent active fonn -ε-ς. 29 In sum, it is safe to conclude that the middle voice is structurally (in particular, phonologically) marked compared to the active 30 • This bήngs us to Croft's second criteήon: behavioural markedness (1990: 77). This criteήon can be divided into two types: a morphological type, which Croft calls injlectional (2a), and a syntactic type, called distributional (2b ). The fonner pertains to the number of fonns in aπ inflectional paradigm. The latter pertains to the number of syntactic contexts in which a grammatical element can occur. Croft defines the inflectional cήterion thus (1990: 79):

28 My measure of morae is as fol\ows: a syllable ending in a short vowel has one mora; a syl\able which ends in a long vowel or a consonant has two morae; syllables ending in a long vowel and a consonant have three morae. 29 Like the pήmary indicative -ηι >> -ε-σαι, in later Greek the ending -ου is analogically replaced by the transparent -ε-σο. 30 For a similar conclusion regarding Modern Greek, see Haspelmath (1993: 99). Note further that the aoήst endings (in Ancient Greek) confirm the markedness pattern. In al\ persons but the second (-σα-ς vs. -σω (< *-σα-σο)), the rniddle ending is longer. Ι did not take the aoήst forms into account here because of the complication of the passive aoήst.

24 CHAP'fER 1

(25)

Behaviour (inflectional): if the marked value has a certain number of distinct forms in aπ inflectional paradigm, then the unmarked value will have at least as many forms in the same paradigm.31

With respect to inflectional behaviour, the active and the middle exhibit an important difference. The active voice has two sets of endings (conjugations), the thematίc (l sg. -ω, 2 sg. -εις, 3 sg. -ει, and 3 pl. -ουσι), and the athematίc (l sg. -μι, 2 sg. -ς, 3 sg. -σι, 3 pl. -ασι), whereas there is only one set of middle endings ((-ο)-μαι, etc,). The fact that the active voice has more distinct forms is an indication that it is the unmarked category as compared to the middle. The second type of behavioural criterion discussed by Croft (1990: 8lff.) is the distrίbutίonal type (2b). This relates to the number of environments in which the linguistic element in question occurs: (26)

Behaviour (distributional): if the marked value occurs in a certain number of distinct grammatical contexts (constmction types ), then the unmarked value will also occur in at least those contexts that the marked value occurs in.

According to Croft ( 1990: 89-91 ), the phenomenon that is often called 'neutralization' or 'neutral value-criterion', is to be considered a subtype of the distributional criterion. Contextual neutralίzatίon implies that, although the meaning predicts that both forms should be able to occur in a given context, in fact only one occurs, this being the unmarked form. In Ancient Greek, it can be observed that in a number of contextual environments, the active voice is used, even though, from a semantic point of view, the occuπence of the middle voice would be considered possible. Consider the following examples: (i) Active verbs are used suppletively in passive constructions (K-G, 1: 98-1 00; Smyth-Messing 1956: 397-8; Schwyzer-Debrunner, 11: 226-7). The clearest example of this phenomenon is the active verb (άπο)θν~σκ:ω 'die' which is used as a suppletive passive to (άπο)-

31 Greenberg (1966: 29) gives a more specific criterion: An unmarked form will have at least as many allomorphs or paradigmatic irregularities as the marked form. In the same vein, Comήe (1988: 19-20) notes that the vaήety of conjugationa1 c1asses is larger in the unmarked member of the opposition.

ΙΝτR.οουcτιοΝ

25

So (άπο)θνfισκω ύπό means 'be kίlled by' .32 The actίve voίce cannot possίbly express absence of subject-affectedness, sίnce that would be ίη clear contradίctίon to the passίve semantίcs of the constructίon. Other examples of thίs phenomenon are φεύγω 'flee', used as a passίve of διώκω 'prosecute' ίη juήdίcal contexts. Thus, φεύγω ύπό means 'be prosecuted by' In compounds, -πίπτω 'fall' functίons as a passίve of -βάλλω 'throw', e.g. έκπίπτω ύπό 'to be thrown out by' Another example of an actίνe form wίth a passίve meanίng ίs the aorist form έάλων 'was caught', whίch has a middle present form άλίσκομαι 'be taken, be caught' (see also sectίon 5.0). (ίί) The actίve voίce may be used in contexts ίη which ίt is clear that the subject benefits from the action. Clear examples are those ίη whίch actίve and middle forms are used alternately. If ίt ίs ίnferrable from the context that the action ίs performed in the ίnterest of the subject, the use of the mίddle form ίs not obligatory (Gίlder­ sleeve 1900: 66, Garcίa Gual 1970: 12). For ίnstance, κτείνω 'kίll'

(27)

[The Peloponnesians] ( ... )

άναγαγόμενοι &μα eφ tπλεον, έπ1.

τεσσάρων ταξάμενοι τaς ναυς, παρa τitν έαυτ&ν Ύfιν εσω έπ1. του κόλπου δεξιφ κέρ~ ήγουμένφ, rοσπερ και rορμουν. έπ1. δ' αύτφ εϊκοσιν έtαξαν τaς &ριστα πλεούσας, (... ) (τh. 2.90.1-

2)

(... ) so they put out to sea at dawn, and, after lining up their ships four deep, sailed along their own shore towards the inner part of the gulf, in the same order as they had lain at anchor, their ήght wing leading the way. Upon their ήght wing they had placed their twenty best sailing ships, (... ). 33

The

expresses that the Peloponnesίans lίne up theίr for theίr own use. In the case of the actίve form (ταξαν, selfprofίt ίs left implicίt. mίddle ταξάμενοι

shίps

The regular passive form κτείνομαι does not occur in Attic. The translations accompanying the examples are taken from the Loeb-editions. In places, they have been adjusted. 32 33

26

(28)

CΗΑΡΊΈR

1

a. [The Egyptians] ( ... ) (Hdt. 2.36.2)

άπο όλυρέων ποιεUνται σιτία,

(... )

( ... ) they make their bread from spelt (... ).

b.

[The Egyptians]

άρτοφαγέουσι

δe

έκ

τ&ν

όλυρέων

ποιεUντες liρτους, (... ) (Hdt. 2.77.4) They eat bread which they make from spelt (... ).

Here again, the active form (ποιε\:ιντες) is used in a context in which it can be infeπed frorn the verb άρτοφαγέουσι 'they eat bread' that the bread is rnade for the benefit of the subject (see also Cock 1981: 16).

Additionally, a cornparable type of neutralization is found in the verbs of eating and dήnking. This class of verbs are rnostly active, e.g. έσθίω, βιβρώσκω, τρώγω, πίνω (see Schwyzer-Debrunner, Π: 225-6). Since these activities are ίnherently for the benefit of the subject, there is no need to code thern with the rniddle voice. For a sirnilar reason, an expression such as άνοίγω τitv θύραν 'Ι open the door' is always active. The (indirect reflexive) rniddle form is never used, because it is only natural that one opens the door for oneself. (iii) Α third type of neutralization occurs in contexts with reflexive pronouns. In the following exarnple, the active voice is used in cornbination with a dative reflexive pronoun αύτψ 'for oneself The reflexive pronoun is used - instead of an indirect reflexive rniddle without pronoun - in cases in which the reflexivity is ernphasized (i.e.: 'for hirnseif'). Frorn a sernantic point of view, the rniddle voice would also have been possible in these cases (K-G, Ι: 110-1): (29)

[Philip weakened Macedonia] ( ... ), tτ' έπισφαλεστέραν ύπftρχε φύσει κατεσκεύακεν αύτψ (Dern. 2.15)

η

[Philip] has rendered [Macedonia] for his own benefit even more insecure than it was by nature.

That also the rniddle voice would have been possible is shown by the following exarnple, where we find both a rniddle verb and a reflexive pronoun:

INrRODUCΠON

(30)

27

τί τftν πόλιν, Αίσχίνη, προσfικε ποιεtν άρχην και τυραννίδα τ&ν 'Ελλήνων όρ&σαν έαυτφ κατασιcευαζόμενον Φίλιππον;

(Dem. 18.66) What should the city have done, Aischines, when she perceived that Philip was establishing a tyranny over the Greeks for himseif? Α similar phenomenon is found in direct reflexiνe contexts: sometimes the actiνe form is used in combination with a reflexive pronoun, in cases in which a middle form would have been possible as well, e.g.: 34

(31)

'Άδρηστος δε (... ), ο.Οτος δη ό φονεuς μεν του εωυτου άδελφεου γενόμενος, φονεi>ς δε του καθήραντος, (... ) έπικατασφάζει τφ τύμβφ έωυτόν (Hdt. 1.45.3) But Adrastus (... ), he who had become the slayer of his own brother and the slayer of his puήfier, (... ), slew himself over the grave.

The following example shows that the middle voice may also be employed35 : (32)

και οί μέν φασι βασιλέα κελε\>σαί τινα έπισφάξαι αύτον Κύρφ, οί δ' έαυτον έπισφάξασθαι σπασάμενον τον άκινάκην

(Xen. An. 1.8.29) And one report is that the Κing ordered someone to slay him upon the body of Cyrus, while others say that he drew his dagger and slew himself.

In both cases mentioned above, the reflexive pronoun is used emphatically in a contrastive context (see also section 2.1.7). (iv) Another indication that the active voice is the unmarked value relates to the passive aorist form (see also Ruijgh 1991: 691 ). If we judge the passive aorist form solely by its endings (1 sg. [-θη]-ν, 2 sg. -ς, 3 sg. -0, etc.) it should be regarded as belonging to the active voice. However, in the system of voice oppositions, in the aoήst stem the suffixes -σα- and -(θ)η- also play aπ essential role. Therefore, it is justified to take the passive aorist as a distinct voice. The active

34 Gonda (1979: 51) mentions that the same ru\e (i.e. when a reflexive pronoun is present the middle form is optional) applies for Rigvedic Sanskrit. This ru\e was already articu\ated by Ρίίηίηί (1.3. 77). 35 The direct reflexive middle -σφάττομαι without reflexive pronoun expresses 'to ki\1 oneselr without emphasis or contrast (see also section 2.1.7).

28 CHAP'fER 1

endings are used unproblematically in the passive aoήst, since the 'passive' meaning (i.e. subject-affectedness) is expressed by the portmanteau morpheme -θη- which codes both aspect and voice. In my view, the case of the passive aoήst forms clearly demonstrates the semantic neutrality of the active voice endings. Apparently, active endings can even be used in contexts of high subject-affectedness such as the passive aoήst. 36 As opposed to the passive aoήst formations, the passive future forms do show middle endings (λυθήσομαι 'Ι will be made loose'; not: *λυθήσω). This redundant marking of subject-affectedness can be βήσομαι, εστην explained analogically with pairs like ί::βην στήσομαι, and perhaps under influence of the middle futures with passive meaning (e.g. αίρήσομαι 'Ι will be taken'). The redundancy of the middle endings in the passive future in Attic and other dialects, is shown by the fact that the Doήc dialects have active endings in the passive future (-θη-σέω). Το conclude the discussion of distήbutional behaviour, we have seen that the active voice readily occurs in contexts in which the element of subject-affectedness is clearly present. This, subsequently, entails that the active voice cannot be thought of as a marking of absence of subject-affectedness. Rather, the active voice must be considered as neutral to the element of subject-affectedness. Croft's next criterion regards text-frequency (3a) (1990: 85): (33)

Frequency (textual): if a marked value occurs a certain number of times in frequency in a given text sample, then the unmarked value will occur at least as many times in a comparable text sample.

The table in Rijksbaron (2002 3: 142) shows that the middle voice is,

36 In ch. 3, Ι wίll argue that the passίve aoήst fonn conveys a high degree of subject-affectedness. Α contrary vίew is advocated by Bakker (1994). He sees the actίve endίngs of the passίve aoήst as desίgnating the absence of affectedness which is due to the punctual aspect of the aorist stern: "Now when the passive event is construed, not ίη ίts inherent duratίon, but as an objectίve punctual event (e.g. punishrnent as the transίtίonal point between the sιate of being guίlty and the sιate of having been punished [as ίη the case of ιcολασθηναι, RJA]), the affectedness disappears, and this is signaled by -the- + actίve (=non-rnίddle) rnorphology)" (Bakker 1994: 40).

INrRODUCΠON

29

indeed, less frequent. In Herodotus 39 % of the counted verb forrns were middles; in Plato 33 %. The ratio active versus middle forrns in a given text is approximately 2 1. The slight difference between Herodotus and Plato is largely due to a difference in the use of the passive. As is well known, the use of passives is dependent ση genrefactors.37 summarize, all markedness-cήteria point in the same direction: the active voice is the unmarked member in the opposition with the marked middle voice. Since the active voice can occur in environments in which the subject is affected (contextual neutralization), it can be concluded that the active is unspecified as to the semantic feature subject-affectedness. 38 Conversely, the middle voice is semantically marked with respect to affectedness of the subject. As a consequence, event types that do not involve subject-affectedness cannot be expressed by a middle verb. The two major event types that do not involve subject-affectedness are the prototypical transitive and the stative event type. In the prototypical transitive event, the subject is an unaffected volitional agent, while the object is the sole participant undergoing the effect of the event. In stative events there is no affectedness, since affectedness can only be the result of a change that is taking place or has taken place. Examples of active stative verbs are: είμί 'be'' ζω 'live'' καθεύδω 'sleep'' μένω 'stay, remain'' οζω 'smell' That stative verbs tend to be active is also shown by the denominative verbs meaning 'be NOUN/ADJ', e.g. άσθενέω 'be weak' (άσθενής 'weak'), βασιλεύω 'be king' (βασιλεύς 'king'), δουλεύω 'be a slave' (δΟ'uλος 'slave), έρυθραίνω 'be red' 39 (έρυθρός 'red'), εύτυχέω 'be prosperous' (εύτυχής 'prosperous'), ήσυχάζω 'be quiet' (ilσυχος 'quiet'). Το

37 An interesting example of this phenomenon is the fact that passives are fairly rare in the oral narrative of Homer (e.g. the passive of δίδωμι appears only twice, against hundreds of active forms), whereas passive verbs are wel\-attested in the administrative Mycenaean texts (e.g. passives of δίδωμι: di-do-to, de-do-me-na). 38 The asymmetry of the contrast between active and middle voice can be compared to the contrast between unmarked lion and marked lion-ess. The former term can a\so be employed in contexts in which the contrast is neutralized, as in We saw some bored lions in the zoo, where the group of Iions may well contain some fema\e specimens (see also Cruse 2000: 173). 39 Note, however, that active έρυθραίνω can a\so mean 'make red'. Compare, by contrast, middle έρυθραίνομαι 'become red'

30

CΗΑΡΊΈR

1

1.2 Polysemy and the Middle Voice

of five differeπt middle usage chapter 2, Ι will argue that it is useful to distiπguish eveπ more middle uses. Οπ the other haπd, as we have seeπ above, it has beeπ the communis opinio that it is possible to assigπ aπ abstract meaπiπg to the middle voice. How do these two facts relate? The questioπ is whether we should pursue a polysemous approach, that is, attempt to distiπguish the differeπt specific meaπiπg variaπts, aπd to aπalyze the way they are iπteπelated. Or, should a monosemous approach be pursued by tryiπg to defiπe a core-meaπiπg that is shared by all middle uses? Ιπ the latter, the differeπt middle uses are to be coπsidered πο more thaπ variaπt interpretatίons that are completely determiπed by the coπtext. 40 Απ objectioπ to the moπo­ semous approach iπ geπeral is that defiπitioπs of the abstract meaπiπg of lexical items or grammatical categories teπd to be uπspecific to such aπ exteπt that also the meaπiπg of oppositioπal forms (iπ our case active forms) are covered by them. However, although οπe caπ raise some objectioπs to the moπo­ semous approach, it is iπdeed possible to defiπe aπ abstract meaπiπg that is iπhereπt to all variaπt rniddle uses. As we have seeπ iπ sectioπ 1.1.3, this abstract meaπiπg caπ be characterized as subject-affectedness. If we dispeπse with this observatioπ, we would miss aπ importaπt geπeralizatioπ. Ιπ the followiπg sectioπ, it will be argued that the moπosemou aπd the polysemous approach are ποt πecessarily mutually exclusive. Απ attractive syπthesis is offered by Laπgacker's theory of complex network categories. Ιπ sectioπ

types was

(1.0),

Rijksbaroπ's distiπctioπ

preseπted. Ιπ

1.2.1 Langacker's Complex Network Category

The complex category model, as it is developed by Laπgacker (1987), builds οπ Rosch's psycholiπguistic work οπ semaπtic categoήes (e.g. Rosch 1973, 1975, 1978). Rosch's psycholiπguistic expeήments have showπ that semaπtic categoήes teπd to have a highly complex iπtemal

40

With respect to the middle voice in Ancient Greek, this is the position taken

by Andersen (1993).

INτR.ODUCΠON

structure.

31

are σf a flexible and graded members are 'better' members than σthers. Membership σf a categσry is determiηed accσrdiηg tσ the degree σf resemblance tσ a ceηtral member, σr prototype. This way σf determiηiηg categσry membership is thus radically differeηt frσm the 'Aήstσtelian' way which requires a member σf a categσry tσ pσssess sσme essentίal Bσuηdaries betweeη categσήes

ηature, aηd sσme

attrίbute.

The prσtσtype σf a categσry is geηerally defiηed as the best exemplar σr the typίcal ίnstance. Prσtσtypical members have the largest ηumber σf attήbutes ίη cσmmση with σther members σf the categσry aηd the smallest ηumber σf attributes which alsσ pertaiη tσ members σf ηeighbσuήηg categσries. Fσr example, as was shσwη by Rσsch (1975), the mσst typical member σf the categσry BIRD tumed σut tσ be ROBIN (at least, fσr Nσrth Americaη studeηts, whσ were her test-subjects). Οη the categσry bσuηdary were birds like OSTRICHES (which dσ ησt fly) and PENGUINS (which dσ ησt fly and dσ ησt have clearly distiηguishable feathers). Eveη BATS were sσmetimes iηcluded ίη the BIRD categσry, which shσws the gradedηess and flexibility σf categσry bσuηdaήes. 41 The view that categσήes are ηση-discrete can alsσ be fσuηd ίη Wittgeηsteiη's Philosophische Untersuchungen (1999: 277-8). Ιη a passage ση the categσry SPIEL, he reaches the cσηclusiση that there is ησ siηgle prσperty that is cσmmση tσ all games. Rather, the categσry is structured as a netwσrk σf σverlappiηg similaήties ("ein kσmpli­ ziertes Netz νση Ahnlichkeiten"), which Wittgenstein called "Familieηahnlichkeiteη" (family resemblances). The categσry SPIEL is like a family. lt is impσssible tσ pσint tσ a feature that is shared by all family members. Instead, what σne can say is that member Α has a certain feature χ ίη cσmmση with member Β, member Β shares with member C a feature y, and sσ fσrth. Α nσtable differeηce betweeη Wittgenstein's categσry and the categσry as it is viewed in prσtσtype theσry is the idea that, in the latter theσry, ησt all members are equal: sσme members are 'mσre member' than σthers. Rσsch' s idea σf categσήzatiσn as hingiηg ση the perceived resemblance σf categσry members tσ a prσtσtypical member has Νσw

what is a

categσry prσtσtype?

41 Α nice example of how peήpheral members of one category can switch to another category is Dutch wal-vis 'whale'. The name shows that this animal once was considered a member of the category FISH (Dutch vis), rather than a mammal. Clearly, of both categoήes, FISH and MAMMAL, the whale is a peήpheral member.

32

CΗΑΡΊΈR

1

inspired a number of cognitive linguists to develop models of categoή­ zation. Important examples are the radίal network model (Lakoff 1987), the complex network model (Langacker 1987), and the famίly resemblance network (τaylor 1989). These models are concemed with all linguistic categoήes, including both lexίcal ίtems (e.g. Gennan Spίel) and morphosyntactίc structures (e.g. the transitive construction). Linguistic categories typically display a polysemous structure of inteπelated and conventional values, centered around a prototype. 42 Now, Langacker's model differs from the two other cognitive models in one important respect. Besides the level of the concrete, polysemously inteπelated meanings of a linguistic unit, he posits a higher-level abstract schema. The notions prototype and (abstract) schema are contrasted by Langacker in the following way: (34)

prototype is a typical instance of a category, and other elements are assimilated to the category οη the basis of their perceived resemblance to the prototype; there are degrees of membership based οη the degree of similaήty. Α schema, by contrast, is an abstract characteήzation that is fully compatible with all the members of the category it defines (so membership is not a matter of degree); it is an integrated structure that embodies the commonalities of its members" (Langacker 1987: 371). ''Α

The notion of ;ιbstract schema is roughly comparable with the traditional notion of abstract meanίng or core meanίng. Langacker' s complex network model can therefore be considered a synthesis of a purely polysemous approach (involving one fonn with different, yet related meanings) with a monosemous approach (involving one fonn with one meaning). The essential building blocks of Langacker' s complex category

42 Important works on linguistic prototypes are Givόn (1984, 2001a), who discusses the cognitive and bio1ogical basis of 1inguistic prototypes; Lakoff (1987), who discusses the deve1opment of the notion prototype in psycho1ogy and linguistics, and presents some convincing prototype-analyses; Langacker (1987, 1991a, 1991b); Taylor (1989), on prototype categoήes; Croft (1991), on syntactic categoήes; Geeraerts (1993), on Iexical semantics; and Goldberg (1995), on constructions. Prototype-based studies on the middle voice are Croft, Shyldkrot & Kemmer (1987), Kemmer (1993), Maldonado (1999), and Manney (2000).

INτRODUCΓION

33

model are prototype, extensίon and abstract schema. In order to grasp the idea of a complex category network, it may be useful to have a look at an example of a lexical category given by Langacker himself (Langacker 1987: 373-385). 43 Consider a child in the process of learning the various senses (conventional usages) of the word tree. ln his early expeήence, the word is first applied to familiar specimens like oaks, elms, and maples. These concrete applications of the word tree may be called usage events (Langacker 1987: 66). Α usage event can be characterized as an actual instance of language use, involving a full, context-dependent (and therefore richly detailed) understanding, that is paired with an actual vocalization. On the basis of a series of usage events, the child will extract a conception that embodies the commonalities of these trees, while properties that vary from one instance to the next wίll be ignored. Thίs abstraction of the concept TREE 144 , though ίt excludes subtler and more contingent properties, wίll nevertheless be fairly concrete and specific as the concept probably involves ίntrinsίc, characteristic, and cognitively salient properties, such as shape, size, color, brachiation, and leaves. Since this concept of TREE 1 is an abstraction from vaήed concrete usages, it constίtutes a low-level abstract schema. Wίth repeated usage, thίs conception will become more deeply entrenched. 45 This ίnitial conception can be recognized as the category prototype. Subsequently, suppose the child is confronted with a pίne. He wίll easily learn to call it a tree, either because somebody calls the pίne a tree, or because ίt is the most applicable term at his disposal. Thus, the original prototypical use of the word tree is extended to a new use, namely ΡΙΝΕ. This semantic extension is based on the judgement of similarity of the new concept ΡΙΝΕ with the highly salient and highly entrenched concept TREE 1 Ι refer to the passage cited for a considerably rnore elaborate treatrnent of the exarnple. For a cornparable discussion of the expansion of a lexical category, see Langacker (2000). 44 Words in capitals here represent conceptual content (roughly the signijil'), whereas srnall italics represent the phonological content (signijiant). 45 Every single use of a linguistic structure (e.g. a lexical itern or a rnorphosyntactic structure) has a positive effect ση its degree of entrenchrnent. Conversely, disuse has a negative irnpact. With repeated use, a novel structure will becorne progressively entrenched, and it will becorne a conventional unit. Α unit is a cognitive structure rnastered by the speaker so that it can be ernployed in a largely autornatic fashion, without requiήng attention to its individual parts or their arrangernent. Since entrenchrnent is dependent on frequency of use, it is a rnatter of degree (cf. Langacker 1987: 59-60). 43

34

CHAPτER ι

which was already acquired by the child. The observation of similarities between the older concept TREE 1 (roughly a broad-leaved tree) and the new concept ΡΙΝΕ takes the foπn of an abstract concept TRE~ that embodies the commonality of TREE 1 and ΡΙΝΕ. This concept TRE~ abstracts away from the conflicting properties of the two lower level concepts, e.g., it will have to be neutral as to the difference between leaves and needles. This higher level concept τRΕ~ is the abstract schema of which the two more concrete uses of the word tree are elaborations. The process descήbed here results in a simple schematic network depicted in figure 4: Fig. 4: Extension and Schematization: TREE

'fREEι

.Γ-\

----------------~~

Note that the concept TREE 1 is depicted by a square. This is a notational convention used by Langacker to indicate that the concept at issue is entrenched. The circles indicate that these concepts are not (yet) entrenched. The solid arrows indicate that the low-level uses are elaborations/specifications of the abstract schema. The dashed arrow indicates that ΡΙΝΕ is an extension of the prototype TREE 1• The process descήbed above can now be repeated. For instance, the next stage could involve the acquisition of the concept of ΡALM as a variant use of the word tree, which, in tum, results ίη an even more abstract schematic concept TREE3• This abstract schema should be neutral with respect to, for example, the degree of brachiation, and the size and nature of the foliage. This expanded network category is

INτRODUCΠON

35

Fig. 5: Expansion of the network: TREE

Earlier in this section, the major ingredients of the complex category model were presented, namely prototype and extension, abstract schema and elaboration. The complex category can be thought of as a network. The nodes of the network consist of particular linguistic structures. These might be allophones of a phoneme, variant senses of a lexical item (as in the tree-example), or variant forms of an elaborate grammatical construction. In addition, the nodes may vary as to the degree of specificity of their content. The nodes are linked to each other in pairwise fashion, thereby indicating their relatedness. Α full-blown network category may look like the following diagram (from Langacker 2000: 14):

36

CΗΑΡΊΈR

1

Fig. 6: Complex Category Network

The prototype of the category represented here is node Α: almost all the other nodes are either extensions of Α (indicated by the dashed arrows), or elaborations (the solid arrows). Node Α is also the most entrenched node, which is indicated by the thickness of the box. Node C could be considered a secondary prototype. Furthermore, three levels of schematicity are represented in the diagram. On the lowest level are the structures that are specified in most detail: Cι, Αι, Α2 , Α 3 • On the mid-level the structures are located that are more schematic than the lower-level structures: Α and C are abstract schemas of which Αι, Α 2 , Α 3 and C 1 are elaborations (note the solid arrows). The highest level consists of structures Α' and C' These are 'superschemas' of the category. In tum, the mid-level subschemas are elaborations of these superschemas (hence the solid arrows). Another dimension that is represented in the diagram is the aspect of 'distance' The nodes are depicted at greater or smaller distances from one another, depending on the difference in degree of elaboration or, in the case of extension, on the number of salient features two nodes have in common. lt is important to keep in mind that the image of a network is a metaphor. lt can capture (and visualize) a number of essential properties of the structure of a complex category, as was explained above, but, just like any metaphor, it has its limitations. For example, the discreteness of the boxes in the network may create the false impression that the boundaries between the nodes in the network are clear-cut. In fact, the transition between the nodes (e.g. semantic

INrRODUCIΊON

37

variants of a lexical item) must be viewed as gradient. In a complex category, there are typically many cases which, having features in common with two nodes (or even more than two ), fall between the two of them. Nevertheless, the image of a network can be considered adequate since most instantiations do tend to center around a typical case. If this were not so, and in-between cases were more than a relatively marginal phenomenon in our category, we would have to reconsider our analysis, and to see whether we should postulate an additional node in the network. Το improve our understanding of the nature of Langacker's model, it is important to note that it is characteήzed as a usage-based model. Langacker describes this model as follows (1987: 494): (35)

"Substantial import is given to the actual use of the linguistic system and a speaker's knowledge of this use; the grammar is held responsible for a speaker's knowledge of the full range of linguistic conventions, regardless of whether these conventions can be subsumed under more general statements. [It is a] nonreductive approach to linguistic structure that employs fully articulated schematic networks and emphasizes the importance of low-level schemas."

In Langacker ( 1991 a, chapter 10), the model is descήbed as a maxί­ malίst, non-reductίve, and bottom-up approach, which is consistent with the nature of cognitive linguistics. Now what do these labels mean? The usage-based model is maximalist and non-reductive in the sense that it recognizes that the linguistic system is a massive, highly redundant inventory of conventional units. In order to become a fluent speaker, a human being has to do a vast amount of actual learning. As Langacker (2000: 2) puts it: 'Ίf one aims for psychological reality, it cannot be maintained on purely methodological grounds that the most parsimonious grammar is the best one.' ' Related to this issue is the non-reductive aspect of the model. This concems the idea that a grammar includes both general rules or schemas as well as expressions that instantiate these rules. For instance, a speaker typically will have learned a general rule for the formation of plural nouns (Ν + -s) as well as certain specific instantiating expressions such as dogs, houses, cars. In some approaches to

38

CHAPτER

1

grammar these instantiating expressions are excluded from the grammar on grounds of economy, since these expressions are regularly deήvable by rule: to list them individually would be to miss a generalization. This reasoning can be called the rulellist fallacy which is based on the assumption that rules and lists are mutually exclusive (Langacker 1987: 29, 2000: 3). Therefore, abstract schemas and specific expressions are part of the grammar, provided that they have become conventional units, through entrenchment by frequent, repeated occuπence. As was shown before, abstract schemas are representations of commonalities observed across a number of specific expressions. For instance, the abstract schematic pattem of noun-pluralization is leamed through the observation of specific forms like dogs, houses, etc. These specific expressions must have become entrenched units before the abstract pattem is extracted. lt is implausible that these specific plural forms will have suddenly lost their status as entrenched, conventional units the moment the general rule is acquired, and that from that moment onward the specific expressions must be computed from scratch οη the basis of the rule (cf. Langacker 1991 a: 263). It is more plausible that abstract schemas/general pattems and specific instantiating expressions coexist as stored units. 46 Ι will now move on to the last characteήstic of the usage-based approach: the bottom-up oήentation. This basically amounts to a difference in emphasis that is given to the different levels of a complex category. Low-level schemas, which are extracted when a language user comes into contact with a fully contextualized instance of a category, are considered of more importance in language processing than higher-level schemas that are extracted from the low-level schemas. ''If high-level schemas are extracted, they may be only of secondary significance, serving more of an organizing function than an active computational one" (Langacker 1991b: 265). For instance, when somebody hears the word dogs, he will immediately access from memory the highly entrenched low-level unit dogs as a pre-packaged 46 The consequence of this view of grammar is that there is no pήncipled division between lexicon and grammar. The grammar of a language is viewed as a structured inventory of conventional linguistic units. Linguistic units may vary in degree of compositionality and specificity. The more specific and the less compositional a fixed unit is, the more it can be viewed as a lexical item in the traditional sense. The point is that this is a matter of degree, just as the determining parameters compositionality and specificity.

INrRODUCΠON

39

whole, instead of computing its structure on the basis of the higherlevel schema of noun-pluralization (Ν + -s). Moreover, many complex categories, both lexical and grammatical, lack a single abstract superschema that subsumes all the lower-level values shown by a particular form. In that case, we are dealing with a classic instance of family resemblances (as for instance Wittgenstein's SPIEL). The view that lower-level schemas are more important to language structure than high-level schemas is in accordance with the observation that it is generally not possible to predict the actually occuπing instantiations of the high-level schema. This entails that the language user has to know which of the potential low-level instantiations of the high-level schema actually do exist, and which don't. For instance, the passive form *κτείνομαι ύπό 'be killed by' could, potentially, be sanctioned by the abstract schematic meaning of the middle voice, namely subject-affectedness. However, a speaker of Attic Greek had to learn that this instantiation is ungrammatical, and that the construction θν~σκω ύπό was used instead. 1.2.2 The Middle Voice as a Complex Category

One of the main objectives of the present study is to show that the grammatical category of middle voice can be insightfully analyzed as a complex network category. In section 1.0, Ι described the five different middle usage types as identified by Rijksbaron (2002\ These are the passive use, the direct reflexive use, the indirect reflexive use, the pseudo-reflexive/pseudo-passive use, and the middle-only use. These usage types can be identified with the concrete, low-level senses of the complex category. These variant uses of the middle voice are semantically related to one another in a polysemous fashion. In the complex category models, these relationships are called extensions. This means that, although they share certain semantic features, there are salient features in which they differ. As was noted above, the abstract meaning of the Greek rniddle voice can be defined as affectedness of the subject. This abstract meaning can be identified with the abstract schema in Langacker's model. In accordance with the usage-based model of grammar, it is conceivable that this abstract schema is less entrenched, and only of secondary importance in actual language use. In speaking and hearing, the language user is more likely to activate the more concrete rniddle

40

CΗΑΡΊΈR

1

usage types, than the rather abstract superschema of subject-affectedness. For example, it is plausible that, when a Greek heard the word ϊσταμαι ίη a context without a direct object or extemal agent, the low-level 'node' of the pseudo-reflexive41 , that specified that the subject undergoes a self-initiated change of state, was activated first and foremost. The abstract schema, with the single implication that the subject is affected, may have been activated less strongly, or not at all. 48 Obviously, this assumption cannot be tested in a dead language like Ancient Greek, and it is, therefore, not crucial to the purposes of my argument. In chapter 2, Ι will give aπ overview of the different middle usage types which Ι distinguish, and Ι will describe the structure of the semantic network of the middle voice. In several respects, my classification will be finer-grained than Rijksbaron' s presented above. This refinement is pήmarily supported by typological evidence (esp. Kemmer, Croft, Shyldkrot 1987; Kemmer 1993), and by morphological evidence from the aorist stem. After this classification of middle uses, the semantic relationships among them will be discussed. 49 This discussion will result in a 'semantic map' of the Greek middle voice. This semantic map represents the precise structure, as Ι see it, of the semantic network of middle uses and their interconnections. The semantic map approach and its implications for distributional pattems and diachrony Ι owe to Anderson (1982) [on the perfect]; Haspelmath (1987) [on reflexives]; Croft, Shyldkrot & Kemmer (1987)

In chapter 2, Ι will rename this middle usage type "body motion middle". 1t can also be assumed that these concrete middle usages are acquired first. The abstract schema is then extracted from the variant usage types ση the basis of the semantic commonalities observed (unconsciously) by the language user (see 1.2.1). 49 These semantic relationships can be identified with the extensions of Langacker's model: one polysemous meaning is an extension of a more prototypical one ifalthough they share one or more semantic commonalities there is at least one salient semantic property in which they differ. Ι shall use the more neutral term relation(ship), because it is often unclear which of the connected uses is the more prototypica\, and which is the extension. From a diachronical point of view, one can only guess as to the exact way in which the midd\e voice emerged since there already must have been a full-blown middle voice in Proto-Indo-European. Synchronically, frequency of use can be an argument for prototypicality. In section 2.3.2, Ι will try to establish the category prototype of the midd\e voice in Greek. 47

48

INτRODUCIΊON

41

[ση the middle vσice]; and Kemmer (1993) [ση the middle vσice] 50 • These are all typσlσgical studies which attempt tσ establish a uηiversally valid semantic map51 fσr each σf the grammatical categσ­ ries cσηcemed. The semaηtic map σf the middle vσice iη Aηcieηt Greek, which will be based ση a semantic analysis σf the differeηt middle usage types, will serve as a basis fσr twσ claims that will fσrm the pσiηt σf departure iη chapter 3. Bσth claims relate tσ mσrphσlσgy, the first frσm a syηchrσηical pσiηt σf view, the secσηd frσm a diachrσηical pσiηt σf view:

(I)

Α form will always cover a connected region of variant middle uses in the semantic network. 52

(Π)

Α form will only spread from one variant use to another when these uses are directly semantically related. 53

These claims are pήmarily relevaηt tσ the aηalysis σf the cσmplex mσrphσlσgy σf the aσrist stem, which will be the σbjective σf chapter 3. The first claim is based ση the geηeral fuηctiσηalist assumptiση that differeηt meanings that are expressed by the same fσrm are related, until prσven σtherwise. With respect tσ the mσrphσlσgy σf the aσrist

° Kemmer's study will be discussed in more detail in section 1.2.3.

5

Croft (2001) distinguishes the notion conceptual space from that of semantic map. The former is defined by Croft as follows: "Conceptual space is a structured representation of functional structures and their relationship to each other" (Croft 2001: 93). Conceptual spaces are hypothesized to be language-universal. Α semantic map is a map of the distήbutional pattem of a language-specific category on a conceptual space. In his thought-provoking book, Croft shows the importance of the notion of conceptual space for the understanding of language-specific grammatical categoήes. He insightfully analyses the different conceptual spaces that relate to parts of speech, clausal syntactic roles, the voice continuum, and the subordinationcoordination continuum. Since the present study is not concemed with a crosslinguistic compaήson of voice systems, but is restήcted to the Ancient Greek rniddle voice, employing the term conceptual space would not be appropήate. Instead, Ι will use the term semantic map or semantic network in reference to the particular polysemic structure of the Ancient Greek middle voice. 52 This principle is related to Croft's Semantic Map Connectivity Hypothesis, which runs thus: "any relevant language-specific and construction-specific category should map onto a CONNECΊΈD REGION in conceptual space" (Croft 2001: 96). 53 This is essentially Crofts's dynamicized version of the Semantic Map Connectivity Hypothesis: diachronic changes in the distribution of α construction should follow connected paths in conceptual space (Croft 200 Ι: 105). 51

42 CHAP'fER 1

stem, it predicts that every aoήst form - i.e. sigmatic middle, thematic middle, root middle, and passive aoήst - will map onto a contiguous portion of the semantic network. The second claim predicts that histoήcal expansion of an aorist form through the network will follow the pathways of the semantic links. In other words, the extending form will not 'jump over' from middle use Α to middle use C, without affecting the intermediate use Β. These two claims can be seen as two sides of the same coin, since the synchronical contiguity of a form in the semantic network results from the diachronical tendency of forms to spread only to semantically adjacent meanings. Furthermore, these claims can serve as a powerful empiήcal tool to test the accuracy of our network structure as it results from a semantic analysis only. For example, if an aorist form does not cover a contiguous region in the network, or if an aoήst form does not extend gradually through the semantic links in the network, then the map of our semantic network should be seήously reconsidered, and modified in such a way that violations of the two claims are avoided.

1.2.3

Keιnmer, τhe

Middle Voice (1993)

In the previous section, a network conception of the middle voice was set forth. The major advocate of this approach to the middle voice has been Suzanne Kemmer. Especially her typological study The Mίddle Voίce (1993) - a revision of her doctoral thesis - has been an important contήbution to our understanding of middle voice phenomena, and it has also been fundamental to the present study. Therefore, it is more than justified to devote a section of my study to a discussion of her book. This section will focus on those analyses and conclusions in the book that are of importance to the argument of the present study. Kemmer's The Mίddle Voίce is a large-scale typological compaήson of middle voice systems in 30 languages. After aπ introductory chapter, Kemmer starts off (chapter 2) by giving an inventory of the middle situation types that are frequently marked by middle morphology across the world's languages. She enumerates 13 types of middle situations (Kemmer 1993: 16-20) 54 :

54

The middle markers are in bold-face.

INrRODUCΠON

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8) (9)

(10)

(11)

43

Grooming or body care: Djola (Niger-Congo) -p;,s-;) 'wash'; Latin lavo-r 'wash'; Bahasa Indonesia ber-dandan 'get dressed'; Old Norse klaόa-sk 'get dressed'; Hungarian borotvάl­ koz- 'shave' Nontranslational motion: Kanuή (Nilo-Saharan) tan-t-in 'stretch one's body'; Old Norse snUα-sk 'tum'; Latin reverto-r 'retum'; Classical Greek trepe-sthai 'tum'; Gennan sich verbeugen 'bow' Change in body posture: Djola lak-~ 'sit down'; Bahasa Indonesia ber-lutut 'kneel down'; Gennan sich hinlegen 'lie down'; Guugu Yimidhiπ (Australian, Pama-Nyugan) daga-adhi 'sit down'; Hungaήan emel-ked- 'rise, get up' lndirect middle: Turkish ed-in 'acquire'; Old Norse eigna-sk 'acquire, lay claim to'; Classical Greek kta-sthai 'acquire for oneself; Latin apϊsco-r 'obtain'; Changana (Niger-Congo, SE Bantu) ku ti-tekela 'take for oneself Naturally reciprocal events: Old Norse hitta-sk 'meet'; Hungaήan δlel-kez- 'embrace'; Latin amplecto-r 'embrace'; Bahasa lndonesia ber-gumul 'wrestle'; Guugu Yimidhiπ yirrgaadhi 'converse, agree'; Translational motion: Pangwa (Niger-Congo, CE Bantu) i-nuxa 'climb up'; Guugu Yimidhiπ madha-adhi 'climb up'; Old Norse ganga-sk 'go, leave'; Bahase Indonesia ber-jalan 'walk, stroll'; Classical Greek pete-sthai 'fly'; Latin veho-r 'travel' Emotion middle: Guugu Yimidhiπ dumba-adhi 'get a shock or fήght'; Gennan sich fiίrchten 'be/become fήghtened'; Mohave (Hokan, Yuman) mat iθα:ν 'be angry'; Latin iriisco-r 'become angry'; Hungaήan bάn-kod- 'grieve, moum' Emotive speech actions: Latin quero-r 'complain', Gennan sich beschweren 'complain'; Classical Greek olophUre-sthai 'lament', Turkish dδv-un 'lament'; Sanskrit krpa-te 'lament'. Other speech actions: Mohave mat kuna:v 'confess'; Pangwa -i-lumba 'admit one's guilt'; Latin fateo-r 'confess'; Kanuri awulo-t-in 'be boastful, be a braggart'; Hungaήan dicse-ked'boast'. Cognition middle: Bahasa Indonesia ber-pikir 'be cogitating'; Old Norse pykkja-sk 'think'; Pangwa -i-sala 'think over, consider'; Latin medito-r 'ponder, meditate'; Mohave mat ahay 'believe' Spontaneous events: Changana ku-ti-milela 'genninate,

44 CHAP'ΓER 1

(12)

( 13)

sprout'; Old Norse grόa-sk 'grow'; Bahasa Indonesia ber-henti 'come to a stop'; Mohave mat ico: 'become, change into'; Turkish dinl-en 'recover' Logophoric middle: the rniddle marker appears on verbs of saying, belief, and perception in construction with a complement clause, e.g. lcelandic peir sάus-st hlaupa 'They saw themselves run' Passive, lmpersonal, Facilitative middles: Kanuή t-uruk-zn 'Ι am seen' (passive); German Der Artίkellίest sich leίcht; French Le lίvre se vend bίen (facilitative).

Apart from (12), the logophorίc mίddle, and the facίlίtatίve middle, each of the enumerated rniddle uses seems to be instantiated in Ancient Greek. In (1) groomίng or body care, we can recognize the direct reflexive middle use. In (2) nontranslatίonal motίon, (3) change ίn body posture, and (6) translatίonal motίon, Rijksbaron's (2002 3) pseudo-reflexίve use, which almost exclusively consists of verbs of motion, can be recognized. The ίndίrect mίddle (4) can be compared to the ίndίrect reflexίve mίddle in Greek. The uses (5) naturally recίprocal events, (8) emotίve speech actίons, (9) other speech actίons, and ( 1Ο) cognίtίon mίddle are typical media tantum classes in Greek, e.g. μάχομαι 'fight' (5), όλοφύρομαι 'lament' (8), μυθέομαι 'speak' (9), λογίζομαι 'calculate; consider' (10). The emotίon mίddle (7) and the spontaneous event middle (11) coπespond to Rijksbaron's (2002) pseudo-passίve use, e.g. φοβέομαι 'fear' (7) and τήκομαι 'melt (intr.)' (11). Generalizing from these seemingly very diverse situation types expressed by the rniddle voice, Kemmer argues that the one crucial semantic property of the middle voice is the relatίve distίnguίshabίlίty of partίcίpants. Kemmer (1994: 211): (36)

"Relative elaboration of events can be thought of as the degree to which different schematic aspects of a situation are separated out and viewed as distinct by the speaker. The speaker in effect can choose to "turn up" or "turn down" the resolution with which a particular event is viewed in order to highlight its internal structure to a greater or lesser extent.''

This property, according to Kemmer, subsumes the notion of 'subject-

INτRODUCΠON

45

affectedness' Thus, the middle voice can be located on a gradient scale between two extremes, i.e. between two-participant events (high distinguishability) and one-participant events (low distinguishability) (Kemmer 1993: 73): Fig. 7: Scale of degree of distinguishability of participants: Two-partίcίpant

+

Event >

Reflexίve

>

Mίddle

>

One-partίcίpant

Event

. Degree of dίstίnguίshabllίty of partίcίpants

The two-participant event is located at the one extreme of the scale. According to Kemmer, in a two-participant event the participants are maximally distinguishable in that the two participants, the Initiator and the Endpoint, are completely separate entities. In a reflexive event the distinguishability is lower. The reflexive marker (the pronoun έαυτόν in Ancient Greek) signals the co-referentiality of the participants in events in which the two participants are normally distinct entities. In the reflexive, the separation of the two participants is thus to some extent maintained. In the middle type the dual roles, Initiator and Endpoint, are conflated in a single participant. Thus, the distinguishability of participants is minimal, yet higher than in prototypical oneparticipant events. In the case of the middle, a certain degree of intemal complexity is extant by virtue of the initiating and affected aspect that is evoked. Ι already mentioned, in passing, the notions Initiator and Endpoint. According to Kemmer, the notions Initiator and Endpoint are important for aπ understanding of middle semantics. Το get a grasp of these notions, it is best to consider the transitive clause, which is, according to Kemmer (1994: 51), the basic conceptual model for these notions. We have seen that the prototypical transitive clause involves a physical transmission of energy, initiated by a volitional agent and resulting in a change of state of the patient. However, it is a trivial observation that transitive clauses generally also denote other types of interactions. For instance, in section 1.1.1 Ι mentioned mental event types that are very often coded as transitive events, even though they involve neither a volitional agent nor a state-changing patient. The coding of these non-prototypical transitive event-types in a morphologically transitive clause is conditioned by the degree to

46

CΗΑΡΊΈR

1

which such events can be construed as an asymmetήcal interaction between two participants. ln this asymmetrical interaction, the first participant is construed as initiating the event (hence Initiator); the second participant is construed as the endpoint of the action chain (Endpoint). 55 The notions of Initiator and Endpoint are general semantic roles, subsuming various other, more specific, semantic roles. The Initiator role subsumes those roles that involve a conceptualization of a 'starting point' of an event, such as agent, experiencer, and mental source. Endpoint, on the other hand, encompasses the 'downstream' roles, such as patient, recipient and bene.ficiary. As they subsume the more concrete semantic roles, Initiator and Endpoint are called "macro-roles" by Kemmer. 56 The semantics of the middle voice can now be descήbed elegantly in terms of these macro-roles: the middle voice expresses that the subject is conceptualized as both the Initiator and the Endpoint. With regard to Ancient Greek, however, this characteήzation of the middle voice is not entirely adequate. As was set out above, the Greek middle voice can be characterized by the notion of subjectaffectedness. The idea that the subject is affected by the event is roughly equivalent to the idea that the subject is the Endpoint in the event. The notable difference between Greek and many other middle voice systems is that the middle voice in Greek also covers the canonical passive meaning. This means that, in Greek, the subject of a middle verb cannot be conceived of as an Initiator in all cases. This is a crucial difference between the Greek middle voice and many other middle voice systems (esp. those of modem European languages) which oήginate from reflexive markers. 57 In these languages, the middle category prototype relates to grooming actions (Kemmer 1993:

ss The notions of Initiator and Endpoint are comparab\e to Langacker' s head and tail of aπ action chain (see section 1.1.2). 56 Α related notion, going by the same name of 'macro-role', a\so features prominently in Ro\e and Reference Grammar (Foley & Van Valin 1984, Van Valin & LaPol\a 1997). Κlaiman (1988, 1991) has a notion very similar to macro-roles, that of conceptual status. Instead of Initiator and Endpoint, she works with the controller and affected entity. The notion of controller is, however, more restricted than that of Initiator, since it roughly only includes animate entities. 57 The same observation ho\ds for Latin, and perhaps more ancient Indo-European languages. In Latin, most of the verbs in -or, -ris, -tur, etc. have a passive meaning.

