W. Sidney Allen-Vox Graeca_ a Guide to the Pronunciation of Classical Greek -Cambridge University...

8/20/2019 W. Sidney Allen-Vox Graeca_ a Guide to the Pronunciation of Classical Greek -Cambridge University Press (1968)

1/174

VOX GR EC


2/174


3/174

V O X G R A E C A

\S

G U I D E T O

T H E P R O N U N C I T I O N OF

C L S S I C L G R E E K

B Y

W . S I D N E Y A L L E N

Professor of ompara tive Philology in the

University of ambridge

C A M B R I D G E

A T T H E U N I V E R S I T Y P R E S S

1968


4/174

Published by the Syndics of the Cambridge University Press

Bentley Hou se, P.O. Box 92, 200 Euston R oa d, Lo ndon , N .W . 1

American Branch: 32 East 57th Street, New York, N.Y. 10022

© Cambridge University Press 1968

Library of Congress Catalogue Card Number: 68-10327

Printed in Great Britain

at the University Printing House, Cambridge

(Brooke Crutchley, University Printer)


5/174

F O R E W O R D

In its purpose, principles, and general arra nge m ent, the present

book forms a companion volume to Vox Latina (Cambridge,

1965), to which there are several cross-references (abbr. VL .

It does not, however, assume a prio r read ing of the earlier book,

and a certain amount of duplication on some of the more

general topics is thus inevitable; in particular, the Phonetic

Introduction is repeated, though with some modification.

A select bibliog raphy is add ed (apart from detailed references in

text and notes, which, though more numerous than in VL, are

limited to the m ost relevant studies) ;

x

as in VL, the classifi-

catory arrangem ent of the contents makes an alphabe tical index

superfluous—the items most likely to be consulted in such an

index would be the individual Greek letters, and full references

to the detailed discussion of these are given in the summary of

recommended pronunciations; straightforward statements of

classical or recommended values are further picked out by

underlining in the text.

As in the case of Latin, there prevailed until quite recent

years a peculiarly English pronunciation of ancient Greek,

which has now been generally superseded by a reform which

approximates to that of the original language, but seldom

transcends the limitations of native English speech-habits. In

some cases there are practical pedagogical advantages in

replacing the correct rendering by a more familiar sound; but

it is desirable in such cases that the proper value should be

known—and this usually is known within limits as narrow as

those which apply to our phonetic reconstruction of Latin.

In genera l the conclusions agree with those of Stu rtev an t s

Pronunciation

of

reek and

Latin, and particular attention is paid

to any points of difference. A book intended not only for the

1

Wo rks appe arin g in the bibliography are elsewhere referred to by au tho r s

name only, with an identifying let ter where necessary.

V


6/174

F O R EWO R

academic scholar but also for the general reader and student is

not the place for presenting the results of new and possibly

controversial lines of research ; it did, however, seem reasonable

to incorporate a revised description of the Greek tonal accent,

which rationalizes rather than contradicts previous accounts;

and also to refer briefly to the results of a study, recently

published elsewhere, on stress in ancient Greek , a subject w hich

has hitherto been virtually ignored but which may be particu

larly relevant to certain metrical phenomena.

In making practical recommendations, realism has seemed

a better counsel than perfection, and, with one exception, no

revolutionary proposals will be found. The exception concerns

our English treatment of the Greek accents, where the balance

of argument seemed to favour the abandonment of present

practice and the adoption of one which enjoys w ider acceptance

and better historical precedents. Such a recommendation is, of

course, only made after detailed historical, analytical, and

practical discussion.

The results of any historical study are only as valid as the

evidence upo n which they are based; a nd a major po rtion of the

book is therefore taken up with the presentation and evaluation

of this. The principal types of data employed in phonetic

reconstruction are: (i) statements by contemporary or near-

contemporary grammarians and other writers, (2) word-play

of various kinds, contemporary etymologies, and onom atopoeia,

(3) representations in other ancient languages, (4) subsequent

developments, (5) spelling conventions and variants, (6) the

internal structure of the language itself including its metrical

patterns. These are the same classes of evidence as were used

for Latin; but in one respect the two tasks of reconstruction are

very different. Variations in Latin are largely a function of the

time-dimension (early—classical—late), and the time-span of

the language is relatively short. At any given period of its life

one can say without gross inaccuracy, and more particularly of

the written language, tha t L at in is Latin is La tin

5

regardless

of where it is found. The end of its life as a verna cu lar language

is marked by a process of fission into a num be r of progressively

vi


7/174

F O R E W O R D

diverging dia lec ts which qui te soon acquired the s ta tus of d is

t inc t l anguages ; and the t echn iques o f compara t ive l ingu is t i cs

often en ab le us to ut i l ize this diversi ty t o establish th e etat de

langue

im m ed iate ly pri or to f ission.

Greek, on the other hand, presents a very di f ferent p ic ture .

At the t ime of our earl iest records i t is already far advanced in

the process of d ivergence,

1

be ing represen ted by a number o f

widely di f fer ing dia lec ts—all cer ta in ly recognizable as Greek,

2

but some of them very unl ike one another , even a t the same

pe r io d; as M ei l le t (p . 79) has co m m en ted , i t m ust ha ve bee n

difficult for G ree ks from d ifferent cities, sp ea ki ng different

dialects , i f not to grasp the general sense, at least to understand

one ano ther exac t ly

5

.

3

For example , an unsophis t ica ted At t ic

vis i tor to Gortys in Crete might wel l have perused the famous

L aw Co de w i tho ut i t be in g c lear to h i m th at , if he were u n

for tuna te enough to be caught in adu l te ry and remain un-

ransomed, h is captors could do wi th him

s they pleased

—in

th e w ord s of th e C od e , επί τοις ελονσ ι εμεν κρεθθαι

ο π α ι κ α

λ ε ι ο ν τ ι In some cases, moreover, as Meil let also observed,

wri t ten forms might conceal yet fur ther d i f ferences in speech—

θ, for ex am pl e, in the C re ta n κρεθθαι p ro b ab ly stood for a

sound unfamil iar to At t ic ears .

La te r a s ing le fo rm of speech , t h e K o in e , becomes d om ina nt ,

and the other d ia lec ts , wi th rare except ions (as Laconian) ,

gradual ly d ie out . The survivor fo l lows the normal processes of

l inguis t ic change,

4

i nc lud ing b or ro w in g , bu t does no t i tself

branch out in to a ser ies of new languages—some dia lec ta l

var ia t ion has of course occurred,

5

but it is relatively slight

compared wi th that of the Romance f ie ld , and there is a

genera l ly accep ted norm.

1

Even M yce nae an, in spite of i ts ear ly da te, comes now here near to represen ting

an undifferent ia ted Pro to-G reek .

2

Cf. H ero do tu s , vi ii. 7 5 : το Έ λληνικόν έόν δμαιμόν τε καΐ όμόγλω σσ ον.

3

Greek sources themselves, however, scarcely refer to the question of mutual

(un)intel l igibi l i ty: as a n isolated ex ception M r J. B. Ha insw orth draws my

attention to Pausanias, ix. 22. 3 (referr ing to Gorinna) .