INτRODUCΠON

47

55), or possibly to the related body action types such as change in body posture, nontranslational and translational motion. The spontaneous event type and the passive58 are to be considered peήpheral usage types. In Ancient Greek, however, the category's gravitational point lies elsewhere. In section 2.3.2, it will be argued that the especially the rnental process type intransitive rniddle types 59 constitute the centre of the rniddle voice in Greek. The groorning type in Greek is to be considered of rnarginal irnportance as both type and token frequency are very low. In chapters 3 and 4, Kernrner gives a thorough descήption of the different reflexive and rniddle types found in the languages of the world. As we have seen above, rnany of these can also be found in Ancient Greek. In chapter 5, the historical ernergence of the rniddle voice frorn a reflexive οήgίη in a nurnber of language-farnilies (Rornance, Gerrnanic and Nilo-Saharan) is discussed. Kernrner descήbes the spread of the reflexive rnarker se in Latin through the rniddle sernantic dornain in French and Surselvan (a Rhaeto-Rornance language), accornpanied by the disappearence of the older rniddle rnarker, the inflection in -or, -ris, -tur, etc. Kernrner descήbes the expansion of the reflexive rnarker se as a process of grarnrnaticalization: the rneaning shifts frorn aπ oήginally ernphatic direct reflexive sense in Latin to a rnore abstract rniddle rneaning in the Rornance languages. Eventually, the rnarker could also be used as a passive. This sernantic generalization process is accornpanied by typical syrnptorns of grarnrnaticalization: cliticization (pan-Rornance), erosion (loss of phonological substance), affixation (in Surselvan), and the spread of se frorn the 3rι1 person to all other persons (in Surselvan). The description of this diachronic process, involving a reflexive rnarker developing into a rniddle rnarker, as it occuπed in Rornance and Gerrnanic languages, is not directly relevant to the Greek rniddle voice and its origins. If the Indo-European rniddle voice did develop frorn a reflexive rnarker, it rnust have been in an earlier stage of Note that in many modem European languages the prototypica\ passive is expressed by an auxiliary plus particip\e, e.g. French La porte est ouverte par Roger. 59 Although, strictly speaking, the direct reflexive midd\e is also intransitive, Ι will use intransitive middle as a cover-term for the body motion midd\e (roughly Rijksbaron's pseudo-reflexive), the mental process midd\e and the spontaneous process midd\e (Rijksbaron's pseudo-passive). The intransitive midd\e is ca\led decausative by some authors (e.g. Geniu§iene 1987, Gerήtsen 1990), since they systematically coπespond to active causative verbs. 58

48

CΗΑΡΊΈR

1

Proto-Indo-European (PIE), since we can already reconstruct a reflexive marker for ΡΙΕ, *s(u)e, alongside a full-blown inflectional middle voice category (ls *-mh2 , 2s *-sthρ, 3s *-to, etc.). In other words, in PIE we find a situation in which the (emphatic) reflexive is expressed by a marker (*s(u)e) that is not genetically related to the middle inflection (ls *-mh 2, etc.). 60 However, for students of the Greek middle voice the primary importance of Kemmer's descήption of the spread of the reflexive marker from one use to another lies elsewhere. The first point of importance is of a methodological nature. Kemmer' s diachronical analyses are based ση the assumption that two meanings Α en Β are related, if one can observe that a form spreads from meaning Α to meaning Β (1993: 5). This makes diachronic evidence as important to a semantic analysis as synchronic evidence. In section 1.2.2, Ι have expressed my adherence to this principle, and in chapter 3, Ι will argue that this principle can improve our understanding of the morphological distribution of the different aorist formations and its semantic implications. The second point of importance concerns the histoήcal development of the reflexive marker. Kemmer descήbes the frequently occuπing development from a reflexive pronoun (used with verbs of grooming, and many verbs of motion) to a more general middle marker that also includes mental (emotional and cognitive) events and spontaneous events. In some cases, this middle marker also develops a passive meaning. 61 If we transpose these findings to Ancient Greek, we can assume a semantic relatedness between different middle uses in the following way: at the one extreme, the reflexive-like uses (the direct reflexive middle and the pseudo-reflexive middle) are located, and at the other, the passive is located. In between these uses lie the emotional, cognitive and spontaneous uses (Rijksbaron's pseudopassive). In chapters 2 and 3, Ι will produce evidence that this configuration is, in pήnciple, coπect, and Ι will, furthermore, propose a number of refinements. In chapter 6, Kemmer presents a 'semantic map' which repre-

The situation we find in Ancient Greek, with both a middle inflection and a non-related emphatic reflexive marker (e.g. 3s acc. έαυτόν 'himself), appears to be directly inheήted from ΡΙΕ. Note that έ-αυτόν contains the morpheme έ­ which is a direct descendant from ΡΙΕ *sue (see also Petit 1999). 61 This development has also been descήbed by Haspelmath (1987: 35). 60

histoήcally

INτR.ODUCΠON

49

seπts the πetwork coπsistiπg of the various middle situatioπ types and the semantic relatioπs existiπg amoπg them (1993: 202). This map of the iπteπelatioπs betweeπ middle types is based οπ the shared semaπ­ tic properties that were revealed by the typological aπd diachroπic data that were analysed iπ the precediπg chapter of her study. Siπce this map is based οπ a large collectioπ of typological data, it is claimed to have uπiversal validity. Ιπ other words, the semantic relatioπs established iπ the map are claimed to be relevaπt to all languages. Ιπ particular, the set of uses fouπd iπ any οπe laπguage is predicted to form a coπtiguous regioπ οπ the map: "( ... ) we would ίπ pήπciple expect the semaπtic range of a giveπ form to exteπd oπly over a set of uses iπ which each use is directly related to at least οπe other use iπ the set'' (Kemmer 1993: 222). The diachroπic coπelate of this predictioπ is that the spread of a giveπ form is predicted to follow the pathways which are formed by the semantic relatioπs, that is, from οπe use to aπother, directly liπked, use. As was said at the outset of this sectioπ, Kemmer' s book is aπ iπspiriπg aπd rich Fundgrube for aπy studeπt of voice pheπomeπa. Ιπ the discussioπ above, it must be πoted, Ι have restricted myself to those aspects of the work that are of particular iπterest to my οwπ study of the Greek middle voice.

1.3 Two General Issues the precediπg sectioπs, Ι have set out the major problems that will be dealt with iπ this study, aπd Ι have preseπted the basic assumptions that will uπderlie my aπalysis. Now Ι will address two geπeral issues that are important to aπ uπderstaπdiπg of the middle voice. First, we will coπsider the status of the media tantum withiπ the middle voice system, and secoπdly, we will discus the ποtίοπ of valeπce reductioπ as aπ explaπatioπ for middle voice pheπomeπa. Ιπ

1.3.1 The Status of the Media Tantum Media tantum are

treated as a distiπct class (see for 1956: 393; Schwyzer-Debruππer, Π: 228-9). This approach, however, is ποt eπtirely uπproblematic. First, it is clear that the class of media tantum is semantically highly diverse. Α traditioπally

iπstance Smyth-Messiπg

50

CΗΑΡΊΈR

1

superficial glance at a selection of media tantum may demonstrate this point: άγωνίζομαι 'contend', άκροάομαι 'listen', &λλομαι 'jump', βούλομαι 'want', γίγνομαι 'be bom, become', δέχομαι 'accept', επισταμαι 'know', Ερχομαι 'go', εϋχομαι 'pray', λογίζομαι 'calculate', πυνθάνομαι 'leam, hear; inquire' The diνersity within this class is also manifested by the differences in aorist morphology. Some of these verbs may have a sigmatic middle aoήst, others a passive aoήst, and yet others thematic or root aoήsts. The rationale behind the distήbution of the different aoήst formations will be the subject of chapter 3. Α second difficulty in treating the media tantum as a separate class is that, in doing so, two inconsistent criteήa are aplied. On the one hand, the vaήous oppositional middle types 62 are distinguished purely on the basis of semantic cήteria. For instance, the indirect reflexive middle is distinguished by the semantic feature subject profits from action, and the passive middle is defined by the fact that the subject is semantically a patient. On the other hand, the media tantum are distinguished by a completely different criterion, namely the non-existence of an active form. This cήteήon stands orthogonal to the semantic criteήon that is applied to the oppositional middles. From a methodological point of view, it is preferable to aim at a unified account of both oppositional middles and media tantum. Here and in chapter 2, Ι shall argue that such a unified account is indeed possible on the basis of purely semantic cήteήa. For a great number of media tantum it is possible to classify them under the already established middle uses in a straightforward way. Many of these media tantum are semantically very close to oppositional middles. lt would, therefore, be highly unnatural to treat them as belonging to a separate class for the sole reason that they do not have an active counterpart. Α nice case in point is the medium tantum tlδομαι 'enjoy' which is semantically very close to the oppositional middles εύφραίνομαι 'be glad' (active causative εύφραίνω 'make glad') and Homeήc τέρπομαι 'amuse oneself (active causative τέρπω 'amuse').

62 The oppositional middles are those middle verbs that stand in opposition to an active verb. Two typical examples are: (i) middle παρασκευάζομαι (indirect reflexive, direct reflexive, or passive) vs. active transitive παρασκευάζω, and (ii) middle ϊσταμαι (indirect reflexive, intransitive, or passive) vs. active transitive

ϊστημι.

INτRODUCΠON

51

lt therefore makes good sense to classify flδομαι as a pseudo-passive (in Rijksbaron's terminology). 63 Supporting evidence for the view that media tantum have to be treated on a par with the oppositional middles is based on their morphology. The media tantum generally make the same choice for either the sigmatic middle aorist or the passive aoήst form as do the oppositional middles of the same semantic type. For example, flδομαι has a passive aoήst flσθην just like εύφραίνομαι - ηύφράνθην. Ι will go into the morphology of the aorist stem extensively in chapter 3. In the same way, other media tantum can be classified under the already familiar middle uses. On closer inspection, it becomes clear that most media tantum belong to a restricted number of semantic classes. These are: (37)

Media tantum: semantic types

Media tantum with a beneficiary/recipient-subject: 'accept, receive', έργάζομαι 'work, make', κ:τάομαι 'acquire' b. Media tantum of body motion: &λλομαι 'jump', eπομαι 'follow', tpχομαι 'go', πέτομαι 'fly' c. Media tantum of emotion and cognition: &γαμαι 'wonder', &χθομαι 'be grieved', βούλομαι 'want', eραμαι Ίονe', flδομαι 'enjoy', οϊομαι 'think', πυνθάνομαι Ίearn, hear' d. Media tantum of volitional mental activities: λογίζομαι 'calculate', μηχανάομαι 'contήve' e. Reciprocal media tantum: άγωνίζομαι 'contend', μάχομαι 'fight' f. Media tantum of perception: άκ:ροάομαι Ίisten', δέρκ:ομαι 'gaze, look at', θεάομαι 'gaze', όσφραίνομαι 'smell', a.

δέχομαι

σκέπτομαι Ίοοk'

g. Speech act media tantum: αίτιάομαι 'accuse', άράομαι 'pray', όλοφύρομαι Ίament', μαρτύρομαι 'call to witness', μυθέομαι 'speak' The first three types (a-c) can be readily recognized as being

63 Note that, after Homer, an active causative verb f\δω was created. Ι would take this as confirmation that the border between media tantum and oppositional middles is not a very important one, since it is easily crossed.

52

CΗΑΡΊΈR

1

related to oppositional middle types. The media tantum with a beneficiary/recipient-subject (a) can be identified as indirect reflexive middles. This identification is also supported - or at least not contradicted - by the aoήst morphology as both the indirect reflexive oppositional middles and the indirect reflexive media tantum consistently have a sigmatic middle aoήst, and never a passive aoήst, as for example, έδεξάμην, είργασάμην, έκτησάμην. The media tantum designating body motion can be classified as pseudo-reflexive middles on a par with middles like άπαλλάττομαι 'depart', ϊσταμαι 'stand up, stand still', πορεύομαι 'go, march' Lastly, the media tantum designating emotion and cognition can be identified with Rijksbaron's (20023) pseudo-passive middles (type φοβέω 'frighten' φοβέομαι 'fear'). Again, aorist morphology suggests that uniting the oppositional middles and the media tantum designating mental processes is justified since both always have a passive aoήst form, and never a sigmatic middle aorist: e.g., oppositional middles: έλυπήθην, έφοβήθην, ώργίσθην; media tantum: έβουλήθην, ησθηv, φήθην. 64 The last four types (d-g) cannot be compared to the already established oppositional middle uses. Υet Ι will argue that they should not be set apart completely from the other middle uses; they should, instead, be integrated into the polysemous structure of the middle voice, just like the other (oppositional) middle usage types. In other words, these types of media tantum can be seen as nodes in the semantic network linked to the other, more familiar, nodes. In chapter 2, Ι will show that the four types at issue are semantically closely connected to the other middle uses, and that the boundaries between them are only a matter of gradiency. In chapter 3, it will be shown that there is also evidence from aoήst morphology for the existence of polysemous relations between the media tantum and the oppositional middles.

Rijksbaron, too, rernarks that the fact that the agentive rnedia tanturn have a sigrnatic rniddle aoήst is in accordance with the agentive direct and indirect reflexive uses. Furtherrnore, the passiva tanturn sernantical\y resernble the pseudo-passives of causative verbs. Both are rnost\y non-agentive and intransitive and have passive aoήst forrns (see Rijksbaron 2002 3 : 160-1). 114

INrRODUCIΊON

53

1.3.2 Valence reduction

Earlier in this chapter Ι argued that the middle voice as a category can be characterized by a semantic feature, namely subject-affectedness. There is, however, a substantially different approach in the literature on the middle voice and reflexive verbs. This more syntactically oriented approach sees the middle voice as a mark.ing of valence reduction. Valence reduction implies, in bήef, that by means of the middle voice morpheme transitive predicates are detransitivized. Απ early example of this approach is Margulies (1929-30: 116): "Wenn wir bedenken daB es keine idg. suffixalen Intransitivbildungen gibt, dann scheint es nicht zu gewagt, anzunehmen, daB die ursprϋngliche Funktion der Medialendungen die Intransitivierung war.'' More recent attempts along these lines are Aissen (1982), Dik (1982) [on reflexive verbs], Geniu~iene (1987) [on reflexive verbs], Risselada (1987) [on Ancient Greek], Dik (1997b: 9-15). There are, however, a number of serious objections to this approach. Firstly, in probably all rniddle voice systems, as in Greek, there is a class of media tantum - middle verbs that do not have an active counterpart. Clearly, these verbs cannot be deήved from active predicates by an argument reduction rule. As a possible solution for this problem one could 'condernn' the media tantum to the lexicon. This solution is, however, not very elegant since it means that the attempt at a unified account of the middle voice is abandoned, and that a highly frequent class of middle verbs is basically left unaccounted for. Moreover, this view contravenes the usage-based approach adopted here, since the latter approach claims that storage in the lexicon and the existence of a general grammatical rule are not mutually exclusive (the "non-reductionalist" view). Therefore, though it is clear that media tantum are individually stored in the lexicon, it is possible that they can be subsumed under a more general schema together with the oppositional middles. Furthermore, oppositional middles and media tantum cannot be distinguished by the criterion that the former are grammatical, and the latter lexical, since many frequently occuπing oppositional middles are probably stored in memory (i.e. "in the lexicon") as well. 65 For example, it is improbable that a

6 ~ The point that both grammatically iπegular forms (e.g. the English strong verbs) and regular forms are lexically stored is also made by Bybee (1985: 113-4, and passim).

54

CΗΑΡΊΈR

1

highly frequent verb fonn like ϊσταμαι 'stand still, stand up' was computed from scratch by a grammatical derivational rule on the basis of the active fonn ϊστημι 'make to stand' every single time it was uttered. lt is more plausible to assume that the fonn ϊσταμαι had become entrenched as a fixed unit (because of its high frequency), and was retrieved automatically from memory as a pre-packaged whole, which clearly cuts down on processing efforts. Α second objection to the valence-reduction approach is related to the previous one. lt is a fact that most intransitive middles have a higher frequency than their active transitive counterparts. The following table shows token-frequencies of active transitive and middle intransitive verbs66 in Herodotus (based on Powell's lexicon (1938)):

6/i Note that most middle intransitive verbs have passive aoήst forrns in -(θ)η-. These forrns are included in the frequency-figures.

INτRODUCΠON

Table 3:

Frequencies of transitive active vs. intransitive middle in Herodotus

Transitive Active

'disgrace' 'set free from' αύξάνω 'grow (trans.)' έγείρω 'stir up' ί::λπω 'cause to hope/expect' ϊζω 'seat' ϊστημι 'set, put' κήδω 'trouble, distress' κομίζω 'carry, fetch, take' λείπω Ίeave behind' λυπέω 'harass' μιμνήσκω 'remind' όρμάω 'urge, dispatch' παύω 'stop' πείθω 'persuade, convince' πορεύω 'make to go, convey' σήπω 'make rot' στέλλω 'send' τήκω 'melt (trans.)' τρέπω 'tum (trans.)' τρέφω 'grow (trans.), nurture' φαίνω 'show' φοβέω 'alarm' φύω 'produce, grow (trans.)' φέρω 'carry, bήng' ψεύδω 'deceive' αίσχύνω

άπαλλάσσω

55

Intransitive Middle

1 2 1 2 ο

1 47 ο

40 36

6 ο

2 14 47 ο ο

9

2 8

15 14 2 8

118

ο

αίσχύνομαι

'feel shame'

άπαλλάσσομαι 'depart' αύξάνομαι 'grow up'

'wake up, arise' ί::λπομαι 'hope, expect' ϊζομαι 'sit down' ϊσταμαι 'stand' κήδομαι 'care about' κομίζομαι 'travel' λείπομαι 'stay behind' λυπέομαι 'be angry' μιμνήσκομαι 'remember' όρμάομαι 'set off παύομαι 'cease' πείθομαι 'believe, obey' πορεύομαι 'travel, march' σήπομαι 'rot' στέλλομαι 'set forth' τήκομαι 'melt (intr.)' τρέπομαι 'tum (intr.)' τρέφομαι 'grow up' φαίνομαι 'appear' φοβέομαι 'fear' φύομαι 'grow (intr.)' φέρομαι 'fall, dήft, charge' ψεύδομαι 'be mistaken' έγείρομαι

3 86

13 5

9

39 85 7 26 16 2

30 93 45 95 75 4 20 3 94 11

141 21 28 16

8

In general, the middle intransitive forms occur far more frequently (printed in bold-face) in Herodotus than the active transitives. It stήkes me as rather counter-intuitive to consider the intransitive middles as derived from, or somehow secondary to, the active transitive counterparts, consideήng that intransitive middles generally have a higher token-frequency. 67 Although, in general, one could maintain that the

In fact, there are well-known cases in which aπ active transitive is created next old intransitive midd1e, e.g. i)δω to i)δομαι, μαίνω to μαίνομαι, πεύθω to πεύθομαι. These active causative vs. middle intransitive pairs contravene the global markedness of the middle voice. This phenomenon can be called local markedness 67

to

aπ

56

CΗΑΡΊΈR

1

middle voice is marked in comparison with the active voice (in terms of frequency, semantic content, and morphological markedness: see my discussion of markedness above ( 1.1.4)), in the case of these causative active and intransitive middle verbs the markedness-relation seems to be reversed. The frequency figures above clearly attest to that. Furthermore, the intransitive middles can be considered unmarked from a semantic point of view, as well. The event denoted by the intransitive middles (roughly: aπ autonomous change of state of the subject) is semantically more simple than the event denoted by the active causative verbs, since the latter add a causer-participant to the event. Α third objection to the valence-reduction view on the middle voice is that many middle voice verbs are transitive, just as their active counterparts, like, for instance, the frequent and productive class of indirect reflexive middles. In conclusion, the valence-reduction view on the middle voice is unattractive. lt is preferable to seek a purely semantic explanation instead. In section 1.1.3, Ι argued that the middle voice can be explained adequately as a marker of subjectaffectedness. 68

(see Tiersma 1982, Bybee 1985: 74-6). 68 Sometimes, a more restricted view is adopted, namely that the passiνe aorist form exhibits valence reduction. τhis view is advocated by Andersen (1989, 1993). Yet, two of the objections that were discussed above are equally applicable. First, there are many media tantum with passive aorists (passiva tantum), e.g. έβουλήθην Ί wanted', έδυνήθην Ί could', ~δέσθην Ί was ashamed', ιj>ήθην Ί thought'. Obviously, these passive aoήsts could never be deήved from transitive counterparts. Secondly, the passive aorists often have a higher token-frequency than their active counterparts (e.g. ίόρμησα Ί urged, dispatched' vs. ώρμήθην Ί started off), which makes it unnatural to consider them as deήvations of the active transitives. The third objection does not apply since passive aoήsts are not transitive. In chapter 3, Ι will argue that the passive aoήst form expresses that the subject is similar to a prototypical patient.

CHAP'fER

2

The Middle Voice as a Complex Network Category

In the first chapter, Ι set out Langacker's Complex Network Category Model. The most important components of this model were abstract schema, elaboratίon, extensίon, and prototype (sections 1.2.1, 1.2.2.). In section 1.2.2, it was argued that the Greek middle voice category may be analyzed insightfully as a complex network category. In this approach, the middle voice is seen as a polysemous network of inteπelated meanings. The abstract schema, embodying the semantic commonality of all middle meanings, can be characteήzed as affectedness of the subject. The different middle meanings can, in tum, be viewed as elaborations of this abstract schema. In this chapter, the structure of the network of the middle voice in Ancient Greek will be analyzed. Ι will propound a classification of middle uses, and Ι will descήbe the semantics of each middle meaning in more detail. We will see that the most essential semantic properties relate to the subject, and especially to the subject's semantic role. Furthermore, for each middle meaning Ι will try to ascertain to which other middle meaning it is related. The main criterion for establishing a relation between two middle types is to assess whether or not they have salient semantic properties in common. In some cases, subsidiary criteήa are used. For example, semantic changes can indicate semantic relatedness; when it can be established that a certain verb first belonged to type Α, and then to type Β, it can be assumed that type Α and Β are semantically related. This analysis will finally result in a semantic map of the Greek middle voice, representing a network consisting of the various middle uses and their semantic inteπelations. The final issue that will be addressed in this chapter concems the category prototype. It will be investigated which of the middle uses can be regarded as the most salient member of the category. The middle uses to be discussed are the passive middle (2.1.1 ), the spontaneous process middle (2.1.2), the mental process middle (2.1.3), the body motion middle (2.1.4), the collective motion middle (2.1.5), the reciprocal middle (2.1.6), the direct reflexive middle (2.1.7), the perception middle (2.1.8), the mental activity middle

58

CΗΑΡΊΈR

2

(2.1.9), the speech act middle (2.1.1 0), the iπdirect reflexive middle (2.1.11 ). The uses that are discussed first (2.1.1 - 2.1.5) predomiπantly have a passive aoήst form ίπ -(θ)η-, whereas the uses discussed last (2.1.6 - 2.1.11) geπerally have a sigmatic middle aorist. 2.1

Α

Classification of Middle Usage Types

2.1.1 Passive Middle Ιπ a typical passive clause, the patieπt is assigπed subject-status. 69 The ageπt is rarely expressed by aπ explicit πouπ-phrase70 • Esseπtial is, however, that aπ ageπt-participant is conceptually preseπt, but pragmatically deemphasized. έγrο δ' ύπ' έχθρου (... ) ύβριζόμην was assaulted by an enemy.

(1)

(Dem. 21.74)

Ι

(2)

ό μη δαρείς &νθρωπος ού παιδεύεrαι (Μeπ. Mon. 422) The man that has not been thrashed gets no education.

Example (2) is a proverb, and it has a geπeήc ageπt. Geπeric ageπts are typically left implicit. Yet, the preseπce of aπ ageπt is still coπcei­ ved of, due to the iπhereπt lexical semaπtics of the verbs δέρω aπd

69

In Classical Greek, there are passiνes όf which the subject coπesponds with a or datiνe complement in the actiνe construction, e.g. έπιβουλεύω τινί,

genitiνe

ίiρχω τινός, βοηθέω τινί, καταγελάω τινός, όλιγωρέω τινός, πολεμέω τινί, φθονέω

τινί (cf. K-G, Ι: 124-5, Mulder 1988, Rijksbaron 2002 3: 139, Conti Jimenez 1998). It is noteworthy that the second arguments are typically human, i.e. entities that are natural discourse topics. 'Γhis possibly explains that they are capable of being selected as subjects. In addition, the semantic feature of affectedness may be releνant. Many νerbs that allow this type of passiνes designate a 'negatiνe' emotional attitude towards the second participant (φθονέω τινί), or imply an exertion of force/power (πολεμέω τινί, ίiρχω τινός). Thus, the second participant can be said to be affected ('patient-like'). 70 The infrequency of explicit agent-phrases is bom out by the figures giνen by Rijksbaron (20023: 141): of the 22 passiνe instances in Herodotus, the agent is expressed only twice. For a discussion of the different morphosyntactic ways of expressing the agent in a passiνe clause, Ι refer to Schwyzer (1943) and Jankuhn (1969: 101-11).

ΤΗΕ MIDDLE VOICE AS Α COMPLEX CA'fEGORY

59

which both denote events that are extemally initiated. Passive clauses can also contain an instrument or a cause.

παιδεύω

(3)

το\>τον δi] c1v τον 'Άτυν σημαίνει τ νυμφεύματα (S. ΟΤ Have no fears ίη the direction of a maπiage with your mother. b. άλλ' εί φοβ!\ προς τοi>το, (... ) (S. Tr. 1211) But if you have fears with respect to that, (... ).

980)

example (a.), the marriage with his mother is, of course, the entity that is feared in a direct way by Oedipus. lt is not merely the source of his fear. The meaning of this construction is, subsequently, similar to one with a pure accusative. The difference is that the direction of his anxious thoughts is refeπed to in a more concrete way by means of είς 'toward' In a similar fashion, in (b.) the phrase προς τοi>το seems to convey the image of the mind directed towards the object of fear (in this case, Hyllus is frightened by the idea of having to set Heracles ση fire ). In this section, Ι have descήbed the mental process middle type. Additionally, the relation of the intransitive middles denoting mental processes with their transitive active counterparts was discussed. Finally, a model was proposed to account for the various case constructions that the mental process verbs occur with. lt was argued that the notions of construal and imagery are of crucial importance to the semantic structure of mental events. Ιη

2.1.4 Body Motion Middle Body motion involves an animate entity that volίtionally brings about a change of state to himself. Thus, the subject is both agent and patient. ι ι 3 This change of state typically consists of either a change of location or a change in body posture. ι ι 4 Many middles that desig-

The body motion type is semantically middle (as opposed to active and passive) since the subject is both the initiating entity (lnitiator) and the affected entity (Endpoint). 114 In ch. 5, Ι will argue that a tendency can be discemed for middle verbs of motion to designate change of body posture and non-translational motion, whereas active verbs of motion tend to designate manner of motion verbs. This tendency may be explained by the fact that in the case of manner of motion verbs the element of 113

ΤΗΕ MIDDLE VOICE AS Α COMPLEX CAτEGORY

77

nate body motion have a transitive causative counterpart. Examples of middles of motion are: 115 Nontranslational motion [+ gen.] 'keep oneselfback from' 'stretch oneself out, reach out' 116 στρέφομαι 'tum around (intr.)' τάνυμαι 'stretch oneself out' τείνομαι 'stretch oneself out' 117 τρέπομαι 'tum (intr.)' (άπ-)έχομαι

όρέγομαι

[+ gen.] 'keep s.o. away from' 'stretch out (one's arm(s))' στρέφω 'tum around (trans.)' τανύω 'stretch out' τείνω 'stretch out' τρέπω 'tum' (άπ-)έχω όρέγω

change is deemphasized to a certain degree. Therefore, the subject is conceived of as less affected by the motion. 115 Translational motion is motion along one dimension in space; nontranslational motion is motion that is neither translational, nor change in body posture (see Kemmer 1993: 67-70). Note, further, that almost all body motion verbs, at least in Classical Greek, have a root, thematic or passive aoήst form. Only a few middleonly verbs have a sigmatic rniddle aoήst, e.g. ίiλλομαι ήλάμην, χάζομαι έχασάμην. However, both aλλομαι and χάζομαι have a poetical and archaic ήng. For aλλομαι, this point will be made in ch. 5. Ι will retum to aoήst morphology extensively in ch. 3. There is also a considerable number of activa tantum that designate body motion, for instance, βαίνω 'make a step, go', εtμι 'go', t:ρπω 'creep, go', τρέχω 'run' In chapter 5, it will be argued that there is a semantic difference between active and middle verbs of body motion. 116 lt is interesting to note that many verbs of motion have a metaphoήcal mental meaning, e.g., άπέχομαι 'keep oneself back from' > 'abstain from'; ϊεμαι 'speed' > 'desire'; όρέγομαι 'stretch oneself out' > 'yeam for'; όρμάομαι 'start off > 'be eager'; σεύομαι 'speed' > 'be eager'; τρέπομαι 'tum' > 'change one's mind'; φοβέομαι 'flee' > 'be afraid' Mental processes are understood as physical motions. Notice also the etymology of the word e-motion and the figurative sense of 'be moved'. 117 The difference between τάνυμαιlτείνομαι and όρέγνυμαι is that the latter is restricted to the arms, whereas the two former verbs involve the entire body.

78

CΗΑΡΊΈR

2

Change in body posture 'stand still, stand up' (καθ)ίζομαι 'take one's seat, sit down' 118

ϊστημι

κλίνομαι Ίean'

κλίνω

ϊσταμαι

'set up, raise, make to stand' 'make to sit down' 119 'cause to lean'

(καθ)ίζω

Translational motion 'raise'

αϊρομαι 'ήse'

αϊρω

'put out to sea, set sail' 120 άπαλλάττομαι 'depart from'

up to sea' 'free s.o. off s.th., κομίζω 'carry, bήng' όρμάω 'set in motion' πορε\ιω 'make to go' σε\ιω 'set in swift motion' στέλλω 'make ready, send'

άνάγομαι

κομίζομαι 'traνel'

'start off 'go, walk' σε\ιομαι 'rush, speed' (Hom./poetry) στέλλομαι 'set out, joumey' σcpζομαι 'get back safe, escape' όρμάομαι

πορεύομαι

άνάγω Ίead

άπαλλάττω

remoνe' 121

σcpζω 'saνe'

The νerb (κ:αθ)ijμαι synchronically functions as a perfect to (κ:αθ)ίζομαι. There are more rniddles that serνe as suppletiνe perfects: είμαι (perfect of [ννυμαι 'clothe onese\f, be c\othed'), κ:εϊμαι (perfect of τίθεμαι 'be set'), and κ:ρέμαμαι (perfect of κ:ρεμάννυμαι 'be hung'). From a synchronica\ point of νiew, the midd\e endings of these νerbs express the affectedness the subject previously underwent - as a resu\t of which the subject is in a certain state (cf. the stative-confective meaπing of the Greek perfect). Histoήcal\y, these verbs are relics of the ΡΙΕ verbal category stative (for the ΡΙΕ stative, see Narten 1968, Rix 1977, 1988, Oettinger 1976, 1993, Beekes 1995: 244). Other relics of this category in Greek are δέατο (Hom.) 'seem', δέχαται 'await' (Hom.), στε\>ται 'boast'. Possibly also tpαμαι 'be in \ονe', [ρχαται (Hom.) 'be shut ίη', εύκ:το (Thebais Fr. 3.3) 'pray, boast', ονομαι 'b\ame', aπd πέταμαι 'fly' Middles belonging to this category caπ be recognized by the eνocalism in the stem, aπd their (oήginally) stative semaπtics. 119 In Homer, the active ϊζω often has its oήginal intraπsitive meaπing 'sit down' 'Ίζω acquired its causative meaning after the emergence of the middle form ϊζομαι 'sit down', which was created by aπalogy with middle forms such as aoήst εζετο 'sat down' aπd 'perfect' f,μαι 'be seated' 120 This midd\e started off as aπ el\iptical indirect reflexiνe rniddle άνάγομαι τfιν να\>ν Ίead the ship up to the high sea' (e.g. Hdt. 6.12.1), but it was subsequently reinterpreted as aπ intraπsitiνe which, in turn, led to a rep\acement of the o\der aoήst άνηγαγόμην by the passive aoήst form άνήχθην (e.g. Hdt. 3.137.5). 121 The midd\e άπαλλάττομαι does not coπespond to the typical meaning of the active άπαλλάττω + acc. + gen. 'free s.o. of s.th.' Ι am aware of only one instance of the active with the meaπing 'make s.o. go away from s.th.', this being the true semaπtic counterpart to the middle 'go away from': ( ... ) έκ:έλευε (... ) τους Πελο­ ποννησίους άπαλλάξαι έκ: τijς χώρας (Th. 8.46.4) (Ήe urged [him] to remove the Peloponnesians from this country'). 118

ΤΗΕ MIDDLE VOICE AS Α COMPLEX CAΊΈGORY

79

Media tantum of body motion 122 ίiλλομαι 'jump' 123 tpχομαι 'go' νέομαι 'retum, go' (Hom./poetry) 124 οϊχομαι 'go away' ι;ι, χάζομαι 'draw back' (Hom./poetry) 126

The label body motion middle is used in this study because it is the most specific of the whole array of teπns that is encountered in the literature. The teπn endoreflexive used (and coined?) by Haspelmath ( 1987) is also attractive, as it underpins the close relationship the body motion middle has with the semantics of reflexivity. On the other hand, it also reveals the crucial property that distinguishes it from the typical direct reflexive (type shave oneselj), that is, intemality (cf. the prefix endo-). By a number of scholars, the teπn pseudo-reflexive is employed (e.g. Dik 1983 [on reflexive verbs in modem European languages], Risselada 1987 [on Greek], Rijksbaron 2002 3 [on Greek]). For Greek, Ι find this teπn inadequate because it suggests that this middle type is somehow secondary to the 'true' (direct) reflexive (see section 2.1.7). This, however, seems unjustified, consideήng that the body motion middle has a considerably higher text-frequency than the direct reflexive middle (for frequency-figures, see section 2.3.2). Furtheπnore, the teπn body motion middle is attractive since it is specific and self-explanatory, unlike some of the aforementioned teπns. 127

The medium tantum [πομαι 'follow' may also be considered a body motion C.J. Ruijgh pointed out to me (pers. comm.) that the νerb might haνe had a causative active counterpart, like many other body motion middles. On one of the new Mycenaean tablets from Thebes (Fq 254+255.1), a form a-pi-e-qe can be found, which can be read as active άμφί-hεσκwε 'make s.o. follow'. 'Έπομαι is a very old medium tantum, cf. Skt. sάcate 'accompany', Lat. sequor 'follow' 123 The different distήbutions of άλλομαι and its synonym πηδάω are investigated in ch. 5. 124 Probably an ancient medium tantum, cf. Skt. nάsate 'join, unite'. 125 Often, οϊχομαι is used in a perfect-like sense 'be gone, be away' 126 An active causative άνα-χάζω 'make to recoil' occurs once at Pi. Ν. 10.49; the simple active is found in Hesychius. The old (Homeήc) causative formation was έκέκαδον (with iπegular -κ-). 127 Α marginal class of media tantum that is related to the middle verbs of body motion are the verbs that designate physical functions such as έρεύγομαι 'spit, disgorge, belch', πέρδομαι 'break wind', πτάρνυμαι 'sneeze', χρέμπτομαι 'cough, clear one's throat', χασμάομαι 'yawn' These verbs are semantically similar to the 122

νerb.

80

CHAYrER

2

Body motioπ verbs are typically volitioπal. However, maπy body motioπ verbs can also be used πoπ-volitioπally. Απ example is the iπtraπsitive middle κ:ινέομαι that can have aπ aπimate subject (someoπe moviπg νolitioπally) as well as aπ iπaπimate subject (somethiπg mονίπg spoπtaneously). For example, (16)

[The Greeks] (... ), οί δε ώς έκινήθησαν, eφευγον liσμενοι την (... ) (Hdt. 9.52)

ϊππον προς την Πλαταιέων πόλιν,

(... ), and once they rnoved, they fled gladly frorn the cavalry to the city of the Plataeans, (... )

(17)

[τhe island Chemmis] (... ), λέγεται δε ύπ' Αίγυπτίων είναι αϋτη ή νilσος πλωτή. αύτος μεν eγωγε οϋτε πλέουσαν οϋτε

κινηθεϊσαν εtδον, (... ) (Hdt. 2.156.2) (... ), and it is said by the Egyptians that this island floats. seen it float nor rnove, (... )

Ι

rnyself have not

These examples make clear that the body motioπ middle is related to the spoπtaneous process middle. Before Ι proceed to the πext middle use, Ι wish to devote some space to a rather peripheral but iπterestiπg member of the body motioπ class. This coπstructioπ-type is exemplified by the followiπg Homeήc liπes:

(18)

άμφοτέρυσι δε χερσ1.ν έπεσσυμένος λάβε πέτρης, τί\ς Εχετο στενάχων, (... ) (ε 428-9)

On he rushed and seized the rock with both hands, and clung to it, groaning, ( ... )

body motion verbs in that they also involve, to some extent, a movement of the body. The difference is that the verbs of body moιion are completely volitional, whereas the middles of physical acιions are often induced by an intemal or extemal physical stimulus. Therefore, although the subject performs the acιion, it is not a prototypical agent. Another difference between these verbs and the typical body moιion middles is that the former do not involve a change of sιaιe, or more specifically, there is no change of Jocation or body posture. There are also a number of active verbs belonging to this semanιic class, e.g., βδέω 'break wind', βήττω 'cough', έμέω 'vomit', όμείχω 'make water', ο-l>ρέω 'make water', πταίρω 'sneeze' (secondarily after aoήst επταρον, cf. πτάρνυμαι), πτύω 'spit, throw up', χάσκω 'yawn', χέζω 'ease oneself

ΤΗΕ

( 19)

MIDDLE VOICE AS Α COMPLEX CAΊΈGORY

άλλCι. μεθορμηθε1.ς έν1. κύμασιν έλλάβετ' αύτfις (ε

81

[= σχεδίης]

325)

( ... ), but lunged after it arnid the waves, and laid hold of it [=

ιhe

raft].

Normally, actiνe (χω and λαμβάνω goνem the accusatiνe, markiπg the direct object. This direct object is a patieπt iπ that the ageπt­ subject exercises complete power over it. Ιπ the examples above, however, (χω and λαμβάνω have geπitival complemeπt. The verb may be active as well as middle, e.g. (active) λάβε πέτρης (ε 428), and (middle) έλλάβετ' αυτης (ε 325). Ιπ this coπstructioπ, the complemeπt is ποt a patieπt, siπce the subject does ποt exercise complete power over it. The geπitive case expresses that the ageπt­ subject caπ oπly take hold of a part of the object (partitive geπitive). The fact that both the active aπd the middle form occur iπ this geπiti­ ve coπstructioπ, is aπ iπdicatioπ of the uπmarkedπess of the active (see sectioπ 1.1.4 ), i.e., the active form can ofteπ be used wheπ the affectedπess of the subject is iπfeπable from the coπtext. 128 Now the questioπ remaiπs as to what the semantic coπtributioπ is of the middle voice iπ this type of coπstructioπ. Ιπ the coπstructioπ at issue, the subject uπdergoes the major effect of the eveπt. lt is the subject that, by performiπg the actioπ, becomes steady aπd stabilized, as opposed to the active coπstructioπ with aπ accusative-object. Thus, the subject may be coπsidered a (albeit somewhat uπtypical) patieπt. 129

128 Notice that ιhe construction of [χομαι + gen. in ιhe sense 'take hold or must be sharply distinguished from ιhe construction of [χομαι + gen. that means quite ιhe opposite 'keep oneself back from', e.g.: αύτοl μt:v έχώμεθα δηιοτίlτοςΙ έκ βελέων (Ξ 129) ('Then we will keep ourselves back from ιhe fight, beyond ιhe range of missiles'). The Jatter construction is an instance of the body motion midd\e, which can a\so be concluded from the existence of a causative active construction (e.g. Β 275, Ν 687). 'Έχομαι + gen. in the sense 'be next to' has probab\y developed from the meaning 'take ho\d or 129 This type of reflexive, mainly occuπing with verbs of holding and grasping, is a\so known from modem Janguages that have reflexive systems. Compare, for instance, German den Stuhl festhalten vs. sich am Stuhl festhalten. The Russian counterpart of this construction is called consequential rejlexive by Gerritsen ( 1990: 88ff.). Note further that in German, as in Greek, the construction does not govem a direct object (cf. am Stuhl).

82

CΗΑΡΊΈR

2

2.1.5 Collective Motion Middle This middle usage type concems verbs that designate motion types that are naturally and necessarily performed by groups of (typically animate) individuals, namely gatherίng and dίspersίng. Examples are: (20)

πυθόμενοι δε οί Κ&ρες συνελέyοντο έπ1. Λευκάς τε στήλας καλεομένας κα1. ποταμον Μαρσύην

(... ) (Hdt. 5.118.1)

And when the Carians heard this, they assembled at the place called the White Pillars, by the ήver Marsyas (... )

(21)

rος τ&ν μισyομένων γένετο ίαχή τε πόνος τε (Δ

456)

So from them, while they were joining in battle, came shouting and toiling.

The active transitive counterpart of the collective motion middle - like that of the body motion middles and the spontaneous process middles - has causative meaning. Examples are: 130 'gather (intr.)' 'gather (intr.)' άθροίζομαι 'gather (intr.)' άολλίζομαι 'gather (intr.)' λύομαι 'dissolve (intr.)' 131 μίσγομαι 'mingle (intr.)' 132 σκίδναμαι 'disperse (intr.)' (συλ)λέγομαι 'gather (intr.)'

'gather (trans.)' 'gather (trans.)' άθροίζω 'gather (trans.)' άολλίζω 'gather (trans.)' λύω 'dissolve (trans.)' μίσγω 'mingle (trans.)' σκίδνημι 'disperse (trans.)' (συλ)λέγω 'gather (trans.)'

άλίζομαι

άλίζω

άγείρομαι

άγείρω

Unlike the reciprocal event type (see next section), the participating

Ι notice in passing that all middles of collective motion have passive aoήst forms in Classical Greek. Ι will retum to this matter in ch. 3. 131 E.g.: λ\iτο δ' άγών (Ω 1). Obviously, this is a meιaphoήcal extension of the physiological meaning 'dissolve' (sponιaneous process middle). 132 Μίσγομαι is often construed with a dative-complement, e.g., Τυδείδης δ' έξα\iτις ίrον προμάχοισιν έμίχθη (Ε 134) ('The son of Tydeus retumed again and mingled with the foremost fighters'). The semantic difference between the construction with dative and the construction without dative is that, in the former case, there is an asymmetrical relationship between the participants involved. In the construction with dative, one individual is singled out as the subject-referent. The dative case can be explained as a comiιative dative (see Schwyzer, Π: 160), cf. English mingle with, meet with. This type of construction is a departure of the collective motion middle since it may involve only one moving participant. However, it still resembles the typical collective motion middle in that it necessarily involves more than one individual. 130

ΤΗΕ MIDDLE VOICE AS Α COMPLEX CA'fEGORY

83

individuals do not direct the action towards one another. Nevertheless, the collective motion type has in common with the reciprocal middle that the action can only take place by virtue of the individuals behaving in an identical and joint fashion. The difference between them is that in the reciprocal event rnicro-events are distinguishable to a certain degree: Α acts on Β as Β acts on Α. Collective motion cannot be analyzed into distinct rnicro-events: the entities involved necessarily act as an inseparable collective. Α further difference is that a collective motion event involves a change of state of the subject: a number of separate individuals tums into a unified collective, or vice versa. 133 The subject of a collective motion is an agent since it initiates and performs the action. Yet, with respect to volίtίon the collective subject ranks lower than the subject of individual motion, since the actions of the individual participants in a collective motion event are not conceived of as separate actions. The collective event is viewed as a single action carried out jointly by a group of individuals in which the individuals are completely dependent on one another if the action is to succeed. Clearly, the volition of such a collective is restήcted as it cannot simply 'make up its mind' to caπy out an action. Accordingly, collective motion events can, to some extent, be conceived of as processes which take place spontaneously. As a result, semantically, the collective motion rniddle is located in a continuum between the (individual) body motίon mίddle (section 2.1.4) and the spontaneous process mίddle (2.1.2) 134 • The collective motion middle is also related to the passive (2.1.1). In both types, the subject undergoes the effect of the event, and is thus a patient. The difference between the two types is the degree in which the subject can be regarded as the Initiator of the event. However, in many cases it is unclear or iπelevant who or what is the initiating entity. For example, (22)

Λi>το δ' άγών, λαο1. δΕ: θοιΧς έπ1. νfιας eκ:αστοι έσκ:ίδναντ' ίέναι (Ω 1-2) Then the assembly was broken uplbroke up, and the men scattered, each man to go to his own ship.

133 Because of their sirnilarity, Kernmer (1993: 123-5) treaιs the naturally collective event type as a subtype of the reciprocal situation type (in her chapter 4.1). 134 Manney treaιs the collective type as a subtype of spontaneous changes of state (2000: 175).

84

CΗΑΡΊΈR

2

The verb λ\>το can both be interpreted as a collective motion 135 and as a true passive 'be broken up' since, in epic at least, assemblies are normally broken up by the leader. This is shown by a parallel in the active voice: (23)

λ\>σεν δ' άγορηv αίψηρήν. οϊ μεν &ρ' έσκ:ίδναντο, έa προς δώμαθ' εκ:αστος (β 257-8) And he broke up the assembly, bήef as it was. They then scattered, each one to his own house; (... )

In the case of λ\>το at Ω 1, the role of the leader - probably Agamemnon and his order to break up the assembly remain completely implicit. If, however, we interpret λ\>το as a collective motion, the participants themselves are viewed as taking the initiative to break up the assembly. Of course, this difference in interpretation is very important to the overall interpretation of the narrative. What is relevant to the story is the fact that the assembly broke up, not who did it. This example shows, however, that the difference between passive middle and intransitive collective motion middle is a matter of degree, depending on the saliency of the role of the leader of the collective. The more salient the role of the leader as an Initiator is, the more we are entitled to regard the verb as passive.

2.1.6 Reciprocal Middle According to Lichtenberk (1985: 21), reciprocal events involve contexts in which "there are two participants, Α and Β, and the relation in which Α stands to Β is the same as that in which Β stands to Α". 136 In Greek, reciprocal middle verbs are rare. All reciprocal s Note that also έσκίδναντ' 'they scattered' denotes a collectiνe motion. The reciprocal middle is semantically middle as the subject is both Initiator and Endpoint of the eνent. Reciprocals are semantically related to reflexives. In both types the subject (typically animate entities) performs the same action as it undergoes; the subject is both Initiator and Endpoint. The difference between the reciprocal middle and the direct reflexive middle is that the former involves a second participating entity, while the latter does not. The semantic similarity between reflexives and reciprocals is manifested by the fact that they are formally marked alike in many languages, as in French ils se regardent. This utterance is interpretable as 'they look at themselves' and as 'they look at each other' Langacker (199la: 370) puts it in these words: "What motivates se's extension from reflexive to 13

136

ΤΗΕ MIDDLE VOICE AS Α COMPLEX CAτEGORY

85

middle verbs denote natural reciprocal events, that is to say, events that normally involve more than one participant acting on each other. Most reciprocal middle verbs are media tantum, and, for the lion's share, they relate to events that have to do with contending. Conversely, events that are normally not performed in a reciprocal configuration are expressed by an active verb and the reciprocal pronoun άλλή­ λους. This phenomenon is analogous to the use of the direct reflexive middle vs. the reflexive pronoun (see section 2.1.7). The first type of reciprocal rniddle verbs Ι will discuss concerns the reciprocal verbs that have one argument. For example, 'contend, fight' 'fight with missiles' δηρίομαύδηριάομαι (Horn.) 'contend' άγωνίζομαι

άιφοβολίζομαι

Α

typical example is the following:

(24)

καΙ. τα μεν στρατόπεδα άμφότερα οϋτως ήγωνίσατο

(Hdt.

1.76.4) So hard did the two arrnies fight.

The two armies are viewed as one holistic referential entity both initiating and perforrning the action (agent) and undergoing the action (patient). The second type of reciprocal concerns two-participant verbs. The individual entities involved are not fused into one subject-participant, but are expressed by separate constituents. In Greek the second argument has the dative-case. 137 Examples are:

reciprocal use is the shared conception of the sarne participant (necessarily plural in the case of reciprocals) functioning as both the head and the tail of an action chain or its analog." The relationship between reflexives and reciprocals across Ianguages is discussed in several papers collected in Frajzyngier & Curl (2000a). 137 Kίihner-Gerth (1: 432) and Schwyzer (11: 161) treat the dative with verbs of fighting as a subtype of the cornitative dative. The dative rnay also express that the participant has the semantic role experiencer in the sense as it is used by Langacker. According to Langacker the expeήencer role has a dual nature. In some expressions the expeήencer only passively registers sensations. However, there are also expressions in which the expeήencer can be attributed some kind of initiative role (Langacker 1991a: 327-9, 2000: 31). In the sarne way, the dative participant with verbs of fighting has a dual character. On the one hand, s/he registers the acts of violence performed by the subject; on the other hand, s/he is actively involved in that slhe

86

CHAP'ΓER

άμιλλάομαι

2

(+ dat.) 'contend with'

διακοντίζομαι (+ dat.) 'contend with others at throwing διαλέγομαι(+ dat.) 'conνerse with' 139 έρίζομαι (+ dat.) 'stήνe, quarrel with' (Hom.) 140

the jaνelin' 138

(+ dat.) 'fight, contend with' (Hom./poetry) (+ dat.) 'fight, quarrel with' πληκτίζομαι (+ dat.) 'exchange blows with' (e.g. Φ 499) ώστίζομαι (+ dat.) 'push s.o. and be pushed' μάρναμαι μάχομαι

An example in context: (25)

οtα ποιε\>μεν, άνδρες Σκύθαι. δούλοισι τοϊσι ήμετέροισι μαχόμενοι

(... ) (Hdt. 4.3.3)

Men of Scythia, what are we doing? While we ar.e fighting with our own slaves ( ... ).

The two reciprocal types mentioned above (with and without dative complement) involve a different conceptualization of the event. When the entities are coded as one participant (i.e. the subject), the individual entities involved are viewed as a single holistic entity (cf. the cats in The cats fight). The event itself is conceived of as a cluster of symmetrical causal relationships. Νο prominence is given to either one of the micro-events, i.e. individual Α acting ση Β = Β acting on Α. The construction with a dative-complement, however, presents the subject-participant as being the most prominent participant (generally the discourse topic ), and, as a consequence, the outgoing causal chain from subject to dative is highlighted. Consider the clause The cat fights with the dog. The incoming causal chain, the fighting of the dog

reciprocates these acts of violence. Interestingly, Apollonius seems to give a similar analysis of the meaning of the dative-case with verbs of fighting. He explains the occuπence of the dative by the idea that the dative-participant, like the subjectparticipant, is both actively and passively involνed in the action (A.D. Synt. 428 Uhlig; Lallot 1994; Lallot 1997, 1: 267). 138 Simplex άκοντίζω 'hurl a javelin' is always active. The preposition δια­ contήbutes a sense of dividedness and mutuality. Compare also δια-λέγομαι 'converse' and λέγω 'speak'. 139 Unlike the other reciprocal verbs διαλέγομαι has a passive aoήst form, διελέχθην. In section 3.2.3.3, Ι will make aπ attempt to explain this rather remarkable form. 140 The actiνe έρίζω has the same meaning, and occurs more frequently. Possibly, the actiνe presents the event as a neutral activity 'make Εpις', whi\e the middle profiles the reciprocal facet.

ΤΗΕ MIDDLE VOICE AS Α COMPLEX CAτEGORY

87

which affects the cat, is somewhat backgrounded. ln Greek, the incoming action chain is, however, explicitly coded by means of the middle inflection of the verb. 141 As was said before, the reciprocal middles are typically verbs of fighting, and most of them are media tantum. However, there appear to be sporadic cases of middle verbs with reciprocal meaning that coπespond to non-reciprocal active verbs. An interesting example is: (26)

άντ1.

γCι.ρ

του

προσαγορεύειν

άλλήλους

φιλέουσι

το'iσι

στόμασι, ην δe ~ οϋτερος ύποδεέστερος όλίγφ, τας παρειας φιλέονται

(... ) (Hdt. 1.134 .ι)

Instead of greeting, they kiss one another on the mouth. If one of the two is a little humbler, they kiss on the cheek.

Remarkably, Herodotus first uses the expected construction of the active verb φιλέω 'kiss' with the reciprocal pronoun άλλήλους, but in the next clause the middle φιλέονται is used, apparently with a reciprocal meaning (the accusative τCι.ς παρειάς must be taken as a accusative of the body part affected). The use of the middle form to express a reciprocal meaning is possible here because a direct reflexive interpretation is precluded: kissing, like fighting, is not an activity

With semantically related verbs like έρίζω 'quarrel', νεικέω 'quarrel', 'quarrel', and παλαίω 'wrestle', which can also have a dative complement, the reciprocity is not coded grammatically, and, as a consequence, deemphasized, e.g. παλαίω Φιλομηλείδυ ('Ι am wrestling with the son of Philomeleus'). It is, however, still an inherent facet of the lexical semantics of the verb. If the speaker wishes to present the event as symmetrical, the construction with the reciprocal pronoun (in the dative) must be used, e.g. παλαίουσι άλλήλοις ('they are wrestling with each other'). Note, further, that there is also a group of verbs with an accusative complement that may well be considered reciprocals, e.g., άΎΙCαλίζομαι + acc. 'embrace', άσπάζομαι + acc. 'welcome, greet', δεξιόομαι + dat./+ acc 'greet with the ήght hand' (άΎΙCαλίζομαι and δεξιόομαι have a passive). The middles διανέμομαι and δαίομαι!δατέομαι 'divide (among one another)' are often claimed to be reciprocals (K-G, 1: 90, Gildersleeve 1980: 67). However, Ι prefer to consider these verbs indirect reflexive middles since (i) they are - unlike typical reciprocals - transitive, and (ii) the subject also receives a share (i.e., the subject is beneficiary), e.g., (... ) 141

πολεμέω

διενείμαντο τfιν άρχfιν ό Ζεύς, ό Ποσειδίί':ιν καΙ. ό Πλούτων, έπειδη παρa του

(PI. Grg. 523a) ('( ... ) Zeus, Poseidon, and Pluto divided the sovereignty among themselves, after they took it over from their father').

πατρος παρέλαβον.

88

CΗΑΡΊΈR

2

one usually perforrns on oneself. 142 Another case of a reciprocal middle verb in opposιtιon to a non-reciprocal active is άμείβομαι in Homer. For example, (27)

οί μtν άμειβόμενοι φυλακιΧς Εχον (I

471)

They kept watch altemating with one another.

Consider further forrnulas such as αύτaρ έγώ μιν άμειβόμενος προσέειπον. These middle verbs cannot be interpreted as passives, since their subjects are actively involved. The active άμείβω means 'exchange s.th.'; the reciprocal middle originally must have meant 'exchange one another', i.e. 'swop over' This meaning then developed into the idea of altemating with one another, as, for instance, in changing the guard and in answering one another (see also section 2.1.1 0). 143 2.1.7 Direct Reflexive Middle

The direct reflexive middle type involves a human agent that volitionally perforrns aπ action on him or herse1f. 144 Many direct reflexive middle verbs relate to grooming activities such as 'bathing', 'dres-

142 Besides the reciproca\ pronoun, the reflexive pronoun can a\so be used with a reciproca\ meaning (see K-G, 1: 573-5). The reflexive pronoun is prefeπed in cases where there is a conιrast with a third group of individua\s, e.g. φθονο\Jσιν έαυτοiς μfιλλον i\ τοiς ίiλλοις άνθρώποις (Xen. Mem. 3.5.16) ('they are more envious and bitter against one another than against the rest of the world'). But often the reflexive pronoun seems to be used merely for the sake of variation, e.g. φθονο\Jντες έαυτοiς μισουσιν άλλήλους (Xen. Mem. 2.6.20) ('envying each other, they hate one another'). 143 Note that keeping watch in shifts may involve what Kemmer cal\s a chaining situation: "a chain of paired relations in an ordered seήes of participanιs. (... ). Each participant is the Initiator in one relation and the Endpoint of a second relation of the same type" (Kemmer 1993: 100). In other words, in the chaining situation Α acιs on Β, Β acιs on C, and so forth, whereas in the typical reciproca\ situation Α acιs on Β and vice versa. 144 The direct reflexive event type is semantically middle (as opposed to passive) in that the subject is both the Initiator of the action, and the Endpoint (affected entity).

ΤΗΕ MIDDLE VOICE AS Α COMPLEX CA'ΓEGORY

89

sing', etc. 145 The action is typically executed on the surface of the body by rneans of the hands. Exarnples of direct reflexive rniddles are (see also K-G, 1: 103-4): 146 Grooming 'anoint oneself 'blow one's nose' άποψάομαι 'wipe one's nose' εvνυμαι 'dress oneself ζώννυμαι 'gird oneself κείρομαι 'cut off one's hair, have a hair-cut' κοσμο\>μαι 'adorn oneself λο\>μαι 'bathe oneself νίζομαι 'wash one's hands/feet' ξυρο\>μαι 'shave oneself άλείφομαι

άπομύττομαι

Other types of direct reflexiνe actions: άπάγχομαι 'hang oneself γυμνάζομαι 'exercise oneself δέφομαι 'masturbate' 147

145 Α difference between the direct reflexive middle and the body motion rniddle is that the direct reflexive middle does not denote a total physical change of state. When one shaves oneself, for instance, one's body does not undergo a complete change of state, the affectedness is restricted to a particular zone, typically on the surface of the body. Conversely, activities like 'stand up', 'move' necessaήly imply a complete change of state, or more precisely, a change of body posture or location. The direct reflexive middle καθαίρομαι 'puήfy oneself (aor. έκαθηράμην) may be the proverbial exception to the rule since the verb does denote a change of state ('become καθαρός'). This is, however, not a very striking exception to the rule, if we consider that λο\>μαι, although it does not denote that the subject becomes clean, will generally imply that the subject becomes clean. Therefore, an utterance such as Ι washed myself, but Ι did not become clean is somewhat odd - though not impossible. Furthermore, a perfect like λελουμένος 'be washed' necessaήly implies that the subject is clean at the moment of utterance. 146 Note that all direct reflexive middle verbs have sigmatic middle aoήst forms, e.g. έκειράμην Ί had a hair-cut', άπηγξάμην Ί hung myself Ι will return to the aorist morphology extensively in ch. 3. Note, further, that the active counterparts of direct reflexive middles do not have a causative meaning, as opposed to the body motion middles (ίστημι 'make s.o. stand' vs. ϊσταμαι 'stand still'). Thus, λούω does not mean 'make s.o. wash himself, but 'wash s.o.' Causation is expressed by the construction of ποιέω with an infinitive, e.g., άπάγξασθαί με ποησεiς (Theocr. 3.9) ('Thou'lt make me hang myself). The same point is made by Rijksbaron (2002 3 : 154nl). 147 The original meaning of active δέφω was 'make supple with the hand, knead', cf. the derivative δέψω (Hdt. 4.64.2).

90 CHAP'fER 2 έντύνομαι (Hom.) 'prepare oneself 148

oneself 'scratch oneself κόπτομαι 'beat oneself 149 παρασκευάζομαι 'prepare oneself στεφανόομαι 'crown oneself σφάττομαι 'kill oneself τύπτομαι 'beat oneself καλύπτομαι 'coνer ιcνάομαι

Direct reflexive middle designate actions that are normally performed on oneself. Actions that are not normally performed on oneself are coded in aπ active construction with a reflexive pronoun (see also KG, 1: 111): 150

(28)

ρίπτει αύτον είς τitv θάλατταν (Dem. 32.6) He throws himself into the sea.

148 Cf. active plus pronoun: εt έντύνασαν ε· αύτήν (Ξ 162) ('( ... ), when she had beauteously adorned herself, (... )'). 149 As an act of mourning (likewise τύπτομαι). The frequent use of these νerbs in contexts of gήef eνentually effected a lexical semantic shift as they gradually came to be used as actual verbs of mourning that are construed with an object of the person being mourned for (cf. όλοφύρομαι), e.g. τύπτομαί τινα Ί mourn for s.o.' 150 More examples in Gildersleeve (1900: 67-8). Α look at the frequency ratios of actives vs. direct reflexives is illustrative. Δίδωμι 'give' denotes an event of which subject and object are usually not co-referential. Therefore, direct reflexivity is expressed by an actiνP- verb form and a reflexive pronoun. Α quick glance in Powell's lexicon to Herodotus shows that only 20 instances (of a total of 270 active forms) of δίδωμι concern the reflexive construction with έαυτόν, i.e. less than 10%. Compare this figure with the ratio of active forms of άλείφω, άπάγχω, γυμνάζω, λούω, νίζω vs. direct reflexive middles of the same verbs. The following figures are based on an index search with Pandora in the ΊLG. Ι have counted the 3rd person aoήst forms of the simple and compound verbs. The aoήst forms are used in order to be able to filter out the passive forms: άπίlγξε 2χ (2%) vs. άπήγξατο 81χ (98%), ilλειψε 98χ (76%) vs. ήλείψατο 31χ (24%), έγύμνασε 45χ (78%) vs. έyυμνάσατο 13χ (22%), ελουσε 24χ (39%) vs. έλούσατο 60χ (71%), Ενιψε 53χ (48%) vs. ένίψατο 58χ (52%). lt can be concluded that the direct reflexive middles occur quite frequently as compared to their active counteφarts (ranging from 98% [άπήγξατο] to 22% [έyυμνάσατο]), whereas the actiνe (pronominal) direct reflexive has a frequency of less than 10% (δίδωμι). The verb άπάγχομαι 'to hang oneself is almost a medium tantum: the - rarely occurήng active never means 'to hang s.o.', but more generally 'to strangle' Ι have not taken κόπτομαι and τύπτομαι into account (of which the active forms are far more frequent than their direct reflexive middles) because of their highly restricted, idiomatic use 'beat oneself as a token of mourning'.

ΤΗΕ MIDDLE VOICE AS Α COMPLEX CAτEGORY

91

Although in the latter construction the subject-referent is affected (so in pήnciple the middle form would be semantically possible), the middle form is not used. This is economically motivated: the speaker need not use the middle voice to express reflexivity, as this is already expressed by means of the reflexive pronoun. 151 An interesting example is found in Xenophon: (29)

γCx.p (... ) προαποθν~σκουσι ύπο του φόβου, οί μ(ν pιπτουντες έαυτοi>ς, οί δε άπαnόμενοι, οί δ' άποσφαττό­

EV\01

μενοι

(Xen. Cyr. 3.1.25).

Some (... ) die in teπor before their time - some by hurling themselves over a precipice, others by hanging themselves, others by cutting their own throats.

Here, both an active verb with a reflexive pronoun, and two middle verbs occur side by side. Το explain these different strategies, Gildersleeve (1900: 64) remarks "The only middle for self-murder is άπάγ­ ξασθαι, to hang oneself [but what about άποσφαττόμενοι?, RJA], which seems to have been the most natural form of suicide" This conclusion is not entirely coπect. The conclusion should not be drawn from the fact that άπάnομαι is a middle that, to Greeks, hanging oneself is the natural way to commit suicide, as opposed to other means (e.g. pίπτειν έαυτόν). Rather, we should infer that it is equally or more natural to hang oneself than to to hang someone else, just as it is equally or more natural to wash oneself than to wash another (λούω vs. λο\>μαι). Obviously, the same cannot be said of pίπτω 'throw' lndeed, this point is clearly supported by the frequency figures: the ratio active άπάnω vs. middle άπάnομαι is about 2 % vs. 98 %! (see note 150). Likewise, σφάζομαι 'kill oneself' seems to be an action that is naturally performed with respect to oneself. 152 The example in Herodotus (1.45.3), which was already mentioned in section 1.1.4, of the active plus reflexive pronoun is a special case: 153

151 As Kemmer puιs it (1993: 235): "Don't bother to mark semantic properties that are inherent in the event". In section 1.1.4, Ι have argued that the active foπn can be used in these reflexive contexιs because the active voice is unmarked with respect to the feature of subject-affectedness. 152 For a similar view, see Cock (1981: 8). 153 In combination with a reflexive pronoun, the middle voice may also be used, e.g. Xen. An. 1.8.29: έαυτον έπισφάξασθαι. Again, the reflexive pronoun emphasizes that the subject (Artapates) slew himself, rather than that someone was ordered

92 CHAP'fER 2

(30)

'Άδρηστος δε (... ), οi>τος δη ό φονεi>ς μεν του εωυτου άδελφεου

γενόμενος,

φονευς

δε

του

καθήραντος,

( ... )

έπικατασφάζει τφ τύμβφ έωυτόν (Hdt. 1.45.3) But Adrastus (... ), he who became the slayer of his own brother and the slayer of his puήfier, (... ), slew hirnself over the grave.

The fact that Adrastus kίlls hίmself ίs explίcίtly and emphatίcally contrasted wίth hίs kίllίng hίs brother and the son of hίs benefactor, whίch was mentίoned ίη the first part of the sentence. Another type of context ίη whίch the actίve νerb plus the reflexίve pronoun occurs ίnvolves contrastive emphasis. Consίder (31)

(... ) όπότε -yυμνάσαι βούλοιτο έαυτόν τε και τους ϊππους (Xen. An. 1.2. 7) ( ... ) whenever he wished to give himself and his horses exercise.

By means of the reflexίve pronoun, the patίent ίs presented as ίf ίt were a separate entίty. Thίs makes ίt possίble for the patίent - ίη spίte of the referentίal ίdentίty wίth the agent-subject - to be contrasted wίth a second, extemal, patίent-partίcίpant. In the dίrect reflexίve middles (λο\>μαι), however, the Inίtίator­ aspect of the subject ίs also, to some extent, conceptually dίstίn­ guίshed from the Endpoίnt-aspect. Α separatίon ίs conceίved between the agent-mίnd of the subject, and the patίent-body. Thίs conceptual dίstίnction of Inίtίator-aspect and Endpoίnt-aspect ίs facίlίtated by the salίent outward manίfestatίon of the actίon: the clearly vίsίble actiνίty of the hands on the body. By contrast, ίη the case of the body motίon mίddles, the Inίtίator-aspect and the Endpoίnt-aspect are less clearly dίstίnguίshable. In aπ actίon such as 'sίt down', the body as α whole ίs both performίng and undergoίng the event. Above, Ι have dίscussed the dίfferences between the dίrect reflexίve middle, the dίrect reflexίve active plus pronoun construction, and the body motίon mίddle. These three constructίon types can be located on a scale that relates to the relatίve dίstίnguίshabίlίty of the agent and patίent partίcίpants (see also Kemmer 1993: 73). Thίs scale also ίncludes the actίve transίtίve constructίon and the actίve ίntransίtί­ ve construction. At the one extreme of the scale, the prototypical

to slew Artapates, as was mentioned in the preceding sentence. The middle voice here redundantly expresses reflexivity (see section 1.1.4).

ΤΗΕ MIDDLE VOICE AS Α COMPLEX CAΊΈGORY

93

transitive is located. The prototypical transitive clause is characterized by maximal distinguishability of agent and patient, since they are clearly distinct participants refeπing to different entities in reality. Lower distinguishability is inherent in the pronominal direct reflexive since the two participants are coreferential. Positioned still lower οη the scale is the direct reflexive middle. The distinguishability of the participants is lower because the agent and patient are fused into one participant (the subject). However, there still remains some conception of separateness of the agent and patient due to the middle inflection that codes the patient-role of the subject. At the other extreme of the scale, we find the active intransitive. In the active intransitive clause there are ηο agent and patient to be distinguished. One sole participant remains: the subject. This participant can be either like an agent (e.g. εtμι 'Ι am going'), or more like a patient (άποθν'(tσκω 'Ι am dying'). Interestingly, the conceptual distinguishability of participants is iconically reflected in the formal structure of the respective expressions: from a full noun that expresses the patient in the prototypical transitive, via a reflexive pronoun that expresses the patient in the direct reflexive active, through an inflection that expresses the affectedness of the subject in the direct reflexive middle, and finally to the total absence of a separate patient-role in the active intransitive verb. This coπelation between conceptual distinguishability and formal structure is represented in the following table: Table 1:

Correlation between conceptual distinguishability and formal structure

Distinguishability of patient-participant

Construction

Foπnal expression of patient-participant

HIGH

Transitive Active Direct Reflexive Active Direct Reflexive Middle Intransitive Active

FULL NOUN PRONOUN (έαυτόν) INFLECτiON (middle endings)

LOW

The coπelation between, on the one hand, the degree of conceptual separation between the two participants, and, on the other, the degree to which they are lexically or grammatically autonomous, has already been noted by Haiman (1983), Haiman (1985: 105), Langacker (1991a: 370), Croft (1990: 174), and Kemmer (1993). The direct reflexive middle is related to the indirect reflexive middle. This is shown by the existence of a special construction that

94

CΗΑΡΊΈR

2

seems to occupy aπ inteπnediate posιtιon between the two middle types. This construction is exemplified by the following lines from Homer (taken from Chantraine 1963: 178): 154 (32)

a.

230)

νίψατο δ' αύτος χε'iρας (Π

He himse1f washed his hands. b. χρόα κ:αλΟν άλειψαμένη (Ξ 175) After she had anointed her beautifu1 sk.in, (... ) c. τεύχεα τ' έξεδύοντο (Γ 114) 155 (... ), and they took off their armour.

d.

κ:εφαλitν δ' ο γε κ:όψατο χερσί (Χ

33)

He beat on his head with his hands. 156

The middle verbs are to be interpreted as direct reflexives, i.e. νίψατο = 'he washed himself' The active counterpart supports this analysis, compare: (33)

τφ σε πόδας νίψω (τ 376) Therefore Ι will wash your feet [lit.:

Ι

will wash you the feet]

The active verb νίψω takes a complex direct object, consisting of a whole (σε) and a part (πόδας). The fact that the possessor of the feet is expressed in the accusative case (i.e. as a patient/direct object) confiπns the analysis that the middle is to be viewed as a direct reflexive, and not as aπ indirect reflexive. In that case, we would expect a construction with a dative, i.e. *σοι πόδας νίψω (lit.: 'Ι will wash the feet for you'). 157

154 Chantraine (1963: 178) notes: "Les categoήes du moyen transitif ου intransitif ( ... ) ne peuvent pas toujours se separer de maniere tranchee." For this construction, see also Munro (1891: 131-2), Gi1ders1eeve 1900-11: 68-9), and Jacquinod (1989: 29-30). 155 Note that the use of the construction also extends to clothing (e.g. τεύχεα). After body parts, clothes are the things that are closest to the body. 156 Other examples are ιcαταμύξατο χεiρα (Ε 425) and δάιcρυ' όμορξάμενον (θ 88). 157 K-G appear not to have understood this construction coπectly. They give a concocted example, λούομαι τους πόδας, and render it as "ich wasche mir die Fίisse". Apparently, they take this type of construction as an indirect reflexive middle, with τους πόδας as a direct object. It is, however, better to interpret the accusative in this type of construction as an accusative of respect. Furthermore, the coπect Attic form should have been contracted λο\Jμαι, and the verb only refers to

ΤΗΕ MIDDLE VOICE AS Α COMPLEX CAΊΈGORY

95

However, direct reflexives of the type νίψατο χεΊρας resemble reflexive middle iπ that it is coπstrued with aπ accusative (χεΊρας). Oήgiπally, this accusative-coπstitueπt was a direct object (as iπ the active couπterpart with a double (whole-part) object above). However, the accusative may have beeπ reaπalyzed as aπ accusativus respectus, i.e. 'Ι wash myself with respect to the haπds' This reaπalysis possibly occuπed after Homer wheπ the active coπstructioπ with a double (whole-part) object had disappeared. Besides the preseπce of aπ accusative, the direct reflexive coπstructioπ of the type νίψατο χεΊρας resembles the iπdirect reflexive coπstructioπ iπ aπother respect. Ιπ geπeral, the subject caπ be said to benefit from the actioπ. Ιπ sum, although this special Homeήc constructioπ should, strictly speakiπg, be iπterpreted as a direct reflexive coπstructioπ, it resembles, iπ some respects, the iπdirect reflexive. the

iπdirect

2.1.8 Perception Middle The middle of perceptioπ relates to eveπts iπ which aπ aπimate subject perceives aπ object through οπe of the seπsory orgaπs. The perceiver is meπtally affected by the perceptioπ. The subject caπ, therefore, be coπsidered aπ experίencer. Perceptioπs caπ be both volitioπal aπd πoπ­ volitioπal. However, it appears as Ι shall argue below that the middle perceptioπ verbs iπ Greek teπd to express a volίtίonal activity. The perceiving subject is typically actively iπvolved iπ the perceptioπ. Examples are γεύομαι 'taste' aπd σκέπτομαι 'look at, coπsider' The subject caπ thus be coπsidered aπ agent. Most middle verbs of perceptioπ are media taπtum, with the exceptioπ of γεύομαι, which has aπ active causative couπterpart γεύω 'make s.o. taste' .158 Before

washing the who\e body, whi\e washing hands or feet is expressed by the verb (άπο)νίζω (see Ruijgh 1991: 696). Unfortunate\y, due to K-G's authoήty this example made its reappearance e\sewhere seνeral times, cf. Smyth-Messing 1956: 390, Schwyzer, Π: 229, Garcίa Gual 1970: 25, Kemmer 1993: 1. 158 The νerb φράζομαι 'notice, consider, observe' also has a causatiνe active φράζω 'indicate, tel\' However, φράζομαι is not a true perception verb since it more often denotes purely cognitive activities. The medica\ author Galen (12.795) has an iso\ated causative όσφραίνω 'make one smell at a thing' Yet, there is a c\ear tendency for mental process middles, as opposed to perception middles, to haνe active causative counterparts. So we haνe menta\ processes like μιμνflσκ:ομαι

96

CΗΑΡΊΈR

2

Ι continue with a discussion of the function of the middle voice in the class of perception verbs, Ι shall dwell briefly on the various caseconstructions that occur with verbs of perception. The perceived object can be marked by the accusative and the genitive. With verbs of seeing, the perceived object is in the accusative; with verbs of hearing and smelling the perceived object is in the genitive. This difference relates to different conceptualizations of the events. The accusative-case expresses that the event is viewed as being analogous to the prototypical transitive clause, i.e. the perceiver is viewed as an Initiator that directs his view to the perceived object (see section 1.1.2). 159 By contrast, the genitive case with verbs of hearing and smelling can be interpreted as a genitive of source. The genitive case evokes the image that the perception emanates from the perceived entity towards the perceiver. 160 Verbs of heaήng may also be construed with an object in the accusative case which expresses the content of what is heard, e.g.,

'remember' - μιμν~σιcω 'remind', φοβέομαι 'be afraid' - φοβέω 'frighten', etc., but only γεύομαι γεύω. The explanation for this difference may be that it is more difficult to force a person to perceive something, than to cause someone to expeήence a change of mental state, since a perceiver (esp. in the case of middle perception verbs) has control of the action, and will be able to offer resistance to an extemal causer, wherea.~ in the case of mental processes the expeήencer is only passively involved in the event. The causer thus has more chance to succeed in his manipulation of the causee (see also Givόn 200Ia: 45, 48-9). Therefore, it is more Iikely that there is a lexicalized causative verb. 159 ΊΊιίs idea of perception is related to one that is often found in folk-theories, namely that the eye is thought of as emanating rays of light. This kind of "naive physics" is found in Empedocles (Fr. 84). He compares the eye with a lantem from which rays of light are leaping forth. This passage is cited by Aήstotle (de sensu 437b23ff.). Aristotle adds that Empedocles, in a passage now Iost, expresses a contrary view, namely that vision is to be explained by effluences of the objects seen. Apparently, Empedocles thought of seeing as a two-way interaction. Apollonius observes that there is a semantic difference between the verbs of that are construed with a genitive, and the verbs of seeing that have an accusative. With regard to heaήng, the subject, according to Apollonius, is passively involved. The body is, as it were, penetrated by the sensory expeήence. Conversely, seeing involves an active subject. The subject can control the event by closing his eyes. These semantic differences are reflected in their respective case-constructions (A.D. Synt. 417 Uhlig; see Lallot 1997, Π: 262-3). 160

heaήng

ΤΗΕ MIDDLE VOICE AS Α COMPLEX CAΊΈGORY

(34)

τα\>τα δ' έγrον Καλυψο\>ς i\κουσα (μ

97

389)

This Ι heard from Calypso.

The distinction between genitive and accusative with verbs of hearing could be explained semantically by the idea that the content of the perception is 'captured' completely by the perceiver (hence the accusative), as opposed to the source of the perception, which is 'captured' only partially. The genitive can therefore be interpreted both as a partitive genitive, and as a genitive of source. The actual sound heard can be expressed both in the genitive and in the accusative, e.g άκ:ούω φωνfιςΙφωνήν 'Ι hear a voice' Apparently, the perceived sound can be viewed either as partially or as completely 'taken in' Having digressed on the various case-constructions of perception verbs, Ι return now to the role of the rniddle voice within the class of perception verbs. On the basis of lexicalization patterns in different languages, Viberg (1984) makes a tripartite distinction in the domain of perception verbs. The first involves perception in which the experiencer is actively involved: controlled actίvίty in the terms of Viberg (e.g. 'look at', 'listen'). The second relates to perception in which the experiencer is more passively involved: non-controlled experίence ('see', 'hear'). The third type includes verbs in which the source is marked as the subject: source-based copulatίve state construction), e.g. 'the painting looks very old' . 161 It appears that this trichotomy is also relevant for the voice marking of perception verbs in Greek. In relation to the voice-marking of verbs of perception, Kemmer (1993: 136) remarks: ''lt remains to be seen whether the difference between more active and less active Experiencer verbs is supported by a difference in middle-marking patterns.'' Ι believe that, in Greek, a difference in middle-marking can indeed be discerned. Consider the following table:

161 It is hard to find examples of source-based perception verbs in Greek (i.e. 'Χ looks/ sounds/ smells/ feels/ tastes good'). The few examples Ι found seem to display a tendency to be active which may be connected to their stative semantics, cf. οζω 'smell' (sporadically middle οζομαι), πικρίζω 'taste bitter'. έγγλύσσω 'taste sweet'. Ι am not aware of source-based verbs meaning 'taste' or 'feel' ('Χ tastes/feels good') in Greek.

98 CHAPτER 2

Table 2: Greek verbs of perception: volitional vs. non-volitional +VOLffiON

- VOLffiON

άιcουάζομαι Ίisten' ιοz

άιcροάομαι Ίisten,

obey'

αίσθάνομαι 'perceiνe,

notice, leam'

όσφραίνομαι 'smell' 163

'taste' όράομαι (esp. Hom.) 'see' (Hom./poetry) 'fix one's eyes, gaze' θεάομαι Ίοοk at (wondeήng)' μαίομαι (Hom./poetry) 'seek for s.th. by touching' σκέπτομαι Ίοοk at' aιω (HomJpoetry) 'hear' άθρέω Ίοοk at' άιcούω 'hear' ΙΙΙS βλέπω Ίοοk at' (δι!έπ)οπτεύω 'watch' βλέπω 'see' ιcλύω (HomJpoetry) 'hear' 1.. ιcλύω (Homlpoetry) Ίisten' λεύσσω (Hom./poetry) Ίοοk at' λεύσσω (Hom./poetry) 'see' όράω (aor. εΙδον) 'see' σιcοπέω Ίοοk at' 164

γεύομαι

δέριcομαι

lt can be observed that there appears to be a tendency for middle perception verbs to be volitional. Exceptions are αίσθάνομαι, όσφραίνομαι 167 and the special middle verb όράομαι (see below). Cf. Boeder (1961-2: 103). Perhaps also όσμάομαι 'smell' belongs to the [- Volition] class, but this is difficult to assess due to its low frequency. 164 In Classical Greek, σκοπέω replaces σκέπτομαι in the present stem. There is also (albeit marginally) a middle form σκοπέομαι, probably contarninated by σκέπτομαι. Σκοπέω and -οπτεύω are denominatives (from σκοπός 'watcher' and -όπτης 'id.' respectively). The oήginal meaning of both verbs was probably stative 'be a watcher'. Note that stative verbs are typically actives. 165 The etymologies of άκούω and άκροάομαι are similar to one another. The former is a compound of the root άκ- 'sharp' and the root of ο~ς 'ear'; the latter is a compound of iiκρος 'pointed' and ο~ς. The verbs must, therefore, oήginally have meant 'have pointed/sharp ears, pήck up one's ears' (see Chantraine 1968). 166 The presents κλύω and cΗω are novel formations, created after a reinterpretation of the oήginally thematic aoήsts eκλυον and iiϊον as imperfects. 167 Interestingly, όσφραίνομαι has a passiνe aoήst ώσφράνθην (besides a thematic aoήst ώσφρόμην), whereas the volitional middle perception verbs al\ have a sigmatic middle aoήst (with the exception of eδρακον), e.g. έγευσάμην, έσκεψάμην. In ch. 3, Ι will argue that the sigmatic middle and the passive aoήst forms express, respectiνely, a low degree and a high degree of subject-affectedness. Incidentally, although the νolitional middle perception verbs that have an accusative object resemble prototypical transitives, it is clear that they are not. In Classical Greek, none of these middle perception verbs can be passivized, which can be explained by the fact that their objects are not patients (but rather zeros). Perhaps also the tendency to avoid ambiguity (e.g. σκέπτομαι Ίοοk at', as well as *'be looked at') may be operative. Note that an exceptional passive aoήst form σκεφθηναι 'to be 162

163

ΤΗΕ MIDDLE VOICE AS Α COMPLEX CA'fEGORY

99

Non-volitional perception tends to be expressed by active verbs. There are, however, also active verbs that express volitional activities. Note that three of them also have a non-volitional use: βλέπω, κ:λύω, λεύσσω (the latter two are restricted to poetry in the Classical language). The phenomenon that active verbs are also used to express meanings that are typically expressed by the middle voice is in accordance with the claim that the active form is semantically unmarked (see section 1.1.4). The verbs are classified according to their typical meaning. 168 For instance, although άκούω and ά"iω sometimes denote a volitional activity (in the sense 'listen, obey'), their prototypical meaning is nonvolitional ('hear'). Likewise, γεύομαι has a marginal non-volitional meaning 'experience', yet its prototypical meaning is volitional 'taste' (English taste is ambiguous in this respect). As to βλέπω, κλύω, and λεύσσω, both usages (+1- volition) are approximately equally frequent. Δέρκομαι is a Homeήc word, typically meaning 'fix one's eye, gaze, look at' (+ volition). In classical Greek, it is only rarely used in poetry, mostly in the sense 'see' (- volition). 169 The question remains as to how the tendency of middle perception verbs to designate volitional perception should be explained. At first glance, the pattem seems to contravene our intuitive convictions. We would expect that the more 'passive' or non-volitional the meaning of a verb is, the more likely it is to be marked by the middle voice. However, it should be kept in mind that the semantics of the middle voice is not directly related to the notion of volition. Instead, as we have seen in chapter 1, the crucial element of middle semantics is the notion of subject-affectedness. The proper question, therefore, regards why there should be a relation between volitional perception and affectedness. Ι will suggest here that the mental affectedness of the subject is the crucial element that explains the middle verbs of perception. The element of volition is to be regarded only as a sideeffect of the mental affectedness. For example, σκέπτομαι unlike active όράω - typically implies that the subject looks at something in order to acquire infonnation about it. lt is, therefore, not surpήsing observed' (which is foπnally distinct from έσκεψάμην 'Ι observed') is found in Hp. de Arte 11. 168 Based on two studies in the verbs of perception by Prevot (Prevot 1935b, 1935c). 169 For δέρκομαι and its remarkable passive aoήst foπns έδράιcην and έδέρχθην, see section 3.2.3.2.

100

CHAP'ΓER

2

that its meaniπg shades iπto the seπse 'coπsider, thiπk about'. Ιπ other words, the perceptioπ briπgs about a lastiπg effect οπ the cogπitive state of miπd of the perceiver. Aπother example is άκ:ροάομαι which can mean 'listeπ' as well as 'obey' The latter meaniπg implies that the listeπer' s meπtal attitude is affected to such aπ exteπt that he obeys the persoπ listeπed to. The stroπger meπtal iπvolvemeπt of the subject ίπ the case of volitioπal perceptioπ caπ also be of aπ emotional kiπd. For example, θεάομαι ofteπ implies that οπe looks at something with a sense of wondering. By contrast, the interaction between the perceiving subject and the perceived object of non-volitional perception verbs ('see', 'hear') is of a more mechanical character. As was noted above, volition is only a side-effect of the mental involvement of the subject, and not an essential property that explaiπs the middle voice of verbs of perceptioπ. This can clearly be seen in the case of όράω - όράομαι. The middle form όράομαι is the marked member of the oppositional pair. lt occurs less frequently 170 , and it marks an increased mental affectedπess of the perceiving subject. In his in-depth investigation of the opposition ίδεtν vs. ίδέσθαι ίπ Homer, Bechert comes to the following conclusion: (35)

"Das Aktiv steht, wenn die Aktion hervorgehoben ist; (... )" (Bechert 1964: 424); "Das Medium steht, wenn das Subjekt und seine besondere Beteiligung an der Handlung hervorgehoben ist: etwa "selbst sehen", (... ), "bei sich sehen", (... ), femer, wenn die Wechselwirkuπg mit dem Objekt betont ist, ( ... ), - und schlieBlich, wenn das Objekt und seine Rϋckwirkung auf das Subjekt hervorgehoben wird: bei ϋbermachtigen, besonders gδttlichen Objekten; lebhaft vergegenwartigend zur Bezeichnung des starken Eindrucks, deπ das Objekt auf das Subjekt macht; bei erwarteten Objekten" (Bechert 1964: 426).

According to Bechert, the middle όράομαι expresses that the object leaves a strong (mental) impression on the subject. 171 For example, 170 In Homer, the active fonn occurs 471 times, the middle times 114 (see Bechert 1964: 42). In Classical Greek the middle fonn occurs only marginally, and in the fixed expression ίδού 'look!' 171 Bechert's characteήzation of the uses of active and middle with ίδε'iν readily fits in with the model set out by Croft (1991), in which the conceptualization of mental events involves two causal relations (see section 2.1.3). In the case of active ίδε'iν the causal relation from perceiving subject to perceived object (die Aktion) is

ΤΗΕ MIDDLE VOICE AS Α COMPLEX CAΊΈGORY

(36)

ro πόποι, ή μέγα θαuμα τόδ'

101

όφθαλμοϊσιν όρmμαι (Ν 99)

Oh! Surely a great rnarvel is this that Ι look upon with rny eyes.

In Homer, the verb always has middle ending when the object is θα\>μα 'wonder', as in the formula θα\> μα ίδέσθαι ('a marvel to look upon'). In conclusion, the middle voice of verbs of perception marks that the subject is mentally affected. Active verbs of perception are unmarked with respect to the feature of affectedness. Middle verbs of perception are often volitional. This tendency may be explained by the fact that their subjects intend to acquire information about or from the perceived object. 2.1.9 Mental Activity Middle

The mental activity middle involves an animate subject that volitionally performs a mental activity, whereby the subject itself is mentally affected. In other words, the subject has two semantic roles: agent and experiencer. In some cases, moreover, the subject benefits from the mental activity so that the subject can also be assigned the role of beneficiary. Typical examples are: 172 'take counsel, plan, resolve' 'calculate, reason, consider' (ΜΤ) μέδομαι/μήδομαι 'plan, contrive' 173 μητίομαι 'devise, contrive' (Horn.) (Μτ) σημαίνομαι 'infer frorn signs, conjecture' σταθμάομαι 'rneasure, estirnate, conjecture' τεκμαίρομαι 'judge frorn signs, conjecture' (Μτ) βουλεύομαι

λογίζομαι

rnore prominent (hervorgehoben), whereas in the case of rniddle ίδέσθαι the reverse relationship frorn object to subject is highlighted: the object sornehow leaves a strong irnpression on the subject. 172 Note that rnental activity verbs tend to have a sigrnatic middle aoήst (e.g. έβουλευσάμην). For a study of the syntax and sernantics of the verbs of thinking in Horner, Ι refer to Bertolίn Cebήan (1996). 173 The active participle μέδων functions as a substative 'ruler' (i.e. sorneone who plans things). The active fonn rnay be explained by the fact that a ruler does not plan for his own sake (at least he is not supposed to), but for the sake of his subjects.

102

CΗΑΡΊΈR

2

middles of meπtal actiνity are media taπtum (Μτ). Α πumber of howeνer, do haνe aπ actiνe form such as βουλεύω aπd σημαίνω. These actiνe forms, howeνer, do ποt haνe a causatiνe meaπiπg with respect to the middle forms. It appears that the middle form expresses aπ iπcreased meπtal iπνolνemeπt of the subject. Thus, actiνe σημαίνω meaπs 'giνe a sigπ', whereas middle σημαίνομαι may be roughly characteήzed as 'giνe oπeself a sigπ, get a sigπ' (similar to aπ iπdirect reflexiνe middle, see sectioπ 2.1.11), heπce 'iπfer from sigπs' The distiπctioπ betweeπ βουλεύω aπd βουλεύομαι is less clear-cut. Το all appearaπces, the actiνe form has the same meaπiπg as the middle. Oπly the stative meaπiπg 'to be a member of the βουλή' is limited to the active form βουλεύω. 174 Ιπ geπeral, meπtal activity verbs desigπate that the subject coπceives aπ idea by reasoπiπg. This idea caπ relate to a thiπg (expressed as a direct object) or to aπ eveπt (expressed by aπ iπfiπitival complemeπt). For example, Maπy

them,

(37)

( ... ) οί 'Έλληνες (Hdt. 8.75.2)

δρησμον βουλεύονται καταρρωδηκότες,

(... )

(... ) the Greek are planning flight in their fear, (... )

(38)

οί δt δη έπτCι τrον Περσέων ώς έβουλεύσαντο έπιχειρέειν τοiσι μάγοισι, ( ... ) (Hdt. 3.76.1) The Seven, having resolved to attack the Magians at once, (... )

αύτίκα

There is πο sharp bouπdary betweeπ the meπtal actιvιty type aπd the meπtal process type (sectioπ 2.1.3). The crucial differeπce is that meπtal activities are volitioπal. There are, howeνer, seνeral verbs that caπ desigπate both meπtal processes aπd activities, for example, έννοέομαι 'uπderstaπd, iπteπd, take thought for, be aπxious' (meπtal process) aπd 'coπsider, reflect upoπ, iπveπt' (meπtal activity); φράζο­ μαι 'thiπk, believe' (meπtal process) aπd 'coπsider, coπtήve' (meπtal activity). This kiπd of polysemy caπ emerge easily siπce, iπ maπy coπtexts, whether the subject is more actively iπvolved iπ the cogπitive process or not is iπelevant. What matters most, then, is that s/he conceived aπ idea, not in what way s/he conceived it. 174 The compounds συμβουλεύω - συμβουλεύομαι do show a clear contrast; the former means 'advise (+ dat.), recommend (+ acc.)', whereas the latter means 'to take counsel with s.o. (+ dat.)' That is, when the subject is the 'receiver' of the advice, the verb takes the middle voice.

ΤΗΕ MIDDLE VOICE AS Α COMPLEX CAΊΈGORY

103

Like perceptioη 175 aηd meηtal processes, the semaηtic structure of meηtal activities iηvolves two simultaηeous causal relatioηs; (a) a cogηizer that focuses his atteηtioη ση, aηd makes meηtal coηtact with, a cogηized eηtity (an 'idea'), and (b) the cogηized eηtity which, by comiηg iηto beiηg, briηgs about a meηtal affectedηess withiη the cogηizer. The middle-markiηg of the meηtal activity verbs is motivated by causal relatioη (b): the meηtal activity has aη effect ση the state of miηd of the cogηizer. Meηtal activities are semantically middle (as opposed to active and passive) ίη that the subject is both aη Iηitiator (beiηg aη agent), and an Eηdpoiηt (beiηg an experiencer). The meηtal activity middle is related to the iηdirect reflexive middle. This is showη by the fact that quite a ηumber of iηdirect reflexive middle verbs acquired a meηtal meaniηg. For example, the origiηal meaniηg of the verbs τεχνάζομαι and μηχανάομαι probably was iηdirect reflexive 'make s.th. for oηeself by art', coηstrued with a coηcrete, physical direct object (e.g. πλο\α 'ships', Hdt. 1.94.6). Theη, through metaphoήcal exteηsioη, they were also used ίη the seηse to 'coηtrive', i.e. a purely meηtal activity with an abstract direct object that oηly exists ίη the meηtal sphere, e.g., (39)

α συν'(lδει (... ) τitν yύναικα ταύτην, ( ... ), τφ πατρi τφ ήμετέρφ θάνατον μηχανωμένην φαρμάκοις, (... ) (Aηtipho 1.9) (... ) who knew that this woman, (... ) had been contήving death against our father by poison, (... )

Other examples of a semaηtic exteηsioη from crete meaηiηg) to meηtal activity (abstract) are:

iηdirect

reflexive

(coη­

Mental activity is similar to perception. Verbs designating visual perception tend to evolve into verbs of menιal activity. Sweetser explains this phenomenon thus: "This rneιaphor is probably based on the strong connection between sight and knowledge, and also ση the shared structural properties of the visual and the intellectual dornains our ability to focus our mental and visual attentions, to rnonitor stimuli rnenιally and visually" (1991: 33). She gives the following exarnples of verbs that either have a physical or a rnental sense: look down on, look up to, look forward to, look back on, overlook, look after. Exarnples frorn Greek are θεάομαι 'gaze at' and σκέπτομαι Ίοοk at' which acquired a rneιaphoήcal sense 'conternplate' and 'consider, exarnine', respectively. See also note 91. 175

104

CΗΑΡΊΈR

2

μετατίθεμαί τι 'change s.th. for oneseιr > μετατίθεμαι Πιν δόξαν 'change one's opinion' >μετατίθεμαι (abs.) 'change one's mind' 176 όρίζομαί τι 'mark s.th. out for oneselr > 'determine for oneself; define' (cf. K-G,l: 111) συντίθεμαί τι 'put s.th. together for oneself > 'agree on, conclude' συμβάλλομαί τι 'bήng s.th. together for onese\f, contήbute' > 'suppose, reckon' (K-G, 1: 111) Απ

reflexive (40)

example that shows activity is:

aπ iπteπnediate

stage

betweeπ iπdirect

aπd meπtal

&λλο δέ τοι έρέω, σi> δ' έν1. φρεσ1. βάλλεο σftσι (Α

297, et

al.). And another thing Ι wil\ tel\ you, and do you lay it to heart.