4

So far as the colloquial langu age is con cern ed: we are not here concern ed with

the artificialities of the K ath are vu sa \

5

The phonetic detai ls are best studied in Thumb, Part 1.

vii


8/174

F O R E W O R D

In a much simplified diagram

1

the p attern s of development in

Latin and Greek, from their earliest attested stages, may be

contrasted as follows:

Latin

Greek ^

In describing the pronunciation of ancient Greek a choice

thus has to be ma de not only of time bu t also of place; a nd, not

surprisingly, it is fifth-century Attic that we select as the goal of

our inquiry—though, as an aid to focusing upon this point in

the continuum, we shall often have occasion to refer to other

dialects and to earlier and later stages of Attic. It is not of

course suggested that literature of other periods and dialects

should not be read aloud—but it is assumed that it will be

read approximately as it would have been by a fifth-century

Athenian; in the case of later literature this is inevitably an

artificial procedure, but the differences between fifth- and

fourth-century Attic are in any case negligible, and for phonetic

purposes both may be included under the cover-term of

classical Greek

5

. For later stages a reasonable amount of

information is given, so that the purist who is so inclined may

take the necessary precautions to avoid anachronism; such

information may also be of interest as providing links with the

modern language. In one case, however, rather more attention

has been paid to a non-Attic form of speech in its own right—

namely the Homeric dialect

5

, for the reason that an Attic

rendering in some respects fails to account for certain metrical

peculiarities; it is not proposed that a H o m er ic pronunciation

1

E.g. disregarding phen om ena of convergence in Greek wh ich may ha ve been

particularly marked in the period preceding elimination of dialects in favour of the

K oin e; cf. Chadw ick p. 4.

viii


9/174

F O R E W O R D

should be attem pte d, b ut sufficient exp lana tion is given to make

these phenomena intelligible.

For the Attic inscriptional m aterial I have relied primarily on

the examples in Meisterhans-Schwyzer, cross-checked in the

CIA

; but as the SEG inter alia bears witness, a mass of new

material is now available, which often provides better examples

and evidence for more accurate dating of phonetic changes.

I have in some cases been able to incorporate such findings, but

until we have a N ew Meisterhans* the exploitation of much

of the newer material is a time-consuming and haphazard

business.

With regard to inscriptional evidence in general, it should be

mentioned that a change of sound must commonly have ante

dated its first indication in spelling, let alone the general

adoption of a new spelling; for, as English orthography most

eloquently demonstrates, spelling tends to conservatism and to

fossilization by grammarians. For this reason, and also because

many changes are likely to have been resisted longer in actual

speech in the more literate circles of the community, it is to the

less well educated of ancient scribes that we are indebted for

much of our knowledge of pronunciation.

I am grateful to the Syndics of the Cambridge University

Press for encouraging me to undertake this further study; to

John Chadwick for reading the whole of it in draft and sug

gesting a num ber of im prov em ents; a nd to Professor H om er A .

Thompson and the American School of Classical Studies at

Athens for the photographs facing p. 67 and permission to

reproduce them.

Cambridge W . S. A.

January ig6y

ix


10/174


11/174

C O N T E N T S

Foreword

page ν

Abbreviations xiii

Phonetic Introduction ι

(i) Syllable, vowel and consonant, ι. (ii) Consonants,

2.

(iii) Vowels, 3. (iv) Accent, 5. (v) Speech and

writing, 6.

1 CONSONA NTS 10

(i) Voiceless plosives, 12: a) unaspirated, 13: τ, 14;

κ, 15; 'ko pp a' , 15;

b)

aspirated, 16: development to

fricatives, 20; φ θ , χθ , 24. (ii) Voiced plosives, 27.

(iii) Labio-velars,

30.

(iv) Nasals,

31: [Q] 33.

(v) Liquids ,

38: λ, 38; p, 39· (vi) Fricatives (σ), 43. (vii) Semi

vowels, 44 : 'd ig a m m a ', 45 ; [y], 49. (viii) The aspi

rat e [h], 50. (ix) Consonant-groups represented by

single symbols, 53: α ) ζ, 53; b) ξ and ψ , 56. (x) ττ/σ σ ,

57:

'sampi ' , 57.

2 VOW ELS 59

(i) Simple vowels, 59: a, 59; €, o, 60; 1, 61; υ, 62;

η and ei, 66 ; ω and ου, JI. (ii) Diphthongs, 75:

αι, 7 5; α υ, 76 ; ευ, 76 ; οι, 77 ; 'dip h th ongs' before

vowels, 77. (iii) 'Long

5

diphthongs, 80.

3 VO W E L -L E NG T H 8 4

'Hidden quantity', 86. Loss of length-distinctions,

88.

XI


12/174

C O N T E N T S

4

V O W E L - J U N C T I O N page

90

' H i a t u s

5

, 90 . 'C or re p t io ep ica ' , 9 1 . Cras is and syn-

ae resis , 9 2 . σννί̂ η σ ις, 9 3 . E lis ion, 94 . ν έφ ελκνστικόν,

95 .

Prodelision, 96.

5 Q U A N T I T Y 97

Length and quant i ty , 97. Syl labic d ivis ion, 98.

Quan t i t y and du ra t i on , 99 . 'Co r r ep t io A t t i ca ' , 100 .

6 A C C E N T 106

A cce ntu al m ark ing , 114. T h e ch an ge to a s t ress-accent ,

119. Stress in classical Greek, 120.

A P P E N D I C E S

A

1 T h e pro nu nc ia t ion of Greek in En gla nd 125

2 T h e ora l ac cen tua t ion of Gre ek 134

Β

ι Selected qu ota t ion s f rom anc ient gr am m ar ia ns

an d oth er wri ters 145

2 C hro no log y of sources 150

Index of Greek technical terms, etc.

152

Select Bibliography

153

SUM M ARY OF RE CO M M E NDE D PRO NUN CI AT I O NS 1 5 5

I L L UST RAT I O NS

Fig. i . T h e orga ns of speech xvi

2.

Classical A tt ic vowel-systems 59

3.

De velo pm ent o f At t i c long vowels an d

c

sho r t '

d iph thongs 74

4 .

D eve lopm ent o f the ' l o n g ' ι-d iph tho ngs 82

5. Ty pe s of vow el- junct ion 92

6 . Inc iden ce of ' co r rep t io A t t ic a ' 105

7.

T h e ep i tap h of S ic ilus n o

8 . D eve lopm ent o f the t rad i t ion a l Engl i sh p ro

nu nc ia t i on of Greek vowels an d dip hth on gs 132

Plate. Early and la te At t ic inscr ip t ions facing p. 67

xii


13/174

A B B R E V I A T I O N S

ι . Journals , ser ies , e tc .

AC

AJA

AL

ArchL

BCH

BSOAS

BZ

CIA

CIL

cj

CP

Cd

CR

DAWB

GG

GL

ICPS

IG

IGA

JA

KZ

MF

NTS

PhW

REG

RhM

RL

SbAWW

SC

Uantiquite classique

American Journal of Archaeology

Acta Linguistica

Archivum Linguisticum

Bulletin de correspondance hellenique

Bulletin of the School of Oriental and African Studies

Byzantinische Zeitschrift

Corpus Inscriptionum Atticarum

Corpus Inscriptionum Latinarum

Classical Journal

Classical Philology

Classical Quarterly

Classical Review

Deutsche Akadem ie der Wissenschaften zu Berlin

Grammatici Graeci

(Teubner , ι867-1910: repr . 1965)

Grammatici Latini

(Teubner , 1857-1880)

International Congress of Phonetic Sciences

Inscriptiones Graecae

(Berlin)

Inscriptions Graecae Antiquissimae

(ed. Roehl)

Journal Asiatique

Kuhns Zeitschrift =

)

Zeitschrift

für vergleichende

Sprachforschung

Le Maitre

honetique

Norsk Tidsskrift for Sprogvidenskap

Philologische Wochenschrift

Revue des etudes grecques

Rheinisches Museum fur Philologie

Ricerche linguistiche

Sitzungsberichte der k. Akademie der Wissenschaften

zu Wien

phiL hist.