The miπd is metaphorically coπceptualized as a coπtaiπer ίπ which οπe can store thiπgs to remember. 177 This is aπ inteπnediate case because the subject (σύ) can be equally viewed as a beneficiary and as aπ expeήencer. The subject is a beneficiary siπce he has the items that are stored in the contaiπer of his memory readily at his disposal; the subject is an experiencer in that he is mentally affected as a result of the event. These semantic roles are, however, related. 178

176 The simple νerb τίθημι is more frequently used in the midd\e when it is used 'in reference to a menta\ action', as LSJ notes (under Β Il). The image of putting something in some place (at one's own disposal, cf. the middle voice) is metaphoή­ cal\y employed in a meaning 'assess', e.g. έκ τούτων τa δίκαια τίθενται (Dem. 8.8) ('they assess justice on these grounds'). The frequent construction of this meιaphoήca\ use of τίθημι with loca\ adverbials (e.g. τίθημtl-μαί τι ί:'ν τινι 'regard s.th. as s.th.') shows that the concrete meaning put something somewhere stil\ shines through the metaphoήca\ meaning. 177 Α very common meιaphor (see Lakoff 1987: 450). 178 More examp\es of this type of meιaphoήca\ shift can be found in Kίihner­ Gerth (1: 111-2). Note that Langacker ιreaιs beneficiaήes and expeήencers as manifestations of one archetypal role experiencer (Langacker 1991a: 327-9), see a\so section 1.1.2.

ΤΗΕ MIDDLE VOICE AS Α COMPLEX CA'fEGORY

105

2.1.10 Speech Act Middle

The speech act middle involves a subject that is involved ίη the speech act ίη a special way. Since speech acts are volitional the subject can be considered an agent. In addition, the subject can be interpreted as a beneficίary or an experίencer. Examples of middles designating speech acts are: 179 'speak in the assembly, speak' (Μ'Γ) 'pray' (Μ'Γ) άπολογέομαι 'speak in defense' (Μ'Γ) μαντεύομαι 'devine, prophesy' (Μ'Γ) μυθέομαι 'speak, say' (ΜΤ) προφασίζομαι 'allege as an excuse' (Μ'Γ) φθέγγομαι 'utter a sound or νoice; speak loud and clear' (Μ'Γ) ψεύδομαι 'cheat by lies, speak false' άγοράομαι άράομαι

Verbs of emotional speech 'accuse' (Μ'Γ) άναίνομαι 'refuse, reject' (Μ'Γ) άρνέομαι 'refuse, deny' (ΜΤ) λοιδορέομαι 'revile, rebuke' μέμφομαι 'rebuke' (Μ'Γ) αίτιάομαι

μύρομαι Ίament'

μωμέομαι

'blame'

όδύρομαι Ίament' (Μ'Γ)

όλοφύρομαι Ίament' (Μ'Γ)

Verbs of promising 'vow, pray, boast,' 180 (ΜΤ) στευται (Hom.) 'promise, boast' (Μ'Γ) ύπισχνέομαι 'promise' (Μ'Γ) εϋχομαι

Most of the speech act middles have a sigmatic middle aoήst, e.g. ήγορη­ number of them also have a passive vaήant form with identical meaning, e.g. άπελοyήθην, ήρνήθην, διελέχθην, έλοιδορήθην. Δέομαι only has a passive form, έδεήθην. There is, however, a semantic distinction between the variant formations έμεμψάμην 'Ι blamed, rebuked' νs. έμέμφθην 'Ι became angry with', and έψευσάμην 'Ι lied' vs. έψεύσθην 'Ι made a mistake'. In these oppositional pairs, the passive aorist forms do not denote speech acts, but mental processes. Ι will elaborate on the issues of (aoήst) morphology in ch. 3. 180 Probably a very old medium tantum, cf. Skt. όhate 'speak solemnly' 179

σάμην, fιτιασάμην. Α

106

CHAP'ΓER

2

ύφίσταμαι 'promise' 181

Verbs of commanding 'command' έντέλλομαι 'command' ιcέλομαι 'command' (Μτ) έφίεμαι

Verbs of asking and answering 'answer' ύπο-/άποιcρίνομαι 'answer' 182 δέομαι 'beg' (Μτ) ήρόμην (aor.) 'ask (information)' 183 λίσσομαι 'beg, pray' (Μτ) πυνθάνομαι 'ask (information)' (Μτ) άμείβομαι

Most middle verbs of speech are media tantum. lf they do have aπ active form, there is not always a clearly detectable semantic distincέντέλλομαι, λοιδορέω λοιδορέομαι). 184 If tion (e.g. έντέλλω there is a clear semantic distinction, it is not predictable, as in άμείβω 'exchange' vs. άμείβομαι 'answer' (see below), and έφίημι 'send' vs. έφίεμαι 'command' (original meaning 'send s.o. for oneself'?). In these cases, the meaning of the middle verb has drifted away from its active counterpart. Most middle verbs of speech are rather specific with respect to their lexical meaning. lt appears that verbs with a general, more neutral meaning like 'speak, say' tend to be active: άγορεύω 'speak', αύδάω 'speak, say', εϊρω 'speak, say' (Hom.), καλέω 'call (by name)', λέγω 'speak, say', and φημί 'say'. 185 In Homer, however, we also find middle forms of φημί (φάτο), and the general verb μυθέομαι 'speak, say', which appear to be semantically neutra1. 186 literally means 'stand under s.th. (as support)' which acquired a sense 'take upon oneself, submit to, promise' Similarly, ύπισχνέομαι (epic and Ionic vaήant ύπίσχομαι) literally means 'take upon oneself 182 The active form άποιcρίνω means 'separate, select, choose'. 183 In Classical Greek, this form functions as a suppletive aoήst to the active present έρωτάω. In Homer, there is still a coπesponding present form εϊρομαι 'ask'. 184 For a further discussion of this issue, see chapter 5. 185 For a study of the verbs of saying in Greek, see Foumier (1946). 186 Cf. also rniddle αύδάομαι in classical poetry. According to Foumier, there is no semantic difference beteen active tφη and rniddle φάτο. In Homer, they are used as metήcal altematives (Foumier 1946: 34). Boeder (1961-2: 20-5, 74-6), however, claims that the rniddles φάτο and μυθέομαι are not semantically neutral. Instead, according to Boeder, they involve a subject that has a special interest in the speech 181

'Υφ-ίσταμαι

metaphoήcal

ΤΗΕ MIDDLE VOICE AS Α COMPLEX CA'fEGORY

107

These middles are possibly the result of semantic geπeralizatioπ (or 'bleachiπg'), i.e. their origiπal, more specific meaniπg uπderweπt a process of geπeralizatioπ. 187 The middle iπflectioπ of verbs of speech can be explaiπed semaπtically ίπ the followiπg ways. Firstly, ίπ maπy cases the subject iπteπds to gaiπ bene.fit from the speech act, as ίπ άράομαι, εϋχομαι, λίσσομαι. These middle verbs are similar to iπdirect reflexive middles ίπ that the subject can be assigπed the semaπtic role bene.ficiary. This probably also applies to verbs of commanding: the subject beπefits ίπ that aπ act is performed accordiπg to his will. Also speech acts like άπολογέομαι, προφασίζομαι, ψεύδομαι (and αίτιάομαι?) will geπerally be profitable to the speaker. Secoπdly, maπy middle verbs of speech imply a stroπg emotional or at least meπtal involvemeπt οπ the part of the speaker: αίτιάομαι, άναίνομαι, άρνέομαι, λίσσομαι, λοιδορέομαι, μέμφομαι,

μύρομαι, μωμέομαι, όδύρομαι, όλοφύρομαι (also ψεύδομαι?). 188 The subject of verbs of emotive speech acts can be assigπed the semaπtic role of experiencer. As for ύπισχνέομαι aπd ύφίσταμαι (aπd εϋχομαι ίπ the seπse 'vow') the coπsequeπce of the speech act is that the speaker is bouπd by his promise, and as such s/he is meπtally affected by the speech act (similarly, Kemmer 1993: 141). More difficult to accouπt for, however, is the middle iπflectioπ of verbs of askiπg and answeriπg. The middles of asking πυνθάνομαι, δέομαι, and aoήst ήρόμην caπ be explaiπed ίπ various ways. The former two imply that the asker will get, as a result of the speech act, the iπformatioπ s/he asked for. Ιπ this way, the speaker-subject can be coπsidered aπ expeήeπcer. The semaπtic developmeπt of πυνθάνομαι makes this poiπt clear. Ιπ Homer the verb has the meaπiπg 'hear that,

act in that he wishes to convey his point of view to the listener. The evidence in support of his claim, however, is inconclusive. 187 As for φάtο (cf. Lat. farϊ 'say'), this process was as follows: the root *bheh 2 originally had a more concrete meaning 'shine, be bήght', e.g. Skt. bhd-ti 'shine' (which coπesponds with Greek φησί). Thus, the oήginal meaning of the middle may have been 'make oneself clear', hence 'explain oneself, speak' (see Chantraine 1968). 188 The fact that many emotive speech acts are actives(-only) attests to the semantic unmarkedness of the active endings (see chapter l), e.g. αίάζω 'to cry αία'i, wail', γοάω 'wail ', γρύζω 'to say γ ρ\>, grumble, mutter', θρηνέω 'wail, mourn', κλαίω 'weep, wail' (also κλαίομαι in poetry, perhaps modelled after future κλαύσομαι), οίμώζω 'to cry οϊμοι, wail' See for the issue of active and rniddle (quasi-)synonyms chapter 5.

108

CΗΑΡΊΈR

2

learn that' (a mental process middle). The subject undergoes a cognitive change of state. This is still the central meaning in the Classical language (especially in the aorist stem). In Classical Greek, however, it acquired aπ additional meaning 'ask' (especially in the present stem), which can be paraphrased by try to acquire information189 In a similar way, the middle inflection of the aorist ήρό­ μην might be explained by the fact that the subject acquires information as a result of the question. 190 Δέομαι (+ gen. + gen.) 'beg s.th. of s.o.' is a semantic extension of the original meaning Ίack, need, want', which pertains purely to a passive emotional or physical affectedness. This mental and physical affectedness is still present in the meaning 'beg for' . 191 This can be nicely illustrated by contrasting δέομαι with active αί­ τέω. 192 Compare the following instances where both verbs are construed with aπ infinitive: (41)

[Adrastus] παρελθrον δε οi>τος ές τa Κροίσου οικια κατa νόμους τοi>ς έπιχωρίους καθαρσίου έδέετο έπικυρfισαι (Hdt.

1.35.1). This rnan came forward into the house of Croesus and begged to win puήfication of Croesus after the customs of that country.

(42)

ό δε Κώης, οtά τε ού τύραννος δημότης δε έών, αίτέει Μυτιλήνης τυραννεuσαι (Hdt. 5.11.2). But Coes, as he was not a ruler but an ordinary citizen, asked to become ruler of Mytilene.

Note that the new rneaning 'ask' probably developed frorn the conative use of the present stem. 190 Απ intήguing question is why the present forrn έρωτάω is active, but the aorist forrn ήρόμην is rnidd\e (although the expected forrn ήρώτησα a\so occurs). In rny view, the answer rnust be sought in the semantic unrnarkedness of the active (see chapter 1). In chapter 5, the issue of active and rniddle (quasi-)synonyrnous verbs will be dea\t with. 191 This sernantic change can be nicely explained as a generalized invited inference. For instance, an utterance such as I need help pragrnatically irnplies that one asks for help. That δέομαι involves a high degree of subject-affectedness is a\so shown by the fact that it has a passive aoήst forrn έδεήθην. In ch. 3, Ι shall argue that the passive aoήst forrn expresses a high degree of subject-affectedness. 192 Marginally, a\so rniddle αίτέομαι occurs, though apparently without distinct rneaning. The rnidd\e forrn rnay however ernphasize that the subject profiιs or hopes to profit as a resu\t of the request (see Rijksbaron 1991: 59 [ση Ε. Βα. 360]). 189

ΤΗΕ MIDDLE VOICE AS Α COMPLEX CAτEGORY

109

In (41), the subject (Adrastus) is in deep trouble, and the purification he begs for is his only chance of salvation. In (42), on the other hand, Daήus allows Coes to propose a reward for his good advice. Coes is clearly not in urgent need of the city of Mytilene when he asks for the tyranny over it. The semantic contrast between δέομαι and αίτέω is also reflected ίη the different case-markings of the object that is asked for. The genitive of δέομαι 193 can be interpreted as an ablatival genitive (i.e. expressing the idea of being separated from s.th.), or, altematively, as a genitive of source (viz. of the emotion of desire). Conversely, αίτέω can be construed with a double accusative - of the person asked and of the thing being asked for. These accusative complements with αίτέω neutrally denote the direct objects. Both the person and the thing can appear as subject in a passive construction. The last middle verb of asking to be discussed here is λίσσο­ μαι (+ acc.). The middle voice can be explained by the emotional undertone of the act of begging, and by the benefit that the subject hopes to receive as a result. Thus its subject can be interpreted both as an experiencer and as a beneficiary. Verbs of answerίng, άμείβομαι and ύπο-/άποκρίνομαι, are more problematic with regard to their middle voice. The explanation may be found in the history of these verbs. In Homer, the middle άμείβομαι means 'altemate with one another' (see section 2.1.6). This reciprocal meaning shifted into the meaning 'answer' Απ interrnediate stage may haνe been 'altemate with words', as in (43)

&ς τω μεν χαλεποiσιν άμειβομένω έπέεσσι eστασαν (γ 148) So these stood alternately retorting with harsh words.

As for άποκρίνομαι, this verb appears to be an Attic replacement (first occuπence in Thucydides) of the older (Homeήc) verb ύποκρί­ νομαι. ln origin, simple κρίνομαι had an indirect reflexive meaning 'select (for oneself), distinguish (for oneself)' (cf. active κρίνω 'select, distinguish'). This meaning is still well-attested (e.g. δ 408 έταίρους 'select companions'). Subsequently, this meaning extended to the

193 Both the person from whom something is asked, and the object that is asked for are marked by the genitive. Both are illustrated by σέο δέομαι μη δέεσθαι ιΧνόμων (Hdt. 1.8.4) ('Ι beg you not to request unlawful things from me').

110

CΗΑΡΊΈR

2

metaphoήca1 meaning 'exp1ain, interpret'. 194 Both the simple verb

and the compound (44) (45)

ύποκρίνομαι

can have this meaning. For example,

άλλ' ίiγε μοι τον ονειρον ύπόκριναι και ίiκουσον (τ But come now, hear this dream of mine, and interpret it for me.

( ... )

ό γέρων έκρίνατ' όνείρους (Ε

535).

150)

( ... ) that old man interpreted dreams.

Finally, the meaning 'explain, interpret' deνeloped into 'answer', probably because, in general, giνing an explanation occurs as aπ answer to a question. The last issue Ι would like to discuss in this section concems the pair ψεύδω ψεύδομαι. The formal contrast between active ψεύδω 'deceive' and middle ψεύδομαι 'cheat by lies' (aor. έψευσά­ μην) can be explained adequately ίη terms of absence vs. presence of subject-affectedness. 195 The active appears to denote 'cheating, deceiving' in a general sense, i.e. not necessarily by lying. This is shown by the occuπence of inanimate entities as subject. For instance, (46)

a. άλλα πιστεύω (... ) μη ψεύσειν με ταύτας τας άγαθας έλπίδας (Xen. Cyr. 1.5.13) But Ι feel confident (... ) that these sanguine hopes will not deceive me.

b. Ναi τrο σιώ,

c1

Άργε'iοι, ψευσε'i ύμ[ τα σίγμα ταυτα (Xen.

HG 4.4.10) By the twin gods, Argives, these Sigmas will deceive you. Ιη example (a.) ψεύσειν refers

the subject is έλπίδας, which makes it unlikely that to a speech act. Ιη example (b.) the subject is τα σίγμα, which refers to the sigmas written οη the shields of the Sicyonians. Pasimachos took these shields from the Sicyonians to deceive the Argives. Again, written letters on shields cannot speak (in the literal sense). The construction allows a second complement in the

In this metaphor, explaining things is viewed as a process of taking something apart in order to understand it better. Α similar metaphor is found in English expressions such as analyse (from Greek άναλύω 'make loose, take apart'), explicate (from Latin expliciire 'unfold'), unravel, and untangle. 195 There is also an intransitive mental process middle ψεύδομαι 'be deceived' (with a passive aoήst έψεύσθην), of which the active ψεύδω 'deceive' is the causative form. 194

ΤΗΕ MmDLE VOICE AS Α COMPLEX CAΊΈGORY

111

genitive, expressing the thing with respect to which one is deceived. This (separative) genitive is probably related to the genitival complement with verbs of depriving. That this active construction has a causative meaning is shown by the occuπence of coπesponding intransitive mental processes like: (47)

ώς δε ψεuσθηναι τfις έλπίδος, ( ... ) When he was deceίved ίη hίs hope, (... )

(Hdt. 1.141.1).

The middle verb ψεύδομαι (aor. έψευσάμην) absolutely, or with aπ object, for example,

caπ

be construed

lie' (absolute) say that which is untrue' (cognate object) ψεύδομαι αύτόν Ί deceive him by lying' (human object) ψεύδομαι ορκια Ί belie the treaty' (inanimate object) ψεύδομαι Ί

ψεύδομαι το\>το Ί

The middle voice of ψεύδομαι can be explained in two ways that do not exclude each other. First, the middle voice may express that the subject generally deήves benefit from the act of lying (i.e. the subject is beneficίary). Second, the middle may also express that the subject is mentally involved to a greater extent (i.e. the subject is aπ experiencer), since the liar consciously utters words which he knows to be untrue. In sum, active ψεύδω 'deceive' is the causative counterpart of ψεύδομαι 'be deceived' (aor. έψεύσθην). Ψεύδω does not imply mental involvement of the subject. Middle ψεύδομαι (aor. έψευσά­ μην) 'lie, cheat by lies', ση the other haπd, is a speech act, aπd it does imply that the subject is mentally affected. Το conclude, in this section Ι have offered a number of explanations for the occuπence of the middle voice with speech act verbs. Ι have argued that in some cases the middle voice marks that the subject benefits from the speech act (subject is beneficiary), aπd that, in other cases, the subject is mentally (emotionally or cognitively) affected. These different explanations do need necessarily exclude one another. There are, however, also middle verbs whose middle endings do not appear to have a special semantic contribution (e.g. φάτο). These cases may be explained by semaπtic generalization (or 'bleaching'). In the course of history, the middle endings lost their

112

CΗΑΡΊΈR

2

specific middle meaning. 196 2.1.11 lndirect Reflexive Middle

The indirect reflexive rniddle involves transitive events performed by a volitional subject (aπ agent). The subject is affected ίο that s/he derives benefit from the action performed, i.e. the subject has the semantic role of beneficiary. It is, however, often quite difficult to distinguish between the beneficiary-role and the recipient-role because the indirect reflexive middle designates typically that the subject, like a recipient, has the effected or affected object at his/her disposal as a result of the event. Recipient-subjects, however, are only found with media tantum. 197 Α typical example is (see also Cock 1981, Rijksbaron 20023 : 147): (48)

a. (... ),

δασάμενοι σ1τον έν τflσι πόλισι οί άστο\ liλευρά τε

καi liλφιτα έποίευν πάντες έπi μfινας συχνούς

(Hdt. 7.119.2)

(... ) the citizens divided up corn in the cities and all of them for many months made wheaten and barley.

b. [The Egyptians] (... ) (Hdt. 2.36.2)

άπο όλυρέων ποιεUνται σιτία,

(... )

196 Α class of rniddle verbs that is related to the speech act rniddles are the rnedia tanturn that denote sounds rnade by anirnals, e.g. βληχάομαι 'bleat', βρυχάομαι 'roar, bellow', μηιcάC'μαι 'bleat', μυκάομαι 'bellow', and φθέγγομαι 'whinney, cry (of birds), speak (hurnans)' Most of these verbs are onornatopoeic. The polysernous verb φθέγγομαι that can apply to anirnal sounds as well as to hurnan speech shows that these classes are related. As with hurnan speech, the rniddle voice of anirnal sounds rnay be explained by the ernotional affectedness of the subject: anirnals typically bleatlbellowlbark etc. when they are agitated/afraid/angry/hungry/need to be rnilked, etc. There are, however, also active verbs of anirnal sounds: γρύζω 'grunt', ύλακτέω 'bark', and χρεμετίζω 'whinney, neigh'. Μηκάομαι and μυκάομαι have active aoήsιs, ί::μακον and ί::μυκον. As was argued in ch. 1, the active voice is not constrained to expressing absence of subject-affectedness. Rather, it is neutral as to the sernantic feature of affectedness. 197 This rneans that verbs such as δίδωμι 'give' and λέγω 'say' do not have indirect reflexive rniddles δίδομαί τι Ί give rnyself s.th.' or λέγομαί τι Ί say s.th. to rnyself Because it is unusual to give or to say sornething to oneself, the indirect reflexive is expressed by the pronoun έαυτφ (see also below). For aπ extensive analysis of the indirect reflexive use of ποιέομαι, Ι refer to Cock (1981).

ΤΗΕ MIDDLE VOICE AS Α COMPLEX CAΊΈGORY

113

(... ) they make their bread from spelt, (... ).

In example (a.), the corn is ground for the Persian king, Xerxes, whereas in (b.) the Egyptians themselves profit from the food they prepare. In addition to indirect reflexive middles that contrast with active verbs 198 , there are also a number of media tantum: 199 'heal, cure; mend, repair' 200 (Hom.) 'take, seize' 201 Cipνυμαι (Hom.) 'acquire' άκέομαι αϊνυμαι

There are also a number of indirect reflexive middles whose meaning has away from that of their active counterparts. E.g., αίpέω 'take' νs. αίpέομαι 'choose' (esp. in the aoήst stem: είλόμην Ί chose'). In Homer, the middle αίpέομαι still exhibits the regular meaning 'take for oneself; άποδίδωμι 'give back' νs. άποδίδομαι 'sell' (< *give back with the aim of profit'). Other examples are cited by Rijksbaron (1994 2: 147). 199 There is also a group of νerbs that designate emotionally motivated actions that seem to be related to the indirect reflexive middle. Often, the subject can also be said to benefit from the activity in that the subject tries to exercise power over the object through the activity. Examples are: αίκίζομαι 'maltreat', βιά(ζ)ομαι 'press hard, wrong, ravish', δηλέομαι 'hurt, damage', δωpέομαι (+ acc. + dat.) 'present s.o. with s.th.', λυμαίνομαι Όutrage, maltreat', φείδομαι (+ gen.) 'spare', φιλο­ φpονέομαι 'treat kindly, show favour to', χαρίζομαι(+ dat.) 'do s.th. favourable to s.o' The subject is either ill-disposed or well-disposed towards the object. Some of the verbs expressing violent actions (βιά(ζ)ομαι, δηλέομαι) semantically shade into middle-only verbs of plundering, like σίνομαι and λη·{ζομαι, that can be classified as indirect reflexives. Like the indirect reflexive middle, these verbs all have sigmatic middle aoήsts (e.g. έβιασάμην, έχαpισάμην). 200 Assuming that the oήginal meaning of άκέομαι was 'mend, repair', the middle inflection can be explained by the fact that you generally repair something for your own benefit. The middle inflection of the near-synonymous ίάομαι may be analogous to άκέομαι. If we do not accept the explanation of ίάομαι as analogous to άκέομαι, the middle inflection becomes somewhat problematic since it is not clear how the meaning 'heal' relates to middle semantics. Healing benefits the healed person, rather than the healer. Quite possibly, the meaning 'heal' developed from an indirect middle meaning such as 'save someone/something (so as to have himlher at one's disposal)', cf. the Homeήc medium tantum pύομαι 'rescue, protect'. 201 The active *αϊνυμι must have existed in Mycenaean Greek where we find a passive perfect participle a-ja-me-no (άyαιμένος) 'be inlaid'. The active *αϊνυμι oήginally may have meant 'provide with' The middle αϊνυμαι, when it was still in opposition with the active form, must have meant 'provide oneself with' 198

dήfted

114

CΗΑΡΊΈR

2

δέχομαι 'receive, accept' 202 δίζημαι (Hom.) 'seek for' έργάζομαι 'work at, make,

perform' 'treat, heal, cure' κτάομαι 'acquire' 203 λάζομαι (Hom.) 'take, seize' λη·{ζομαι 'seize, caπy off booty, plunder' pύομαι 'protect, rescue' σίνομαι 'harm, plunder, waste' σιτέομαι 'feed on, eat' χειρόομαι 'master, subdue' ώνέομαι 'buy' (aor. έπριάμην Ί bought') ίάομαι

The use of the indirect reflexive middle is an unemphatic way of expressing that the subject is the beneficiary. If one wishes to emphasize that the subject benefits from the action, the reflexive pronoun is used, and mostly with an active verb form (comparable to the direct reflexive construction with reflexive pronoun, see 2.1.7). This can be seen as a form of neutralization (see also section 1.1.4). Απ example is (49)

[Philip weak:ened Macedonia] (... ), tτ' έπισφαλεστέραν η ύπfJρχε φύσει κατεmc:εύακεν αύτφ (Dem. 2.15) [Philippus] has rendered [Macedonia] for his own benefit even more insecure than it was by nature.

However, the (redundant) use of a middle form is another option, e.g., (50)

τί τ-hv πόλιν, Αίσχίνη, προσfJκε ποιεiν άρχην κα1. τυραννίδα τ&ν Έλλt'ινων όρ&σαν έαυτφ καταmc:ευαζόμενον Φίλιππον;

(Dem. 18.66) What should the city have done, Aischines, when she perceived that Philip was establishing for himself a tyranny over the Greeks?

202 Thematic δέχομαι is the successor to the older athematic verb *δέγμαι, of which forms such as έδέγμην, δέγμενος survive in the Homeήc language. This athematic medium tantum with e-vocalism in the stem possibly belonged to the ΡΙΕ class of stative middle verbs, cf. also εtμαι 'be clad', κεiμαι 'lie'. Likewise, *δέγμαι had a stative meaning 'wait for', e.g. ό.λλ' αίεί τινα φίi>τα μέγαν κα1. καλΟν έδέγμην/ ένθάδ' έλεύσεσθαι (ι 513) ('But Ι always waited for a tall and handsome man to come here'). See also note 118. 203 The media tantum that have a meaning 'acquire' can be compared to the indirect reflexive middle πορίζομαι 'furnish oneself with' (cf. active πορίζω 'fumish').

ΤΗΕ MIDDLE VOICE AS Α COMPLEX CA'fEGORY

115

Apart from the indirect reflexive middle with a reflexive pronoun in the dative, we also find indirect reflexive rniddles in combination with an external (non-coreferential) indirect object. Consider (51)

[Amasis to the Greeks] ( ... ) eδωκε χώρους βωμΟ'uς κα1. τεμένεα θεοiσι (Hdt. 2.178.1) (... ), he gave lands where they might set up altars and

ένιδρύσασθαι

sanctuaήes

to their gods.

The pήmary beneficiaήes are the Greeks the (unexpressed) subject of ένιδρύσασθαι - as they can directly and physically dispose of the altars and the cult-places. The gods are the beneficiaήes of the cultplaces in a more indirect way. Another example of this type of expression is: (52)

[Megabazos leads his army through Thrace] (... ),

πασαν πόλιν

κα1. παν eθνος των ταύηι οίκημένων ήμερούμενος βασιλέϊ

(Hdt. 5.2.2) ( ... ) subduing to the king every city and every nation that lived there.

Again, both the subject and the dative-constituent are beneficiaήes of the event. The subject of the middle verb ήμερούμενος (Megabazos) exercises an immediate, physical power over the cities he conquers, while the king is only indirectly, more distantly involved. It is worthy of mention that this greater distance on the semantic level is iconically miπored by the morphological distance. The dίstal beneficiary is expressed by means of a separate dative-constituent, whereas the proxίmal beneficiary is coded by means of a bounded inflection. 204 So far we have discussed the more typical indirect reflexive construction. Now Ι wish to mention two additional construction-types that Ι regard to be subtypes of the indirect reflexive mίddle. Fίrst, the so-called causatίve mίddle and, secondly, what Ι would lίke to call the perspective-changίng mίddle. typίcal

The term 'causative middle' 205 ίs a very unfortunate one. Α example ίs:

204 Compare also υίέι δε Σπάρτηθεν Άλέκτορος ηγετο κ:οΌρην (δ 10) ('but for his son he [= Menelaus] was bήnging to his home from Sparta the daughter of Alector'). Apparently, Menelaus, as the patriarch, is seen as the pήmary beneficiary of the marήage; his son, the actual groom, is only the secondary beneficiary. 205 See e.g. Goodwin (1895: 267), Stahl (1907: 54-5), Smyth-Messing (1956: 392).

116

(53)

CHAP'ΓER

2

η ούκ άκήκοας, οτι Θεμιστοκλf]ς Κλεόφαντον τον υίον ίππέα μεν έδιδάξατο άγαθόν; (Pl. Men. 93d) Have you πever heard how Themistocles had his sοπ Cleophantus taught to be a good horseman?

The subject, Themistocles, is not the actual perfonner of the eνent. He is, rather, the inίtiator of the eνent since he employs a professional teacher to instruct his son. Howeνer, as K-G (1: 108) point out (see also Gildersleeve 1900-11: 67), the active verb can be used ίη a similar way. For example, (54)

[Pericles had his sons taught]

τούτους

(... )

ίππέας

έδίδαξεν ούδενος χείρους 'Αθηναίων, (... ) (Pl. Men. He had them taught to be the foremost horsemeπ of Atheπs, (... )

μεν

94b)

The fact that both the middle and the active construction convey the same causative meaning leads to the conclusion that the middle inflection itself does not express causativity. 206 In fact, any transitive verb can be interpreted as a causative, given the appropriate context. For instance, 207 (55)

Κ\>ρον

δ'

αυτον

[= τον παράδεισον]

έξέκοψε

κα1.

τα

βασίλεια κατέκαυσεν (Xen. An. 1.4.10) But Cyrus cut dowπ the park and bumed up the palace.

In this type of expression, the intennediate agent that physically perfonns the action denoted by the verb, is pragmatically iπelevant (non-topical). lts presence can only be infeπed on the basis of the hearer's world-knowledge, namely that the subject-referent himself would not perfonn the activity. Instead, an anonymous intennediary agent, usually a professional, physically perfonns the actίvity. 208 See also K-G, 1: 108: "Im Medium aπ sich 1iegt der Kausativbegήff ebeπso­ wie im Aktiv." Compare, coπverse1y, Goodwiπ's remark (1895: 267): "The midd1e sometimes has a causative meaπiπg". 207 For more examp1es, see K-G, 1: 99-100. 208 Thus, the iπtermediary ageπt caπ be compared to a mere iπstrumeπt. Iπstru­ meπts, too, ofteπ remaiπ uπexpressed, due to their pragmatic iπe1evaπce. Coπsider aπ expressioπ such as The police shot the burglar. The iπstrumeπt (some kiπd of firearm) is ποt expressed here because it is iπfeπab1e οπ the basis of our wor1dkπow1edge, aπd pragmatically uπimportaπt. 206

weπig

ΤΗΕ MIDDLE VOICE AS Α COMPLEX CA'fEGORY

117

The question now remains what the function of the middle inflection ίη this type of causative clause is. Notice that, ίη (53), it is Themistocles' own son who is taught to be an excellent horseman. The subject, Themistocles, is therefore indirectly the beneficiary of the event. For this reason, the middle can be viewed as a subtype of the indirect reflexive middle use. This brings me to the second subtype of the indirect reflexive middle, which Ι would like to call the perspective-changing middle. With this middle type, the selection of the subject effects a change of perspective on the event descήbed. For example, (56)

a.

(... ) τον

Θρασύλοχον

τριηραρχίαν επεισεν

(... ) he

b.

iπduced

(... )

τ 'coπsult aπ oracle'

118

CΗΑΡΊΈR

2

subject is a recipient. This type of middle use can, therefore, be considered a subtype of the indirect reflexive middle. 211 2.2 Semantic Feature Analysis

In the previous sections, a classification of the middle usage type was set forth. In order to summarize the description of the middle uses, it might be useful to present aπ analysis of the essential semantic features that distinguish the middle uses from one another. These semantic features all relate to the subject of the clause. 212 Table 3: The mίddle uses: semantic feature analysίs213 Aoήst

2.1.1 Passive 2.1.2 Spontaneous Process 2.1.3 Mental Process 2.1.4 Body Motion 2.1.5 Collective Motion 2.1. 6 Reci procal 2. Ι. 7 Direct Reflexive 2.1.8 Perception 2. 1.9 Mental Activity 2.1.1 Ο Speech Act 2.1.11 Indirect Reflexive

211

middle

form

έποιήθην έτάιcην

έφοβήθην άπηλλάχθην

ήγέρθην έμαχεσάμην έλουσάμην έγευσάμην

έλογισάμην έλοιδορησάμην έποιησάμην

Subj.

Subj. Subj. Subj. Subj. Subject = = = undergoes Anim. Ag. Ben. Exp. Pat. Change of State +1- + +1- + + +1+ + + + + + +1- + + + - + + + - + + + + + + + + + +1- + + + + + +

=

=

Note that the middle verbs involving a change of perspective have a sigmatic aoήst form, just like the more typical indirect reflexive middles, e.g. έδανει­

σάμην, έμισθωσάμην. 212 It must be noted that these semantic features relate to the typical examples of each middle use. There will always be exceptions and boundary-cases. 213 The numbers refer to the sections in which the uses were discussed. Abbreviations: Subj. Subject, Anim. Animate, Ag. Agent, Ben. Beneficiary, Exp. Expeήencer, Pat. = Patient.

=

=

=

=

=

χαρασκευάζομαι

λοi!μαι

+Animate +Agent + Patient

ReΩexive

Dluc:t

ποιέομαl τι κτάομαl τι

+Animate +Agent + Beneficiary

:ιndlrect eΩexive

I

μάχομαι

άyωνiζομαι

Reciprocal + Animate +Agent + Patient

μυθέομαι

λοιδορέομαι

\0

!--

I

tι:l

> >

ti1

~

~
μια-), of which the original meaning may haνe been *'become stained' (i.e. a spontaneous process). The form that is found e1sewhere is έμιάνθην. 235 The one form with stem in -σ- and suffix -η- is τερσ-η-. lt is probably based on an archaic form *τρασ-η- (the ε-νocalism is analogical). τhis would semantica11y fit in well since its oήginal meaning 'dry up' is a spontaneous physical process - the prototypical meaning of the formation in -η-. The aoήst forms with stems in -σ­ include the verbs with roots in dentals (e.g. έφράσθη), and the presents in -άζω, -ίζω. Α number of forms, such as δαμά-σ-θην, has an analogically inserted -σ-. 236 Clearly a secondary form, analogically after πήγνυμι, tπηξα and πηκτός. Incidentally, πήγνυμι and πηκτός, in their tum, are probably influenced by their antonyms pήγνυμι and pηκτός (both regularly from zero degree *wrh2g-). Regular forms would be *πάγνυμι and πακτός (preserved in πακτόω).

ΤΗΕ MIDDLE AND PASSIVE VOICES ΙΝ 'fHE AORIST SτΕΜ

131

have, according to the table, a slight predilection for the aoήsts in -θη-. Many forrns have both features, e.g. έρχθείς, θέλχθη. lt must be noted, however, that the table regards token frequency. Qua types, this predilection is much stronger. There are only a few passive aoήst forrns with heavy stem syllables or ε-vocalism that have the suffix -η-. These are θερέω (subjunctive), τερσήμεναι, τάρπημεν, and πλήyη. Apart from πλήyη (to be discussed below), these exceptions can be explained easily by assuming an analogy with their present stems θέρομαι, τέρσομαι and τέρπομαι. The oήginal forrns with zero-grade must have been *θαρ-η- 237 , *τρασ-η- and τραπ-η- (cf. τραπείομεν). The forrner pair took the e-grade, the latter forrn took the position of the -ρ- from their presents. In sum, we can safely add two rules to those mentioned above. In Homer, the aoήst in -θη- occurs with (iv)

heavy monosyllabic stems, e.g. πλάΎΧθη, χρίμφθη, πilχθεν; stems with ε-vocalism, e.g. pεχθέν, περιπλέχθη, στρεφθείς.

(ν)

As to the stem with a phonological structure (C)CCiC-, (C)CtC-, (C)C1:JC-, a tendency can be observed to occur with the suffix -η- (24 7). This predominance of -η- is partly due to the following rule. In Homer, the forrnation in -η- occurs with (vi)

stems of the structure (C)CCiλ-, (C)CCiμ-, (C)CCiρ-. 238

Although in Homer only the forms έάλη, έδάμη and έχάρη 239 occur, there is no evidence against this rule. In Classical Greek, this rule is also operative (see below). 240 *g"hrr·e- (Lindeman's Law), cf. χαρη- (see Ruijgh 1996: 366). One might hesitate to include stems -αν- in ιhis rule. The only example in Homer, φάνη, can also be explained by semantic factors (spontaneous process; see below). Α furιher reason not to include stems in -αν- here might be ιhat, in ιhe Classica\ Ianguage, stems in -αν- do not always take ιhe passive form in -η-, cf. έφάνθην, έρράνθην (see next section). 239 Έχάρη has two a\ternative forms wiιh, apparently, ιhe same meaning, κεχάροντο and χήρατο. The former may well be old; ιhe latter is eiιher analogically formed to present χαίρω (cf. έκάθηρα to καθαίρω), or more speculatively a Mycaenean relic (cf. Arc. φθήρω vs. Ion. φθείρω). The regular lonic form should have been *χέρσατο or *χείρατο. 240 Απ apparent exception is iiρθεν (from aραρίσκω). However, ιhere may have been a ru\e ιhat stems wiιh ιhe structure #VC- receive ιhe suffix -θη-. In Classical Greek, ιhere are more examp\es that coπoborate ιhis rule (see below). 'Αρθείς from 237

23 Η

*θαρη- from

132

CΗΑΡΊΈR

3

Above, a number of morphological rules (i-vi) were formulated that determine the distήbution of the two passive aoήst formations. However, there is a group of forms that cannot be explained by means of these morphological rules. These are the forms that possess a stem with the structure (C)CaC-, (C)CtC-, and (C)C'i>C-. The question remains to be answered as to what factors determine the aoήst of these remaining stem forms. Ι would suggest that these forms can be explained by a semantic factor. This semantic factor can be summaή­ zed by two rules: Forms that denote a spontaneous process have -η-241 ; Forms that have a passive meaning or denote body motion have -θη-.

(1)

(2)

Consider the following table:

probably has a long vowel (< *άfεp- ). The spontaneous process use can be considered the semantic prototype of the aoήst in -η-. Note that this was probably also the oήginal meaning of the formation before it spread to the passive meaning and the body motion meaning (in Homer only έάλη). There is some controversy as to the meaning of the suffix -eh 1 -l-hι- in ΡΙΕ. Was it statiνe or .fientiνe? The alignment of the formation with the aoήst stem in Greek argues in favour of the latter option, as recently advocated by Harδarson (1998). According to Harδarson, the formations with stative meaning found in the daughter-languages (e.g. Latin rub-e-re 'be red') go back to a yod-present in *-(e)h1yέlό- that is deήved from the aoήst in -eh1-. Ruijgh explains the suffix -η- from a complex suffix *-η-σ- (cf. the Hittite fientive formation in -ess-), of which the -σ­ can be identified as the aoήst suffix. The combination *-η-σ- designated the transition to a state. The -σ- disappeared after the reinterpretation of 2nd sg. forms Iike πάγης (< *παγ-ησ-σ) as παγ-η-ς. This led, in tum, to the replacement of 3n1 sg. παγης (*παγ-ησ-τ) by πάγη (see Ruijgh 1996: 187-9, 366-7). Α comparab\e development is found in 3n1 sg. σβfj, which, according to Risch (1937: 209), must go back to *sgwes-t. αϊ ρω

241

ΤΗΕ MIDDLE AND PASSIVE VOICES ΙΝ ΤΗΕ AORIST S'ΓΕΜ

Spontaneous Process έάγη 'broke (intr.)' έπάγη 'got stuck' -ράγη 242 'broke (intr.)' έκάη 'bumt (intr.)' pύη 'flowed' σαπήηι 'rotted' τράφη 'grew up' φάνη 'appeared'

133

Passive φραχθείς (φράσσω) 'was fenced' έτύχθη 'was done' δρύφθη 'was scratched' κρίνθη 'was chosen' κρύφθη 'was hidden'

Body Motion κλίνθη 'bent aside' τράφθη (τρέπω) 'tumed (intr.)'

These fonns are not detennined by one of the morphological rules (i-vi). lt can be observed that fonns that denote spontaneous processes have the suffix -η-, whereas fonns that are passive or denote body motion take the suffix -θη-. The working of these semantic rules can be illustrated by the stems in -αγ- and -απ-. The spontaneous process verbs have -αγ-η- (έάγη, έπάγη, -ράγη), whereas the one fonn with passive meaning has the suffix -θη-: φραχθείς (&χθη- from &γω not attested in Homer). Likewise, the fonn σαπ-η- designates a spontaneous process, while τραφ-θη- (τρέπω) designates body motion (see for this fonn also section 3.2.1). These semantic rules are also reflected in the frequency figures in Table 1. In the table, we can see that passives (19 9) and body motions (20 Ο) more often take the suffix -θη-, whereas spontaneous processes prefer -η- (14 10). The mental process meaning and the collective motion meaning have not yet been considered. Mental processes can have both the suffixes -θη- and -η-. Most fall under the morphological rules (i-vi), e.g χολώθη [stem in vowel], μνησθηναι [stem in -σ-]), and έχάρην [stem in -αρ-]. The two fonns that cannot be explained by these morphological rules, έδάην and τάρπημεν, may predate the emergence of the aorist in -θη-. 243 As to the collective motion verbs, it appears

242

The forrn -ρ/Χγ-η- (pήγνυμι) must haνe been adapted by analogy with (f)ΙΧγ­ The regular deνelopment would haνe been *(f)ρηγ-η- from *wrh 1g-

η- ((f)άγνυμι). ehι-.

Ι η spite of their stem forrns (ending in a νowel and a heaνy syllable, respectthey haνe the suffix -η-. Aboνe, Ι gaνe a histoήcal explanation: their oήginal stems were *δασ-η- and *τραπ-η-. Altematiνe middle forms to these passiνe forrnations exist, apparently with the same meaning (at least, a semantic nuance is not distinguishable): δεδαέσθαι (with actiνe causatiνe δέδαε); ταρπώμεθα, τε­ ταρπώμεσθα, τερψάμενος. The reduplicated formations are probably old. The sigmatic actives with causative meaning are relatively recent, and so are the sigmatic 243

iνely),

134

CΗΑΡΊΈR

3

though there are very few forms attested in Homer that they can have both the form in -η- and in -θη-. The only examples are έμίγην and μίχθη (or μείχ-?) Ί mingled', and τμάγη 'divide oneself, part' .244 It is but a small number of verbs in Homer that cannot be explained by the morphological and semantic factors set forth above. Notable exceptions are the aoήsts in -η- that have a passive meaning, and do not fall under the morphological rules (ί-νί): έβλάβην, έτύπη, πλήγη, άναβροχέν. For these cases, Ι would like to give a tentative explanation. The passive forms έβλάβη, έτύπη and πλήγη may have been formed by analogy with verbs expressing violent, destructive events such as -ράγη, διατρυφέν, έδάμη (cf. also post-Homeήc έκόπην?). The form πληγ-η is exceptional in another respect. After all, one would expect a form in -θη- (e.g. πλάγχθη) for heavy stem syllables. Perhaps *πλaγ-η- (< *plh 2g-) was influenced by the verbs with aoήst stems in -aγ-, e.g. έαγ-η-, pαγ-η-, παγ-η-. Not surprisingly, in later Greek the anomalous form πληγ-η- is sometimes replaced by synchronically more regular forms. Thus, the compounds έκπλαΎflναι and έκπληχθfιναι (only Ε. Tr. 183) occur, used in the special meaning 'to be stricken with teπor' There are two variant forms: βλαβη- and βλαφθη-. The former is probably older. In Homer, βλαφθη- is more frequent, but that could be due to the demands of the hexameter. In Classical Greek it is somewhat less frequent. How can this altemation be explained? Ι would suggest the following explanation. Βλαβ-η- might have orginally designated a spontaneous process 'got stuck, stumbled' There are a number of instances with this meaning in Homer. For example,245

rniddles that have been deήved frorn thern. The occuπence of the suffix -η- with verbs expressing rnental events is probably old. Evidence is the forrnation ΡΙΕ *widih1- ('see'; hence 'have in rnind, know') that can be reconstructed on the basis of Lat. videre 'see', OCS. viditi 'see', and perhaps Greek ~δεμεν < *ή-fίδ-ε-μεν < *-wid-h 1- (Ruijgh 1996: 194). Another very o\d forrnation rnight be ΡΙΕ *mn-ih 1- as in έμάνην 'raged' (not attested in Horner), and OCS. 3 pl. ml,ni-J~ 'they be\ieved' (see Rix e.a. 1998). 244 The forrn έάλη 'thronged together' cannot be used as evidence here, since it falls under the rnorphologica\ ru\e (rnentioned above) that sterns in -αλ- take the forrn in -η-. 245 Cf. a\so Ψ 545.

ΤΗΕ MIDDLE AND PASSIVE VOICES ΙΝ Π1Ε AORIST

(1)

S'fEM 135

αί δέ που αύτου

tβλαβεν έν πεδίφ, αϊ κεiσέ γε φέρτεραι ήσαν (Ψ 461) And the other [mares] have stumbled out there in the plain, they that were supeήor up to that point.

Subsequently, the βλαβ-η- acquired a true passive meaning 'was hindered, harmed'. This, in tum, led to the creation of the altemative form in -θη-. Eventually, the semantic contrast became faint, and the two variants could be used both in the spontaneous meaning and in the passive meaning. The following table roughly summarizes the distήbution of the forms in -η- and -θη- in Homer: Table 3:

The distribution of the passive aorists in -η- and -θη- in Homer •

Sponιaneous

Process • Mental Process

• Passive • Body Motion

• Collective Motion

Polysyllabic stem

θη

θη

θη

Stem in vowel

θη

θη

θη

Stem in

θη

θη

θη

Heavy monosyllabic stem

θη

θη

ε-vocalism

θη

θη

θη

η

η

η

η

θη

η/θη

-σ-

Stems in (C)Caλ-, (C)Caμ-,

(C)C/Xp-

Remaining stems (C)CIXC-, (C)C'ίC-, (C)CiJC-

In this section, Ι have also discussed the exceptions (e.g. βλαβ-η-, πληγ-η-) to this distήbutional pattem. Ι tήed to explain most of the exceptions either by their historical background or by occasional analogical processes.

Ι36 CΗΑΡΊΈR

3

3.1.2 The Distήbution of -η- vs. -θη- in Classical Greek Ιη

the previous section, we have seen that distributions of the aorist

ίη

-η- and the aorist ίη -θη- ίη Homer could be explained by a number of

morphological rules, and a number of additional semantic rules. Το get an idea of the distήbution of the two passive aoήst formations ίη Classical Greek, some figures may once again be helpful: 246 Table 4:

The semantic distribution of the passive aorist in in Classical Greek

-θη­

αnd -η-

θη

η

Passive

54

2

Spontaneous Process

6

5

Mental Process

15

Ι

Collective Motion

6

Ι

Body Motion

9

Ι

Total

90

10

The token-frequency of the passive ίη -η- has decreased clearly (10 90) compared to that ίη Homer (34 66). As ίη Homer, the typical use of the formation ίη -η- is the spontaneous use (5 instances out of ΙΟ). The formation ίη -θη- is relatively infrequent ίη the spontaneous use (6 out of 90).

The sample of 100 instances was taken from Herodotus: from the first book the 12 first occurήng passive aoήsts were taken, and from the remaining books the 11 first occurήng instances. 246

ΤΗΕ MIDDLE AND PASSIVE VOICES ΙΝ 'fHE AORIST

Table 5:

S'fEM 137

The morphological distribution of the passive aorist in -θη- and -η- in Classical Greek θη

η

Polysyllabic stem

42

ο

Stem in vowel

31

ο

Stem in

27

ο

Heavy monosyllabic stem

22

1

ε-vocalism

15

ο

Stem #VC-

8

ο

4

9

-σ

Stems (C)CaC-, (C)CtC-,

(C)CυC-

Α number of morphological rules can be infeπed from the figures. As in Homer, the aorist in -θη- occurs with the following stems:

(i) (ii)

(iii) (iv) (ν)

polysyllabic stems; stems in vowels; stems in -σ-; heavy monosyllabic stems; 247 stems with ε-vocalism. 248

As in Homer, there is a tendency for stems of the structure (C)CaC-, (C)CtC-, (C)CυC- to occur with the aorist in -η-. As in Homer, the foπn in -η- occurs with

247 E.g., άνει\>χθην, έλήφθην/έλάμφθην, and the novel fonn έτήχθην (Eur., Pl.), fonned under influence of the present and the active aoήst fonns. The fonn έπλήγην is the only exception. 248 E.g., έθρέφθην (τρέφω), έπλέχθην (πλέκω), έστρέφθην (στρέφω), έτρέφθην (τρέπω). There are a few exceptions: έλέγην (from λέyω 'gather'), έστέρην, ήγyέλην (Ε. Π 932). These fonns are, however, less common than the vaήant fonns in -θη-, έλέχθην, έστερήθην, ήγyέλθην. Έλέγην and ήγyέλην are younger than the vaήants in -θη-.

138 CHAP'fER 3

stems of the structure

(νί)

(C)Cαλ-, (C)Caμ-, (C)Caρ-. 249

For example, έστάλην (στέλλω), έσφάλην (σφάλλω), έδάμην 250 (δάμνημι), έδάρην (δέρω), έκάρην 251 έσπάρην (σπείρω), έφθάρην (φθείρω).

(κείρω),

έπάρην

(πείρω),

Contrary to Homer, in Classical Greek there is also evidence for the following morphological rule: the form in -η- also occurs with

stems of the structure

(vii)

(C)Caκ-, (C)Caπ-, (C)Caφ-.

Examples are: (-ακ-) έπλάκην (πλέκω), έτάκην (τήκω) 252 ; (-απ-) έκλάπην 253 (τρέπω);

(κλέπτω),

(-αφ-)

έλάπην

έβάφην

(λέπω),

(βάπτω),

έσάπην

έγράφην

(σήπω), (γράφω),

έτράπην έρράφην

(pάπτω), έσκάφην (σκάπτω), έστράφην (στρέφω), έτάφην (θάπτω), έτράφην 254 (τρέφω). 255 As for stems of the structure (C)Caκ- and (C)CιΧ.φ-, for lack of evidence we cannot be certain whether this morphological rule already applied for Homeήc Greek. In the case of stems of the structure (C)CιΧ.π-, however, it can be seen that the rule probably did not exist in the Homeήc language. In Homer, we find the form τραφθfιναι (τρέπω), whereas in Classical Greek the normal form is έτράπην. Above, it was stated that stems of the structure (C)CιΧ.λ-, (C)CιΧ.μ-, (C)CιΧ.ρ-, (C)CιΧ.κ-, (C)CιΧ.π-, (C)CιΧ.φ- receive the suffix -η-.

249 The - at least partia\ - productiνity of these ru\es is showπ by the replacemeπt of the old thematic aoήst επταρον 'sπeezed' (πτάρνυμαι) by the aoήst ίπ -η­ έπτάρην (Arist. Pr. 887b35, Hippocr.) 250 Έδάμην is Homeήc and poetic. Attic prose has έδαμάσθην (Χeπ., Isocr.). 251 The form κερθείς (with -ε- from κείρω, εκειρα or older εκερσα) is fouπd at Pi. Ρ. 4.82 (cf. ru\e ν). 252 Ιπ Piπdar, we fiπd the πονe\ form δρακείς, aoήst of δέρκομαι. 253 Also έκλέφθην occurs (Eur., Hdt.), cf. ru\e (ν). The ε-νocalism is probably adopted from κλέπτω or εκλεψα. 254 There is also a νaήaπt form έτρέφθην, aπ aπalogica\ (ε-νocalism ιakeπ from preseπt stem), aπd youπger (πο deaspiratioπ) formatioπ. The suffix -θη- is due to the ε-νoca\ism (rule ν). lt occurs oπly margiπally, aπd is especially used ίπ poetry (for

metήcal reasoπs?). 255

Exceptioπs to the ru\e are έβάφθην, έγράφθην (oπce ίπ a Milesian iπscήp­

c. BC, SIG 51,5, aπd post-C\assical), and έθάφθην (see also sectioπ 3.1.3). These forms, however, occur oπly margiπally. Maπy νerbs haνe νaήanιs ίπ -θη- with aπ ε-νocalism ίπ the stem (cf. ru\e ν) which is probably iπtroduced by analogy with the preseπt stem (e.g. κλέπτω) or the actiνe sigmatic aoήst (e.g. ί:κλεψα): έκλέφθην, tioπ, 5ιh

έπλέχθην, έστρέφθην, έτήχθην, έτρέφθην.

ΤΗΕ MIDDLE AND PASSIVE VOICES ΙΝ ΠΙΕ AORIST

S'fEM 139

This implies that this rule only applίes to stems that begin with a consonant. Ιη other words, the passive ίη -θη- occurs wίth (νίίί)

stems of the structure #VC-.

Examples are: άρ-θη- (άραρίσκω), άφ-θη- (iiπτω), άχ-θη- (aγω), όρ­ θη- (ορνυμι; Coήnna 1.21), όφ-θη- (όράω). There are ηο counter-examples to this rule. Especially άρ-θη- (from άραρίσκω) and άφ-θη- are significant, as you would normally expect forms ίη -η­ with stems ίη -αρ- and -αφ- (e.g. έδάρην, έγράφην). 256 So far, the distήbution of the two aoήst forms has been explained by means of a number of morphological rules. However, there is a group of forms, that is not covered by these morphological rules. These remaining forms possess a stem with the structures (C)CaC-, (C)Ct:C-, or (C)Ci>C-. With these stems, the following semantic tendency can be observed: ( 1)

Forms that denote a spontaneous process or a mental process have -η-; Forms that have a passive meaning, or denote body motion have -θη-.

(2)

As for collective motion verbs, both aoήst forms occur. Take, for instance, έμίγην vs. έμ(ε)ίχθην, and -έλέγην vs. -έλέχθην. The best example of this semantic tendency is the pair έφάνην vs. έφάνθην; the former expressing a spontaneous process 'appeared', the latter a true passive 'was shown, was discovered' 257 • Other examples of verbs with monosyllabic aorist stems ίη -αν- are έμάνην 256

This rule might be explained by the intransparancy of foπns as *i\γη (aγω),

*i\φη (aπτω), *ίόπη (όράω). 257 In tragedy, the foπn έφάνθη seems to be used sporadically as an intransitive 'appear' (cf. also the intransitive use of έστάθην in Homer and tragedy). The strongest case in point is S. Ant. 103 (lyr.): άκτ1.ς άελίου, το κ:άλ-Ιλιστον έπταπύλφ

φανΕν/ θήβ~ των προτέρων φάος,/ έφάνθης ποτ' αι χρυσέαςΙ άμέρας βλέφαρον

('Beam of the sun, fairer than al\ that have shone before for seven-gated Thebes, finally you shone forth, eye of go\den day, (... )'). Note the coπespondence with φανέν in vs. 101. However, even this case can be interpreted as passive: 'you [the beam] have been shown forth [namely, by the sun]' As to the other cases of έφάνθην, one can likewise hesitate between an intransitive interpretation 'appear', and a passive 'be revea\ed'. Ι prefer to take them as passives where possible: S. ΟΤ 525, 1485, S. Tr. 743, Ε. HF 804 (lyr.); Α. Pers. 264 (lyr.) may be intransitive.

140

CΗΑΡΊΈR

3

(μαίνομαι), which is a mental process (rule (Ι)); passives are έκράν­ θην 'was accomplished' (κραίνω), έρράνθην 'was spήnkled' (pαίνω), έχράνθην 'was defiled' (χραίνω) (rule (2)). The semantic tendency

can also be discerned with the stems in -αγ-. The forms in -η- denote spontaneous processes: έρράγην, έπάγην, έάγην 258 , whereas the forms in -θη- denote passives: έσφάχθην (σφάζω), έτάχθην 259 (τάτ­ τω), έφράχθην (φράττω). 260 Compare, finally, the following stems in -υφ-: έτi>φην (τ-όφω) 'smoked (intr.), smouldered' (e.g Ar. Lys. 221) denotes a spontaneous process, while έκρύφθην 'was hidden' is passi ve. 261 The following table sketches the distήbution of the forms in -η- and -θη- in Classical Greek:

2' 8

In Attic, the compound κατa-yfjναι is the usual fonn: a contraction of *κατα­

FαΎflναι. The Iong α also spread to the augmented fonns, e.g. κατεlχγην. 2' 9 Έτάχθην does not haνe an agentive use 'to post oneself, as this notion is expressed by the fonn έταξάμην. 260 Some of these verbs, however, have variant fonns in -η-: έσφάγην (σφάζω), έμάγην (μάττω), and έτάγην (τάττω). These variants occur Iess frequently than their counterparts in -θη-. The latter two are attested only once in Classical Greek. Έφράγην, furthennore, is post-Ciassical. Ι will retum to this morphological variation below (3.3). 261 Also the fonn έπνίγην (πνtγω) can be explained by its meaning 'choke, drown (intr.)' (e.g. Ar. Nu. 1504, Dem. 32.6), which is a spontaneous process.

ΤΗΕ MIDDLE AND PASSIVE VOICES ΙΝ 'fHE AORIST SΊΈΜ

Table 6:

The dίstribution of the passive aorists in Classical Greek

-η-

and

-θη-

• Spontaneous Process • Mental process

• Passive • Body Motion

• Collective Motion

Polysyl\abic stem

θη

θη

θη

Stem in vowel

θη

θη

θη

Stem in

θη

θη

θη

Heavy monosyllabic stem

θη

θη

ε-vocalism

θη

θη

Stem in #'VC-

θη

θη

Stems in

η

η

η

η

η

θη

-σ-

141

in

θη

(C)Cδ.λ-, (C)Cδ.μ-, (C)Cδ.p-

Stems in (C)Cδ.κ-, (C)Cδ.π-, (C)Cδ.φ-

Remaining stems (C)Cδ.C-, (C)CIC-, (C)Ci>C-

η/θη

Α number of exceptions to this pattem have already been discussed in this section. In the next section, Ι will discuss those verbs that show a variation between the two aorist forms.

3.1.3 Morphological

vaήation

In the previous section Ι have set forth the morphological and semantic principles that determine the distήbution between the aorist in -η­ and the aorist in -θη-. However, the boundary between the two formations is not entirely clear-cut. Many verbs exhibit a variation between the two forms. The aim of this section is to take a closer look at those variant forms that are not covered by the morphological and

142

CΗΑΡΊΈR

3

semantic rules that were proposed in the previous section?62 Ιη most cases, the variation can be attήbuted to a conflict between the morphological rules and the semantic rules. Generally, the alternative form that is not expected οη the basis of the aforementioned rules occurs less frequently, or even only sporadically. Additionally, stylistic factors may be relevant. Ιη comedy there is a statistical preference for the form in -η- for verbs that show vaήant aoήsts, while tragedy prefers the forms in -θη-. 263 Here follows an inventory of the verbs that display morphological variation. 264 άπηλλάγην vs. άπηλλάχθην (άπαλλάττω). Since these forms have polysyllabic stems, and the meaning of these forms is 'depart' (body motion) or 'be removed' (passive), the passive aorist form in -θη- is to be expected according to our rules. However, in Attic several stems in -αγ- show morphological variation, for example, άπηλλάγην, έμάγην (only in Plato), έσφάγην, and έτάγην (once in the comic Phrynichus. The regular form is έτάχθην). Outside Attic we find forms in -θη-. 265 Thus, άπηλλάχθην is the commonest form in Herodotus (Did he take άπηλλαΎflναι (8.84.1) and άπαλλαγήσε­ σθαι (2.120.3), from Attic?); έσφάχθην is found in Herodotus, Pindar, and Euήpides (once, at Π 177 [lyr.]).

έβάφην vs. έβάφθην (βάπτω). According to the rule stating that stems in -αφ- take a passive in -η-, έβάφην is the regular form (vii). Βαφ-θη- only occurs at Ar. Fr. 416 [άπεβάφθη], and might be explained by the strong association of passive meaning with the form in -θη- (see table 4: 54χ -θη- vs. 2χ -η-). There may also have been an analogy with βλάπτω έβλάφθην = βάπτω Χ.

262 In other words, vaήant fonns such as έστρέφθην vs. έστράφην will not be discussed since they can be explained by rule (ν) and rule (vii), respectively. 263 This can be concluded from the figures given by Lautensach (1911: 228). He collected the fonns that occur in a metrically equivalent environment. With the stems in labials and gutturals, the ratios are as follows: -η- tragedy 58χ, comedy 29χ vs. -θη- tragedy 69χ, comedy l3x. 264 The verbs are collected by Prevot (1934: 24 [Homer/Herodotus], 45-6 [drama], 63 [Piato]). 2 Μ Remarkably, in Attic prose the fonn άπηλλάχθην only occurs in the 7111 letter of Plato (335a).

ΤΗΕ MIDDLE AND PASSIVE VOICES ΙΝ 'fHE AORIST

S'fEM 143

έζ{yyην vs. έζε6χθην (ζεύγνυμι). The latter foπn is regular because it has a stem with a heavy syllable. lt occurs more frequently. However, it does not occur in comedy, which may suggest that έζεύχθην is of a higher register (cf. Lautensach 1911: 255). Έζύγην was probably foπned by analogy with μεί­ γνυμι

έμείχθην

έμίγην = ζεύγνυμι

έζεύχθην

Χ (=

έζύγην).

έθλίβην vs. έθλίφθην (θλίβω). Since these foπns have a passive meaning 'was squeezed, compressed', the latter foπn is in accordance with the semantic tendency. lt is attested slightly earlier (in Plato, whereas the foπner appears in Aristotle). The aoήst in -η- was probably foπned by analogy with έτρίβην: έτρίφθην, which is both phonologically and semantically similar. 266 έκάην vs. έιαχύθην (καίω). After the disappearance of the F· the stem ended in a vowel. Έκαύθην was foπned in accordance with the rule which states that stems in vowels get the -θη- foπn. However, the older vaήant did not disappear. -έdίνην vs. έdί(ν)θην (κλίνω). The passive foπn is either volitional 'lie down' or passive 'be lain down', so the suffix -θη- is regular. Originally, -έdίνην (only in compounds) was probably a somewhat colloquial variant (as mentioned above, the comics prefer the vaήant foπn in -η-) since it is attested first in Aristophanes and Pherecrates, then also Plato. Elsewhere the older έκλί(ν)θην occurs. The more recent variant is possibly foπned by analogy with foπns like eφηνα έφάνην = eκλϊνα χ (= έκλiνην ). έκόmtν vs. έκόφθην (κόπτω). The passive meaning favours the latter foπn. Έκόφθην however, is post-Classical. Έκόπην may have been created by analogy with the synonymous foπns έπλήγην and έτύπην (see section 3.1.1). Furtheπnore, it is possible that the rule which states that stems in -απ- take -η­ (rule vii) spread incidentally to this stem in -οπ-.

266

In the same way, έφλίβην in Hippocrates can be explained analogically to and έθλίβην. *Έφλίφθην is unattested.

έτρίβην

144

CΗΑΡΊΈR

3

έκρuφην vs. έκρύφθην 267 (κρύπτω). The former occurs oπly

at S. Αί. 1145, whereas the latter is highly commoπ (Hom., Aesch., Soph., Hdt., Eur., Th., Ar.), aπd it is also the form we would expect coπsideriπg its passive meaπiπg. The occuπeπce of έκρuφην is possibly due to aπ analogy with the verbs with passive aoήst stems ίπ -αφ-, e.g. ράπτω έρράφην. έμάyην vs. έμάχθην (μάττω). The former occurs oπce ίπ Plato (Tht. 191d), the latter is post-Classical. Οπ the basis of their passive meaniπg ('was kπeaded'), the latter is to be expected, but see uπder άπηλλάγην above. έπλάκην vs. έπλέχθην (πλέκω). Although the latter is attested earlier, ίπ Homer, it is probably more receπt, coπsideήπg its ε­ vocalism (which, ίπ turn, explaiπs the suffix -θη- (rule ν)). Both variaπt forms can have a passive meaπiπg 'be eπtaπgled, iπtertwiπed', as well as a volitioπal meaπiπg 'eπtangle oπeself, embrace' (body motioπ or collective motioπ?). Έπλάκην may be abseπt from Homer because it oπly fits iπto the hexameter with difficulty. Ιπ Classical Greek, it is more frequeπt than έπλέχθην. The ε-vocalism of έπλέχθην must have beeπ iπtroduced uπder the iπflueπce of πλέκω or ί::πλεξα.

έρρiφην vs. έρρίφθην (ρίπτω). Α priori, οπe would expect the form to be ίπ -θη- siπce the verb stem is heavy. However, both forms are fouπd and roughly with the same frequeπcy. The secoπdary -φ- of έρρ~φην, which was takeπ from έρρίφθην (Duhoux 2000: 381π3), or from perfect έρρiφθαι (C.J. Ruijgh, pers. comm.), is aπ iπdicatioπ that it is more receπt. It is possible that the verb was iπflueπced by the verbs haviπg passive aorist stems ίπ -αφ-η-. This rule might also be respoπsi­ ble for the highly margiπal aorist form ήλιφη- (oπly oπce, ίπ Pl. Phdr. 258b; ήλείφθην first attested ίπ Aeschylus). Further, aπ aπalogy with the pair ράπτω έρράφην could have played a role.

Έκρύβην is late, though mss. Eur. Supp. 543 have κρυβήσομαι, also pήnted by Diggle. The increase of passive forms with stems in velar and labial voiced stops is a typical post-classical phenomenon (see Blass-Debrunner-Rehkopf 1979 1s: 60, Ruijgh 1991: 480n8), e.g. έθρύβην, tψυyην.

ΤΗΕ

MIDDLE AND PASSIVE VOICES

ΙΝ ΤΗΕ

AORIST

SΊΈΜ

145

έστέρην vs. έστερήθην (στερέω). The fonner only appears in Euripides, and is possibly a backfonnation of the regularly fonned future στερήσομαι. According to Ruijgh ( 1991: 119n32), this might be an old form, evolved from *strhι-eh 1 -. έσφάγην vs. έσφάχθην (σφάζω). The fonner occurs in Aeschylus and Euripides. Considering the passive meaning of the fonns, έσφάχθην is to be expected. See further under the heading άπηλλάyην above. έτάγην vs. έτάχθην (τάττω). The fonner is attested only once, namely in the comic poet Phrynichus, whilst the latter fonn shows a high frequency. Above, it has been noted that in comedy a tendency exist to use the fonns in -η-. See further under άπηλλάyην above. έτάφην vs. έθάφθην (θάπτω). The latter fonn violates the rule which states that stems in -αφ- have aorists in -η- (vii). It occurs only very infrequently and is clearly more recent, considering the absence of aspirate dissimilation ('Grassmann'). It was probably created by analogy with έβάφην έβάφθην, or with βλάπτω έβλάφθην.

έτρiβην vs. έτρίφθην (τρίβω). We would expect the latter fonn since the stem is heavy (cf. τρίβω). Έτρtβην, however, occurs more frequently and is probably older (see Ruijgh 1991: 480). Possibly, έτρtβην originally refeπed to a spontaneous process 'waste away, wear out (intr.)', whereas έτρίφθην had a stήctly passive meaning 'be rubbed, be worn out' In historical Greek, however, this semantic distinction had disappeared. The -1:must be analogical, possibly by analogy with verbs like *τιiκω

έτ&κην ήyyέλην

= τρίβω

Χ

(>> έτρtβην).

vs. ήyyέλθην (άyyέλλω). The former occurs only at Ε.

ιτ 932, and in an inscήption (1 3• 78.19 [ca. 422 BC. ?], see Threatte 1996: 557), whereas ήyyέλθην is very common, and

regular - it is polysyllabic (rule i) and it has ε-vocalism (rule ν). Ήyyέλην may have been fonned under influence of the stem in -αλ- such as στέλλω έστάλην.

146

CΗΑΡΊΈR

ήλίφην

3 vs. ήλείφθην (άλείφω), see έρρίφην above.

συνελέyην vs. συνελέχθην (συλλέγω). In Attic, συνελέγην is more frequent than συνελέχθην. This form, in tum, is prefeπed by Herodotus. The occuπence of the suffix -η- with this verb could be the result of one of several factors: (a) the need to bring about a contrast with έλέχθην, passive of λέγω 'say'; (b) the influence of synonymous (συν)εμίγην; or conversely, (c) there rnight have been some influence of the stems in -αγ- that have -η-, e.g. άπηλλάγην, έπάγην.

έψ\>χθην vs. έψ{>ΧΙ'\ν (ψ\Ίχω). The former is the oldest (e.g. Κ 575), and it is also the expected form since the verb stem is long (cf. ψ-όχω). In the creation of έψ-i>χην, its meaning 'grow cold' (a spontaneous process, e.g. Ar. Nu. 151, Pl. Phdr. 242a) may have played a role. 268

conclude, in sections 3.1.1 and 3.1.2 Ι argued that the distήbution of the aoήsts in -η- and -θη- can be explained pήmarily by means of a number of morphological rules. With a number of stems, however, a semantic tendency appeared to be relevant to the distήbution of the two variant suffixes. In section 3.1.3, verbs were discussed that possess both variant aoήst forms. For each individual verb, the specific morphological and semantic factors were analyzed that may have motivated the emergence of the variant forms.

Το

268

Έψύγην

is post-Ciassical.

Ι

Ι

I I

Ι

ι

1ι1ι1ι1

Ρ

ι

tion

:;:;:;:::::;:::ιι .:::::: : :

:ι:ι:ι:ι:ι:ι:ι

ι ι ι ι ι ι ι ι ιιιιιιιιιιιι

:ι,, ι ι ,ς ι ι ι~;:;:~:;:;:;:;:;:~:;:;:;:;:;:;:;:;:;:;:;::~:~

κείρασθαι ...!

.. :-:-:.___.,. .,. . ._,__

:•:•:•:ι:ι:ι:ι:•:•:•:ι:•:·:·:·:·:·:·:·:·:•ι ·:·:·:·: ι:ι:::ι:ι:•:·:·:::•:•:ι:•:·:·:·:·:·:·:·:·:•, ·:· : ·:·

.•.•,ιι•ι•ιιιι,ι,ι,ι,ιι•ιιιιι•ι•ι•ι•ι•ι•ι•ι

,ι,•ι•ι•ι•,ι,ι,•,•ι•ι•ιι,ι,ι,ι,•,•ιιιιιι,ι,ι .· . ·.·.

'ι'ιιιιιιιιι•ιιιιιιιι,ι,ι,•ι•ι•ι•.ι,ι,ι,•ι•ιιιιιι,ι,ι,ιι

1 ι 1 ι 1 ιιιιι 1 ι 1 ιιιιιιιιιιιιι ι

ι:ι:ι:ι:ι:•:•:ι:ι:ι:ι:ι:ι:

:::::::::::: χολώθη .. :.:.:.:.:.:.:.:.:.:.: · · · · · · · · · ..1..-.-_---,,....fiiii.'-_J .·.·.·.·.·.·.·.·.·.· ι:·ι,ι,ι,ΙJΙ,ιιι 1 ι,ιι•:ι:ι:ι:ι:ι:ι:ι:ι:ι:ι:ι:ι:ι:ι:•:•:•:ι:ι:ι:ι;ιι :-:-: · · · · · · .·.·.·. Reflexίve ·:·:·:·:·:•:ι:ι:ι:ι:ι:ι:ι:ι:•:•:•:•:ι:ι:•: ::::::::

'•'•'•'•'•'•'•'•'•' ····- ·.·.·.·.·.·.·.

::::::::::::;:;:;:::;::::: μυθήσατο

•'•'•'•'•'•'•'•'•'•'•'•'•'

•'•'•'•'•'•'•'•'•'•'•' .·.·.·.·.· .· .· Process 1 ι 1 ι ι ι ι ι ι ι ι ι . : : : : : : : : χολώσατο

::::::::,.,.,.,.,.,.,., . . ·.·.· .. -.

ι; ι; ι; ι; ι; ι; ι; ι; ι; Ί; ι;, ~ι~;ι~;~ι;~ιΖk!Ε:Ί:j;;;-;~~

1 ι 1 ι 1 ι 1 ι 1 ι 1 ι 1 ι 1 ι 1 ι 1 ι 1 ι 1 ι 1~

1

·:::·:·:·:·:·:·:·:·:·:·:·:·:·:·

~ι, ι: ι: ι: ι: ι: ι: ι: ι :ι: ι: ι: ι: ι: ι:

ι ι ι ι ι ι ι ι 1 1 1 1 1 1 1 1

1 ι 1 ι 11 ι 11 ι 11 ι 1 ι 1 ι 11 ι 11 ι 11 ι 11 ι 11 ι 'ι 1 1 Ί 1 1 •'•IPerce

ι;ι·ι·ι·ι;ι;ι·ι·ι·ι·ι·ι·ι·ι·ι·ι

::::::::::::::::::::::, :: Speech '•'•'•'•'•'•'•'•'•'•'•'• , Α t :•:•:•:•:•:•:•:•:•:•:•:•, c

ι

έλύσατο

Reflexίve

Indίrect

Homer .

Distribution of the sigmatic middle aoήst

ίn

ίάνθη

Spontaneous Process

Π

έβλάφθην

Passίve

Intersection ofthe distributions ofthe U sigmatic middle aoήst and the aoήst in ~­ §Ξ§ Distribution ofthe aoήst in ~-

ι; ι; ι

'•'•'

Fig. 1: The Dίstrίbutίon of the Sίgmatίc Mίddle Aorίst and the Aorίst ίn -θη-

~

-

~

o-j

{I)

§

>

~

2!

8

~
μαι can also be used in aπ ingressive-aoristic meaning, and φανήσομαι also in a durative meaning. 317 In alliance with Stahl ( 1907) he states that the ''treibende Moment'' behind the creation of the morphological distinction middle vs. passive future is the need for specific (semantically unambiguous) middle and passive forms. The aspectual distinction, according to Wackemagel, could have played a secondary role in the development of the distinction. In the process of the development of the passive future form, Wackemagel distinguishes four stages (1920: 203). Stage (i): Stage (ii): Stage (iii): Stage (iv):

The formation φανήσομαι did not yet exist. Φανο\ιμαι could be used both with a durative aspect and an aoήstic aspect. Beside the older form φανο\ιμαι, expressing both aspects 318 , φανήσομαι is created, which specifically expresses aoήstic aspect. The aspectual contrast between the two forms fades away. Φανήσο­ μαι is used increasingly to express durative aspect as well. The middle future is no longer used in an intransitive or passive meaning.

In sum, according to Wackemagel the aspectual distinction only existed in stage (ii). Moreover, according to Wackemagel, this 317 Ingressive-aoristic and durative are Wackemagel's terms. These correspond with the current terms perfective and imperfective, respectively. 318 The one insιance in Homer seems to confirm this: Ενθεν γάρ μιν έδέγμην πρ&tα φανεiσθαύ Σ!Cύλλην (μ 230) ('For from there Ι expected that Scylla would first appear'); φανο\ιμαι here is used with an ingressive ('aoήstic') meaning.

ΤΗΕ MIDDLE AND PASSIVE VOICES ΙΝ 'fHE FuτuRE SΊΈΜ

181

distinction was not complete since the middle future form could be used in both meanings. 319 Or, to put it in somewhat anachronistic terms, the middle future is the unmarked member in a pήvative opposition. This stage lasted, according to Wackernagel, from Homer until the 5ιh century BC. Now let us consider the state of affairs in Homer. In Homer, only two verbs have a passive future formation: δαήσεαι en μιγή­ σεσθαι. Both future formations appear to express ingressive meaning, and can therefore be regarded as futures of the aorist stem: a. οσσα δ' ένi μεγάροισι καθήμενος ήμετέροισι

(5)

πεύθομαι, η θέμις έστί, δαήσεαι, ούδέ σε κεύσω (γ 187) But what tidings Ι have heard as Ι abide in our halls you shall hear, as is ήght, nor will Ι hide it for you.

b.

π&ς γ(φ έμευ σύ, ξε'iνε, δαήσεαι, εί

(... )

(τ

325)

For how shall you learn of me, sιranger, whether (... ). c. άλλ' οτε δη τάχ' (μελλε μιγήσεσθαι φυλάκεσσι (Κ

365)320

But when he was going to mingle with the sentinels (... ).

However, the middle future, too, could express ingressive meaning. Consider the following example: (6)

rος Όδυσεi>ς κούρnσιν εύπλοκάμοισιν (μελλε μίξεσθαι, γυμνός περ έών (ζ 136) Even so Odysseus was going to mingle with the fair-ιressed maidens, naked though he was.

Further, the middle form sense (i.e. 'presentic'). (7)

μίξομαι

could also be used

ίη aπ

iterative

θυμος δ' (τι ν&ιν έώλπει μίξεσθαι ξενίn ήδ' άγλα&. δ&ρα διδώσειν (ω 314) And our hearts hoped that we should yet meet as host and guest and give one another gloήous gifts.

Wackemagel summaήzes it thus: ''Aiso nur bei φανήσεται und nur ganz bei den ersten, die die Form brauchten, ist eine Spezialbedeutung nachweisbar: darauf reduziert sich in diesem Fall Blassens Theoήe" (Wackemagel 1920: 203). 320 Note that Iliad Κ (the Doloneia) might be an interpolated episode. 319

vorίibergehend,

182 CHAP'fER 4

Notice the coordination with the anomalous future of the present stem διδώσειν. lt can be concluded that, as Wackernagel claimed, the middle future is aspectually neutral in Homer. Note that, from a morphological point of view, the middle formation μίξομαι is also neutral since it neither possesses the present morpheme -σγ- (< *-γ­ σκ- ), nor the passive aorist morpheme -η-. 321 W ackernagel states that the distinction was already fading away in Plato (stage ίίί). Το support this claim he cites the following instances:

(8)

άλλ' έγrο δια παντος του βίου δημοσία τε εϊ πού τι επραξα

τοιουτος φανοομαι, κα1. ίδί~ ό αύτος οtτος (... ) (Pl. Ap. 33a) my 1ife, in any public activity Ι may have engaged in, found to be the same man as Ι am in pήvate life (... ).

τhroughout

a.

(9)

Ι

will be

η τουτο μεν οϋ, αύτου δε τρεφόμενοι σου ζ&ντος βέλτιον

θρέψονται κα1. παιδεύσονται, μη ξυνόντος σου αύτο\ς (Pl. Crί.

54a)

Or perhaps you will not do that, but if they are brought up here while you are living, will they be better brought up and educated if you are not with them?

b. θρέψονται δε δη ήμ\ν οi>τοι κα1. παιδευθήσονται τίνα τρόπον; (Pl. R. 376c) How will they be brought up and educated by us?

Wackernagel paraphrases example (8) by "es wird sich an mir herausstellen", thus interpreting it as ingressive. However, a perfect alternative would be to interpret φανουμαι as an event taking place repeatedly, i.e, "each time, Ι will be found to be, etc." As for the latter instances, in (9a) θρέψονται coπesponds with the present participle τρεφόμενοι. In other words, both θρέψονται and παιδεύσομαι designate non-completed events. As for (9b), it is conceivable that θρέψονται relates to growing up as an ongoing process, while παιδευ­ θήσονται relates to the completion of the education. 322 In sum, it is In Homer, we a1so find a contrast between the active future forms [ξω and The former coπesponds with the present stem έχ-, and means, according to Chantraine (1958: 446) 'avoir, garder' (e.g. Ρ 232, Σ 274, ζ 281). The 1atter coπe­ sponds with the aoήst stem σχ-, and means 'aπeter, contenir, suspendre' (e.f. Λ 820, Ν 151, Ξ 100, Ρ 182, Ω 670, χ 70, 248). 322 In connection with this examp1e, Kiihner-Gerth (1: 117n2) p1ead in defense of the aspect theory: "Wenn in einze1nen Fiillen, z. Β. Ρ\. civ. 376, c θρέφονται και παιδευθήσονται der Bedeutungsunterschied der beiden Futura nicht k1ar zu Tage 321

σχήσω.

ΤΗΕ MIDDLE AND

PASSIVE VOICES

ΙΝ 'fHE

FuruRE

SΊΈΜ

183

possible that the aspectual distinction was still present in the 4th century. In the next section (4.1), Ι will argue that it did indeed exist. First Ι would like to note that the aspect hypothesis was not accepted by everyone however. Stahl (1907: 83-7) rejects it, and advocates a diathetical distinction instead: ''DaB es sich bei den Futura dieser Art eher um ein Vorwiegen passiver oder medialer Bedeutung als um eine Zeitart handelt, zeigt sich auch bei άπαλλάσ­ σεσθαι" (Stahl 1907: 85). 323 Stahl renders passive άπαλλα-γήσεσθαι with 'befreit werden, etwas los werden' (as examples he cites Hdt. 2.120. 3, Th. 4.28.5, Pl. R. 445b, Dem. 22.37), and rniddle άπαλλάξεσθαι with 'loskommen, sich einer Sache erledigen' (τh. 4.128.5, 8.2.1, Pl. Euthphr. 15e, R. 465d). In other words, middle άπαλλάξομαι is volitional, whereas passive άπαλλα-γήσομαι is non-volitional. Ι will retum to Stahl's view at the end of this chapter. According to Schwyzer & Debrunner (Π: 265-6), the outcome of the debate seems to be beyond any doubt. They consider the aspectual hypothesis untenable. 324 Amongst others, Wackemagel and Chantraine are mentioned as adversaries of the aspect hypothesis. However, as we have seen above, Wackemagel has a finely tuned judgement on the matter. Chantraine, too, takes an intermediate position (1928: 13-4). Chantraine admits that the passages cited by Blass are not very convincing, but he concludes, "( ... ) Demosthene n'emploie pas φανο\>μαι et φανήσομαι sans nuance de sens. (... ) φανήσομαι indique un resultat obtenu, une conclusion qui apparait apres examen.'' Schwyzer & Debrunner do not do justice to this intermediate position. Instead, they adduce altemative factors that may explain the choice between the middle and passive future form as, for example, the length of the verb stem (following, in this respect, Wackemagel, who mentions it as a secondary factor), and word rhythm (following Hartmann 1935: 128). Schwyzer & Debrunner do not refer to a diathetical distinction between the two future forms. Could it be that Schwyzer & Debrunner tacitly presuppose the validity of this explanation, and that they use the length of the stem and word tritt, so ist zu bedenken, dass auch die entsprechenden Tempora der Vergangenheit, Imperfekt und Aoήst, mannigfache zweife\hafte Grenzfιi.lle aufzuweisen haben." 323 Magnien (1912: 279-80), in his monograph on the Greek future, a\lies with Stahl. For a rejection of Blass' hypothesis, see a\so Hartmann (1935). 324 Cf. a\so Moorhouse (1982: 176-7), who denies the existence of an aspectual distinction in Sophoc\es.

184 CΗΑΡΊΈR 4

rhythm only as secondary explanations? In any event, Schwyzer & Debrunner' s treatment of the matter is highly unsatisfactory. When we overlook the battlefield, an analogy with the battle for Thyrea, as descήbed by Herodotus, forces itself onto us. Like Alcenor and Chromios, Schwyzer & Debrunner claim victory prematurely. The aspectual hypothesis has been rejected before it has been refuted adequately. Before going into the aspect hypothesis any further, Ι would like to make some remarks on the morphological aspect of the hypothesis. It is obvious that the passive future form is connected to the passive aoήst form since both are formed with a stem in -(θ)η-. Conversely, the morphological relationship between the middle future and the present stem is tenuous. With regard to a large number of verbs, the present and the future stems are different. For example, φαίνομαι contains the stem φαιν- (< *φαν-yεlο-), whereas future φανο\>μαι is formed with the stem φαν(ε)- (< *φανε-σεlο-). This difference in stem form applies to almost all oήginal yod-presents (i.e. also to verbs in -άζω en -ίζω). Α difference in stem form is also found with most reduplicated presents (δίδωμι, τίθημι) 325 , nasal presents (δείκνυμι, λαμβάνω, τέμνω), presents in -σκ- (εύρίσκω), and suppletive verbs (όράω vs. οψομαι). In other words, it is not selfevident to assume a connection between the future middle and the present. lt is impossible to determine whether the future stem is derived from the present stem or from the verbal root for thematic presents such as παιδεύω (without yod, -σκ-, etc.) only, because the two are identical. 326 With respect to the morphological relationship between the future stem and the present and aoήst stems one can take two altemative positions: (1) lt can be assumed that the middle future has specialized to a 'presentic' meaning under the influence of the semantic

32 s Interestingly, in Homer the future formation διδώσομεν (ν 358), διδώσειν (ω 314) occurs (see also example (7)), formed on the present stem of δίδωμι. This future form has an iterative ('presentic') meaning. The formation can be regarded as a paral\el to the future that is based on the aorist stem, cf. δαήσεαι (Ύ 187, τ 325), μιγήσεσθαι (Κ 365). The future of the present stem probably vanished, contrary to the thriving 'aoήst' future, because it could only yield morphologically transparent forms with few present types (namely the reduplicated present, and potentially also the nasal present). 326 From a synchronical point of view, the future morpheme -σε/ο- is both related to the present morpheme -ε/ο- and to the aoήst morpheme -σ(α)-.

ΤΗΕ MIDDLE AND PASSIVE VOICES ΙΝ ΤΗΕ FuτuRE

S'fEM 185

with the 'aoristic' passive future, or (2) οπe might assume a other words, the middle future is semantically πeutral, whereas the passive future is semaπtically marked with a feature + perfective. This is approximately Wackemagel's (1920) view, as we have seeπ above. Notice that this semaπtic aπalysis is reflected ίπ the morphological system ίπ a very πatural fashioπ. The semaπtically πeutral φανο\Jμαι is πeither formed with the preseπt stem φαιν-, ποr with the aorist stems φηνα- (active and rniddle) or φανη­ (passive), but, iπstead, it coπtaiπs a separate stem φαν(ε)-. Coπversely, the 'aoristic' future φανήσομαι coπtaiπs the aorist stem φανη-. Ιπ the πext sectioπ, the oppositioπal pair φανο\Jμαι aπd φανήσομαι will be discussed, aπd it will be iπvestigated if they, as Blass coπteπds, express aπ aspectual coπtrast. Ιπ sectioπ 4.2, the questioπ will be raised whether this aspectual coπtrast also occurs with other verbs. Fiπally, ίπ sectioπ 4.3 Ι will argue that Stahl was ποt eπtirely iπcoπect ίπ his claim that the middle aπd passive future form expressed middle aπd passive meaπiπg. oppositioπ

privative

oppositioπ. Ιπ

4.1 Aspect Revisited: the case of φανο\Jμαι vs.

φανήσομαι

The aim of this sectioπ will be to revive Blass' aspectual hypothesis. Followiπg Blass, Ι will start my iπvestigatioπ with the most frequeπt pair of middle aπd passive future forms, these beiπg φανο\Jμαι aπd φανήσομαι.

feature of φαίνομαι is that there appears to be a teπse stem (preseπt aorist) aπd syπtactic complemeπt (iπfiπitive participle). This coπelatioπ is showπ ίπ the followiπg Table 1. The iπstances of the 3'd persoπ siπgular secoπdary iπdicative (i.e. έφαίνετο and έφάνη) were couπted ίπ Sophocles, Euripides, Herodotus, Thucydides, aπd Plato. Απ iπterestiπg

coπelatioπ

betweeπ

186

CΗΑΡΊΈR

4

Table 1: Έφαίνετο and έφάνη: frequency figures of constructions tota\

+ dat.327

10 (12.8%) 46 (59.0%)

78 (100%)

34 (43.6%)

28 (24.8%) 81 (72.6%)

113 (100%)

22 (19.5%)

+inf.

+ part.

έφαίνετο

22 (28.2%)

έφάνη

3 (2.7%)

ellipsis/abs.

lt is clear from the figures in the table that the imperfect form έφαί­ is construed more frequently with aπ infinitive complement, whereas έφάνη prefers the construction with the participle. This distribution caπ be explained by the semaπtics of the two altemative cσnstructiσns. The cσnstructiσn with infinitive invσlves a subjective view. The state σf affairs denoted by the infinitive seems tσ be true tσ (a) particular person(s), whereas the speaker is not certain about its truth. In σther wσrds, the state σf affairs expressed by the infinitive is nσn-factive. Fσr σur present purpσses, it is impσrtaπt tσ note that this subjective view ('seem tσ') is inherently stative. The constructiσn with a participle-cσmplement, ση the σther haπd, invσlves aπ σbjective view. The state σf affairs expressed by the participle is considered tσ be true by the speaker (factive value). 328 Additiσnally, the difference with the fσrmer cσnstructiσn is that a chaπge is invσlved: φαίνομαι cσnstrued with a participle refers typically tσ the mσment at which a particular state σf affairs is recσgnized as being σbjectively true. Since the cσnstructiσn with a participle ('be found σut tσ') invσlves a change σf state, it inherently has a telic aspect. The σbjective versus subjective contrast alsσ has cσnsequences fσr the frequency of a dative cσnstituent. Dative cσnstitutents σccur mσre σften wit!ι the infinitive cσnstructiσn since, in general, it is σf mσre impσrtaπce whσ is the particular hσlder of the subjective view (e.g. When Ι was younger, the world seemed α happy place to me). The subject σf aπ σbjective view is mσre σften geneήc σr easily understσσd ση the basis σf the cσntext (e.g. Oedipus turned out to be his wife's son). This difference in inherent aspect between the two constructiσns explains the different frequencies of present and aoήst stem.

νετο

327 The dative complemenιs occuπing with all constructions are counted: infinitive, particip\e, ellipsis and absolute. 32R For an analysis of the factive and non-factive values of the respective constructions with participles and infinitives, see Rijksbaron (1986).

ΤΗΕ MIDDLE AND PASSIVE

VOICES

ΙΝ 'ΓΗΕ F'uτuRE SΊΈΜ

187

Telic φαίνομαι cum participle tends to occur in the aoήst stem, whereas statiνe (atelic) φαίνομαι with an infinitive prefers the present stem. Now, consider the following table: Table 2:

Φανο\>μαι

and

φανήσομαι:

frequency

figures

of

constructίons

+ inf.

+ part.

e\l./abs. 329

tota\

+ dat.

φαν ου μαι

2 (2.7%)

9 (12.3%)

62 (84.9%)

73 (100%)

16 (21.9%)

φανήσομαι

ο

(0%)

8 (21.6%)

29 (78.4%)

37 (100%)

5 (13.5%)

In the table above Ι have collected all instances of middle and passive futures occuπing in the same authors as those of the previous table. Α notable difference between the figures of the two tables is the lower frequency of the construction with an infinitive. From this we may infer that it is generally more relevant for something to be found to be the truth at some time in the future, than that something will seem to be so. There are, however, also some important similarities between the two tables. First, φανο\>μαι occurs more often with the infinitive construction than φανήσομαι does (2.7% vs. 0%), although, obviously, these low numbers do not provide definitive proof. Second, φανήσο­ μαι more often occurs with the participle-construction than φανο\>μαι does. The percentages, furthermore, are very similar to those of έφαίνετο and έφάνην: 12.3% vs. 12.8%, and 21.6% vs. 25.0%. The third and final similarity is that dative constituents occur more often with φανο\>μαι than with φανήσομαι (likewise έφαίνετο vs. έφάνη). These three similarities suggest that φανο\>μαι coπesponds with έφαί­ νετο, whereas φανήσομαι coπesponds with έφάνη. In the following section Ι will coπoborate the evidence provided by these figures by an examination of a number of passages, taken

When no predicatiνe noun is present, it is hard to distinguish the auxiliary with el\ipsis of the participle or the infinitiνe from the non-auxiliary usage of φαίνομαι ('become νisible'). When the subject is a concrete noun, an interpretation 'become νisib\e, appear' is more likely. In the case of an impersonal subject, el\ipsis of participle or infinitiνe is the more likely interpretation ('it appears that'). The cases with an abstract noun are highly ambiguous, e.g. Ε. lon 1063 έλπίς (Hope will appear νs. There will appear to be hope), Th. 3.20.1 σωτηρία. The same holds for metaphoήcal\y used nouns, e.g. Hdt. 3.156.2 πόρος. 329

φαίνομαι

188

CΗΑΡΊΈR

4

from Sophocles, Euήpides and Plato. Νο orators were chosen because, as will be shown below, they have an overwhelming preference for the passive future. Apparently, the passive future forrn was selected by them in a more or less routinized manner. This can be considered an indication that, at least in oratoήcal prose, the semantic contrast between the rniddle and the passive future was already fading away in the 4th century. Let us first consider some typical instances of φανήσομαι in the meaning 'be found, tum out'. Oedipus will be found out to be a bom Theban and, moreover, the brother of his own children.

( 10)

είτα δ' έγγενης φανήσεται Θηβα\ος (S. ΟΤ 453) But later he will tum out to be a native Theban (... ).

(11)

φανήσεται δε παισ1. τοις αύτου ξυνrον άδελφος αύτος καΙ. πατήρ (S. ΟΤ 457) And he will tum out to Iive together with his own children, as both a brother and a father (... ).

In Plato, φανήσομαι is often used in anticipation of the conclusion of an argument. In Lg. 892c, for example, the Athenian anticipates the final conclusion that the soul is one of the first creations. As an interrnediate step in the argument it must be proven that the soul is older than matter (σ&μα). This is the starting-point of the investigation. (3)

ΑΘ. ( ... ) εί δΕ: φανήσεται ψυχη πρ&τον, ού πυρ ούδΕ: άήρ, ψυχη δ' έν πρώτοις γεγενημένη, σχεδον όρθότατα λέγοιτ' αν

είναι διαφερόντως φύσει. ταυτ· εσθ' οϋτως (χοντα, αν ψυχήν τις έπιδείξτι πρεσβυτέραν οi>σαν σώματος, &λλως δΕ: ούδαμ&ς.