KL

Studii Clasice

x


14/174

A B B R E V I A T I O N S

SEG Supplementum Epigraphicum Graecum

SIFC Studi italiani di ilologia classica

Τ Α Ρ Α Transactions of the American Philological Association

TCLP Travaux du Cercle Linguistique de Prague

TPS Transactions of the Philological Society

WSt Wiener Studien

ZPh Zeitschrift fur Phonetik

2 E d i t i o n s o f g r a m m a t i c a l a n d t e c h n i c a l w o r k s

C M . Co nsb ruch (Hephaes t ion ,

Enchiridion, cum comm en-

tariis veteribus.

T eu bn er , 1906).

Η A . H i lg a r d

Scholia in Dionysii Thracis Artem Gramm ati-

cam

=

GG,

1. iii)

A. H i lga rd (Theodos ius ,

Canones;

Choeroboscus,

Scholia

in Theod Can., i = GG, iv. i . Choeroboscus , Scholia,

ii = GG, iv. i i ) .

Κ Η . K eil

{GL,

i-vii)

L A . Len tz (He rod ianus Techn icus =

GG,

in. i/ii)

Μ Η . S . M a c r a n ( A ris to xe nu s,

Harmonics.

O xfo rd, 1902)

S R . Sch neid er (Apol lonius Dyscolus ,

Scripta minora = GG ,

11. i

U G. U hl ig (Dionys ius T hr ax ,

Ars Grammatica; Supplementa

Artis Dionysianae vetusta = GG, 1. i)

G . U hl i g (Apol lonius Dy scolus , De Construction = GG,

11.

ii)

U R H . Usen e r L . R ad e rm ac he r (D ionysius of Ha l i ca r -

nassus , Opuscula, ii. Te ub ne r , 1904-29)

W I R . P . W inn ing ton - In g ra m (Aris ti de s Q u in t i l ianus , De

Musica.

Teubner , 1963)

xiv


15/174

J e ne croy pas qu il y ait personne qu i n e

voye quelle misere c est de ne rien com-

prendre en cette Langue que par les yeux.

G . LANCELOT, Nouvelle Mitkode pour apprendre

facilement la Langue Grecque


16/174

Fig. ι. The organs of speech.

Β

Back of tong ue

Ε

Epiglottis (draw n over

windpipe when swallowing)

F

Food-passage

G G um s (alveoli)

Η H ard pa la t e

L

La rynx , wi th Ada m s app le

y

Μ M iddle of tongu e

Ν

Nas al cavity

Ρ

Pharynx

S Soft pa late (ve lum ),

in lowered position

Τ Tongue- t ip

U

U v u l a

V V oca l cords (glottis)

W

Windp ipe

[After Ida G. Ward, he

Phonetics

of

nglish]


17/174

P H O N E T I C I N T R O D U C T I O N

i ) S y l l a b l e , v o w e l a n d c o n s o n a n t

In any ex tended u t te rance , in any language , the re i s an a l t e r

nat ion of sounds having more and less acoust ic power , or

son o r i t y , so t ha t a d i ag r am m at i c r ep re sen t a t i on o f t he

ut terance would comprise a success ion of h igh and low points .

These points would occupy various levels on a scale of sonori ty,

but i t i s only thei r re la t ive posi t ions compared wi th preceding

and fo l lowing sounds tha t a re immedia te ly re levan t .

1

T h e

n u m b e r o f S Y L L A B L E S i n an u t te rance genera l ly cor responds

to the nu m be r of h igh po in t s . T h e sounds w hich hab i tua l ly

occur a t these points are termed

V O W E L S ,

whi ls t those which

habi tua l ly occur a t low po in t s a re t e rmed C O N S O N A N T S .

Some types of sound, however , may occupy e i ther h igh

or low points re la t ively to thei r neighbours ; such sounds

ar e classified as vowels in th eir forme r ( n u c le ar ) func tion,

bu t a re genera l ly t e rmed

S E M I V O W E L S ,

and classified with

the con sonants , in the i r l a t t e r ( m ar g i na l ) func t ion . M a n y

languages employ di f ferent symbols to indicate th is d is t inct ion

of functions ( thus English y a n d w for the consonants corre

spo ndin g to the vowels

i

a n d

u)

; in classical Greek, however, the

m ar gin al funct ion of ι an d u is very res t r ic ted , m uc h m ore so th an

in En gl ish or L at i n , an d no specia l symbo ls are used to ind icat e i t .

F inal ly , two success ive vowel-sounds may occur as inde

pendent sy l lab ic nuc le i , the necessa ry marg in be ing c rea ted by

some d imin ut ion of energy be twe en them , even tho ug h they

m ay h av e the sam e degree of in he ren t sonor i ty , as e .g. in

όγδοος, Δ ιί, insc r. αθη ναα ( th ou g h this situ ati on is ra re in A tti c,

bein g general ly resolved by c o n tr a ct io n in to a s ingle syl lable ,

as

a

Α θήνα) .

1

I t should also be mentioned that we are at present concerned only with the

in r nt sonority of th e sound s, igno ring such pros odic * factors as stress, pitc h, an d

du rat i on , which also con tr ibute to overal l promin enc e (cf. Jon es Λ ) , §§2o8fF.;

Gimson, pp. 216 ff .) .

1 I A V G


18/174


i i ) C o n s o n a n t s

A primary classif icat ion

of

c onsonan t s

is

i n to

the

ca tegor ies

of

V O I C E D

and

V O I C E L E S S .

Voiced sounds involve an approxi

m a t i o n

of the two

edges

of the

vocal cords ,

so

t h a t w h e n

air

passes th rough them it sets u p a cha rac te r i s t i c v ib ra t ion , known

technica l ly as g l o t t a l t one

5

or

V O I C E ;

voiceless sounds involve

a c lear separa t ion

of the

co rds ,

so

t h a t

no

such vibra t ion occu rs .

The difference may be exemplif ied by the En glish (voiced)

ζ

and (voiceless)

s. If the

e a r s

are

closed,

the

v i b r a t i on

of th e

fo rmer

can be

c l ea r ly heard

by the

s p e a k e r ;

the

v i b r a t i on

can

also be felt by p l a c ing a finger on t he p ro tub era nc e of the thyro id

c a r ti la g e ( A d a m s a p p l e ) .

S o u n d s

ma y be

furth er classif ied ac co rd ing

to the

pos i t ion

or

organ invo lved in t h e i r a r t i cu l a t i on . Thus

L A B I A L

or B I

L A B I A L )

involves the a r t i cu la t ion of t h e two l ips e.g. En gl ish

p),

L A B I O - D E N T A L

the

a r t i cu la t ion

of th e

up pe r t e e th

and

lower

lip e.g.

Engl ish

f)

9

D E N T A L

the

a r t i cu la t ion

of the

t ongue - t i p and uppe r t e e th e.g. En g l i sh th)

y

A L V E O L A R

the

ar t i cu la t ion

of

t he tongue- t ip

and

u p p e r g u m s

e.g.

Eng l i sh

t),

P A L A T A L

the a r t i cu l a ti on of the m i d - p a r t of the t o n g u e and

t he ha rd pa l a t e ,

VE L AR

the ar t ic ula t ion of the back of the ton gu e

a n d the soft palate or v e l u m e.g. Engl ish

k).