ΚΑ. 'Αληθέστατα λέγεις. ΑΘ. Ούκουν τα μεττα έπ' αύτο δη το1>το στελλώμεθα;

(Pl. Lg. 892c) ΑτΗΕΝΙΑΝ: If it will be found that soul came first, not fire or air, and it was one of the first things to be created, it will be quite coπect to say that soul is preeminently natural. This is true provided you can demonstrate that soul is older than matter, but not otherwise. CLINIAS: Very true ΑΤΗΕΝΙΑΝ: So this is precisely the point we have to tackle next?

ΤΗΕ

MIDDLE AND PASSIVE VOICES

ΙΝ τΗΕ FuτuRE SΊΈΜ

189

Eventually, in 896b Clinias concludes that the soul is indeed the oldest thing there is. Besides the construction, the sernantic property of iterativity or genericity is relevant to the distribution of the future forrns at issue. Iterativity and genericity rnay be indicated by the nature of the subject of the clause. In Sophocles, Euripides and Plato, there are 6 instances of φαvουμαι with a clausal subject which point to aπ iterative or geneήc interpretation, these being οστις eστ' άνήρ χρηστός (S. Ant. 662), πάντα άνδρα (Pl. Lg. 738e), τίς eκαστος (Pl. R. 537b), οσοι πάντες (Pl. Tht. 206e), πάντα (Pl. Chrm. 172b), and άλλα μυρία (Pl. Sph. 245e). Significantly, no subjects of this type were found in cornbination with the passive forrn φανήσομαι. In two cases (Lg. 738e, Chrm. 172b), φανουμαι has the subjective rneaning 'seern'; the three other instances involve φαίνομαι in the objective rneaning 'tum out, be found' For exarnple, (12)

έν τοις γ(φ οίκείοισιν οστις eστ' άνήρ

χρηστός, φανεiται κάν πόλει δίκαιος rον (S. Ant. 662) The man who acts ήghtly in family matters will be seen to be ήghteous in the city also.

(13)

κα\. οϋτως οσοι τι όρθον δοξάζουσι, πάντες αύτο μετα λόγου φανοϋνται eχοντες (... ) (Pl. Tht. 206e) And that being so, anyone at all who makes a coπect judgement will tum out to have it 'together with an account' (... ).

As was rnentioned aboved, in general, φαίνομαι construed with a participle ('be found out') tends to have a passive future φανήσομαι. In these generic contexts, however, φανουμαι is found. We rnay conclude that the 'presentic' future φανουμαι can also be used to express the rneaning 'will be repeatedly/always found out' (objective plus iterative/generic rneaning). Other indications of iterativity or geneήcity are the use of particular adverbs or the occuπence of a negation. In our corpus, φανουμαι occurs 4 tirnes accornpanied by aπ adνerb with iterative or generic value. These are άεί [2χ] (Pl. R. 591d, Lg. 871a), έκάστοτε (Pl. Lg. 689e), and τοτε μέν τοτε δέ (Pl. Lg. 860b). The forrner three cases inνolve the objective φαίνομαι 'tum out, be found' .330 330

These kind of adverbs are not found in combination with φανήσομαι. The in the Republic (59ld) will be discussed below.

occuπence

190 CΗΑΡΊΈR 4

For example, (14)

ό γιlρ νόμος άπαγορεύει και άπαγορεύων ύπtρ πάσης της πό­ λεως άει φαίνεται και φανεiται (Pl. Lg. 871a) The reason is that the law itself makes the proclamation, and it is always seen to make a proclamation on behalf of the state, and it will always be.

Once again, even though we are dealing with the objective meaning of φαίνομαι (cf. part. άπαγορεύων), we find the future form φανεiται due to its iterative/geneήc meaning (άεί). Notice also the coπes­ pondence with the present φαίνεται. Α typical kind of generic expression are γνrομαι. The following γνώμη is from Heraclitus, quoted by Plato: (15)

άνθρώπων ό σοφώτατος προς θεον πίθηκος φανεiται και σοφί~ και κάλλει και τοiς άλλοις π&σιν (Pl. Hi. Μα. 289b) The wisest of men will be seen to be a monkey compared to god in wisdom and fineness and everything else.

Here too, it can be seen that the 'presentic' future form occurs with a generic value. 331 There appears to be a tendency for negations to occur with φανο\>μαι, rather than with φανήσομαι. Of all the instances with a negation, 13 occur with φανο\>μαι, whereas 4 occur with φανήσο­ μαι.332 This tendency can be explained by the link between negation and iterativity. 333 Furthermore, in a number of instances featuring 331 For use of the future tense in gnomic expressions, see Magnien (1912: 16970). One of his examples coπoborates our point: μισ& πολίτην οστις ώφελεiν πάτραν/ βραδiις φανεiται μεγάλα δε βλάπτειν ταχύς (Ar. Ra. 1427) ('Ι hate any citizens who will tum out to be slow to aid, and swift to hurt, his town'). There is, however, for φανεiται a vaήant reading πέφυκε. 332 The instances of φανο\'>μαι with negation are the following: S. OC 51, 804; Ε. Heracl. 769, Ph. 1623 [μή], Hipp. 332; Ρ\. Lg. 738e [μή], 744a, 770d [μή], Hi. Μα. 289b, 291d [μή], Prm. 13ld, 132a, Alc. 1 109a, Sph. 245b; φανήσομαι with negation: Ρ\. R. 479a, Tht. 190e [μή], Sph. 216b, Μχ. 248a. 333 The phenomenon that the presence of a negation may bήng about an iterative interpretation has a\so been noted by Rijksbaron (1994 2: 15). An illustrative example is: έaν yaρ τί σε φαν& κακον πεποιηκώς, όμολσyω άδικεiν. έ&.ν μέντοι μηδΕν

φαίνωμαι κακον πεποιηκΟΟς μηδε βουληθείς, ού κα1. σiι αu ομολογήσεις μηδΕν

(Xen. Cyr. 5.5.13) ('If it will be found that Ι have done you harm, Ι confess that Ι am in the wrong; but if it tums out that Ι have done you no harm and Ι intended none, will you then ση your part confess that you have suffered

ύπ' έμο\'> άδικεiσθαι;

Ί'ΗΕ MIDDLE AND PASSIVE VOICES ΙΝ 'ΠΙΕ FιJτuιlli SτΕΜ

191

φανο\>μαι, the negation is combined with the adverb ποτέ, which reinforces the iterative or generic interpretation meaning (cf. its positive counterpart άεί). 334

can also have the (original) concrete meaning 'become visible, appear' (without complement). Here, the 'aoristic' passive future form predorninates. lt occurs 8 times in our corpus; the middle future is used only twice. This can be explained by the inherent telic aspect of φαίνομαι in the meaning 'become visible, appear' (an accomplishment). 335 For example,336 Φαίνομαι

(16)

φανήσομαι γάρ

(... )

δούλης ποδrον πάροιθεν έν κλυδωνίφ (Ε. Hec. 47) Ι shall appear (... ) in ιhe sea swell before ιhe feet of a slave girl.

(17)

η γάρ τις άροτηρ η τις οίκέτις γυνη φανήσεται νψν

( ... )

(Ε. ΕΙ.

105)

Some farmer or s\ave woman will appear to us (... ).

( 18)

καίτοι έγrο lρμην ένταυθά που φανήσεσθαι τilν έπιστήμην, -ίlν δη πάλαι ζητο\>μεν

(Pl. Euthd. 289d)

And yet it was in ιhis connection ιhat Ι expected been seeking a\1 ιhis time wou\d appear.

ιhat ιhe

knowledge we have

Το sum up the discussion so far, it is possible to schematize the main features of the distήbution of φανο\>μαι and φανήσομαι as

no wrong at my hands?'). The aoήst φαν& denotes a punctual event, whereas ιhe present stem φαίνωμαι in combination wiιh ιhe negation (μηδΕν) denotes an iterative event ('it will tum out, time and again, ιhat ... ) 334 These instances are: S. OC 804, Ε. Heracl. 169, Ρ\. Lg. 738e, Ηί. Μα. 29\d. There is, however, one exception: κ:α1. πέρας (... ) ούδΕν φανήσεταί ποτε (Ρ\. Sph. 26\b) (Άnd no limit (... ) will ever appear'). 335 As might be expected, φαίνομαι in ιhe meaning 'become νisib\e' occurs more frequently in ιhe aoήst stem ιhan in ιhe present stem. Of instances of ιhe 3π1 sg. imperfect έφαίνετο counted in Tab\e 1, ιhere were no cases wiιh ιhe meaning 'become visible', except when a negation was present. In ιhis case, ιhe c\ause often has an iterative interpretation (cf. Rijksbaron 2002 3 : 15), for example, Hdt. 2.118.17, 3.156.11, e.a.; Th. 5.7.5, 8.8.4. Furιhermore, it is significant ιhat ιhe compound άναφαίνομαι, which is mostly used in ιhe concrete meaning 'become visible', shows a clear preference for ιhe passive future form (see Blass \892: 271). 336 Other examp\es are Ε. Supp. 605, Hipp. 479, Phaί!th. 275, Ρ\. Plt. 209b, Sph. 26\b. Exceptions are: S. El. 1155, Ph. 1064.

192

CHAP'ΓER

4

follows: Table 3: Distribution of φανο\Jμαι and

φανήσομαι

abs./ + part. 'become visible, become clear'

+ inf.

iterative/geneήc

φαν ου μαι

φαν ου μαι

non-iterative/geneήc

φανήσομαι

φαν ου μαι

'seem'

In order to verify the conclusions stated above, Ι will now discuss, in more detail, the nine cases (Ι-ΙΧ) of the middle and passive futures φανο\Jμαι [4χ] and φανήσομαι [5χ] that occur in Plato's Republic. The first instance is: (Ι)

νυν δέ γ', tφην, εϊπερ σοφία τε και άρετή έστιν δικαιοσύνη,

p~δίως οtμαι φανήσεται και ίσχυρότερον άδικίας (... ) (Pl. R.

351a) But, now, if justice is indeed wisdom and virtue, it will easily be shown to be stronger than injustice (... ).

The previous step in the argument is recapitulated in the εί-clause preceding the main clause. The future φανήσεται 'will be shown' introduces the issue that is to be dealt with next, which results finally in the conclusion in 352b: οτι μeν γ(χρ και σοφώτεροι και άμείνους και δυνατώτεροι πράττειν οί δίκαιοι φαίνονται (... ) ('Now that the just are found to be wiser and better and more capable of doing things ( ... )'). (ΙΙ)

Τούτων γaρ δή, ι1 aριστε, φήσομεν, τ&ν πολλ&ν καλ&ν μ&ν τι εστιν ο ούκ αίσχρον φανήσεται; (... ). οϋκ, άλλ' άνάγκη, tφη, και καλά πως αύτα και αίσχρ(χ φανfJναι (... ) (Pl. R. 479a) SOCRAΊΈS: "My dear fellow", we'll say, 'Όf all the many beautiful things, is there one that will not also be found to be ugly? (... )" GLAUCON: There isn't one, for it is necessary that they are found to be beautiful in a way and also ugly in a way ( ... ).

Socrates presupposes the truth of the idea that all beautiful things are, in a way, also ugly. On closer inspection, the beautiful things will indeed be found to be ugly. Thus, φανήσεται is used here in the

ΤΗΕ MIDDLE AND PASSIVE VOICES ΙΝ 'ΓΗΕ FυruRE SΊΈΜ

193

objective sense 'be found' Glaucon then confinns Socrates' μ&ν­ question. Note that the aoήst φανfιναι in Glaucon's answer coπes­ ponds with the passive future in the question. lt is worthy to note that the future tense is used here in a special sense. lt does not refer to an event in the future. Rather, it appears to have a modal, or more specifically, an epistemic value (see Palmer 1986: 61-2). By using the future, the speaker asserts that the truth of the proposition at issue is the logical consequence of an hypothesis that has been accepted earlier in the discourse. 337 The future tense, used thus, occurs quite frequently in Plato's dialogues (see also examples ΙΙΙ, Vll, and ΙΧ below). (111)

οϋτε γάρ που σκοτωδέστερα μη οντος προς το μίiλλον μη

εtναι φανήσεται, οϋτε φανότερα οντος προς το μίiλλον εtναι

(Pl. R. 479d) Surely, they will not be found to be more than what is or not be more than what is not, for apparently nothing is darker than what is not, or clearer than what is.

lt is

appropήate, according to Sς~crates, to place the beautiful things (as well as other things) intennediate between being and not being (479c). Το support this claim, he puts forward the argument that is cited here (οϋτε γάρ κτλ.) as he considers it a self-evident truth. Thus, φανήσεται can here be interpreted as 'be found' (objective meaning). As in example (11), the future tense is used with an epistemic value.

337 This special use of the future tense has also been observed by Magnien (1912, 168-169). For a recent discussion of this special future type, see Bakker (2002). This use of the future tense in Greek may be compared to the use of the auxiliary will in English as in That'll be the postman. According to Langacker, there is a difference between this use of will and the use of must. Whereas must conveys a sense of epistemic immediacy, will suggests a certain epistemic disιance: "( ... ) will implies that confiπnation requires a non-negligible expansion of present knowledge (so that new infoπnation might alter the prediction), wheras must (... ) indicates that confiπnation is regarded as virtually inevitable" (Langacker l99la: 280). This semantic distinction explains the following judgmenιs: (a) They {wilV *must} quite possibly be there now.; (b) Since you saw it with your own eyes, it {mustl*will} exist. This subtle distinction may also be detecιable in our example (II). Socrates, by using the future, Jeaves open - or pretends to leave open - the possibility that his logical conclusion might be altered by new infoπnation or a new argument from Glaucon. In this way, he involves Glaucon in his argument. Glaucon, subsequently, confiπns that Socrates' conclusion is ineviιable by using the word άνάΎΚ1'J.

Π:

194 (IV)

CΗΑΡΊΈR

4

Λαβο\> τοίνυν, ήν δ' έγώ, ολου αύτο\> όρθ&ς, καί σοι εϋδηλόν τε φανεiται και ούκ &τοπα δόξει τα προειρημένα περ1. αύτ&ν

(Pl. R. 491c)

If you coπectly grasp the general point I'm after, it will be clear to you, and what I've said before won't seem so strange.

Socrates suggests that if Adimantus grasps the point wholly and in the right fashion the issue will indeed become clear (not: seem clear) to him. Φανεiται is used here in the objective meaning. The question now is why the 'presentic' middle future occurs as, after all, we are not dealing with an iterative or generic meaning. Ι would suggest that the explanation may be found in the immediative meaning of φανεi­ ται. 338 The case will immediately become clear to Adimantus once he looks into it in the ήght manner. The first event is presented as a necessary condition for the realization of the second. The use of καί expresses the continuity between the two events (λαβο\> AND φανεi­ ται 'grasp and it will be clear straight away'). Α comparable tum is found at S. OC 51: (19)

σήμαινε, κούκ &τιμος εκ γ' έμο\> φανft (S. OC Tell me what it is, and no refusal shall come from me.

51)

Again, the middle future appears to have an immediative value. 339 However, as S.R. Slings pointed out to me (pers. comm.), φανεiται in ex. (IV) may be accounted for in an altemative way. In this passage, a distinction is implicitly made between Adimantus' temporary, incoπect conclusions, and an evidently coπect (εϋδηλον) conclusion which relates to a state of affairs that serves as a background for subsequent questions which Adimantus will have to answer. According to Slings, it is this feature of permanence that deter-

33 κ For a descήption of the immediative use of the present stem Ι refer to Ruijgh 1996: 535-95 (who uses the term inceptive), and Rijksbaron (20023 : 17f.). 339 Another example of the immediative use (cf. αύτίκα) of the middle future φανουμαι is εϊτε καταλείψομέν τινας ήμίi>ν αύτίi>ν φύλακας έπ1. το'iς έφ' ήμ'iν γιγνομένοις, αύτίκα ούδεμίαν ίσχuν Εχοντες άναφανούμεθα (Xen. Cyr. 4.5.15). ('and if we leave some of our own men to guard what falls into our possession, it will immediately be found out that we have no strength'). Although the verb has the objective meaning 'be found' (cf. the participle) the rniddle future form is used wiιh an immediative value.

ΤΗΕ

MIDDLE AND PASSIVE VOICES

ΙΝ 'fHE Fιm.1RE

S'fEM 195

mines the selection of the imperfective middle future form. (V)

και &μα μία και αϋτη τrον βασάνων ούκ έλαχίστη, τίς ί::καστος έν τοiς γυμνασίοις φανεiται (Pl. R. 537b) At the sarne tirne, this is also not the Ieast of the tests, how each one of thern tums out to be in this physical training.

Socrates is speaking about the nature of one of the tests by means of which future philosophers can be selected. Φανεiται can be interpreted as iterative-distributive (cf. εκαστος). (VI)

-,.Αρ' ο-i)ν, ήν δ' έγrο, δς &ν φαίνηται πονηρότατος, και άθλιώτατος φανήσεται; και δς &ν πλεiστον χρόνον και μάλιστα τυραννεύΟΊJ, μάλιστά τε και πλεiστον χρόνον τοιο\>τος γεγονrος τη άληθείμαι νs. φανήσομαι: frequency of occurrence342 φαν ο\> μαι

Aeschylus Sophocles Euripides Herodotus Thucydides Lysias Aristophanes lsocrates Xenophon Plato Aeschines Demosthenes

2

17 12

'

φανησομαι

ι

4

2

9 6

5

1

1

10

8 3 6

5

42 1 17

43 3 30 6 62

The authors are listed in order of their (assumed) date of birth. lt can be observed that no significant shift takes place during the 5th and 4th centuries. Rather, there appears to be a division according to genre. In the orators Lysias, lsocrates, Aeschines and Demosthenes, the use of the passive form predominates overwhelmingly, whereas in the remaining authors the middle form is prefeπed. Only Herodotus shows a different picture. This may, however, not be significant in view of the low numbers. lt may be concluded from these figures that the orators selected the passive form more automatically than the other authors.343 This would mean that, in the orators, the semantic distincti-

342 The data are deήved from the ΊLG ση CD-ROM, and from Magnien (1912, 1: 281, 342). Included are the compounds of φανουμαι and φανήσομαι. Not included are the insιances of φαν ου μαι (esp. in the compound άπο-) with an indirect reflexive meaning. 343 Ι have found the same general tendency with the other verbs that have an opposition between middle and passive future that is similar to φανουμαι vs. φανήσομαι (e.g. λείψομαι λειφθήσομαι). Again, the orators show a general preference for the passive future: Lys. 24χ [passive] 4χ [middle], Isocr. 57χ [p]: 9χ [m], Aeschin. 21χ [p] 4χ [m], Dem. 92χ [p] 47χ [m]. Note that Demosthenes - as in the case of φαvουμαι φανήσομαι - uses the middle forms relatively often. The 'non-orators', conversely, have a predilection for the middle future form: Α. llx [m] 4χ [p], S. 29χ [m] 10 [p], Ε. 37χ [m] 24χ [p], Hdt. 39χ [m] llx [p], Th. 30χ [m] 23χ [p], Ar. 20χ [m] l5x [p], Xen. 59χ [m] 16χ [p], Pl. l01x [m] 82χ [p].

198

CΗΑΡΊΈR

4

ση between the two foπns began to fade away. 344 It appears that, in this respect, the language of the orators was less conservative than the language of the other authors. 345 In conclusion, in this section Ι have attempted to revive Blass' hypothesis by arguing that the middle future foπn φανο\>μαι differs from the passive future foπn φανήσομαι with regard to aspect. Φανο\>μαι, like the present stem, has an imperfective meaning, whereas φανήσομαι, like the aorist stem, has a perfective meaning. In practice, φανο\>μαι has the stative meaning 'seem' (subjective view), and φανήσομαι the telic meaning 'turn out, be found, become clear' (objective view). Φανο\>μαι may also express the iterative or geneήc meaning 'be found repeatedly/generally found to be' Finally, we saw that this semantic contrast may have been fading away in the language of the orators.

344 Only Demosthenes has a considerable number of middle future forms (17χ). Possibly, he still knew the semantic distinction. Απ interesting passage can be found at Dem. 20.80-1: (... ) πανθ' ύπερ ύμ.&ν φανήσεται πράξας Χαβρίας (... ). aξιον (... ) σκοπεiν, οπως μη φανούμεθα φαυλότεροι Χίων περl τούς εύεργέτας γεγενη­ μένοι. ( ... ) εί (... ) των ( ... ) τι δοθέντων άφ!Jpημένοι φανήσεσθε. ('It will be seen that Chabήas did eνerything on your behalf (... ). lt is worth while to see that we do not prove to be less generous than the Chians with regard to our benefactors. If (... ) we tum out to rob tιim of part of the reward (... ).') Even Blass (1892: 269) admiιs, be it reluctantly, that Demosthenes employs the middle and the passive form without semantic distinction, for the sake of vaήation. However, Ι would maintain that there is a semantic difference since φανούμεθα is used in a geneήc sense, whereas the two passive futures refer to single eνenιs. Furthermore, in this speech Demosthenes uses comparable phrases twice: ά~ρημένοι φανούμεθα (20.46), and άφ!Jpημένοι φανεiσθε (20.120). Contrary to άφ!Jpημένοι φανήσεσθε in the passage above, these two middle cases can be interpreted as iterative (cf. the plural παiδας at 20.46), and as geneήc (20.120). 345 Note that in the Koine, which stems from vemacular Attic, the use of the middle forms almost completely disappeared. Thus, in Polybius and the papyή, there are no more verbs that have both the middle form and the passive form with passive meaning (see Mayser 1926-34, Il 1: 212, Blass & Debrunner 1979 15 : 62). By contrast, in the highly conservative epic language only middle futures occur. Thus, Hesiod, the Homeήc hymns, Apollonius Rhodius, Theocήtus, Babήus, Oppian, Quintus Smymaeus, Colluthus, Tryphiodorus, and Nonnus only use the middle future forms.

ΤΗΕ MIDDLE AND PASSIVE

V OICES

ΙΝ 'fHE FuτuRE SΊΈΜ

199

4.2 Other Verbs with an Aspectual Distinction

The question now remains whether the distinction between φανο\>μαι and φανήσομαι can also be found with other verbs. There is a large number of verbs that show a morphological contrast in the future stem that is similar to the contrast between φανο\>μαι and φανήσο-μαι. 346 Απ example is κινέω, which has the passive aoήst form έκινήθην 'Ι moved (intr.), Ι was moνed (pass.)', but two vaήant future forms, κινήσομαι 'Ι will move, Ι will be moνed' and κινηθήσομαι 'id.' Again, at first glance, there seems to be no semantic distinction between the middle and the passiνe future forms. Howeνer, the two vaήant future forms may differ in aspect. Consider the following examples: (20)

ΣΩ. κινε\ται κα1. pε\, rος φατε, τc:'χ πάντα; ή γάρ; θΕΟ. Ναί. ΣΩ. Ούκο\>ν άμφοτέρας &ς διειλόμεθα κινήσεις, φερόμενα τε \

και

άλλοιούμενα; θΕΟ. Π&ς δ' οϋ; εϊπερ γε δη τελέως κινήσεται (Pl. Tht. 182c) SOCRAΊΈS: According to you, all things move and flow; isn't that so? 'fHEODORUS: Yes. SOCRAΊΈS: And they have both the motions that we distinguished, that is to say, they both move and alter? 'fHEODORUS: That must be so, if they are to be wholly and completely in motion.

346 The verbs that display this type of contrast are enumerated and discussed by Blass (1892: 273ff.).

200

(21)

CΗΑΡΊΈR

4

ΞΕ. Κίνησίς τε στήσεται κα1. στάσις αi> nνηθήσεται · περ1. γap

άμφότερα

θάτερον

όποτερονουν

γιγνόμενον

αύτο1ν

άναγκάσει μεταβάλλειν αi> θάτερον έπι τούναντίον τflς αύτου φύσεως, aτε μετασχον του έναντίου (Pl. Sph. 255a) VISΠOR: Then change would rest and rest would change. In both cases, if either change or rest comes to be either the same or different, then it will force the other to change to the contrary of its own nature, since it will share in its contrary.

In (20), the middle future κινήσεται is imperfective. All things are in constant motion (compare present stem κινε1ται τα πάντα). In (21), however, the transition from rest to motion is meant. This can be infeπed from the coordination of κινηθήσεται with its opposite στήσεται 'will stand still, halt' This means that κινηθήσεται has an ingressive (i.e. 'aoristic') value. It can be concluded that the aspectual contrast of the type φανουμαι vs. φανήσομαι may also be found with other verbs. 347

4.3 Middle is Middle and Passive is Passive Recall that, according to Stahl, the middle future form has a middle meaning, and the passive future form has a passive meaning. In other words, he advocated a dίathetίcal view. Does this consequently lead us to the conclusion that Stahl was simply wrong? The situation appears to be more complicated than that. With many verbs, the middle future does, indeed, have a middle meaning, whereas the passive future has a passive meaning. The opposition is therefore comparable to the opposition between the sigmatic middle aorist and the passive aorist form. This situation is presented in the following schema:

347

Compare a\so examples (3), (5), (6), (7), (9), given in section 4.0.

ΤΗΕ MIDDLE AND PASSIVE VOICES ΙΝ 'ΓΗΕ FuΊ1JRE SτΕΜ

Table 5:

201

Middle and passiνe aorists and futures: σφζω and -σκευάζω348 Future

Aoήst

Active transitive

-εσκεύασα

εσωσα

-σκευάσω

σώσω

Indirect reflexive

-εσκευασάμηv

έσωσάμηv

-σκευάσομαι

σώσομαι

Direct reflexive

-εσκευασάμηv

Body Motion Passive

-σκευάσομαι

έσώθην -εσκευάσθην

έσώθην

σωθήσομαι -σκευασθήσομαι

σωθήσομαι

Here we can observe a neat pattern. The middle future foπns correspond with sigmatic middle aorist foπns, and the passive future foπns coπespond with passive aorist foπns. 349 In other words, this system of diathetical distinctions leaves no room for an aspectual distinction. The importance of diathetical distinctions is even clear in a foπn like φανο\>μαι since it can also be used as an indirect reflexive (albeit only in the compound άπο-). In sum, both aspectual and diathetical distinctions are relevant to the middle-passive future. Blass and Stahl, we have to conclude, were both coπect in their assertions. 350 The contrast between the middle and the passive future is to

This verb only occurs in the compounds κατασκευάζω and παρασκευάζω. In some cases, the opposition between middle and passive future forms does not correspond with that in the aoήst stem completely. For instance, the passiνe aoήst έπείσθην 'Ι was persuaded' coπesponds, as we would expect, with the passiνe future form πεισθήσομαι 'Ι will be persuaded'. However, the form έπείσθηv in the meaning 'Ι obeyed, believed' corresponds with the middle future form πείσομαι 'Ι will obey, believe'. This pattem probab\y reflects aπ earlier stage of the Ianguage, in which έπείσθην and πεισθήσομαι were only used with a passive meaning, whereas έπιθόμην and πείσομαι had aπ intransitive meaning. This o\der pattem was disturbed by the rep\acement of the thematic aoήst by the passive aoήst form (see section 3.2.4). Α similar mismatch between future and aoήst is exhibited by the pair βουλήσομαι vs. έβουλήθην (βουληθήσομαι appears Iate). 350 Recal\ that Stahl mentioned άπαλλάττομαι as aπ example of a diathetical contrast. According to Sιah\, άπαλλάξεται had a volitiona\, middle meaning Ίoskommen, sich einer Sache erledigen', whereas άπαλλαγήσομαι had a nonvolitional, passive-like meaning 'befreit werden, etwas los werden' Sιah\'s distinction is probably coπect. However, it is remarkab\e that he did not take into consideration the more frequent meaning 'depart' since it appears to confirm Sιahl's claim. 348

349

202

CHAP'IΈR

4

be explained differently for each individual verb. The contrast rnay involve an aspectual distinction, as in the case of κινήσομαι vs. κινηθήσομαι, or it rnay involve a diathetical distinction, as in -σκευάσομαι vs. -σκευασθήσομαι. 351 The contrast between φανου­ μαι and φανήσομαι has elernents of both since φανουμαι can express the indirect reflexive rneaning - which is never expressed by φανήσο­ μαι -, and it can have an irnperfective intransitive rneaning. These two rneanings differ frorn the perfective intransitive φανήσομαι with regard to diathesis, and with regard to aspect, respectively.

The rniddle άπαλλάξομαι is used 11 tirnes with the volitional rneaning 'depart' (Hdt. 5.67.2, 7.222, Ε. Ion 524, Hel. 437, Th. 8.83.3, Pl. R. 496e; Xen. An. 5.1.15, 7 .1.4, 7.1.6, 7.6.2, Dern. 24.85), whereas passive future άπαλλαyήσομαtl-χθήσομαι occurs only 2 tirnes with this rneaning (Ar. Αν. 940, Pl. Lg. 642e). The rnetaphoήcal sense 'depart frorn life' is not taken into account here since it is non-volitional (Ε. Jon 847, Ε. Hipp. 356, Med. 878). Jsι Deήved verbs in -άζω, -ίζω, -ύνω alrnost never have middle future forrns with passive or intransitive meaning. The reason for this rule rnay be that the socalled Άttic futures' are rnorphologically less transparent. There are only three exceptions (out of several dozens of insιances): άνασκολοπιο\ιμαι (Hdt. 3.132.2, 4.43.2), and όνειδιο\ιμαι (S. ΟΤ 1500). Of the verbs in -αίνω, however, both -ανο\ιμαι (5 tokens) and -ανθήσομαι (9 tokens) occur. In sorne cases, word length might be a factor that deterrnines the occurrence of -σομαι and -θήσομαι (Wackernagel 1920: 204, Hartrnann 1935: 128). Forrns with monosyllabic sterns such as δοθήσομαι, ήσθήσομαι, pηθήσομαι do not have midd\e vaήanιs with passive rneaning, whereas polysyllabic stems such as άδικήσομαι, βουλεύσομαι, πορεύσο­ μαι, φυλάξομαι do not have vaήant forrns in -θήσομαι. For a collection of data, Ι refer to Magnien (1912), and Hauή (1975).

CHAP'fER

5

5 'Synonymous' Active and Middle Verbs 5.0 Introduction One of the problems that might puzzle students of the Greek middle voice is the co-existence of active and middle verbs that are (near-)synonyms. Well-known examples are:

Of different verbal stems: έθέλω - βούλομαι χαίρω - ilδομαι εtμι - eρχομαι 352 ίκάνω, ϊκω - ίκνέομαι άκολουθέω- επομαι353 πηδάω

- δ.λλομαι

σκοπέω- σκέπτομαι θαυμάζω - θεάομαι

Of the same verbal stem: όρμάω (intr.) - όρμάομαι πειράω - πειράομαι πολιτεύω - πολιτεύομαι The first, unavoidable point that must be made is that absolute synonymy is a very rare phenomenon - if it exists at all (see e.g. Lyons 1996: 67-8, Cruse 2000: 156). In languages, it runs against the pήnciple of economy to have two different expressions for only one meaning. The force of economy shows a strong tendency to maintain 3' 2 Utoublon (1985: 71) observes a difference in inherent aspect between ίέναι and t:ρχεσθαι, especially in the non-indicative moods. The former is unmarked as to inherent aspect, while the latter often has a durative or iterative value. How this distinction relates to their different voices Ι cannot explain. 3' 3 Notice that άκοΜιuθέω 'follow' is a denominative verb (from άκόΜιuθος 'follower, attendant'), cf. also σκοπέω from σκοπός 'watcher' The active voice of these verbs may reflect their oήginal stative meaning, which may be rendered by 'be a follower' and 'be a watcher' respectively.

204

CHAP'ΓER

5

(or to achieve) a divisioη of semaηtic (or pragmatic) labour betweeη two 'syηoηymous' expressioηs, at least to some exteηt. This teηdeηcy is sometimes refeπed to as synonymophobia. The questioη ηοw is how to deal with the 'syηoηymous' active - middle pairs above. The most obvious explaηatioη for the existeηce of these pairs would be to state simply that the active verb does ηοt iηvolve subject-affectedηess, while the middle verb does. This explaηatioη is a pήοή possible, but ηοt very likely. If the differeηce betweeη the active and the middle member of these pairs would be a questioη of abseηce vs. preseηce of subject-affectedηess, it would have beeη hard to coηceive of them as (ηear- )syηoηyms. Ιη other words, the semaηtic distiηctioη betweeη the 'syηoηymous' pairs is probably more subtle thaη that. Take, for iηstaηce, the pair βούλομαι aηd έθέλω (to which Ι will return later). Οηe caηηot simply state: έθέλω is actiνe therefore it lacks subject-a.ffectedness, siηce it is clear that έθέλω, like βούλομαι, implies an elemeηt of meηtal iηvolvemeηt ση the part of the subject. The coηclusioη is that we have to coηsider other types of explaηatioηs for our 'syηoηymous' pairs. The questioη Ι would like to tackle ίη this sectioη is whether or ηοt the active and middle eηdiηgs of these pairs coηtήbute to the overall verb meaηiηg. Ιη pήηciple, οηe caη thiηk of three possible sceηari­ os that could apply to each of these 'syηoηymous' pairs:

Neither verb of the pair involves subject-affectedness, i.e. the middle ending of the middle member is lexicalized and meaningless, possibly a relic of aπ older meaning of the verb.

(i)

the status of the media tantum (1.3.1), Ι have argued this type of explaηatioη. Οηe of the ceηtral objectives of the preseηt study is to aim for a uηified accouηt of all types of middle verbs: oppositioηal middles as well as media tantum. As a matter of priηciple, ίη my view, the burdeη of proof lies ση those who would claim that these two types of middles should be treated separately. Ιη the sectioη ση media taηtum, Ι have also showη that most media tantum beloηg to a restricted ηumber of semantic classes, aηd that ίη all these classes subject-affectedηess is aη elemeηt of their meaniηg. Now if we look at the verbs eηumerated above, we can observe that Ιη

the

agaiηst

sectioη ση

'SYNONYMOUS' ACΠVE AND MIDDLE VERBS 205

they are either verbs of motion, or verbs of mental processes. 354 Both semantic verb-classes involve subject-affectedness. The subject undergoes a chaπge of location, or a change of mental state, respectively. Although one caπnot rule out the possibility that, as a result of a radical semantic change, the oήginal middle meaning of a particular middle verb has disappeared without a trace, one should consider such aπ explaπation only as a last resort. In maπy cases it is possible to discem a subtle semaπtic distinction between active aπd middle 'synonyms', involving a difference in degree of subject-affectedness. This brings us to scenario (ii):

Both 'synonymous' verbs inherently involve subject-affectedness. However, the aspect of subject-affectedness is emphasized by means of the middle inflection.

(ii)

Το illustrate this scenaήo, consider the pairs puppy vs. puppy dog, aπd oak vs. oak tree (from Langacker 1987: 294-7). The lexical items puppy and puppy dog do not differ in truth-conditional content. The two expressions are semaπtically very similar since the meaning component 'dog' is already inherent in the lexical item puppy. In the expression puppy dog, the meaning of dog is included in the lexical meaning of puppy. The same holds for oak vs. oak tree. Nevertheless, although the 'objective', truth-conditional content of the two expressions may be identical, there is a conceptual difference. By using aπ analytic expression like puppy dog or oak tree the semaπtic facets dog and tree are highlighted, because the status of the designated entities as members of the category dog aπd tree is made explicit. 355 Α similar relation may hold for 'synonymous' pairs of active and middle verbs: both expressions imply subject-affectedness since this is aπ inherent feature of the lexical semaπtics. However, the semaπtic facet of subject-affectedness is highlighted by the middle

354

Πολιτεύομαι is the odd one out here. Ι will retum to this interesting verb

below. 355 In the verbal domain, examples would be: English rise up (the adverb emphasizes the upward movement already inherent in the lexical item rise), and the auxiliary do as in You DID see it! and Did you see it?. The auxiliary can be analyzed as designating a maximally schematic state of affairs (cf. Langacker 1991: 205). Therefore, using the auxiliary do emphasizes the abstract element of the actual occuπence of the state of affairs - which is also inherent in the lexical semantics of every verbal predicate.

206

CΗΑΡΊΈR

5

ending. Consequently, there is a subtle difference between the active verb - of which subject-affectedness is aπ implicit facet of the lexical meaning - and the middle verb- in which subject-affectedness is made more salient by means of the middle ending. 356 Απ example of this type of subtle contrast is the active-middle pair ίδε\ν ίδέσθαι, of which the middle variant conveys the notion that a lasting mental impression is left ση the subject (see section 2.1.8). In other words, the middle inflection explicates, and thereby emphasizes the subjectaffectedness inherent in the lexical meaning. As Ι will argue below, the pair βούλομαι - έθέλω can also be accounted for by means of this scenario. While both verbs involve a degree of mental affectedness in their lexical meaning, this element is emphasized in the case of βούλομαι by means of the middle ending of the verb. Finally, we come to scenario (iii): (iii)

Both verbs of the pair involve subject-affectedness. The affectedness of the subject of the active verbs is inherent in the lexical meaning of the verb. There is no demonstrable semantic difference between the active and the middle verb.

This is a scenario Ι deem very plausible since it is in harmony with the claim that the active voice is the unmarked member of the opposition, neutral with respect to subject-affectedness. It is possible that, histoήcally, there had been a semantic distinction (as in scenario (ii)) which had disappeared in the course of time. 357 In other words, for this type of active-middle pairs, semantic factors do not determine the Within the Greek verbal system, a comparable phenomenon is found in the that are formed from inherently stative verbs, the so-called intensive perfects (see e.g. Ruijgh 1996: 670, Rijksbaron 20023 : 38). E.g., the perfect form σεσιώπηκα 'maintain complete silence' (from stative σιωπάω 'be silent') emphasizes the stative aspect that is already inherent in the lexical meaning of the verb. Maldonado ( l 999) descήbes a comparable phenomenon in relation to Spanish se. E.g., (i) lA maestra captό la voluntad de los alumnos ('The teacher captured the student's good will') vs. Es un maestro que se capta la voluntad de los alumnos (Ήe is a teacher that captures [for himself] the student's good will'). The lexical semantics of capturar 'capture' inherently implies that the event is beneficial to the subject. The clitic se is used to highlight the benefit implied by the verb (see Maldonado 1999: 167-8). 356

perfecιs

lt can never be excluded that a speaker of Ancient Greek still felt a subtle semantic nuance (consciously or not) which cannot be detected anymore for Iack of native speakers. 357

'SYNONYMOUS' ACΠVE AND MIDDLE VERBS

207

of the active and the middle verb. We have a strong indication that semantic factors are iπelevant if we can point to other factors that adequately explain the distήbution of the 'synonymous' active and middle verb. Other important factors might be diachronical factors, genre-factors, and stylistic factors. There is a number of phenomena that might be explained in a this way. The first case of active-middle pairs without detectable semantic distinction concerns the following verbs (see also Kϋhner-Gerth, 1: 102; Chantraine 1927; Schwyzer & Debrunner, Π: 232-3): distήbution

(prose/poetry) - άκούομαι (poetry) 'hear' (poetry)- αύδάομαι (poetry) 'speak' βρέμω (poetry) - βρέμομαι (poetry) 'roar' δακρύω (prose/poetry) - δακρύομαι (poetry) 'cry' διώκω (prose/poetry) - διώκομαι (poetry) 'pursue' δωρέω (poetry) - δωρέομαι (prose/poetry) 'give, present' ίμείρω (esp. poetry) - ίμείρομαι (poetry/Ion. prose) 'desire' κλαίω (prose/poetry) - κλαίομαι (poetry) 'lament' λάμπω (prose/poetry) - λάμπομαι (poetry) 'shine' μέλπω (poetry) - μέλπομαι (poetry) 'celebrate with song and dance' νήχω (poetry) - νήχομαι (poetry) 'swim' ό"iω (poetry) - ό"iομαι/οϊομαι (prose/poetry) 'think' πέλω (poetry) - πέλομαι (poetry) 'become, be' 358 σπέρχω (prose/poetry) - σπέρχομαι (poetry) 'haste' σπεύδω (prose/poetry) - σπεύδομαι (poetry) 'speed' φημί (prose/poetry) - φάτο (esp. poetry) 'say' άκούω αύδάω

Of these verbs, either one of the variants is restricted to poetry (mostly the middle vaήant, e.g. άκούω άκούομαι), or both vaήants occur only in poetry (e.g. αύδάω - αύδάομαι). In sum, oppositions between the active and the middle variant of these verbs is only found in poetry. Απ example may be illustrative: (1)

a.

αύτ(φ έγώ γε

νηχόμενος τόδε λαiτμα διέτμαγον (η 276) But by swimming Ι made my way through this great gulf of the sea here (... ). b. άλλ' άναχασσάμενος νfιχον πάλιν (η 280)

358 Neuberger-Donath (1980) argues, against Schwyzer, that there is a semantic distinction between πέλω and πέλομαι. Ίbis distinction, according to NeubergerDonath, is in harmony with the general definition of the middle: πέλω expresses a static event and is 'extroverted' (related to είμί), whereas πέλομαι is dynamic and 'introverted' (related to γίγνομαι).

208

CΗΑΡΊΈR

5

(... ) but Ι retreated, and swarn back (... ).

Within a short passage the poet uses the verb twice; once in the middle voice, once in the active. There are no indications of a semantic difference. With Schwyzer & Debrunner (Π: 232), Ι assume that the coexistence of the active and the middle verbs in poetry is a matter of metrical convenience, and that the variants are used indistinctly. 359 However, this assumption does not imply that, in these verbs, the use of voice is completely arbitrary from a semantic point of view. On closer inspection of the list above, it can be observed that the verbs almost all belong to a restήcted number of semantic classes: verbs of motion, verbs of perception, (emotional) speech act verbs, and verbs of mental (emotional) processes. In other words, all verbs inherently involve subject-affectedness. 360 This inherent subject-affectedness motivates the presence of the (semantically redundant, cf. Schwyzer & Debrunner's "Doppelcharakteήsierung") middle inflection. The active inflection can be used in these verbs since it is, as Ι have argued, unspecified as to the feature of subject-affectedness. The second case concerns verbs that have different voices in different aspect stems, e.g.:

Α different position is often taken by Boeder (1961-2). Wit regard to many of the active-middle pairs at issue, he claims that there is a semantic distinction. According to Boeder (p. 78), the middle δωρέομαι involves a degree of generosity on the part of the subject; as to ο{ω - ο{ομαι, the rniddle conveys a (more subjective) suspicion, whereas the active expresses "die Feststellung einer Meinung, fast eine Behauptung" (p. 130); ίμείρομαι, against ίμείρω, expresses the desire as α reαction to α stimulus (p. 131). Cf. further διώκω/-ομαι (p. 135-6); πέλω/-ομαι (p. 163-8). 359

360 Λάμπωl-ομαι can be semantically compared to φαίνομαι 'become visible', both denoting a spontaneous intemal affection.

'SYNONYMOUS' ACIΊVE AND MIDDLE VERBS 209

Middle present • active aoήst άλίσκομαι - έάλων 'be caught' δέρκομαι - eδρακον 'look' δύομαι - eδυν 'enter, plunge into' ϊσταμαι - eστην 'stand still, stand up' πέρδομαι- eπαρδον 'break wind' πτάρνυμαι- eπταρον 'sneeze' τρέφομαι- eτραφον (Hom.) 'grow up' φύομαι - eφυν 'grow; become' 361 Active present • passive χαίρω - έχάρην 'enjoy'

aoήst

The explanation of this remarkable active middle pattern probably lies in the pre-history of Greek. 362 lt is clear that the middle forms Other possible examples are: δίεμαι - δίον (Χ 251); έρεικόμενος (Ν 441) fίριπον; έρεύγομαι fίρυγον; λάζομαι [Hom.] - eμυκον; κίνυμαι - eκιον [Hom.]. Note, further, that most of these verbs have an active perfectum, e.g. έάλωκα, δέδορκα, δέδυκα, eστηκα, πέπορδα, τέτροφα, and πέφυκα. 362 Though Gonda (1979: 90) notes that in Sanskήt the active voice predominates in the root aoήst, there is no clear evidence that this pattern was inheήted from ΡΙΕ, unless one counts as evidence the Sanskήt middle pardate 'breaks wind' (coπespon­ ding with πέρδεται) and the uncertain Late Avestan active aoήst form paraδan (V. 3.32) 'they broke wind' (coπesponding with eπαρδον). In the remaining cases the present middle appears to be a Greek innovation. Thus, ϊσταμαι and πτάρνυμαι coπespond with the active presents Skt. tίsthati 'stand', Lat. sistό 'stand', and Lat. stemuό 'sneeze', respectively. As to δέρκομαι and φύομαι, these verbs probably did not have a present form in PIE, given that in Sanskήt only the aoήst formations seem to be old: dάrsam Ί see' (cf. (e-)δρακον) and άbhat 'has become' (cf. eφυ). As an explanation of the occurrence of an active aoήst (and perfect) vs. a middle present, Ι tentatively suggest the following process: (l) Sigmatic aoήsts with causative meaning (e.g. eστησα, eθρεψα) evolved in opposition to the intransitive active aoήsts (e.g. eστην, eτραφον). (2) This led to the development of coπespon­ ding causative active present formations (ϊστημι, τρέφω), or, altematively, the already existing active formation acquired a causative meaning (e.g. ϊστημι *'stand still' > 'make stand' (3) Then, intransitive middle presents (ίσταμαι, τρέφομαι) were formed. This analogical process is not implausible, since an almost identical analogical process led to the formation of the sigmatic aoήst είσα Ί made sit down' (in opposition to the intransitive middle aoήst έζόμην Ί sat down'), which, in tum, brought about that active present ϊζω acquired a causative meaning 'make sit down', in addition to its oήginal intransitive meaning 'sit down' (cf. Lat. -sϊdo 'sit down'). Now, as a result of the analogical process, we find oppositions such as ϊσταμαι/τρέ­ φομαι vs. eστην/eτραφον. These pairs may have constituted a precedent for the formation of other verbs, e.g., eτραφον τρέφομαι = eδρακον >> δέρκομαι (also 361

fίρικε (Ρ 595); έρείπομαι eλλαβον; μυκάομαι (κ 413)

21 Ο

CΗΑΡΊΈR

5

have a middle meaning. They all expres mental or physical affectedness. The active aoήsts (or active present, in the case of χαίρω), however, do not seem to express a lower degree of affectedness. 363 The subject-affectedness, therefore, must be an inherent property of the lexical meaning of the verb root. The unmarked active voice can be used since the subject-affectedness is already present in the context (in this case the verb root). This phenomenon may be compared to the use of the active άποθν~σκω (ύπό) 'be killed by' as the suppletive passive of κτείνω 'kill' (see section 1.1.4). The subject-affectedness is an inherent property of the lexical meaning of άποθν~σκω. In the two cases Ι have discussed above, there was no indication that any semantic distinction existed between active and rniddle verbs. This phenomenon was explained by means of the neutral value of the active voice, and the inherent lexical middle semantics of the verbs concemed. In the first case mentioned above (type άκούω άκούομαι), the distήbution between active and middle form was explained by genre-factors (the need of metήcal altematives in poetry). έάλων) was The second case of active-middle pairs (άλίσκομαι explained by the pre-history of the Greek language.