I f the speech-organs form a comple te c losure , dur ing which

a ir

is

p re ve nte d f rom pass ing unt i l

the

c losure

is

re leased,

the

resul t ing sound is t e r m e d a S T O P . S tops are fur ther sub divid ed

in to

P L O S I V E S

and

A F F R I C A T E S .

Engl ish

has the

p losives/?,

b

(bi labial , voiceless and vo i ced ) ,

t, d

( a lveo la r ) , an d £,

g

(ve lar) .

For affricates, see u nd er f r ica t ives below .

If

the

vocal cords

are

left open

for a

br ief period after

the

release of a s t op , p roduc ing an a ud ib l e t ype of A - sound

immedia te ly fo l lowing ,

the

s top

in

ques t ion

is

desc r ibed

as

A S P I R A T E D :

t h e r e

is

c lear aspira t ion

of

voiceless stops,

for

e x a m p l e , at the be gin nin g of s tressed ini t ia l syl lables in Engl ish .

I n F r e n c h ,

on the

o the r ha nd , th e voca l cords

are

a p p r o x i m a t e d

a lmos t s imul taneous ly wi th the re lease, and the resul t is a

re la t ively

U N A S P I R A T E D

sound .


19/174

C O N S O N A N T S

Consonants o ther than s tops

are

broadly classif iable

as

C O N T I N U A N T S ,

and may be of va r ious types . If t h e t ongue or

l ips form a c losure , but air is a l lowed to escape via the nasa l

passages

by

l ower ing

the

v e l u m ) ,

the

resul t

is a

N A S A L

con

sonant ( somet imes , as in

VL,

referred to as a nasa l stop on

a c c o u n t

of the

o ra l c losure) .

In

mos t l anguages

the

nasa ls

are

al l inherent ly voiced; Engl ish

has the

nasa ls

m

( b i l ab ia l ) ,

η

(a lveo la r ) , and as ng in sing (ve lar) .

I f

the

o rgans

are not

complete ly c losed,

but if the

c hanne l

be tween them is so n a r r o w as to c ause an audible effect as the

air passes through

it, the

r e su l t ing sound

is

t e r m e d

a

F R I C A

T I V E . E n g l i s h e x a m p l e s a r e / a n d

υ

( labio-dental , voiceless

and

vo iced ) , den t a l as in thin (voiceless) and then (voiced) , s and ζ

(a lveo la r ) ,

and

p a l a t e - a l v e o l a r

as in ash or passion

(voiceless)

a n d pleasure ( vo i ced ) ; a voiceless velar fricative is h e a r d in

Scott ish lock. Th e

A S P I R A T E ,

h, is somet imes cal led a g lo t ta l

f r i ca t ive .

A

fricative effect

is

a l so p roduced

by the

g r a d u a l

re lease

of a

s top ,

a nd it

is th is whic h chara cter izes

the

affricates;

Engl i sh examples are p a l a to -a lveo la r as in chest (voiceless) and

jest (voiced).

If one side

of the

t ongue fo rms

a

c losure,

bu t the

o t he r s ide

pe rmi t s air to flow freely,

1

the r es u lt i s a L A T E R A L c o n s o n an t ,

such

as the

Eng l i sh

l.

Such sounds

are

som etim es classed w ith

the r -sounds as l i q u i d s see p. 38) .

i i i ) V o w e l s

Va r ia t ion s o f vow el -qua l i ty

are

effected pr im ar i ly

by

th e ra is ing

of different port ions

of

the ton gu e s surface tow ard s

the

pa la te ,

a n d by d ifferent degree s of such ra is ing re sul t ing in different

degrees of a pe r tu r e be tw een tongu e and pa la te . Vowels may

t h u s

be

classified according

to

a)

h o w

far

F R O N T

or

B A C K

t hey

are ar t icu la ted ( i.e . involving m ore forward or m o r e b a c k w a r d

areas of the tong ue an d pal ate ), an d

b)

how C L O S E or O P E N

t hey

are i.e.

i nvo lv ing grea te r

or

lesser raising

of

t h e t ong ue ) .

The re la t ions

of

the vowels

to one

a n o t h e r

may

t h e n

be con-

1

Alternatively there

may be a

central closure, with air-flow

on

both sides.

3


20/174


venien t ly represen ted in t e rms of a two-d imens iona l d iagram.

W he n so represen ted they tend to fa ll in to a qu ad r i la te r a l o r

t r i a n g u l a r p a t t e r n ,

1

such as :

Front ack

-

Close

Open

Vowels in te rm ed ia te be tw een fron t an d back a re re fe r red to as

C E N T R A L ,

an d vowels in te rm ed ia te be tw een close an d ope n as

MID

( the so-ca lled n eu t r a l vowel o f s ta nd ard sou th ern Br i t ish

English, as at the end of sofa o r finger, is a mid cen t ra l v ow el ) .

Assoc ia ted wi th the fea tures a l ready ment ioned a re var ious

degrees of l ip-ROUN DiNG; gen eral ly spe aking b ack vow els a re

associa ted wi th rounding and f ront vowels wi th i t s absence

( l ip - spread ing) . Thus the Engl i sh

u

a n d

i

in e .g.

put, pit

a re

respec tive ly c lose back ro un de d an d c lose f ron t u n ro u n d e d .

Somet imes, however , rounding is associa ted wi th a f ront vowel

and sp read ing w i th a back vowe l—thus t he F rench

u

a n d

G e r m a n u a re f ron t rou nd ed vowels , whi l s t back un ro u n d ed

vowels occur in some languages .

Vowels a re normal ly a r t i cu la ted wi th the nasa l passages

closed (by ra is ing the vel um ), bu t if the y are left op en th e resu l t

is a

N A S A L I Z E D

vowel (as e .g . in French

on,

t r ansc r ibed

phonet ica l ly as 5) .

D I P H T H O N G S

a re fo rmed by a r t i cu la t ing a vowel an d th en ,

wi th in the same sy l lab le , making a g radua l change of a r t i cu la

t ion (or

c

gl id e ) in the d i re c t ion of an o th er v owel . M os t

com m on ly , bu t no t inev i tab ly , the first e lement o f a d i ph th on g

is m ore ope n tha n the second . T h us the d iph th on g of En gl i sh

1

Such a pattern, however, applies more exactly to the acoustic effects of the

vowels tha n to their actu al physiological ar t icu lat ion ; cf. M . Joo s, Acoustic Pho

netics { anguage

monograph no. 23, 1948), pp. 49 if.


21/174

V O W E L S

high involves a glide from a towards i, ο ι how from a towards u,

an d of hay from e towards i. Considerations of the general s tructure

of a language sometimes make it approp riate to interpret a diph

thong

as a

combination

of a

vowel

and a

semivowel

j or

w

.

In many languages vowels fall into two degrees of LENGTH,

LONG

and

SHORT.

By and large the difference corresponds to a

greater as opposed to a lesser duration—but not invariably so.

Other features, notably differences of tenseness and quality,

m ay be at least as im po rtan t (they are, for example, in dis

tinguishing the so-called sho rt vowel of English bit from the

so-called long

5

vowel

ο ι

beat .

iv) Accent

In addition

to the

vowels

and

consonants

of

which

a

word

is

constituted, a particular segment of the word e.g. syllable or

vowel) may be characterized by a superimposed feature which

sets

it

off against th e othe r segments not

s

characterized. Such

a

feature is referred to as an ACCENT, and is sometimes said to

have a

ς

culm ina tive function, as forming, so to speak, the

phonetic climax of the wo rd.

The accent may be either F IXED or F REE. The former

type is exemplified in such languages as Czech, Icelandic, or

Hungarian, where the accent norm ally falls on the first syllable

of each word; Armenian, where

it

falls

on the

last syllable;

or

Polish, where it falls on the penultimate. The Latin accent is

also fixed, though it is regulated by a more complex formula

cf.

VL,

p. 83). The

free accent

is

typical,

for

example,

of

English or Russian, where it is not bound to a particular place

in each word; and this freedom of location makes it capable,

unlike the fixed accent , of differentiating the meanings of other

wise identical words: thus

e.g.

English

import

(noun),

import

(verb) ;

x

forbears, forbe rs and four bears ; Russian muka tor

m e n t ,

muka

f lour

5

\

pl cu

C

I weep

5

, pl cu

ς

Ι pay

5

.

1

More often, however, English spelling also masks differences of vowel(s) as

between homographic verbal and nom i na l forms, e.g. in the first syllable of

convict, in

both syllables

of present—and in all

four

of

analyses.


22/174


Physical ly the accentual fea ture may be manifes ted in e i ther

of two ways, by var ia t ion in the

P I T C H

of the voice

( T O N A L

acc en t ) o r by

S T R E S S ( D Y N A M I C

a c c e n t

5

) . Stress, however,

though pr imar i ly ef fected by an increase in muscular ef for t , i s a

com plex ph en om en on , in wh ich fac tors o f p i tch a nd du ra t ion

may a l so p lay an impor tan t ro le .

I t is essential to dist inguish tone from

I N T O N A T I O N .

T h e

former re fe r s to the p i tch-pa t te rns opera t ive wi th in ind iv idua l

words , wh e reas i n to na t io n re fe rs t o t he p i t ch -pa t t e rn op e ra

t ive over the whole c lause or sentence. However , there may be ,

and there usual ly is , considerable in teract ion between these two

pa t t e rns ; t hus t he p i t ch -pa t t e rn o f a g iven word may va ry

grea t ly in acc ord anc e wi th the p i tc h- pa t te r n o f th e sen tence (as

also of o ther words in the environment) ; such an effect i s some

t imes re fer red to as a p er tu rb a t io n of the word- ton es .

v ) S p e e c h a n d w r i t i n g

In the s tudy of a d e a d lan gu ag e the re is inev i tab ly a m ai n

emp has i s on the wr i t t en w ord . But i t is we l l to rem em be r th a t

wr i t ing i s secon dary to speech , an d , how ever m uc h i t m ay

devia te f rom i t , has speech as i t s u l t imate bas is . The wri t ten

symbols correspond, in a more or less complete manner , to

phonolog ica l o r g rammat ica l e lements o f speech ; and , as Andre

M ar t in et ha s po inte d out , voca l qu al i ty is d i rect ly responsible

for the l inea r i ty of speech a nd the co nse qu en t l inear i ty of

s c r i p t .

1

I t is therefore in a sense misleading to speak of writ ten

symbols as be ing pron ou nc ed — rath er i t is the o ther way ro un d ,

the symbols represe n t ing spoken e lem ents . Bu t w hen , as in the

case of ancient Greek, our u t terances most ly involve reading

f rom a w r i t t en tex t , the t rad i t ion a l t e rmino logy of p ro

no un c ing le t t e r s m ay reasonab ly be to le ra ted , an d is in fac t

main ta ined in th i s book .

In anc i en t Greek , a s i n m od e rn E uro pe an l anguages , t he

corre spo nd enc e is betw een sym bols ( le t ters) an d pho nolo gical

e lements , an d is m uc h m ore regu la r th an in some langua ges ,

1

A unctional View of Language p. 25.


23/174

S P E E C H A N D W R I T I N G

such as English or French or Modern Greek (or Irish or

Tibetan), which notoriously use different symbols or combi

nations of symbols to indicate the same sound.

It is sometimes stated that an ideal writing-system would

have a symbol for every sound—that it would in fact be a kind

of vis ib le sp eech . Since, however, the num be r of sounds in a

language is infinite, and the s am e sound prob ably never

precisely recurs, this requirement is quite impracticable. It is

also unnecessary, as alphabets from earliest times have recog

nized. The number of symbols can be reduced to manageable

proportions w ithout any resultant ambiguity by a process which

has long been unconsciously followed, though its theoretical

basis has only been worked out during recent decades.

W ha t is required is not one symbol per sound, bu t one symbol

(or combination of symbols) per P H O N E M E . A phoneme

5

is a

class of similar sounds that are signific ntly different from other

sounds, e.g. the class of t sounds in English tin, hat, etc., or the

class of

d

sounds in

din,

had

etc. The (voiceless)

t

phoneme and

the (voiced) d phonem e are different phonemes in English, and

so require distinct symbols, because tin has a different mean ing

from

din,

h t has a different mean ing from

had

etc .; in technical

terminology, the members of the d and t phonemes are in

pa rall el distribution — i.e. they can contrast significantly with

one another, and with members of other phonemes, in other

wise identical im mediate environm ents, such as

-)in,

ha{-),

etc.

On the other hand, the fact that an initial t in English (as in

tin) is more strongly aspirated than a final t (as in hat) is not

responsible for any difference of meaning, since the two varieties

occur only in different environments, and so cannot contrast

with one another—they are in com plem entary

5

and not

parallel distribution. Th ey are thus both members (or allo-

phones

5

) of the same

t

ph on em e; only one symbol is required to

write them, since the difference in sound is predictable from

their environment, i.e. initial or final position as the case may

be. It should be noted, however, tha t the phonem ic d istribution

of sounds varies from language to lang uage ; in a languag e such

as Hindi, for example, aspirated and unaspirated t sounds

7


24/174


belong to separate phonem es, since the occurrence of one or the

other is not predictable from environment and they may

contrast significantly (e.g. sat ' seven ', sdth 'wi th ' ) .

Th e nu m ber of phonemes in a language varies; the num ber of

consonants, for example, varies from 8 in Hawaiian, through

24 in English and 32 in Sanskrit, to 55 in the East Caucasian

Tabasaran and 80 in the West Caucasian Ubykh. Latin,

according to different analyses, has from 15 to 18 consonant

phonemes in native words, and classical Greek from 14 to 18.

1

In languages with very large consonant systems the number of

vowel phonemes tends to be correspondingly small (1 or 2 in

some Caucasian languages), since numerous environmental

(allophonic) variants are needed for each vowel phoneme in

order to provide additional cues for the recognition of some of

the otherwise very fine consonantal distinctions. The analytical

segmentation into discrete phonemes in fact masks much of the

complexity of actual speech. Hu m an language has been evolved

for use in less than perfect acoustic conditions, and to this end

possesses a high degree of inbuilt 're dun da ncy

5

; so that even in

a language like English the distinction between e.g. cat and

pat

depends not simply upon the consonantal difference but largely

also upon variation in the transitional phases of the following

vowel—to the extent that the wrong vowel-variant is liable to

cause misinterpretation of the consonant and, conversely, the

correct vowel-variant may induce identification of the conso

nant even if the latter is deleted.

2

This 'phonemic' principle, then, is an economic principle,

reducing redundancy and employing the minimal number of

symbols tha t is consistent w ith the un am biguous representation

of speech.

3

And the post-Eucleidean spelling of Greek (see

1

D epe nd ing on whe ther th e η sounds (see p. 37) are established as a sepa rate

phoneme, whether t and υ in their non-syllabic function are treated as consonants

(pp . 44 ff., 77 fT.), an d w he the r th e rou gh b rea thi ng is trea ted as a cons ona nt o r as

a modification of the vowel (pp. 51 f.).