5.1 Some Active and Middle 'Synonyms' In the previous section three possible scenaήos that rnight hold for active - rniddle 'synonymous' pairs were discussed. In this section, a number of case-studies will be presented. For five of the active rniddle pairs, Ι will try to analyze the factors that explain their distribution. These may be semantic factors, but also diachronical or stylistic factors. Three of the pairs concemed are built on the same lexical stem: όρμάω - όρμάομαι (5.1.1), πειράω - πειράομαι (5.1.2),

inflύenced by σκέπτομαι?). The active χαίρω is probably old, witness Skt. hάryati. The fonn έχάρην probably arose when the η-aoήst fonnation did not yet stήctly belong to the verbal conjugational system, and could also be deήved from verb roots with active presents, cf. also pέω 'stream', aor. έρρύην. 363 We cannot exclude (nor prove) that a speaker of Ancient Greek expeήenced a subtle semantic nuance between the rniddle voice of δέρκομαι and the active voice of tδpακον. However, then the question aήses as to why the present stem would involve a affectedness, whereas the aoήst stem would not.

'SYNONYMOUS' ACΠVE AND MIDDLE VERBS 211

πολιτεύω πολιτεύομαι (5.1.3). 364 The two remaιnιng pairs are from different stems: &λλομαι - πηδάω (5.1.4) and βούλομαι - έθέλω (5.1.5). After a discussion of these pairs Ι will devote some attention to the domain of the verbs of motion (section 5.2). It will be argued that it is possible to discem a tendency behind the distήbution of active and middle verbs of motion.

5.1.1 όρμάω vs. όρμάομαι The first near-synonymous pair to be discussed, consisting of an active intransitive and a middle intransitive verb, is όρμάω 'rush, set off (intr.)' and όρμάομαι 'set off'. 365 In Homer, the two verbs seem to be used without a clear distinction in meaning. Compare the two following examples: (2)

αύτος δ' rος τ' ϊρηξ ώκύπτερος c1pτo πέτεσθαι, ος pά τ' άπ' αίγίλιπος πέτρης περιμήκεος άρθε1.ς όρμήτοι τaς νέας τrον βαρβάρων ές φυγftν ίδρμησαν (Hdt. 7 .179) These, when they sighted the foreigner's ships, took to flight.

(8)

[Demaratus to the Persians] "Ηκουσας μεν καi πρότερόν μευ, εi">τε όρμrομεν έπi την 'Ελλάδα (... ) (Hdt. 7.209.2) Ι

have told you already (... ) when we were marching off to Hellas.

(9)

έκ μεν δη τftς άντίης προσπλέειν οϋ κώ σφι έδόκεε τrονδε εϊνεκα, μή κως ίδόντες οί 'Έλληνες προσπλέοντες ές φυγftν όρμήσειαν (... ) (Hdt. 8.6.2) Now they were not yet minded to make an onfall front to front, for fear lest the Greeks should see them coming and take to flight (... ).

(Ι Ο)

[Pausanias to the Athenians about the Persian cavalry] εί μέν νυν ές ύμέας ίδρμησε άρχην ή ϊππος (... ) (Hdt. 9.60.2) If the cavalry would have rushed up to you first (... ).

Twice, the verb is construed with aπ infinitive. 367 In both cases, we are dealing with armies marching (cf. έξελαύνειν and στρατεύεσθαι).

(11)

[Cyrus] πρiν δε έξελαύνειν όρμilσαι τον στράτον, πέμψας κήρυκας ές τους 'Ίωνας έπειρίiτό σφεας άπο Κροίσου άπιστάναι (Hdt. 1.76.3) But before marching off in order to lead out the army, he sent heralds to the

367 The infinitive may be interpreted as an optional final infinitive, or as a complement depending on a semantic element of volition present in the verb όρμάω, cf. Homeric ϊεμαι 'rush', but with inf. 'wish, be eager to'

214

CHAP'ΓER

5

Ionians to try to draw thern away frorn Croesus.

(12)

εστι δε &λλος λόγος (... ) ώς Ξέρξης επεμψε κήρυκα ές "Αργος πρότερον η περ όρμησαι στρατεύεσθαι έπι τi]ν 'Ελλάδα (Hdt. 7.150.1) (... ) but there is another story told in Hellas: That before Xerxes set out to march against Hellas, he sent a herald to Argos (... ).

In 4 of the instances cited above, the speed of the movement is essential, either in attack ((6) and (10)) or in flight ((7) and (9)). In the remaining 3 instances ((8), (10), and (12)) the verb means 'march, advance (against)' In these cases, speed may not be important. Instead, the element of (military) force is relevant. 368 That speed is often essential to the semantics of όρμάω may also be illustrated by two examples from Thucydides, where όρμάω is modified by the adverbs μ&λλον and μάλιστα, thereby showing its gradable character. This would be impossible if the verb would merely mean 'set off, start'. (13)

[Brasidas] χειμrον δε ήν και ύπένειφεν · ~ και μ&λλον ίόρμησε ( ... )

(τh.

4.103.2)

The weather was bad and somewhat snowy, and for this reason he made the rnore haste ( ... ).

(14)

[of ships] αί μεν μάλιστα όρμήσασαι τρεϊς διαφθείρονται (τh.

8.34)

The three that had pursued most hotly were wretched.

Let us now tum to intransitive όρμάομαι. As Ι claimed above, its meaning is neutral as to speed and/or force. Instead, it focuses on the onset of the motion. Το support this claim - without going through all 67 instances in Herodotus Ι would like to point out that it is highly significant that όρμάομαι, 59 times out of 67, is accompanied by an adverbial phrase with έκ or άπό, expressing the starting point of the motion (cf. Powell 1938). Furthermore, it is not surprising that the concrete spatial meaning 'set off, start' has been extended to the metaphorical meaning 'begin to', which is construed with an infinitival complement. Το illustrate how different όρμάομαι is from όρμάω, the

Possibly, the rniliιary rneaning 'rnarch, advance', which is neutral as to speed, has developed frorn the rneaning 'storrn, assail' which irnplies both speed and force. 368

'SYNONYMOUS' ACΠVE AND MIDDLE VERBS

first three instances of όρμάομαι

(15)

occuπing

215

in Herodotus may suffice:

[Arion; passage in acc. plus inf.] όρμ(Χσθαι μέν νυν έκ (... ), μισθώσασθαι πλοiον άνδρ&ν Κορινθίων (Hdt. 1.24.2)

Τάραντος,

He set off from Tarentum and (... ) he hired a ship of men of Coήnth.

Arion leaves to retum to Corinth. Since he is in no particular rush όρμάω would not have been appropήate.

(16)

[Croesus to Adrastus] (... ) φύλακα παιδός σε του έμο\> χρηίζω γενέσθαι ές aγρην όρμωμένου (... ) (Hdt. 1.41.2) (... ) Ι ask you to watch over my son as he goes out to the chase (... ).

Croesus' son goes out hunting.

(17)

Νο

rush nor force involved.

[Croesus sends out the Lydians to consult the oracles]

έντει­

λάμενος δε τοiσι Λυδοiσι τάδε (... ) άπ' ής &ν ήμέρης όρμηθέωσι έκ Σαρδίων, άπο ταύτης ήμερολογέοντας τον λοιπον χρόνον έκατοστfι ήμέρ11 χρρος ώς εΙδεν όρμωμένους, έξορμ~. κα1. αύτος πρ&τος ήγεiτο ταχέως, κα1. ό Κυαξάρης μέντοι έφείπετο (... ) (Xen.Cyr. 1.4.20) So when Cyaxares had ιaken some of the most powerful horses and men he advanced. And when Cyrus saw them leaving, he rushed off and soon took the lead, while Cyaxares followed after (... )

216

CΗΑΡΊΈR

5

Cyrus sees Cyaxares and his men leaving (όρμάομαι), and speeds away (όρμάω) in order to outstήp them. Το summaήze, όρμάω can be adequately rendered by 'rush (up to)' or 'march (against)'; όρμάομαι can be rendered by 'start off, leave'. 369 The question remains how these semantic differences between the two verbs relate to their respective voice markings. My explanation runs as follows: the inherent lexical semantics of both active όρμάω and middle όρμάομαι implies the affectedness of the subject, since in both cases the subject undergoes the effect of the action, viz. a change of location. However, in the case of όρμάομαι the subject-affectedness is conceptualized as stronger: the verb focuses on the moment in time at which the subject undergoes a change from a stationary position into motion. Thus, όρμάομαι is a punctual verb. This change is cognitively more salient than that expressed by όρμάω, which focuses on the ongoing (and essentially unchanging) motion at high speed (atelic/durative Aktionsart). In sum, the middle inflection of όρμάομαι highlights the relatively salient change undergone by the subject. As a consequence, scenario (ii) above is applicable: both verbs involve a degree of subject-affectedness, but in the case of middle όρμάομαι, the affectedness of the subject is emphasized.

369 It is notewonhy that the semantic distinction between όρμάω and όρμάομαι appears to be a reflection of the semantics of the noun όρμή from which both verbs are deήved. In LSJ, we find the following meanings of όρμή: (I) rapid motion forwards, onrush, onset, assault; (ll) impulse to do a thing, effΌrt; (ΠΙ) setting onself in motion, start. 'Ορμή (I) coπesponds with the όρμάω 'rush; march against'; όρμή (ΙΠ) coπesponds to the middle όρμάομαι 'start off; όρμή (ll) is of a more metaphoήcal nature, and coπesponds though Ι have not elaborated on this point largely to the middle όρμάομαι. Incidentally, there seems to be aπ interesting distinction between άπαλλάττομαι and όρμάομαι. Though they both denote the departure from some starting point, the difference lies in the nature of the starting point and the goal of the movement. That is, άπαλλάττομαι seems to designate going away from a place where something has happened or something is done by the person who leaves (e.g. leave after battle), whereas όρμάομαι implies that the subject leaves with the intention of achieving a specific goal elsewhere (e.g leave to go to battle). 'This may explain why όρμάομαι has developed the metaphorical, psychological sense of 'be eager to, be minded to' (often in the perfect stem), which focusses on the intention of the subject while backgrounding the element of spatial motion.

'SYNONYMOUS' AcηVE AND MIDDLE VERBS 217

5.1.2 πειράω vs. πειράομαι Ι will continue now with a discussion of the pair πειράω - πειράομαι. In Homer, the active occurs 10 times, against 66 instances of the middle (including έπειρησάμην and έπειρήθην). lt is possible to distinguish a number of different meanings of Homeήc πειράω πειράομαι. As we shall see, the active and middle verbs are, to some extent, distributed differently with respect to the variant meanings. The active verb πειράω exhibits three meanings in Homer ('Α' as in ΆΙ' stands for 'Active'):

Al. make ( 19)

aπ

attempt on by words (+ gen. [human]) [2xl

μή μευ πειράτω εi> είδότος. ούδέ με πείσει (I 345) [πειράσθω metήcally equivalent] Let him not tempt me who know it well; he will not persuade me.

Cf. also:

Α2.

(20)

Ι

181.

try by force (+ inf./+ ώςj+ oπως-clause 370) [6xl [Telemachus to suitors] πειρήσω rος κ' ϋμμι κακaς έπi κίlρας ίήλω (β (... ) Ι will try to send upon you fates of evil death ( ... ).

316)

Cf. also: Δ 66=71, Θ 8 [πειράται; πειράσθω metr. equiv.], Τ 30, δ 545 [πείρα; πειρ& metrically equivalent].

370 1t is hard to tell whether the ώς/oπως-clause is a complement or an optional satellite. The fact that πειράω does not occur with ϊνα nor οφρα (introducing optional final clauses) may be an indication that the oπως-c\ause is complementary.

218 Α3.

(21)

CΗΑΡΊΈR

5

make an attempt on by force 371 (+ gen. [animal]) [2xl κ:έλεται δέ έ θυμος άyήνωρ μήλων πειρήσοντα κ:αi ές πυκ:ινον δόμον έλθεiν (Μ 301= ζ

134) And his proud spiήt tells him to go even into the well-built fold to make an attempt on the flocks.

Now the middle variant variants:

Μ1.

try (mostly by force) (+ inf./+

This meaning is similar to

(22)

πειράομαι

shows five meaning

ώς-clause/ellips)

[8xl

(Α2).

[Trojans at the gates of the Greek camp] τοi δε κατ' αύτaς

ίστάμενοι πειρίΟντο βίτι pήξαντες έσελθεiν (Μ 341) And before them stood the foe, and tήed to break them by force and enter in.

Also: Δ 5372 , Ε 279 [abs.], Ζ 435 [abs.], Ν 457 [abs.], Φ 459 [πειρίf; πειρίfς metr. equiv.], Χ 381 [abs.], δ 417 [sc. άλύξαι]

Μ2.

make an attempt on by words (+ gen. [human]) [1xl

This meaning is similar to (23)

(Α1).

πειρίf έμεtο, γεραιέ, νεωτέρου, ούδέ με πείσεις (Ω 433)

You make tήal of me, old sir, a younger man than you; but you will not persuade me.

371 Probably, this meaning comes closest to what must have been the oήginal meaning of πειράω. The noun πε"iρα (< *per-yh2 ) is built on the root *per- 'penetrate' Hence, the oήginal meaning of the noun must have been something like 'penetration (through some barrier) by force' Yet, the extended meaning 'test, expeήence' may also be quite old, witness Latin experior 'test, expeήence'. 372 Δ 5: αύτίκ:' έπειρίiτο Κρονίδης έρεθιζέμεν 'Ήρην/ κ:ερτομίοις έπέεσσι ( 'And immediately the son of Cronos attempted to provoke Hera with mocking words'). Although no physical force is inνolved, πειράομαι here still appears to have an aggressive and hostile nuance.

'SYNONYMOUS' ACΠVE AND MIDDLE VERBS 219

Μ3.

(24)

test, trv (mostly by force) (+ gen. [humanl/ellips) [24xl [Zeus challenges the gods to a tug-of-war] γνώσετ' επειθ' οσον είμ1. θε&ν κ:άρτιστος άπάντων. εί δ' aγε πειρήσασθε, θεοί, ϊνα εϊδετε πάντες (Θ 18)

[sc.

μευ?]

Then you will recognize how far the mightiest am just try it, you gods, so that you all may know.

Ι

of all gods. But come on,

That the action has the character of a test (the subject desires to know whether the genitival object is capable of something, cf. English try me), is shown by the many cases in which the implicit question underlying the test is made explicit. In the example above, Zeus trusts that the trial of strength will reveal his supeήority to the gods (γνώσετ' tπειθ' and ϊνα εϊδετε ... ). Other examples are: (25)

[Achilles challenges Agamemnon to take away his gifts] εί δ' aγε μην πείρησαι, ϊνα γνώωσι κ:αl. οϊδε.

αtψά τοι αtμα κ:ελαινον έρωήσει περ1. δουρί (Α 302) Come on, just try it, so these men here also may know: at once your dark blood will spurt around my spear.

Here the expected 'test-result' (ίνα γνώωσι κ:α1. οϊδε ·) is made explicit in the next line (αtψα περ1. δουρί).

(26)

οφρ' tτι κ:α1. Τρώων πειρήσομαι άντίον έλθών, αϊ κ:' έθέλωσ' έπ1. νηυσ1.ν ίαύειν σ 70) (... ) so that Ι may go out against the Trojans and make tήal of them yet again, whether they are minded to spend the night by the ships.

The question to be answered is αϊ κ:' έθέλωσ' έπl. νηυσ1.ν ίαύειν. 373

(27)

πάνηι δ' άμφ1. φάλαγγας έπειρ(Χτο προποδίζων, εϊ πώς οί εϊξειαν (Ν 806) And everywhere he made tήal of the battalions, moving forward step by step, to see if they would perhaps give way before him.

373 1t is also possible to interpret the εi-clause, not as aπ indirect question 'in order to see whether .. .', but as a conditional clause: Ί will test the Trojans, if they will wish, etc.'

220

CΗΑΡΊΈR

5

In this example, the added indirect question (εί + oblique optative) shows that this meaning is related to that of (Μ5) questίon. 374 Other examples with explicit 'test-result' are Κ 444, Φ 225, Ψ 804, π 305, ω 216. In many cases the test-character of the action is not explicitly expressed. However, in these cases, too, we can assume that middle πειράομαι conveys (albeit implicitly) this semantic nuance. E.g.: (28)

και έκτελέσειεν άέθλους πολλούς, τους Φαίηκες έπειρήσαντ' 'Οδυσίlος (θ

22-3)

[τους: intemal accusative] (... ) and that he might accomplish the many feats in which the Phaeacians made tήal of Odysseus.

Cf. also: Β 73 [sc. υί&ν Άχαι&ν], Β 193 [sc. υί&ν Άχαι&ν], Ε 129 [sc. σευ], Ε 220 [sc. Diomedes], Λ 386 [sc. μευ], Π 590 [abs.], Υ 349, Υ 352 [= Τ 70], Φ 580, Ψ 553, θ 120 [sc. one another], θ 126, θ 205 [sc. μευ], θ 213 [sc. everyone], θ 377 [abs.].

Μ4.

(29)

try, test, investigate (+ inanimate gen.) [23xl οφρα μεθ' ύμ\ν χειρ&ν και σθένεος πειρήσομαι, η μοι ετ· εστιν ϊς, οϊη πάρος εσκεν ένι γναμπτοισι μέλεσσιν'

η f1δη μοι ολεσσεν &λη τ' άκομιστίη τε (φ 282) (... ) that in your midst Ι may prove my hands and strength, whether Ι have still vigour such as was formerly in my supple limbs, or whether by now my wandeήng and lack of food have destroyed it.

The disjunctive indirect question expresses the question that is to be answered by the test. The genitival complement can express the particular quality of the subject that is tested (as in φ 282 above: χειρ&ν και σθένεος), or it can express the test-object by means of which one tήes one's powers, e.g. του τόξου in:

374 In many cases, εί-clauses combined with πειράομαι are ambiguous. They can be interpreted both as a final clause ('in the hope that') and as an indirect question (see Wakker 1994: 370-l).

'SYNONYMOUS' (30)

ACΠVE AND

MIDDLE VERBS 221

και δέ κεν αύτος έγrο του τόξου πειρησαίμην.

εί δέ κεν έντανύσω διοϊστεύσω τε σιδήρου (φ 113) And Ι too myself might make tήal of that bow. If Ι shall stήng it and shoot an aπow through the iron (... ).

Again, the question that is to be answered by the test is expressed explicitly (εί δέ κεν έντανύσω). Cf. also: Ο 359, Σ 601 375 , Τ 384, Ψ 432, Ψ 707=753=831, ζ 126 [πειρήσομαι ήδε ϊδωμαι; sc. the unknown people], θ 100, θ 145, θ 149 [sc. άέθλων], θ 184, σ 369, φ 135, φ 159 [cf. ζ 159], φ 180=268, φ 184 [sc. έντανύσαι?], φ 394, φ 410.

Μ5.

(31)

test by questioning (+ gen./ellips)

[8χ)

έλθrον τ&νδ' άνδρ&ν πειρήσομαι, οϊ τινές είσιν (ι will go and make tήal of these men, to leam who they are (... ).

174)

Ι

Also: Ω 390, δ 119 [sc. Telemachus] = 215, ψ 181, ω 240 [sc. Laertes].

Μ6. expeήence

(32)

ω

238 [sc. Laertes],

ν

336,

τ

(+ gen.) [2xl

το\. δ' &λλοι κοίλης eντοσθε φαρέτρης κείατο, τ&ν τάχ' eμελλον Άχαιοi πειρήσεσθαι (φ 418) But the others [aπows] were stored within the hollow quiver, those of which the Achaeans were soon to get expeήence.

Also: γ 23 [abs.; perfect: 'be Greek].

expeήenced';

frequent in Classical

The different meanings of πειράω and πειράομαι are summarized in the table below. We may obserνe that some meanings

s At Σ 601 πειράομαι seems to be construed wiιh an accusative object: ώς οτε τις τροχον aρμενον έν παλάμησιν/ έζόμενος ιcεραμεi:ις πειρήσεται, αϊ ιcε θέJΊσιν. 31

lt is, however, preferable to interpret the accusative as a thematic object in a proleptic construction or as an anacoluιhon in which έν παλάμησιν [χων is to be understood.

222 CHAP'fER 5

occur exclusively with one particular forrn, and that there is an overlap in the meaning try (by means of force). Table 1: Different meanings of πειράω and πειράομαι in Homer πειράω

make an attempt on by fon:e (+ gen. [animal])

[lOx]

πειράομαι [66χ]

Α3.

2

A2/Ml. try (mostly by fon:e) (+ inf/+ ώς/+ οπως)

6

8

Al/M2. make an attempt on by words (+ gen. [human])

2

ι

test (mostly by force) (+ gen. [human])

24

try, test, investigate (+ gen. [inanimate])

23

test by questioning

8

Μ3.

Μ4.

Μ5.

Μ6. expeήence

2

Now what can we learn from the table? First, the meanings test (+ gen.), test by questioning, experience are exclusively expressed by the middle verb. The active forrn, on the other hand, occurs exclusively in the meaning: make an attempt on by force (be it only in two identical lines). 376 Subsequently, the question arises as to the rationale behind The question may aήse whether it is justified to distinguish the meanings + gen. [animal] 'make an attempt on Χ by force' from πειράομαι + gen. [human] 'test Χ' (mostly by means of force). Do they not, instead, express the same meaning even though they differ in voice? My answer would be negative. lt is significant that in the sole instance of the active form construed with a genitive complement, this complement is non-human: μήλων (Μ 301 = ζ 134), whereas in the 23 instances of the rniddle verb with a genitive complement, the complement is always human. Ι would suggest that this pattem is not incidental: attacking sheep can, from the perspective of a lion, hardly be conceived of as a test or a trial of strength, since one cannot expect substantial resistance from such an unequal opponent. Conversely, in the case of a human attacking a human, the outcome is uncertain, and the fight will have the character of a mutual trial of strength. The middle voice here conveys the nuance of a test of each other's powers. Victory or 376

πειράω

'SYNONYMOUS' ACΓIVE AND MIDDLE VERBS

223

this distribution. Ι would like to suggest that there is indeed one: the meanings that are exclusively expressed by the middle form each involve an element of cognition, i.e. the subject acquires information as a result of the event, either by testing or by questioning the object. Or the subject may acquire information about the object in a nonvolitional way. Conversely, the active voice seems to be neutral as to the aspect of cognition. As was noted above, both the active and the middle voice occur ίη the meaning try (by means of force) to Χ (+ inf.l+ ώςΙ+ oπωςlellipsis). It is possible, though hard to prove, that the middle voice conveys the semantic nuance that the subject acquires knowledge as a result of the attempt, which then can be considered as a trial of strength. In this case, the meaning of πειράομαι (+ inf./+ ώςΙ+ oπωςJellipsis) can contain the same element of cognition as the meanings (Μ3) and (Μ4) test (by means of force) somebody/something. Thus, in example (22), το\. δΕ: κατ' αύταςΙ ίστάμενοι πειρ&ντο βί11 pήξαντες έσελθεiν (Μ 341), the middle may express that the Greeks, as a result of the attack, know whether or not they are strong enough to break in. Furthermore, the availability of two variant forms, active and middle, to express the frequently occurring meaning 'try to' is obviously convenient for metrical purposes. Ι will not go into the opposition between πειράω and πει­ ράομαι in Classical Greek in much detail, since the middle πειράομαι has almost been completely generalized at the cost of the active verb (e.g. Herodotus: 3χ πειράω vs. 88χ πειράομαι). Instead, Ι will briefly discuss Thucydides, the only classical author in which the opposition is still alive. In Thucydides, too, it appears that the element of cognition is relevant to the distribution of πειράω and πειράομαι. Thucydides consistently uses the middle form to express the meaning 'experience', where the element of cognition is apparent. 377 On the other hand, he uses the active verb to express the cognition-neutral meanings: (i) 'seduce' 378 ; (ii) 'try something' 379 ; (iii) 'make an

defeat is the "test-result".

Α

nice example illustrating this idea is

Φ

225:

ιcα\. 'Έκτορι

πειρηθηναι/ άντιβίην, η ιcέν με δαμάσσεται, η ιcεν έγrο τόν ['Έιcτορι to be

construed with άντιβίην] (Άnd to make tήal, man to man with Hector, whether he will slay me or Ι him'). 377 E.g., Th. 2.44.2, 2.85.2, 4. 108.6, 4.114.4, 5.111.1. 378 On1y Th. 6.54.3; in fact a passive aoήst to πειράω. 379 With interna1 accusative: πaσαν ίδέαν (2.19), πολλά (6.38.2).

224

CΗΑΡΊΈR

5

attempt on something by force' 380 • Further, he uses the active with an infinitive-complement, always with the special nuance 'try to V by means of force'. 381 Conversely, the middle πειράομαι plus infinitive 'try to V' is neutral as to the element of force, for example, ( ... ) τοtς δε φιλίαν καi ξυμμαχίαν πειρώμεθα ποιεtσθαι (τh. 6.34.1) ('( ... ) and to endeavour to make friendship with others'). 382 Clearly, ίη the case of πειράομαι + inf. ίη the neutral meaning 'try to V', the semantic element of cognition is somewhat diluted. lt may, however, still be present, since ίη attempting to do something, the person who makes the attempt will always be cuήous as to the outcome. Το conclude our discussion of πειράω/πειράομαι, the semantic distinction we have found between active πειράω and middle πειράομαι - the semantic element of cognition - can conveniently be connected with the general meaning of the active and middle voices. Middle πειράομαι expresses that the subject-referent is affected ίη that he undergoes a mental change. In this manner, the verb is semantically related to cognition-verbs like πυνθάνομαι 'hear/leam that; ask' and αίσθάνομαι 'perceive' Conversely, the active πειράω is neutral with respect to subject-affectedness. 383 πειράω

With genitiνal object: Th. 4.25.10, 6.63.2, 7.37.3, 4.100.1, 2.77.2. E.g., πειράσαντες ιcαταλαβε\ν την πόλιν (Th. 2.5.5) ('haνing tried to seize the city'). Cf. also 2.72.2, 4.128.1, 4.9.2, 4.102.1, 7.32. 382 The other instances of πειράομαι + inf. (37) 1.25.1, 1.71.7, 1.81.3, 1.143.1, 1.144.4, 2.33, 2.35.3, 2.65.1, 3.38.2, 3.39.1, 3.54.1, 3.63.1, 4.15.2, 4.41.3, 4.49.4, 4.60.2, 4.61.2, 4.66.4, 4.92.4, 4.92.7, 4.117.1, 4.126, 5.35.5, 5.69.1(bis), 5.112.2, 6.18.6, 6.34.1, 6.38.4, 6.40.6, 6.48, 6.87.3, 6.88.7, 6.89, 8.3.5, 8.63.3, 8.76.6. The fact that the middle πειράομαι + inf. is neutral with respect to the element of (military) force does not exclude the νerb being used in a context in which force is inνolνed. These cases resemble the use of the actiνe form, e.g., άφιιcόμενοι δe μηχανάς τε τ~ Ποτειδαί~ προσέφερον ιcα1. παντ1. τρόπφ έπειρό':ιντο έλε\ν (Th. 2.58.1) (Άnd on their arήνal they brought siege-engines to bear upon Potidaea, and tried in eνery way to take it'). 383 The semantic distinction between πειράω and πειράομαι seems to be a reflection of the semantics of the noun πε\ρα from which both νerbs are deήνed. In LSJ we find the following meanings: (I 1) trial, auempt and experience; (l 2) experiment; (11) auempt on or against one and (abs.) auempt, enterprise. lt appears that πε'iρα (I) coπesponds roughly with the middle νerb, while πε\ρα (11) coπe­ sponds with the actiνe νerb. Note further that the semantic distinction between πειράω and πειράομαι resembles the distinction between ίδε\ν and ίδέσθαι as demonstrated by Bechert (1964): middle ίδέσθαι expresses that a 1asting mental impression is made on the subject. 380 381

'SYNONYMOUS' AcnVE

5.1.3

πολιτεύω

vs.

AND

MIDDLE VERBS 225

πολιτεύομαι

According to Kϋhner-Gerth, the active πολιτεύω means 'bin πολίτης', whereas the middle means 'verwalte den Staat', and the passive 'lebe unter einer Verfassung, in e. Staate' (1 2: 525). In the aorist stem, the three forms are, indeed, neatly distinguished: έπολίτευσα, έπολιτευσάμην, έπολιτεύθην 384 . However, the semantic distinction between active πολιτεύω (aor. έπολίτευσα) and middle πολιτεύομαι (aor. έπολιτευσάμην) is not as cut-and-dry as Kϋhner-Gerth's remark suggests. In fact, the opposition appears to have a very marginal existence. Only Xenophon uses both the active (6 times) and the middle form with middle meaning (about 10 times). Apart from Thucydides, no other classical authors use the active verb. Thucydides, on the other hand, only uses active πολιτεύω 'be/act as a citizen' (17 times), and never the middle form. Only the passive aorist (with passive meaning) occurs once. This situation is summarized in the following table: Table 2: The occurrence of πολιτεύω and ACτiVE

MIDDLE

PASSIVE

πολιτεύω (έπολίτευσα)

πολιτεύομαι (έπο λιτευσάμην)

πολιτεύομαι (έπολιτεύθην)

Xenophon

+

Thucydides

+

Others

πολιτεύομαι

+

+ +

+

+

As was noted above, the semantic distinction between active πολιτεύω 'bin πολίτης' (stative) and middle πολιτεύομαι 'verwalte den Staat' (activity) is not as clear as Kϋhner-Gerth take it to be. Thucydides uses the active form with the meaning 'be a citizen' (stative), but also in a more active sense 'function as a citizen, act politically' E.g. statiνe 'be a citizen': JR4 Passive πολιτεύομαι may be compared to βασιλεύομαι 'be ruled by a king' (to βασιλεύω + gen. 'be king of) and ίiρχομαι 'be ruled' (to ίiρχω + gen. 'rule'). Cf. also δημοκρατέομαι (aor. έδημοκρατήθην) 'be govemed democratically'

226 (33)

CHAP'ΓER

5

( ... ) ό Βρασίδας το\ς μεν μετa τ&ν 'Αθηναίων Τορωναίοις καταπεφευγόσι κήρυγμα έποιήσατο τον βούλομενον έπι τα έαυτου έσελθόντα άδε&ς πολιτεύειν ( ... ) (rh. 4.114.1) (... ) Brasidas made proclamation to the Toronaeans who had taken refuge with the Athenians, that whoever wished might retum to his property and be a citizen without fear; ( ... )

Actίvίty

(34)

'act politically', e.g.:

οί δε ταυτά τε πάντα ές τούναντίον ί::πραξαν και liλλα ί::ξω

του πολέμου δοκουντα είναι κατα τaς ίδίας φιλοτιμίας και ϊδια κέρδη κακ&ς ες τε σφας αύτοi>ς και τους συμμάχους

έπολίτευσαν (... ) (rh. 2.65.7) But they acted not only contrary to his advice in all these things, but also in matters that apparently had no connection with the war they were led by pήvate ambition and pήvate greed to adopt policies which proved injuήous both to themselves and their allies (... ).

These people are not just in a state of being citizens. The active έπολίτευσαν must refer to an active political behaviour (cf. also ί::πραξαν before), namely a behaviour that aims at the benefit of their allies and of themselves. In a comparable expression, Lysias uses the middle verb: (35)

[subject: those that beat the phalluses off the Hermes-statues] ( ... ) και είς liπασαν την πόλιν ήμαρτήκασιν, άδίκως και παρανόμως και προς τοi>ς liλλους πολιτευόμενοι και προς σφας αύτοi>ς διακείμενοι (... ) (Lys. 14.42) (... ) and they committed offences against the wole city, showing injustice and illegality alike in their public treatment of their fellow-men and in their behaviour to each other (... ).

It appears that Thucydides uses the active voice both in the stative meaning 'be a citizen' and in the activity-meaning 'act politically' This observation is, of course, in keeping with the fact that Thucydides does not use the middle verb (save with a passive meaning). In Xenophon, who as we have seen is the only author using both πολιτεύω and πολιτεύομαι (with middle meaning), it is unclear what factor determines the choice of either of the verbs. This obscure situation may well be caused by the circumstance that, in the real world, being a citizen usually implies that one govems the state (esp. in Athens) and/or Iives under a certain constitution, and vice versa. In

'SYNONYMOUS' ACΠVE AND MIDDLE VERBS

227

this fashion the contrast between the neutral active on the one hand, and the middle/passive on the other, may easily become bluπed. Illustrative examples of this confusion are the following: (36)

(37)

εδοξε τφ δήμφ τριάκοντα &νδρας έλέσθαι, οϊ τους πατρίους νόμους συyyράψουσι, καθ' οϋς πολιτεύσουσι (Xen. HG 2.3.2) lt was voted by the people to choose thirty men to frame the ancient Iaws into a constitution under which to conduct the govemment.

αίρεθέντες δΕ: έφ' ο/rε συyyράψαι νόμους, καθ' οϋστινας πολιτεύσοιντο (Xen. HG 2.3.11) ( ... ); although chosen for the purpose of framing a constitution under which to conduct the govemment (... ).

(38)

επειτα δΕ: νόμους θεiναι, καθ' οϋς πολιτεύσοιντο

(Xen. HG

5.3.25) Secondly, they should draw up a constitution under which to conduct govemment.

The active and the two rniddles refer to the same state of affairs in reality. Ι fail to notice any semantic nuance. 385 In the following example, Xenophon uses the active verb in a comparable context as Thucydides above (34): (39)

[Alcibiades] (... ) έπιβουλευθε1.ς δΕ: ύπο τ&ν ελαττον έκείνου δυναμένων μοχθηρότερά τε λεγόντων και προς το αύτ&ν ϊδιον κέρδος πολιτευόντων, έκείνου άε1. το κοινον αϋξοντος

( ... ) (Xen. HG 1.4.13) (... ) but rather because he was plotted against by those who had less power than he and spoke less well and ordered their political doings with a view of their own pήvate gain, whereas he was always advancing the common wheal (... ).

As in Thucydides' examples above, the active πολιτεύω appears to refer to political actίvίty, rather than to a state. Το all appearances, Xenophon does not use the active and middle with any semantic distinction. The question now remains how we must account for the occuπence of both the active and the middle

385 It can, however, not be excluded that the latter two examples must be interpreted as passives: 'according to which they would be govemed'

228 CHAP'fER 5 verb. Tentatively,

Ι

would suggest that there is a stylistic difference:

πολιτεύομαι is the nonnal fonn, in Attic as well as elsewhere, whereas πολιτεύω is restricted to Thucydides. In other words, my

guess is that in using the active verb, Xenophon - consciously or not commits a Thucydideanism. It is, in any case, significant that 5 of the 6 instances of active πολιτεύω (also once in the Anabasis) occur in the Hellenica, which is Xenophon's sequel to Thucydides' Histories. The consequence of the absence of a semantic distinction between πολιτεύω and πολιτεύομαι (with non-passive meaning) is that scenario (iii) above applies: both verbs imply subject-affectedness, and the active inflection is neutral wit respect to subject-affectedness. Απ interesting additional question might be how πολιτεύομαι is to be classified as to middle usage type. Πολιτεύομαι is often mentioned as the prime example of a dynamic middle. Schwyzer & Debrunner (Π: 232) describes the dynamic middle thus: (40)

"Die gefίihlsmaBige Beteiligung des Subjektes am Verbalinhalt, die das Medium gegenίiber dem Aktiv zunachst ausdrίickt, wirkt sich teilweise durch Steigerung des Verbalinhaltes aus: intensives (oder dynamisches) Medium.''

In other words, the dynamic middle heightens the intensity of the lexical verbal content. The problem with such a definition is that it is rather unspecific (not to say vague), and ad hoc. Gildersleeve, in his review-article ση Stahl's syntax, rightly puts it as follows: (41)

now we have come to what Ι would fain call the drippan middle, the πανδέκτης-middle, the middle that is put at the bottom to catch the drippings of the other uses as the ablative is put to catch the drippings of the other cases. It is called the intensive middle, the dynamic middle" (Gildersleeve 1908: 277) 'Άnd

Indeed, Gildersleeve's sarcasm is justified if we consider Stahl's inventory. His enumeration of 'dynamic' middles constitutes quite a mixed bag. Consider, for example, παρέχεσθαι, στρατεύεσθαι, λαβέ­ σθαι, πολιτεύεσθαι, eπεσθαι, ίδέσθαι, and πειρφον έκ νεrος άφήλατο (Α.

Pers. 305)

He jumped a light leap from the ship.

Here, the action noun πήδημα is used as an intemal object to άφή­ λατο. From these examples, it can be concluded that &λλομαι and πηδάω are semantically identical. The question then remains as to what factors determine the use of the two verbs. Let us first examine the distribution in poetry. As was noted above, Homer strongly prefers &λλομαι, which may indicate the antiquity of the verb, as compared to πηδάω. In classical poetry the situation is different. The tragedians prefer πηδάω, which probably reflects the contemporary Attic usage. In tragedy, &λλομαι appears to be used under specific conditions, as can be seen in table 4, which depicts the occurrences of the two verbs (and their compounds) in Aeschylus, Sophocles, Euήpides, and Aristophanes (fragments excluded): Table 4: Occurrences of &λλομαι and

πηδάω

in Attic drama

PRESENT

AORIST

&λλομαι

2

16

πηδάω

34

15

We can observe that the Attic dramatists exhibit a fairly clear tendency to use &λλομαι in the aorist stem390 , and πηδάω in the present stem. This tendency is not discemible in prose: in Herodotus, Thucy-

in Xen. Eq. 3.7 and Eq. Mag. 8.3. forms: ιhematic ήλόμην and pseudo-sigmatic ήλάμην. At Α. Pers. 516 (see comm. Groeneboom), Eu. 372, and Soph. ΟΤ 1311 (see comm. Jebb), ιhere is some doubt between imperfects ήλλο/ε- and thematic aoήsts ήλο/ε-. The ιhematic aorist, howeνer, is semantically and textually preferable in every case. 389 390

Α similar case can be found 'Άλλομαι has two aoήst

'SYNONYMOUS' ACΠVE AND MIDDLE VERBS

233

dides, Xenophon, Plato, and Demosthenes &λλομαι occurs 28 times, of which 21χ present, and only 3χ aorist. 391 The distήbution of πηδάω and ήλάμην/ήλόμην in drama is, one may conclude, not determined semantically. Α possible altemative explanation is that metrical factors are relevant. Assuming that πηδάω is used as the unmarked default-option in drama as it is in Attic prose, the poet will encounter difficulties when trying to fit the aorist forms of πηδάω in the iambic metre, e.g. έπήδησα, έπήδησας, έπήδησε, etc. 392 ln such cases, he will have to take recourse to the synonymous aorist forms of &λλομαι. This mechanism would explain why &λλομαι is virtually confined to the aoήst stem. 393 Now let us tum to the use of the verbs in prose. In prose, the choice of &λλομαι appears to be a matter of register. 'Άλλομαι, as Ι will argue, belongs to an elevated, more poetic, register, whereas πηδάω is neutral from a stylistic point of view. This is especially supported by the use of &λλομαι in Plato. "Αλλομαι appears six times in the works of Plato. In all six cases, we are dealing with a stylistically marked context. The clearest example, in this respect, is Ion 535b which clearly refers to the Odyssey (χ 2).

(50)

( ... ) η τον 'Οδυσσέα ( ... ) (Pl. lon 535b)

οταν έπ1. τον ούδον έφαλλόμενον ίfδ'!lς

(... ) either when you sing of Odysseus - when he sprang to the thresho1d (... )

391 The future occurs once. Exc1uded from these numbers are three cases in Xenophon, where the mss. vary between present and thematic aorist: HG 4.4.11, 4.5.7, An. 4.2.17. 392 Indeed, these forms do not occur in drama, un1ess with e1ision (e.g. άνεπή­

δησ'). 393 It should be noted that, from a semantic point of view, one would expect a preponderance of aorist forms, given the punctual aspect of the 1exica1 meaning. In Homer, indeed, the aorist occurs 56 times, the present only twice. Furthermore, the formation a1so poinιs to a punctua1 (or at 1east telic) Aktionsart, since the root aorist formation &.λτο is morpho1ogically unmarked, as opposed to yod-present &λλομαι (< *s(h 2)l-yo-). The two exceptiona1 cases of &λλομαι in drama used in the present stem (both cases with iterative meaning) are found at Ar. Nu. 145 (Socrates asks Chaerephon about a flea: όπόσους &λλοιτο το\Jς αύτης πόδας 'how many of its own feet a flea could jump'), and at Lys. 83 (the Sparιan gir1 Lampito asserιs: γυμνάδδομαι γap κ:αt ποτt πυγάν &λλομαι ('At any rate Ι do gymnastics and jump hee1-to-buttocks'). Aristophanes may have used the more e1evated &λλομαι here to a comica1 effect.

234

CHAP'ΓER

5

Compare Homer: (51)

&λτο δ' έπι μέγαν ούδον εχων βιον ήδε φαρέτρην (χ 2) He sprang to the broad threshold with the bow and the quiver (... ).

Plato uses the νerb &λλομαι in order to remain close to the original wording of Homer. Notice that he also employs the Homeric form ούδόν, instead of Attic όδόν. Likewise, the case of Lg. 653e τα μΕν άλλόμενα και σκιρτ&ντα about the inclination of young beings to moνe around and cry out has a poetic ήng to it. This may be infeπed from its occuπence in the hendiadys, and from the fact that σκιρτάω is a νerb that is typically used in poetry. 394 Next, consider Phdr. 255c: (52)

και οtον πνε\>μα η τις ήχrο άπο λείων τε και στερε&ν άλλομένη πάλιν οθεν ώρμήθη φέρεται, οϋτω το του κάλλους pε\>μα πάλιν είς τον καλΟν δια τ&ν όμμάτων ίόν,

( ... ) (Pl.

Phdr. 255c) And just as the wind or an echo rebounds from smooth, hard surfaces and retums whence it came, so the stream of beauty passes back into the beautiful one through the eyes, (... ).

The use of &λλομαι is in accordance with the exalted style of this simile (cf. also the use of the poetic word ομμα). In the next case in Plato, Cra. 413b, ύπερ τα έσκαμμένα &λλεσθαι ('to jump oνer the trenches'), we are dealing with a proνerbial expression originating from the language of athletics, which conνeys the id~a '' going too far'' lt is a common phenomenon that proνerbial expressions exhibit a somewhat conservatiνe wording. Hence, the occuπence of &λλεσθαι here, instead of πηδftν, is not surprising. The last two cases of &λλομαι in Plato are related somehow, if howeνer in an unclear manner. Both cases concem a combination of

At Lg. 672c, likewise referring to young beings, Plato writes άτάκτως α{) Again, the difference between ίiλλομαι and πηδάω appears to be stylistic (cf. also the absence of the hendiadys in 672c), not semantic. In the same manner, our example from Herodotus (2.66.3) can be explained by stylistic factors. This may also be infeπed from the appearence of (ύπερ)θρφσκ:ω in the same sentence, a verb that almost exclusively occurs in poetry. 394

πηδ~.

'SYNONYMOUS'

ACΠVE AND MIDDLE

VERBS 235

the verbs θεϊν και &λλεσθαι: (53)

Θεϊν δe και &λλεσθαι και τα του σώματος &παντα tργα, ού τα μΕν όξέως και ταχu γιγνόμενα τα του καλου έστιν, (... )

(Pl. Chrm. 159c) And with running and jumping and all the movernents of the body, aren't the ones that are performed bήskly and quickly the admirable ones?

(54)

[etymology of the word θάλλειν] οtόνπερ oi>v μεμίμηται τφ όνόματι, συναρμόσας άπο του θεϊν και α.λλεσθαι το ονομα (Pl. Cra. 414b) Something of that sort the namegiver has reproduced in the name, which he cornpounded of θείν (run) ίiλλεσθαι (jump).

As regards theίr cooteot, there ίs οο relatίoo betweeo the two passages. lo the cootext of the Charmίdes, Socrates ίs dealίog wίth the questίoo of whether certaio actίvίtίes like playίog the lyre aod wrestliog are more admίrable wheo performed quίckly or quίetly. Lίkewίse, he asks whether ruooίog and jumpίog are coosίdered more admirable wheo they are performed quίckly. There ίs οο reasoo to assume that &λλομαι ίs prefeπed to πηδάω due to the elevated tooe of the passage. Ιο fact, the subject-matter ίs quίte colloquίal: boxίog, wrestliog, etc. How theo cao the use of &λλομαι be explaίoed? Although Ι dίd oot eocouoter addίtίooal ίostances ίο the 'ΓLG, the reappearance of the ίdeotίcal combίoatίoo θεϊν και &λλεσθαι ίο the Cratylus suggests that we are dealίog wίth a fίxed expressίoo that cooveys the geoeral ίdea of vehemeot movemeot. As ίο the example of Cra. 413b, ίο fίxed expressίoos the occuπeoce of less colloquίal words such as &λλομαι ίs oot surpήsίog. Το cooclude our dίscussίoo of &λλομαι ίο Plato, ίt appears that the verb ίs used ίο marked cootexts, eίther ίovolvίog a hίgher, more poetίc regίster (/on 535b, Lg. 653e, Phdr. 255c), or ίovolvίog fίxed expressίoos (Cra. 413b, 414b, Chrm. 159c). There ίs also some evίdeoce that Xeoophoo used &λλομαι as a more elevated altematίve for πηδάω.

236 CHAP'fER 5 Table 5:

Occurrences of Xenophon

&λλομαι

&λλομαι (19)

Hellenica Anabasis Cyropaedia De Equitandi Ratione De Equitum Magistro Memorabilia Cynegeticus

6 4 4 3 ι ι ο

(40.0 (36.4 (25.0 (20.0 (25.0

%) %) %) %) %) (5Ο.ο %) (0.0 %)

and

πηδάω

in the works of

πηδάω

(50)

9 (60.0 %) 7 (63.6 %) ι2 (75.0 %) ι2 (80.0 %) 3 (75.0 %) ι (5Ο.ο %) 6 (ιΟΟ %)

lt can be seen that, of Xenophon's works, the Hellenica and the Anabasis have the highest ratios of &λλομαι as compared to πηδάω (40.0 % and 36.4 % ). This may be due to the rnilitary subject-matter of these works. Especially in contexts of battles, Xenophon seems to be inclined to use &λλομαι: soldiers jumping on and off walls, horses, chariots, etc. The fact that Homer, too, mostly uses &λλομαι in this type of context may have influenced Xenophon's usage of the verb. Through the use of &λλομαι, Xenophon' s descriptions of military events may have evoked aπ association with Homer' s battle scenes. By comparison, in the more technical works the occuπence of &λλομαι is less frequent (cf. de Equitandi Ratione, de Equitum Magistro, Cynegeticus). Το conclude, in this section Ι have argued that there is no semantic distinction between &λλομαι and πηδάω (cf. scenario (iii)). Both verbs are verbs of body motion, both verbs imply subject-affectedness. Distributional differences must be explained by stylistic or, in the case of Attic drama, by metήcal factors. In section 5.2 below, the general class of motion verbs will be dealt with in relation to their voice marking.