2

Cf. C arol D. Scha tz, 'T h e role of context in the perceptio n of

s tops ' ,

Language,

30 (1954), pp. 47 ff.

8

T h e possibili ty of further redu ction by ' m orp ho ph on em ic' m ethod s (cf. Allen,

Sandhi, pp . i6 f . ; E . P . H am p, CP, 62 (1967), p. 44; also p. 37 below) is here

ignored to avoid undue technicali ty.


25/174

S P E E C H A N D W R I T I N G

p .

15) comes reasonably nea r to being phonem ic. T he princip al

shortcomings are (a) in th e vowels, failure to distinguish between

short and long α, ι and υ (but see p p .

85

ff.); and (b) in the con

sonants, the use of special symbols

(

c

m onog raphs') to represent

some combinations of two phonem es, viz. 3, ξ, ψ (pp . 53 ff.).

When indicating particular sounds in a phonetic notation it

is custom ary to enclose the symbols in square brackets, e.g. [t

h

]

to represent the initial sound of English tin; phonem ic symbols,

on the other han d, are conventionally set between obliques, e.g.

/t/ for the phoneme which includes the initial sound of tin and

the final sound of

hat.

In a book intended prim arily for the classi

cal and general reader rather than the technical linguist and

phonetician it has seemed desirable to keep phonetic symbols

to a minimum. Partly for the same reason the conventions of

the International Phonetic Alphabet have in some cases been

modified in the direction of (for English classical readers) more

familiar forms—e.g. by the use of [y] instead of [j] for the

palatal semivowel, and by the use of the macron instead of the

colon for vowel-length.

1

In any case, regrettably or not, the

IPA has no canonical status; it is not in fact true (as stated by

one reviewer of VL) t h a t ' the use of IPA symbols is sta nd ard '—

certainly not, for instance, in the U.S.A.; what matters is not

so much the shape of the symbol as the definition of its value.

Note:

Where English equivalents are given for Greek sounds,

the reference, unless otherwise stated, is to the standard or

c

Received Pronun ciation

9

(RP) of southern British English. T he

choice of this form of English as a basis of com parison

is

m ade on

purely practical grounds. It is impossible to cite examples that

will be equally applicable to all nationalities and dialects of

Eng lish; one must perforce select a stan dard, and ' R P ' is by

far the best documented and most familiar of such standards.

1

In discussing the Greek vowel-systems and their development there are

posit ive advantages in using the same basic symbols, with appropriate diacri t ics ,

for all mid front vowels and for all mid back vowels (rather than e.g. IPA [e:], [ε:]).

9


26/174

C H A P T E R I

C O N S O N A N T S *

Before discussing the individual consonant-sounds in detail it is

necessary to emphasize that wherever the normal spelling

writes a

double

con sonant, it stands for a correspondingly

lengthened sound.

1

This is most clearly seen from its effect on

the quantity of a preceding syllable, the first syllable of e.g.

ίππος or εννέα being always 'heavy

5

(see p. 97) although the

vowel is short. And potentially the distinction between single

and double consonants may be responsible for differences of

meaning, as in ορρός 'rump

5

beside όρος 'mountain

5

, or

έκαλύπτομεν 'we concealed

5

beside έκκαλύπτομεν 'we reveal

5

.

In English double consonants are pronounced as such only

when (as in έκ-καλύπτομεν) they are divided between separate

words or elements of a compound word—e.g.

hip-pocket, leg-

glide,

d isservice,

unnamed

(distinct from

unaimed).

In other con

texts the written double consonants have no function except to

indica te th at the preceding vowel is short—e.g. in sitting, shilling,

penny,

copper.

2

It is, therefore, the compound type of word in

English that provides the model for the pronunciation of

double consonants in Greek.

In early Greek inscriptions the double consonants are written

single (cf.

VL,

p. 11); but at Athens the double writing makes

its appearance by the end of the 6 c. B.C.

In pure Attic dialect the geminate σσ does not occur except

in compou nds such as συσσιτεΐν (from συν-σ ιτεΐν). For in some

* An asterisk after a term indicates that it is explained in the phonetic intro

duction.

1

Inscriptional spellings often show doubling of the first consonant of a group,

pa rtic ula rly if this is σ (e.g. 5 c. B.C. A ttic μαλισσ τα); b u t such do ub lin g is no t

dist inctive; i ts purpose is uncertain, and i t may be intended only to show that the

group is divided between two syllables.

2

In M iddle En glish long vowels were generally shortened before two conson ants

(cf. wisdom beside wise); and in Early Modern English double consonants between

vowels were simplified. Since, however, the double writing served to indicate the

shortness of the vowel, i t was retained and further extended to words which

originally had a single consonant (as peni, coper).

I O


27/174

D O U B L E C O N S O N A N T S

w ords , w he re var ious oth er d ia lec ts ha ve σ σ , A t t ic ( like Ion ic)

has simplified this to σ: e.g. έσονται, κατεδίκασαν, μέσος beside

Le sb ian εσσ οντοα, κα τεδικα σ σ αν, μ εσ σ οσ ;

1

and in o ther words ,

w he re m ost o the r d ia lec ts ha ve σ σ , A t t ic (l ike Boe ot ian) shows

T T : e.g. γλ ώ ττα , τέττα ρες, ιτρά ττειν besid e Io n ic γ λ ώ σ σ α ,

τέσσ ερες, ττρήσ σειν. B ut, l ike m a n y li te ra ry la ng ua ge s, li t er ar y

Attic was subject to influences from outside the restr icted area of

the spoken dia lec t , most par t icular ly f rom Ionic . And one of the

most characterist ic features of this influence is the substi tut ion

of forms with σσ for the ττ of

c

pure ' At t ic as exempli f ied by the

insc r ip t ions .

2

In fac t in t ragedy, and in prose works up to and

inc lud ing T hu cy did es , the ττ of A t t ic i s a lmo st ent i re ly avo ided .

Even though no rma l A t t i c g r ammar was u sed , and A t t i c

phonology genera l ly adopted, i t seems that the ττ was fe l t as

something of a provincia l ism by contras t wi th the σσ of most of

the res t of the G reek-sp eakin g wo rld— all the mo re to be avo ided

as a character is t ic of the speech of th e 'σ υο β οιω τοί

5

; and even

false Ionicism s (no tab ly ήσ σ α σ θαι as ag ain st A tt ic ή ττά σ θα ι

an d Io nic έσσ ουσθαι) were l iable to be pe rp et ra ted in avo idan ce

of this shibboleth.

Al though the At t ic fo rms came more and more to ga in

l i t e ra ry accep tance (and no t on ly in comedy and ora tory , where

loca l fo rms would be par t i cu la r ly appropr ia te ) ,

3

i t was not long

before the inf luence of the Koine began again to re inforce the

c l a ims o f t he gene ra l Greek σσ . Thus , whe rea s Xenophon had

favoured the ττ forms, a l ready in Aeneas Tact icus (4-3 c .

B.G.)

one finds 78 cases of σσ as against 24 of ττ; and, in spite of the

art if icial revival of ττ by the 'Att icists ' , the Koine i tself shows

few exam ples of i t (most no tab ly ή ττά σ θα ι; note a lso m od er n

1

Inscriptional forms are rendered, as in the original , without accents or

brea thing s, or dis t inction of f inal s ; cu rren t word-divisions are however emp loyed.

2

Fr om the beginn ing these show ττ except in non-A ttic nam es such as (5 c.

B.G.)

Λ αλικαρνασσιοι In th e 4 c. th er e beg in to ap p ea r a few forms w ith σ σ : e.g. in 336 o ne

inst an ce of θάλασσα (b ut Θ αλαττα still gen eral in the 3 c ) , and tow ards th e end of the

ce ntu ry the K oin e wo rd βασίλισσα, which is alway s so writte n. Oth erw ise At tic

inscriptions continue to show ττ up to the t ime of Augustus.

3

In ora tory P ericles is said to ha ve b een t he first to ado pt t he ττ forms (Aelius

Dionysius, fr. 298 Schwabe), allegedly for reasons of euphony (cf. Plato Gomicus,

fr. 30 K ock : εσω σας ήμας εκ τώ ν σίγμα τω ν Ε ύριπίδου, w ith c lear ref erence to Medea,

476 f.). See also Stanford , p p . η ϊ., 53 f.

I I


28/174

C O N S O N A N T S

Greek

π ι τ τ ά κ ι

from Attic τηττάκιον); indeed, even the Atticists

were liable to overlook an occasional σσ when their attention

was concentrated on other matters.

T he ττ of pu re A ttic is part of an isogloss having its prob ab le

po int of origin in Boeotian (which even has e.g. μ εττω , εψ α-

φ ιττα το beside Attic μέσου, έψ ηφ ίσατο). This ττ does no t derive

directly from the σσ shown by other dialects; but both ττ and

σσ are separate developments from an earlier more complex

sound , and this fact has given rise to some speculation ab ou t th e

nature of the sounds which they represent. The matter is

discussed in more detail below (pp. 57 f.).

The value of orthographic yy is separately discussed under

Υ = [Q] (PP· 33

ff

-)>

a n d t h a t o f

PP

u n d e r

Ρ

(ΡΡ·

39 ff·)·

1

i ) Voice le ss* p los ives*

In Greek, as in a number of modern languages, there were two

distinct varieties of voiceless plosive, unaspirated*

(ΤΤ,

τ, κ) and

aspirated* (φ, θ, χ). Their distinctiveness is demonstrated by

minimally different pairs such as iroposjφόρος, ττάτος/πάθος,

λέκος/λέχος. Similar oppositions are found in Sanskrit and the

modern languages derived from it (e.g. Hindi hand

c

one-eyed 7

khana 't o e a t ') , and there extend also to the voiced plosives

(e.g. Hindi bat 'thing*jbhat 'cooked rice

5

). Both aspirated and

unaspirated plosives are indeed also found in English; the

initial

t

of top, for exam ple, is clearly asp irated , b ut the

t

of

stop is not. Here, however, the contrast is not distinctive—it is

not 'phonemic' but merely 'allophonic' (see pp. 7f.); for the

two varieties never occur in identical environments, the non-

aspiration being a special characteristic of the position after s

(unlike in classical Attic, where e.g. both στένω and σθένω

occur).

The two varieties were categorized by the Greek gram

marian s as (γράμ μα ) ψ ιλόν ('smooth , plain

5

, i.e. unaspirated)

and δασύ ( 'rough

5

, i.e. aspirated). The expected Latin trans-

1

N ote , how ever, that R P provides no mode l for a dou ble [r] sound the

difference between e.g.

four

elms

dmafour

re lms is comparable with that between

an

oce n and

a notion

cf. p. 94).

1 2


29/174

VOI CE L E S S P L OS I VE S

lation of these term s would be

(littera) lenis

and

aspera

(as in the

case of

spiritus lenis jasper

tran slating πνεύμα ψ ιλόν/δασ ύ for the

' smooth

5

and ' ro ug h' breathings). But in fact the La tin terms,

as found e.g.

in

Priscian,

are tenuis

and

aspirata; and tenuis

is still

occasionally encountered

as a

term

for the

voiceless

un-

aspirated plosives in modern works of a conservative kind.

a) Unasp irated*

The fact that aspirated and unaspirated plosives were distin

guished in Greek m eans th a t a spiration must be suppressed in the

latter if confusion is to be avo ided; such a pron unc iation comes

more readily to native speakers of e.g. French than to those of

English

or

G erm an , w here voiceless plosives, more par ticularly

in initial position,

are

generally aspirated. Apart from

the

evidence of its differential function, the unaspirated pronuncia

tion of π , τ, κ in Greek is strongly suggested by the term ψ ιλόν,

and further supported

by

statements that those consonants

are

' smooth' 'which occur without the expulsion of brea th '

(jPs.-Aristotle,

De Audibilibus, 804 b,

8-11)

1

or

'which gently

propel the air' (jAristides Quintilianus, De

Musica

ii. 11,

p . 76 W L ; cf. |i. 20, p. 41).

All this evidence

is

comparatively late,

but the

same pro

nunc iation is indicated for a very early period by the operation

of w ha t is term ed

'

Grassmann's L aw ',

2

whereby the first

of

two

originally aspirated syllable-initials in a word loses its aspira

tion. In the case of an initial vowel, a form such as (present)

εχω [ekho] involves loss

of the

initial aspiration

[h]

( 'rough

breathing')

by

comparison w ith (future) εξοο [hekso], where

there is no aspirated consonant following. The same law as

applied to an initial voiceless plosive produces contrasts of the

type

(gen.

sing.) τριχός:

(dat.

plu r. ) θριξί. T hus

τ is to θ as

zero is to [h]—in other words τ stands for [t] as

Θ

stands for

[th], i.e. τ is

unaspirated,

and is

therefore app ropria tely

described

by the

same term (ψ ιλόν)

as the

'smooth bre athin g' .

1

Texts

of

references mark ed thus

f)

are given

on

pp. 145

ff.

2

After

its

discovery

in 1862 by the

mathematician

and

linguist Herm ann

Grassmann.

1 3


30/174

C O N S O N A N T S

Fina l ly the unasp i ra ted pronunc ia t ion i s en t i re ly in accord

ance w i th re la ted fo rms in San skr i t : thus e.g. π α τή ρ = Skt. pita

w h e r e p a n d t a re kno wn from th e anc ie n t In d i an pho ne t ic

t rea t i ses to have been unasp i ra ted

1

(Sanskri t in fact also has

i ts own vers ion of Grassmann 's Law, giving a l ternat ions such

as (pres.)

budhyate:

(fut.)

bhotsyati).

The voiceless unaspirated plosives, l ike the other plosive

classes , occur wi th bi labia l* (π) , denta l* (τ) , and velar* (κ)

ar t icula t ion. They are descr ibed by Dionysius of Hal icarnassus ,

for ex am ple , as be ing pro du ce d respect ively ' f rom th e extrem i

t ies of the l ips ' , 'by the tongue being pressed agains t the f ront

of the mo ut h a t the up pe r t ee th ' , an d ' by the tongu e r i sing to th e

p a l a t e n e a r t h e t h r o a t ' { \De Compositione Verborumxiv, p . 5 6 U R ) .

τ T h e descr ip t ion of the de nta ls as bein g pr od uc ed 'κ α τά τους

μετεώ ρους οδ όντας' is ra th er im prec ise a n d cou ld possibly refer

to an a lveo la r* ra the r than a pure ly den ta l con tac t . But modern

Greek shows a den ta l p ronunc ia t ion , and in re la t ive ly anc ien t

t imes th is receives suppor t f rom t ranscr ip t ions in to Prakr i t

(Middle Indian) on coins of the Greek kings of Bactr ia and

In d ia in th e 1 an d 2 c. B.C. F or in Pr ak ri t (as in S ansk ri t a n d

the modern Ind ian languages) the re i s a charac te r i s t i c d i s t inc

t ion be tween den ta l consonants ( romanized as

t

etc .) an d

' re t ro f lex ' consonants (t e tc . ) , the la t ter be ing ar t icula ted wi th

the inver ted tongue- t ip on the gums beh ind the upper t ee th .