5.1.5 βούλομαι vs. έθέλω One of the first issues that will pop up in any Greek linguist's rnind when thinking about the Greek middle voice is probably the contrast between middle βούλομαι and active έθέλω. Why do these verbs have different voice-markings, even though semantically they are closely

'SYNONYMOUS' ACΠVE AND MIDDLE VERBS

237

related? There have been a number of studies describing the semantic differences between the verbs, e.g. Fox (1917), Rtidiger (1917), Wifstrand (1942), Madden (1975). However, none of these studies deals with the difference in voice. The aim of this section, therefore, is to analyze the relation between the semantics of the two verbs and their respective voices. In Homer, the meanings of the two verbs are clearly distinct (see LfgE395 ). Βούλομαι, according to LfgE, means: "( ... ) eine spontane, intuit. Regung, die dem den Vorzug gibt, was sich als vorzϋglicher anbietet (... ) oder als das geringere ϋbel erscheint'' The verb is often modified by πολύ or occurs with the comparative particle f\. In other words, βούλομαι can be rendered by 'prefer' As for έθέλω, LfgE observes that its meaning can be characterized as a spectrum between two extremes: "(1) Verlangen haben, etw. zu tun od. geschehen zu lassen (lmpuls kommt in erster Linie vom Subj.) u. (2) bereίt, wίllίg seίn, etw. zu tun oder geschehen zu lassen (Verhalten ist vor allem Reaktion auf von auBen kommenden Impuls)" Thus, έθέλω can be rendered by (1) 'wish, want' or (2) 'be willing' Further, since it the meaning of βούλομαι 'prefer' 396 is more specific implies the existence of an altemative choice than that of έθέλω which has a more general, neutral sense 'wish, want' This semantic difference coπelates with their frequencies in Homer: βούλομαι 38 occuπences, against έθέλω 292. In Classical Greek, the boundary between βούλομαι and έθέλω has gradually become fuzzier in comparison to the situation in Homer. This development is mainly due to a broadening of the meaning of βούλομαι: the verb extended from the specific meaning 'prefer' to a more neutral meaning 'wish, want' .397 As a consequence, έθέλω lost semantic 'ground' in favour of βούλομαι. This development is reflected in the frequencies of the verbs.

395 Under βούλομαι we find: "Den Bereich von έθέλειν (s.d.) berίihrt β. nur aπ den untyp. Stt. Α 67, δ 353, (... )". 396 Cf. Neuberger-Donath (1975), who renders Homeήc βούλομαι by Latin malle. 397 Thus, βούλομαι did not lose its meaning 'prefer' in the Classical language. In many insιances of βούλομαι aπ altematiνe choice is still implied. However, the verb met with competition from the (post-Homeήc) middle verb προαιρο\>μαι 'prefer'

238

CΗΑΡΊΈR

5

Table 6: Frequencies of βούλομαι and έθέλω 398

Homer Aeschylus Sophocles Euripides Herodotus Aristophanes Thucydides Xenophon Plato Demosthenes

βούλομαι

έθέλω

38 3 38 170

292 68 102 288

256 193 369 788 976 757

188 93 70 292 527 269

The preponderance of έθέλω in the Attic tragedians must be taken as a feature of the poetic vocabulary. In the Attic prose-authors, έθέλω is almost entirely reduced to the meaning 'be willing' ('bereit sein, geneigt sein', cf. ROdiger 1917: 14, Wifstrand 1942: 17). In Herodotus and in poetry, έθέλω has partly preserved its broader use 'wish, intend', cf. Wifstrand (1942: 29) 'wίinschen, beabsichtigen'. 399 Απ example which neatly illustrates the semantic contrast between the two verbs is the following: (55)

εί δε βούλει, σο\ έγώ, ώς τουτο οϋτως εχει, έθέλω λόγον λέξαι

If you

(Pl. Grg. 522e)

Ιίkι-,

I'm willing to give you an account showing that this is so.

In this example, βούλει represents the mental state of wishing that exists autonomously in the subject-referent. Έθέλω denotes that the subject only acts in compliance with the will of the other person. This The data is drawn from the LfgE [Homer], Powell (1938) [Herodotus], (1917) [others]. 399 lt is a remarkable fact that in Koine Greek (έ)θέλω regained the ground it had lost to βούλομαι. In the New Testament θέλω is ιhe general word for 'wish, want', cf. Modem Greek θέλω. Wifstrand (1942: 31-2) explains this fact by assuming ιhat έθέλω preserved its oήginal, more extensive, use in vernacu/ar Attic (as it had in Ionic), i.e., the temporary success of βούλομαι may have been restricted to ιhe literary Attic language. 398

Rδdiger

'SYNONYMOUS' ACΠVE AND MIDDLE VERBS

239

example also shows another important feature of the semantics of the two verbs. ln this example, the subject-referent of βούλει is not in control as the realization of the state of affairs that is wished for. Έθέλω, however, does imply that the subject is in control and also has the intention of carrying out the event designated by the infinitive. Α strong indication of έθέλω's more restήcted use in Classical Greek prose as compared to Homer is the fact that it has almost lost the possibility to be construed with an accusative plus infinitive, a construction which is fairly comrnon in Homer. 400 This means that έθέλω requires that the subject of the main clause is coreferential with the (unexpressed) subject of the infinitive. In other words, έθέλω implies that the subject-referent performs the event himself. In the exceptional cases of έθέλω + acc. c. inf., the subject of the main clause may still be involved ίη the event expressed by the infinitive even though the subject of the main clause and the infinitival subject are not coreferential. E.g.:

(56)

[Alexandros hears the story of the rape of Medea] δευτέρ-n δ[ λέγουσι γενεft μετiχ. τα\>τα Άλέξανδρον τον Πριάμου άκηκοότα ταυτα έθελησαί οί έκ της 'Ελλάδος δι' άρπαrilς γενέσθαι γυνα!κα, (... ) (Hdt. 1.3.1) Then (so they say) in the second generation after this Alexandros son of Pήam, having heard this tale, intended to win himself a wife out of Hellas by ravishment.

Though the subject of the subordinate clause is γυνα!κα, and therefore different from the subject in the superordinate clause ( 'Αλέξανδρον ), the state of affairs descήbed in the subordinate clause clearly implies an action and a degree of control ση the part of Alexandros. 401 And, indeed, imrnediately after this, Herodotus descήbes Alexandros' rape of Helen ( 1.3.2). In other words, έθελησαι does not merely express -

400 E.g., ίη Herodotus 2 times out of 188 occurrences (1.1 %), whereas βούλομαι occurs 16 times out of 256 with acc. c. inf. (6.3 %). 401 The feature of control exercized by the matήx subject may also be relevant to the tendency of έθέλω to occur with divine subject-referenιs, e.g. Dem. 2.20: ( ... ), &ν οϊ τε θεοl θέλωσι και ύμεϊς βούλησθε. For a similar case of contrast, cf. Pl. Alc. Ι 135d. Expressions like έav θεος (έ)θέλτι (θεοl (έ)θέλωσι) are frequently occurring formulas. Note that the alliteration may also have contήbuted to the success of these formulas.

240

CHAP'ΓER

5

as βούλομαι would Alexandros' wish tout sec, but it implies the intention of the subject to realize the event denoted by the infinitive. 402 Furthermore, the subject-referent of the main clause is involved through its role of beneficiary/recipient in the subordinate clause (οί). 403 The syntactico-semantic property of subject-coreferentiality approximates έθέλω to the class of auxiliaries (e.g. δύναμαι, μέλ­ λω404). lt is well known that equi-subject constructions are the main diachronic source of grammaticalized auxiliaries such as tense-aspectmodality markers (see e.g. Gίνόη 2001b: 79-80). It is therefore no surprise that the verb eventually developed into the future particle θα in Modem Greek, e.g. θα γράφω 'Ι will be wήting' This indeclinable clitic particle finds its origin in the syntagm θέλω ϊνα 'Ι want that',

Note that έθελfiσαί οί (... ) γενέσθαι Ύ\)ναtκα is practically equiνalent to a coreferential subject. an acc. c. inf. from Herodotus is: (... ) γίνεταί μοι άναγκαιότατον έντέλλεσθαι τα θέλω μοι γενέσθαι τελευτοον τον βίον (Hdt. 3.65.6) ('( ... ) necessity constrains me to charge you with what Ι want to be done for me as Ι now am dying'). While dying, Cambyses beseeches the Persian nobles neνer to let the empire deνolνe to the Medes. One may wonder whether θέλω implies control on the part of Cambyses, since he is dying. But, perhaps he - as Great Κing of the Persians - presents the situation as if he were still in control. Further, Cambyses is inνolνed as a beneficiary (μοι) in the embedded eνent. Besides the two instances Ι haνe discussed hitherto (Hdt. 1.3.1 [ex. (56)], 3.65.6), Ι am only aware of two additional instances of έθέλω + acc. c. inf. in Classical Greek (see also Wifstrand 1942: 24): Pl. Lg. 628c (a peculiar case where έθέλω appears to mean 'prefer', i.e. as an equiνalent of βούλομαι), and Xen. Cyr. 3.1.43 (where the matrix subject has control). Another insιance of έθέλω construed with an acc. plus inf. is cited by Fox (1917: 636): ( ... ) έπει αύτοον γε και πόδας και χεiρας έθέλουσιν άποτέμνεσθαι οί aνθρωποι, έCχν αύτοiς δοκiJ τα έαυτοον πονηρα ε!ναι (Pl. Smp. 205e) ('For men are prepared to haνe their own feet and hands cut off if they feel these belongings to be harmful'). Fox ιakes και πόδας και χεiρας as accusatiνe subject. Howeνer, as C.J. Ruijgh pointed out to me, it is better to take οί aνθρωποι as the subject of the infinitiνe άποτέμνεσθαι, which should be taken as a permissiνe passiνe ('to let themselνes be cut off as to feet and hands'). The accusatiνe και πόδας και χεiρας, then, is to be interpreted as an accusatiνe of the affected body part. 404 That βούλομαι, έθέλω, μέλλω, and δύναμαι are semantically closely related is also shown by the spread of the remarkable augment ή-. This augment was a product of reanalysis of the pair θέλω (from έθέλω by aphaeresis) and the augmented form fίθελον. This noνel augment ή- then spread through a process of analogy, e.g. θέλω fίθελον = μέλλω Χ(= fίμελλον). In the same fashion, ήδυνά­ μην and ήβουλόμην were created (see Debrunner 1954). 402

έθελf\σαι σχεiν Ύ\)ναtκα with 403 The other insιance of

'SYNONYMOUS' ACΠVE AND MIDDLE VERBS 241

after a typical process of grarnrnaticalization, viz. coalescence and attrition. In this rnanner, we can follow the developrnent frorn έθέλω as a verb of full sernantic content in Horner to a content verb with auxiliary-like properties in Classical Greek, and eventually into a future rnarking clitic particle in Modem Greek. This diachronic pathway full verb > auxiliary > clitic is one that is typically found in language-change (see e.g. Hopper & Traugott 1993: 108).405 My last rernark on the sernantics of βούλομαι and έθέλω concems their use with inanirnate subjects. The use of έθέλω with aπ inanirnate subject is rnore frequent than that of βούλομαι, and there are considerable sernantic differences between the two. In cases in which βούλομαι has an inanirnate subject, the subject typically refers to entities like μ\>θος, τέρας, and the infinitives are verbs like σημαί­ νειν, λέγειν. 406 Though βούλομαι here has an inanirnate subject, the verb still irnplies a rnental involvernent of sorne kind, narnely that of the hurnan that wanted to convey sornething by rneans of the word or sign. Words can be identified rnetonyrnically with the hurnan rnind that produced thern. Now the use of έθέλω with an inanirnate subject is stήkingly different. Unlike βούλομαι, έθέλω with inanirnate subject does not involve the (irnplicit) presence of a rnental entity. Rather than to hurnan intentionality, έθέλω seerns to refer to the rnornenturn that is inherent in the norrnal course of events. (57)

[Harpagus is deliberating whether he should kill Astyages' infant son] εί δ' έθελήσει τούτου τελευτήσαντος ές την θυγατέρα ταύτην άναβfιναι ή τυραννίς (... ) (Hdt. 1.109.4) If after his death the sovereignty should pass to this daughter of his (... ).

40s Έθέλω's development can be compared to that of English will, which originally meant 'want' (cf. be willing, and Dutch willen 'want'), and now functions as a temporal auxiliary. 406 In cases where the infinitive is absent, such verbs can be understood. Then, βούλομαι is usually translated by 'mean', e.g. Pl. Tht. l56c: τί ήμ"iν βούλεται οi>τος ό μ\Jθος; ( ... ) Βούλεται γf:ι.ρ δi, λέγειν ώς (... ) ('Now what does this tale mean? ( ... ) For what it means to say is that (... )'). Το complicate matters, Herodotus appears to use έθέλω in this type of context consistently (e.g. 2.13.3). However, Herodotus' use of έθέλω seems to be generally more extensive than that of the Attic authors. This is also shown by the frequency-figures above.

242

CHAP'ΓER

5

The construction έθελήσει άναβηναι is practically equiνalent to a future άναβήσεται. Έθέλω here contήbutes the idea of the force that is inherent in the natural course of events which brings about the traπsition of the tyraπny. Here, as often, έθέλω irnplies that, unless it is blocked by aπ unforeseen force, the course of events will lead to the occuπence of the event expressed by the infinitive. Consider further: (58)

οίκότα μέν νυν βουλευομένοισι άνθρώποισι ώς το έπίπαν

έθέλει γίνεσθαι εi> (Hdt. 8.60.γ). Success tends to come most often to men that make reasonable designs.

In such contexts, έθέλω caπ be rendered by 'be wont' Here, once again, έθέλω407 refers to the force that resides in the natural course of events. Our rnain question now still rernains, narnely: what is the relation between the sernaπtics of έθέλω aπd βούλομαι and their respective voices? My clairn is that there is indeed such a relation. In general, βούλομαι irnplies a higher degree of rnental involvernent thaπ έθέλω. The forrner verb irnplies a positive choice, by which the subject actively aπd autonornously prefers one altemative to aπother. The latter verb typically denotes an intention that ernerges under the influence of the will of another person. In rnaking up one's rnind (βούλομαι), a person is psychologically involved to a greater extent thaπ in cornplying to sorneone's request (έθέλω). Also the special use of έθέλω 'be wont', aπd its eventual developrnent into a ternporal auxiliary shows that the rnental aspect of its lexical rneaning rnay be sornewhat backgrounded. Surnrnaήzing, the pair βούλομαι and έθέλω caπ be regarded as a case of scenaήo (ii) above. Although in the lexical sernaπtics of both verbs aπ elernent of rnental affectedness is present, this inherent elernent is ernphasized by rneans of the rniddle voice ending in the case of βούλομαι. As for έθέλω, the active ending, being neutral as to subject-affectedness, does not contήbute to the lexical rneaning of the verb.

407 lnterestingly, the verb φιλέω seems to have undergone a similar semantic shift from a verb designating an affection Ίove' to an auxiliary meaning 'to be wont' The change can be explained as follows: 'she (always) wanιs/loves to speak the ιruth' pragmatically implies 'she is wont to speak the ιruth'

'SYNONYMOUS' ACΠVE AND MillDLE VERBS 243

5.2 Active and Middle Body Motion Verbs

In this chapter, Ι have already discussed two active-middle pairs of body motion verbs, όρμάω όρμάομαι and πηδάω - &λλομαι. For these two pairs, Ι have tήed to pin down the factors that determine the distribution of the active and the middle member of the pair. In this section, the general class of body motion verbs will be addresses. What makes this class of verbs interesting for our purpose is the fact that it is relatively numerous, and that it consists of both active and middle verbs. These include activa tantum (e.g. βαίνω), media tantum (e.g. ίκνέομαι), and oppositional middles (e.g. κινέομαι - κινέω). Το all appearances, there is no pattem that explains why one verb is active and the other middle. However, in this section Ι will argue that a number of general tendencies can be discemed within this class of verbs. Το this purpose, consider the following synoptical table of body motion verbs: 408

408 An exhaustive inventory and discussion of the semantics of the media tantum of motion in Homer can be found in Boeder (1961-2: 33-63). For a study of the verbs of motion in Homer, Ι refer to Kun (1966). Recall that the term translational motion refers to motion along one dimension in space; non-translational motion refers to motion that is neither translational, nor involves a change in body posture.

244

CΗΑΡΊΈR5

Table 7: Verbs of body motion Translational Motion Manner of Motion Activa Tantum gαδί;ω 'walk, proceed' ιδρασιcω 'run tpπω 'creep' θέω 'run' θριpσιcω (poetry) 'jump' νέω Swim' πατέω 'walk' 1

πηδάω 'jum~· πλέω

'sail, oat'

Manner-neutral Motion Activa Tantum ~αίνω 'go' 412 λώσιcω (Hom.) 'come' εΙμι ·~σ·

Nontranslational Motion and Change in Body Posture Activa Tantum ιcύπτω 'bend forward, stoop' Other actives (Hom.llon.) 'sit down'

(aor. ηλθον Ί came')413 f\ιcω/ίιcω '(have) come' ίιcάνω (Hom.)414 νοσrέω 'retum' χωρέω 'yield, go'

ϊζω

Media Tantum tρχομαι ·~ο, come' ίιcνέομαι reach' νέομαι 'retum (home), go' οϊ{ζμαι 'depart, be away' χα ομαι 'draw back'

Op~ositional

στείχω 'march' 4011 τρέχω

'run'

Other actives

νήχω (poetry) 'swim' 410 όρμάω 'rush (intr.)' 411

Media Tantum iiλλομαι rtump' πέτομαι' y' Other middles νήχομαι (poetry) 'swim'

Oppositional Middles άπαλλάττομαι 'depart' δύομαι 'enter, plunge into' ϊσταμαι 'stand still, stand up' ιcινέζοαι 'move (intr.)' ιcομί ομαι 'travel' όρμάομαι 'start off (intr.)' ορνυμαι (Hom.) 'arise' πορεύομαι 'go, walk' στέλλομαι 'set out, joumey

Middles

έρειδοζ;αι (Hom.) Ίean ση' ~ιcαθ)ί ομαι 'sit down'

ίσταμαι 'stand still, stand up' κλίνομαι Ίean, lie down' όρέγνυμαι 'stretch oneself out' στρέφομαι 'tum around' τανύομαι 'stretch oneself out' τείνομαι 'stretch oneself out' τρέπομαι 'tum around'

The first tendency that can be observed in the table concems the verbs

According to Utoublon (1985: 166-77), the oήginal meaning of στείχω was 'marcher droit, en ligne', which bleached to 'go' Α similar semantic bleaching can be observed in βαίνω 'make a step' > 'go'; Εpπω 'creep' > 'go' (in some dialects); νέομαι 'return' > 'go'; χωρέω 'yield' > 'go' 410 In section 5.0 it was argued that there was no discemable semantic distinction between νήχω and νήχομαι. 411 'Ορμάω designates motion at high speed (see 5.1.1). With some hesitation, Ι would regard it therefore as a manner of motion νerb. Όρμάομαι, on the other hand, is neutral with respect to the manner of motion. 412 Βαίνω oήginally meant 'make a step', and thus belonged to the class of manner of motion verbs. Βαδίζω can be considered the successor to βαίνω in Attic. 413 For the semantic distinction between present εtμι ('centrifuge') and aoήst ilλθον ('centripete'), see Utoublon (1985: 27ff.). 414 In Homer, the middle νaήant ίιcάνομαι (probably formed after ίκνέομαι) also occurs. 409

'SYNONYMOUS' ACΠVE AND MIDDLE VERBS 245

manner of motίon: the majσήty σf these verbs are actives, e.g. 'walk, διδράσκω 'run' There are a few exceptiσns (e.g. &λλομαι). Frσm this prepσnderance σf actives, σne may assume that this class σf verbs invσlves a lσw degree σf subject-affectedness. Ι wσuld argue that this is indeed the case: verbs denσting manner σf mσtiσn inherently invσlve a lσwer degree σf subject-affectedness than the σther verbs σf bσdy mσtiσn. Since these verbs fσcus ση the manner in which the change σf lσcatiσn takes place, the element σf change (σf lσcatiσn) undergσne by the subject is backgrσunded. Therefσre, mσst manner σf mσtiσn verbs are atelic (durative) as tσ their inherent lexical meaning. 415 They denσte aπ event that is viewed as unchanging thrσugh time, withσut aπ inherent endpσint. 416 Fσr example, the subject σf aπ atelic event like τρέχω may be viewed as less affected than the subject σf a verb like ίκνέομαι, which denσtes aπ event that fσcuses ση the endpσint, and hence ση the change σf lσcatiσn which the subject undergσes. 417 The secσnd tendency that can be σbserved in the table invσlves the verbs denσting nontranslatίonal motίon (mσstly verbs denσting change ίη bσdy pσsture, e.g. καθίζομαι). Verbs designating nσn­ translatiσnal mσtiσn tend tσ be σppσsitiσnal middles. Α priσri, σne can σf

βαδίζω

415 Clear exceptions are verbs of 'jumping': ίiλλομαι, θριj>σκω, and πηδάω (see section 5.1.4). 416 There is some evidence from other \anguages that telicity in the domain of body motion is connected to middle marking. For examp\e, French s'en voler 'f\y away' focuses on the fact that the flying entity departs from its initia\ sιationary position, whereas voler refers to the fact of flying without specific reference to a change of state, nor to the beginning or endpoint (Kemmer 1993: 157). Klaiman (1991: 58) makes a similar observation in re\ation to Fula, a Niger-Congo Ianguage with a midd\e voice system. Atelic verbs of physica\ motion are active tantum, e.g. war- 'come', yah- 'walk', dill- 'go', dogg- 'run', whereas telic verbs are media tantum, e.g. yott- 'arrive', ma ΎΎ- 'climb, mount' 417 The telic Aktionsart of ίκνέομαι is also ref\ected in its morphology: the unmarked formation is found in the aoήst stem (iκ-όμην), whereas the present stem is marked (iκ-νέ-ομαι). Notice further that there are no contrastive middle verbs denoting manner of motion. This may explained thus: the active causative verbs would express that a person makes another person move in a cerιain manner (cf. active causative ϊστημι means to make (mostly command) somebody stand stilVup). However, in practice when a person Χ makes (commands) a person Υ to move, the manner of this motion is generally irrelevant. Therefore, the concept of make somebody move in α certain manner is not often lexicalized (e.g. **ίiλλω 'make jump'). Rather, this meaning is expressed by a lexica\ causative verb (e.g. by means of ποιέω).

246 CΗΑΡΊΈR 5

assume that these verbs involve a higher degree of subject-affectedness. This may be explained by the salience of the element of change (mostly change in body posture), that is inherent in the meaning of these verbs. This salient conception of change implies an increased subject-affectedness. lt must be noted, furthermore, that the middle voice inflection, of course, has the additional advantage that it enables an opposition with an active causative verb (e.g. ϊσταμαι - ϊστημι). This factor alone, however, is not sufficient means to explain why the nontranslational motion verbs show a tendency to be middles. The two tendencies discussed above can be considered as manifestations of a semantic scale within the class of motion verbs. (59)

Motion verbs: Scale of Subject-affectedness LOW Subject-atJectedness

mGH Subject-atJectedness

Manner of Motion < Neutral Translational Motion < Nontranslational Motion

At the left extreme of the scale, verbs tend to be active; at the ήght extreme, verbs tend to be middles. The manner-neutral translational verbs located in between do not exhibit a preference for either of the voices. Although we are not dealing with a pair of verbs in this section, but with a whole class of semantically related verbs, it is interesting to see whether one of the scenarios applies to the distribution of the active and middle voice within the class of verbs of motion. lt appears that scenaήo (ii) accounts for the distributional pattem best. That is to say, all verbs of motion - actives and middles alike inherently involve a degree of subject-affectedness: in all cases, the subject changes location or body posture. However, the function of the middle voice marking is to emphasize the element of subject-affectedness. ln sum, although both active τρέχω and middle καθίζομαι involve, to a certain degree, subject-affectedness, the middle ending of the latter makes the element of subject-affectedness more salient as compared to τρέχω. 5.3 Conclusion In a number of semantic studies on (near-)synonymous active and middle pairs, we have seen that very often one can discem a semantic

'SYNONYMOUS' ACΠVE AND MIDDLE VERBS 247

distinction between active and middle near-synonymous verbs, namely in the cases of όρμάω - όρμάομαι, πειράω - πειράομαι, βούλομαι έθέλω. In these cases, the lexical semantics of the active aπd middle verb alike involve, to some extent, aπ element of subject-affectedness. The middle ending makes this inherent element conceptually more salient, whereas the active ending being neutral as to subjectaffectedness - does not contribute to the meaning of the verb (scenario (ii)). In two other cases, πολιτεύω πολιτεύομαι and &λλομαι πηδάω, Ι argued that there is no discemable semaπtic distinction. In these cases, too, the lexical meaπing of both the active and the middle verb involves aπ inherent element of subject-affectedness. The middle voice redundaπtly codes this inherent subject-affectedness (scenario (iii)).

CΗΑΡΊΈR

6

General Conclusion

In this dissertation, an analysis was given of the meaning of the middle voice in Ancient Greek. In section 1.1, Ι argued that the notion of prototypical transivity is crucial to an understanding of the semantics of the middle voice. The notion of prototypical transitive event was described with reference to Langacker's Billiard-ball Model. The prototypical transitive event is coded by the unmarked active voice. Ι argued that the middle voice can be defined as a marked coding of a departure from the prototypical transitive event. Contrary to the prototypical transitive, the middΊe voice codes that the subject is affected by the event. In other words, the subject, in some way or other, undergoes an effect of the event. That the middle voice is marked can be shown by means of a number of markedness-cήteria. These cήteria pertain to aspects of phonological and morphological structure, as well as to distήbutional behaviour, and frequency. lt was argued that, in Ancient Greek, the rniddle voice is the marked member in a privative opposition with the active voice. In sum, the middle voice expresses the presence of the semantic property of subjectaffectedness. The active voice, on the other hand, is neutral with respect to subject-affectedness. In section 1.2, it was argued that the middle voice can be analyzed insightfully within the framework of Langacker's Complex Network Category Model. Essential notions of this model are abstract schema, elaboration, prototype, and extension. Following this model, the rniddle voice can be descήbed as a complex, polysemous network of inteπelated middle uses. The semantic property of subject-affectedness can be considered the abstract schema of the Ancient Greek middle voice. In turn, the various middle uses can be seen as elaborations of this schema. The structure of the semantic relations among the vaήous middle uses can be shown in a semantic map which, itself, can serve as a basis for two predictions: (I) Α form will always cover a connected region of vaήant middle uses in the semantic network, and (11) a form will only spread from one variant use to another if these uses are directly semantically related. Next, Ι discussed Kemmer's typological study The Middle Voice. Α number of central notions in her book, such as semantic map, lnitiator and Endpoint, and relative distinguishability of participants were shown to be relevant to the Ancient Greek middle voice.

GENERAL CONCLUSION 249

The middle voice in Ancient Greek can be said to code that the subject is the Endpoint of the event. Two general issues conceming the middle voice were addressed in section 1.3. Firstly, it was argued that the media tantum (middle-only verbs) should be integrated into the polysemous structure of the middle voice. Secondly, Ι put forward a number of objections against the valence reduction approach to the function of the middle voice. In chapter 2, the various middle uses and the relations among these rniddle uses were analyzed in detail. Eleven middle uses, in all, were distinguished: passive middle, spontaneous process middle, mental process middle, body motion middle, collective motion middle, reciprocal middle, direct rejlexive middle, perception middle, mental activity middle, speech act middle, and indirect reflexive middle. This semantic analysis resulted in a semantic map of the middle voice in Ancient Greek depicting the network that consists of the different middle uses and their semantic inteπelations. Finally, it was argued that the mental process middle could be considered the category prototype. The body motion middle, spontaneous process middle, and indirect reflexive middle can be viewed as secondary prototypes. Chapter 3 deals with the middle and passive forms in the aorist stem. In section 3.1, Ι have argued that the distήbution of the two variant passive forms in -η- and -θη- can be explained by a number of morphological rules, and a number of additional semantic rules. Subsequently, Ι tried to account for each of the verbs that possess both passive aoήst formations. In section 3.2, the distribution of the sigmatic rniddle aorist and the passive aoήst was discussed. It can be observed that the sigmatic middle aoήst and the passive aoήst cover a connected region of middle uses ίη the semantic network (cf. prediction (I) above). It was argued that the occuπence in Homer of sigmatic middle aoήsts designating mental processes, body motion, and collective motion is a relic of an earlier language stage, preserved as metrical altematives to the passive aorist forms. In this earlίer stage of the language the passive aorist form in -θη- was only used with passive and spontaneous process meaning. The passive aorist in -θη-, subsequently, extended to the mental process meaning, the body motion meaning, and the collective motion meaning. In Classical Greek, the passive aorist formation continued to expand, albeit ίη a more sporadical fashion. It was shown that, in Classical Greek, the passive formation sporadically extends to verbs of perception, verbs of speech, verbs of mental activities, and reciprocal verbs. It can be

250

CΗΑΡΊΈR

6

observed that, throughout the history of Greek, the passive aoήst fonn always spreads through the sernantic network frorn one related use to another (cf. prediction (11) above). In section 3.3, Ι proposed a definition of the abstract rneaning of the passive aoήst and the sigrnatic rniddle aoήst fonnations. Ι clairned that the passive aoήst rnarks that the subject is, or is at least sirnilar to, a prototypical patient. By contrast, the subject of the sigrnatic rniddle aoήst is sirnilar to a prototypical agent. Finally, Ι rnade aπ atternpt to account for the fact that there is a distinct passive fonn in the aoήst stern, as opposed to the present stern. This phenornenon was explained by the relation between the cornpletedness of the event and the degree of affectedness of the subject. In chapter 4, the rniddle and passive voices in the future stern were analyzed. Ι argued, reviving a theory of Friedήch Blass, that rniddle φαvουμαι and passive φανήσομαι express different aspects. Φανουμαι expresses a 'presentic' (irnperfective) aspect, whereas φανήσομαι expresses aπ 'aoήstic' (perfective aspect). This sernantic contrast rnay also be relevant to other rniddle and passive future fonns. Nevertheless, there are also verbs which do not have this aspectual distinction. With these verbs, the rniddle future fonn sirnply has a rniddle rneaning, and the passive future a passive rneaning. Chapter 5 was devoted to the issue of 'synonyrnous' active and rniddle verbs. Three possible scenaήos were sketched that rnay be applicable to each of the individual active-rniddle pairs. Following on frorn this, five case-studies of 'synonyrnous' active-rniddle pairs were presented. With regard to each of these pairs, Ι have tήed to detennine whether the distήbutions of the active and the rniddle verb in question could be explained by a sernantic distinction, or by other factors, such as diachrony, genre, and register. In the last section of this chapter, Ι tήed to account for the occuπence of active and rniddle verbs in the class of rnotion verbs. Ι argued that a tendency can be detected for rnanner of rnotion verbs to be active, and for nontranslational rnotion verbs and change in body posture verbs to be rniddle. These tendencies were explained by differences in the degree of affectedness of the subjects.

BffiLIOGRAPHY

Aissen, J. 1982 'Valence and coreference', in Studies in Transitivity (Syntax and Semantics 15), P.J. Hopper & S.A. Thompson (eds.), New York: Academic Press, 7-35. Allen, J.T., & G. Italie 1954 Α Concordance to Euripides, Berkeley. Andersen, Ρ.Κ. 1989 'Remarks on the Oήgin of the Term 'Passive", Lingua 79:1, 1-16. 1993 'Zur Diathese', HS 106, 177-231. Anderson, L. 1982 'The 'perfect' as a universal and as a language particular category', in P.J. Hopper (ed.), Tense and Aspect: between semantics and Pragmatics (τSL 1), Amsterdam: John Benjamins, 227-64. Bader, F. 1972 'Parfait et moyen en grec', Melanges Ρ. Chantraine, 121 : Κlincksieck. Bakker, E.J. 1994 'Voice, Aspect and Aktionsart; Middle Passive in Ancient Greek', in Β. Fox & Ρ. Hopper (eds.), Voice: Form and Function (τSL 27), Amsterdam/Philadelphia: John Benjamins. Bakker, S. 2002 'Futura zonder toekomst', Lampas 35, 199-214. Barber, E.J.W. 1975 'Voice - Beyond the Passive', in C. Cogen, e.a. (eds.), Proceedings of the first annual meeting of the Berkeley Linguistics Society: BLS, 16-24. Barnes, J. 1995 τhe Complete Works of Aristotle, 2 Vols., Pήnceton: UP. Barrett, W.S. 1964 Euripides Hippolytus, Oxford: UP. Basile, Ν. 1998 Sintassi storica del greco antico, Βaή: Levanti editoή. Bechert, J. 1964 Die Diathesen von ίδεiν und όρκτο 78, 170

ϊζομαι

160

ηύξάμην

130

'Ιαίνω ίάνθη

εύνήθην 127 Εϋχομαι 105, 107

78

'Ίημι

106

77 170 203, 244

ϊεμαι ετο

'Έχω

182 77, 80 εσχετο 169 έσχέθην 169 έσχόμην 170 σχήσω 182 εξω

Ίκάνω

Εχομαι

244 203, 244, 245 "Ικω 203, 244 Ίμείρομαι 207 Ίόομαι 63 "Ισταμαι 77, 208, 244 στάθητι 157 έστάθη 127 εστην 208 Καίομαι 61, 63 έκάη 130, 132 έκάην 143 έκαύθην 143, 155 Καλύπτομαι 90 Κατασκευάζομαι 27, 114 Καθαίρομαι 89 Καθίζομαι 78, 244 Κείρομαι 89 κείρασθαι 148 κερθείς 138 έκάρην 138 έκειράμην 154 Κέλομαι 106 Κεiμαι 67, 78 Κήδομαι 65, 72 Κινέομαι 80, 244 ίκάνομαι

Ίκνέομαι

Ζεύγνυμι

143 143

έζεύχθην έζύγην

89

Ζώννυμαι

Ήγέομαι iJγησάμην

166

51, 55 ηδομαι

51,65, 73,203 160 202

ησθην 155, ησθήσομαι ?Ημαι

έθλίβην έθλίφθην Θν'(Ισκω

Εύνάζομαι

'Ήδω

131

Θλίβω

θωρηχθηναι

'Ε ρωτάω ηρόμην

Έφίεμαι

61

θερέω

ηρείσατο

67

Ήμερόομαι

115

Θάπτω έτάφην

έθάφθην Θαυμάζω 203 Θεάομαι

138, 145 138, 145

51, 98, 203

θηήσατο 148 έθεασάμην 154, Θέλγω

θέλχθη

131

158

INDEX OF GREEK WORDS 273

199, 200 200

κινήσομαι

κινηθήσομαι 199, έκινήθην 127

λαμβάνομαι λήψομαι

Λάμπομαι

207

κλινάμενος

Λανθάνομαι 65 λασθη μεν 158 έλαθόμην 158, 170

έκλί(ν)θην

Λέ:yομαι

έκλέφθην

78, 244

Κλίνομαι

έκλίνην Κνάομαι

148, 150 133, 143, 150, 155 143

'gather'

90

λέξασθαι ι 48 έλέ:yην

κοιμήσασθε 153 κοιμήσατο 150 κοιμήθημεν 150 Κομίζομαι

82

λέ:yομαι

Κοιμάω

έλέχθην

137, 139 137, 139

Λέ:yω

pηθήσομαι 202

78, 244

έλέχθην

Κόπτω

146

Λείπομαι

90, 94 134, 143 έκόφθην 143

λίπεν 128 έλείφθην 171,

κόπτομαι έκόπην

έλιπόμην

Κορέννυμι

Λέπω

150 έκορέσσατο 150 Κοσμέομαι 89

Λευκαίνομαι

κορέσθην

Κ ραίνω

έκράνθην Κρέμαμαι

140

78

Κρίνω κρίνομαι

109

κρινάσθων 149 κρίνθη 133 Κρύπτω

144 133 έκρύβην 144 έκρύφην ι 44 έκρύφθην 140, 144 Κτάομαι 51, 114 έκτησάμην 154 έκτήσατο 148 κρυβήσομαι

κρύφθη

Κ τείνω

έκτάθην Κυκλέω

81, 228, 229

178

έλάμφθην 137 έλήφθην 137

Κλέπτω

138 138

l 14, 209

Λαμβάνω

209 Κίχραμαι 117 Κλαίομαι 107, 207 Κίνυμαι

έκλάπην

158

έκκυκλήθητι Λάζομαι

ι 69

173 171, 173

138 61 Λη"(ζομαι 113, ι 14 Λίσσομαι 106, 107, 109 λιτόμην 170 Λογίζομαι 51, 101, 167 έ~σάμην 154, 166, 167 Λοιδορέομαι l 05-107 έλοιδορησάμην 154, 160, 164 έλοιδορήθην ι 64 Λοiιμαι 89, 94 έλουσάμην 154 Λυμαίνομαι l 13 Λύομαι 82, 83 λiιτο 170 έλύσαο 148 Λυπέομαι 65 έλυπήθην 160 Μαίνω 55 μαίνομαι 67 έμάνην 134, 139 Μαίομαι 98 έλάπην

274 INDEX OF GREEK WORDS

105

Μαντεύομαι

Μωμέομαι

μαντεύσατο Μάρναμαι 86

148

Νεμεσάομαι νεμεσήσεται 150 νεμεσσήθη 150

Μάττω

140, 142, 144 έμάχθην 144 Μάχομαι 51, 86 μαχέσθητι 169 μαχήσατο 148 έμαχεσάμην 154 Μέδομαι 101 Μέλπομαι 207 Μέμφομαι 67, 73, 105, 107, 161, 162 έμεμψάμην 161, 162, 163 έμέμφθην 161, 162, 163 έμάγην

148

112 101

νοσφισάμην

νοσφισθείς Ξηραίνομαι

150 150

61

έξηράνθη

127

Ξυμβάλλομαι

ώδυράμην

148 51, 103

έμηχανησάμην

166

Μιαίνω μι(f

Νοσφίζομαι

160

51, 67, 207 ιj>ήθην 166 ώίσθην 150 οίσσατο 150 Cffχομαι 79, 244 Cffομαι

μητίσαντο Μηχανάομαι

207, 244 89, 94, 95

νήχομαι Νίζομαι

Ξυρέομαι

μήσατο Μητίομαι

Νήχω

ξύμβλητο

101

Μηκάομαι

79, 244 244

Νέομαι

170 89 'Οδύρομαι 105, 107

Μετατίθεμαι 104 Μήδομαι

130

νάσθη

51

Μαρτύρομαι

105, 107

Ναίω

130 65

'Όλλυμαι

Μιμνήσκομαι

169, 170 51, 105, 107 ώλοφύρθην 164 Όμόργνυμαι 94 ώλόμην

127, 133, 150 έμνησάμην 152, 161 έμνήσατο 150 έμνήσθην 152, 155, 158, 161, 164, 165, 166 μνησθηναι

'Ολοφύρομαι

'Ονειδίζομαι όνειδιοuμαι

202

'Ονίναμαι

Μίσγω μίσγομαι

82

μιγήσεσθαι 181 μίξομαι 181 μίχθη 127, 134 έμ(ε)ίχθη 139 έμ(ε)ίχθην 155

128, 134, 139, 155 117 Μυκάομαι 112, 209 Μύρομαι 105, 107 Μύσσομαι 94 Μυθέομαι 51, 105, 106 μυθήσατο 148 έμίγην

Μισθόομαι

ίόνητο "Ονομαι

170

78

Όράω

98, 100, 228 139 157 'Οργίζομαι 65, 73 ώργίσθην 161 Όρέγνυμαι 244 Όρέyομαι 77 'Ορμάω 203, 211-213, 214-216, 244, 247 όρμάομαι 77, 78, 203, 211, όράομαι

όφθηοφθητι

INDEX OF GREEK WORDS 275

'Ορμή

212, 214, 215, 216, 244, 247 όρμήσασθαι 153 όρμήσασθε 153 ώρμήσατο 148, 150 ώρμήθη 127, 150 ώρμήθην 155 ώρμηθήτην 153 όρμηθηναι 153 216

'Όρνυμι

139 170 ίhρτο 170 'Οσμάομαι 98 'Οσφραίνω 95 όρθη-

ι'όρετο

επαρδον Πέταμαι 78

170 55, 67 Πήγνυμαι 61, 63 πf\χθεν 130 έπάγη 128, 132 έπάγην 140, 155 Πηδάω 203, 230, 232, 233, 235, 236, 244, 247 έπτόμην

Πεύθω

Πημαίνω

πημάνθη Πίμπλαμαι

51, 98 159, 160 ώσφρησάμην 159 ώσφρόμην 158, 170 "Οσσομαι 66 ώσφράνθην

παρακελευσάμην Παρασκευάζομαι

154

2,90

παρεσκευασάμην

πλf\το

170

Πλάζω

πλάγχθη

130

Πλέκω

130 138, 144 έπλέχθην 137, 138, 144 Πληκτίζομαι 86 περιπλέχθη

154

Πλήσσω

228

έκπλα'γίjναι

17 Πειράω 203, 217, 221, 222, 223, 224, 247 πειράομαι 203, 217, 218, 220, 221, 222, 223, 224, 228, 229, 247 πειρήσατο 150 πειρήθη 150 έπειρασάμην 152, 154 έπειράθην 152, 154 Πείρω έπάρην 138 Πείθομαι 59, 65, 68, 73 πείσομαι 201 πεισθήσομαι 201 έπείσθην 160, 171, έπιθόμην 170, 171, Πέκω

149 149

172, 173 172, 173

134

έκπληχθηναι 134 πλήγη 128, 131, 134

Πάθος

έπέχθη Πέλομαι 207

130

61

έπλάκην

Παρακελεύομαι

έπέξατο

51, 244

Πέτομαι

όσφραίνομαι

Παρέχομαι

79, 208 208

Πέρδομαι

Πνίγω έπνίγην

140

Ποιέω

26, 112 154 έποιήθην 155 Πολιτεύω 203, 225-229, 247 πολιτεύομαι 203, 225, 226, 228, 229, 247 Πορεύομαι 78, 244 Πορίζομαι 114 Πρεσβεύω 211 πρεσβεύομαι 211 Προφασίζομαι 105, 107 Πτάρνυμαι 79, 80, 208 έπτάρην 138 επταρον 208 Πυνθάνομαι 51, 106, 107 έπυθόμην 158, 170 ποιέομαι

έποιησάμην

276 INDEX OF GREEK WORDS

Στρέφομαι

'Ραίνω

ι 3 ι, ι 40

έρράνθην 'Ράπτω

130 138 έστρέφθην 137, 138 Συλλέγομαι 82, 168, 169 συνελέχθην 146, 155 συνελέγην 146, 155 Συμβάλλομαι 104 Συμβουλεύομαι 102 Συμφέρεται 61 Συντίθεμαι 104 έστράφην

138

έρράφην

'Ρέζω

pεχθέν 127, 130 'Ρέω

pύη 130, 133 'Ρηγνύμαι

59, 61, 63 132, 134, 140

έρράγην 'Ρίπτω

144 144 'Ρύομαι l 13, l 14 έρρίφην

Σφάζω σφάζομαι 27,90, 91 έσφάyην 140, 142, 145

έρρίφθην

eσβη

128

Σφάλλω

67 77, 78

Σέβομαι Σεύομαι

εσσυτο

έσφάλην Σrpζομαι

170

101 60, 63

Σκάπτω

138

Σκέπτομαι 51, 95, 98, 99, 203 έσκεψάμην 154, 158

82 203

Σκίδναμαι

138 207 Σπεύδομαι 207 Σταθμάομαι 101 Στείχω 244 Στέλλομαι 78, 244 έστάλην 138 έσπάρην

Στερέω

137, 145

έστερήθην 137, 145 Στεφανόομαι 90 στεφανωσάμενον ι 49 Στευται

78, 105 228, 229

Στρατεύομαι

Τανύομαι

201 201

77 244

τανυσσάμενος

150

τανυσθείς 150 Ταραττόμαι 67 Τάττω τάττομαι 25 έτάγην 140, 142, 145 έτάχθην 140, 142, 145 Τείνω τείνομαι

Σπέρχομαι

έστέρην

σώσομαι Τάνυμαι

133 έσάπην 138 Σίνομαι ll3, l 14 Σιτέομαι 114 σαπήηι

έσκάφην

138

78

σωθήσομαι

Σημαίνομαι Σήπομαι

140, 142, 145

έσφάχθην

Σβέννυμι

Σκοπέω Σπείρω

77, 244

στρεφθείς

77, 244

ταθείς 130 Τεκμαίρομαι 101 Τελεiται Τέρπομαι

61 66

131, 133, 150 133 τερψάμενος 133, 150 τετάρπετο 153 τεταρπώμεσθα 133 τραπείομεν 150 έτάρπην 128 έτάρφθην 153 έτερψάμην 153 έτέρφθητε 150 Τέρσομαι 61 τάρπημεν

ταρπώμεθα

INDEX OF GREEK WORDS 277

τερσήμεναι Τεύχομαι

131

61 133 103

έτύχθη Τεχνάζομαι

έτεχνασάμην Τήκομαι

154

61, 63

έτάκην

έτήχθην

138 137, 138

Τιμάω

έτιμήσασθε τίνομαι

148

117

Τίθημι

τίθεμαι

104

183, 185, 187, 188, 191, 192, 193, 195, 198, 202 φάνηθι 157 φανο\Jμαι 178, 179, 183, 185, 187, 189, 192, 194, 195, 196, 201, 202 έφάνην 139 έφάνθην 131, 139 Φαντάζομαι 61 Φείδομαι 113 φάτο

106, 111, 207

τραφθηναι

Φθέγγομαι 105, 112 φθέγξατο 148 Φθείρομαι 60 έφθάρην 138 Φθίνομαι φθίμην 170

τρεψάμενοι

Φιλέομαι

128 134 Τρέπομαι 77, 244 τράφθη 133 τμάγεν

τμάγη

138, 151 151 έτράπετο 153 έτράπην 138, 172, 173 έτραπόμην 170, 172, 173 έτράφθην 153, 172, 173 έτρέψαμην 153 έτρέφθην 137, 138, 172, 173 Τρέφομαι 60, 208 τράφη 128, 133 έτράφην 138 [τραφον 208 έτρέφθην 138 έθρέφθην 137

87

Φιλοφρονέομαι 113 Φοβέομαι 66, 69, 70,

έφοβήθην

74, 75, 77 155

95

Φράζω

φράζομαι

66, 95, 102, 166

φράσατο 150 έφρασάμην 166

έφράσθην έφράσθης

158, 166 150

Φράσσω έφράγην

έφράχθην φραχθείς

Τρίβω

140 140 133

60, 209 209 Χάζομαι 79, 244 τύπτομαι 90 Χαίρω 203, 209 έτύπη 128, 134 έτύπην 155 έχάρη 128, 131 έχάρην 133, 155, 209 Τύφω έτύφην 140 κεχάροντο 131 χήρατο 131 Ύπισχνέομαι 105, 107 Ύ φίσταμαι 106, 107 Χαρίζομαι 113 Φαίνομαι 61, 184-187, 189, 190, 191 Χασμάομαι 79 φάνη 128, 131, 133 Χειρόομαι 114 φανήσομαι 178, 179, 180, 181, Χέομαι 61 έτρίβην

Τύπτω

145

180, 182, 190, 191, 197, 198,

Φημί

εθετο 170 Τμήγω

189, 190, 196, 197,

Φύομαι

[φυν

278 INDEX OF GREEK WORDS

Χολόομαι

73

149 127, 133, 149, 150 έχολώσατο 148, 150

χολωσάμενος χολώθη Χ ραίνω

140

έχράνθην

117 Χρέμπτομαι 79 Χράομαι

Χρίμπτω

χρίμφθη Χώομαι

Χωρέω Ψεύδω

130

67 244 110, 111 Ψεύδομαι

66, 105, 107, 110,

111 έψευσάμην έψεύσθην

160 159, 160,

167 Ψύχω

61 144, 146 έψύχθην 146 Ώνέομαι 114, 117 Ώστίζομαι 86 ψύχομαι Εψυγην