When Engl ish words conta ining a lveolar p los ives are spoken

by Ind ians o r bo r rowed in to mode rn Ind ian l anguages , t he

Engl ish sounds in quest ion are normal ly rendered by the

Ind ian re t ro f lexes : thus e .g . Eng .

station

b e c o m e s H i n d i

stesan.

But the Greek τ , θ , δ regular ly appear as Prakr i t denta ls and

not re t rof lexes—e.g. Evukratidasa, Agathukreyasa, Diyamedasa =

Ε υκρα τίδου, 'Α γαθο κλέους, Δ ιομ ή δου ; the y are therefore l ikely

to have been t rue denta ls , as e .g . in French, and not a lveolars

as in English.

1

The Sanskrit grammarians describe the aspirated and unaspirated plosives as

'mahapranc? an d

i

alpaprana\ i .e. 'having big/l i t t le breath' respectively: cf . Allen,

p p . 37 f.

1 4


31/174

V O I C E L E S S P L O S I V E S

κ As in m an y lang uag es , the precise po int of ar t ic ula t ion of

the velar ser ies may have var ied to some extent according to the

following vowel, i .e . further forward before a front* vowel and

fur ther back before a back* vowel . Such var ia t ion would, of

course , be non-d is t inc t ive and so , by phonemic pr inc ip les ,

would no t demand symbol iza t ion , bu t would be l i ab le to be

indicated i f , by his tor ica l accident , a symbol happened to be

available (cf. VL, p p . 14 f .) . T h u s in the oldes t A t t ic inscr ip t ions

one f inds before back vowels the symbol 9 (κόππα) , which had

represented the Semit ic uvular p los ive [q] ( '

Qof )

: e .g. (p re-

550 B.C.) ευδκροσ, b u t ανδοκιδεσ . T h is pr ac tic e, ho w ev er, cease d

at an ear ly date , and wi th the off ic ia l adopt ion of the Ionic

alphabet in the archonship of Eucle ides (403-2 B.C.) the s ign no

long er existed (except as a n um er al = 90,

1

w he re i t re ta ine d i ts

o r ig ina l a lphab e t i ca l pos it ion be tween π = 8 o a n d ρ = io o , w i th

va rio us late r shape s, as e .g. S, q, l | ) . I t survived in th e west G reek

al ph ab et , a n d t he nc e as th e ς) of L at in (cf. Q ui nt i l i an , i . 4 . 9) .

T he re is no ev idence in anc ien t times fo r the 'p a l a t a l i z ed '

p ronunc ia t ion of κ as [k

y

] before front vowels which is normal

in mode rn Greek .

κ oc cu rrin g a t the en d of the pre po sit ion εκ seems to h av e

been ass imila ted to the type of consonant which fol lowed,

i .e . voiced* or aspira ted. Hence we regular ly f ind in 5 c . At t ic

in sc ri pt io n al sp elling s of th e ty p e εγ β υζά ντιο , εγ δελφ ον, εγδοι

( = έκδω ), εγ λινδο , εγλεγεν ( = εκλέγειν), a n d , less re g u la rl y ,

e.g . εχ θετον ( =

εκ

Θ ητώ ν) , εχ φυλεσ . T h e la t t e r p rac t ice , h owev er ,

ceases at th e be gi nn in g of th e 3 c . B.C., an d εκ also becom es n o rm al

before voiced ini t ials from th e 1 c. B.C. T h e w rit in g of εκ before b ot h

voiced and aspira ted consonants i s l ike ly to be normat ive ra ther

th a n p ho ne t ic ( just as in Eng l ish we genera l ize th e use of s for th e

pl u ra l , even after voiced sou nds , e .g. in

dogs,

w here i t

is

p r o n o u n c e d

[ z ] ) ;

th is normat ive spel l ing is regular in our texts , but prob

ably misrepresents the ac tual pronuncia t ion, v iz . as [eg] before

voiced

2

an d [ekh] before asp ira ted co nson ants (o ther th an χ) .

3

1

See further p . 45 , n. 1.

2

Oth er than ρ—but in fact as an init ial this was pr ob ab ly voiceless (see p p . 39 f.).

Before σκ th e

κ

seems to ha ve be en lost alt og eth er (t hu s εσκυρου =

εκ

Σκύρου, 329 B .C.),

but was also analogically restored (hence e.g. έκσκαλεύω).

3

See pp. 24 f.

15


32/174

C O N S O N A N T S

(b ) Asp irated*

The evidence for this category is required primarily to show

that in classical Attic the sounds written φ, θ, χ were aspirated

plosives, like the

ph

9

th, kh

of Sanskrit and the modern Indian

languages (and similar to the initial

p, t, k

of English or

German), and not fricatives* as in modern Greek (where

φ = labio-dental* [f

]

as in English

foot,

θ = dental [Θ] as in

English

thin,

and χ = velar

[x]

or palatal* [ς] like the German

'ach'

and

'ich

9

sounds respectively). There is no doubt that at a

later date the aspirated plosives did develop to fricatives (see

pp .

20 ff.), and so the main task will be to prove that this had

not happened as early as the 5-4 c. B.C.

The earliest evidence from ancient descriptions lies in the use

of the te rm δασ ύ, as against ψ ιλόν for the unasp ira ted series

(see p. 12). It is first found in the passage from the

De Audib.

cited ab ove,

1

wh ere th e sounds to which it app lies are described

as ' expelling the air immediately w ith the so un ds'

;

2

bu t the use

of the term may well go back further than this. An interesting

poin t ab ou t the choice of the term s δασ ύ and ψ ιλόν is th at the

same binary opposition is found in non-technical, material

uses—e.g. Hd t., iv. 175, where a wooded ridge is co ntrasted

with the treelessness of the rest of Libya; similarly iii. 32 con

trasts a lettuce-stalk with and without its leaves, and iii. 108

the presence and absence of fur on an animal. In all such cases

it is a 'privative

5

opposition, contrasting the presence with the

absence of an additional discrete feature, rather than one

inherent quality with another; Dionysius

(De Comp.

xiv, p. 57

UR) does in fact refer to the category of δασέα as having

ς

τη ν

του πνεύματος

π ρ ο σ Θ ή κ η ν \

Such a terminology would be

eminently appropriate to the opposition of aspirated and un

aspirated consonants, but hardly to the distinction between

1

T h e term s δασύτης an d ψ ιλότης ar e inde ed fou nd in Aristo tle,

Poetics,

1456 b ,

but the passage is probably an interpolation.

2

Th e words used a re '

ευθέως

μετά τω ν φ θόγγω ν \ If th e wo rk is of early au th or sh ip

(? Stra ton ), μετά with the genit ive should me an 'w it h ' , n ot (as Stu rtev ant, p . 77)

'after*, an d this mig ht be interpreted as implyin g simultaneous b rea th, i .e. fr ict ion.

But

the use of the adverb ευθέως makes this interpretation improbable (the genitive

is found with μετά meaning

*

after ' in Byzantine G reek ).

1 6


33/174

A S P I R A T E D P L O S IV E S

f r ica t ive and plos ive , i .e . be tween incomplete and complete

c losure of the organs . Moreover , the same terminology is

employed to d i s t ingu ish the ' rough ' f rom the ' smooth '

b r e a t h i n g

1

(cf . p. 13), and there is no doubt that this is a

pr iva tive o pp osit io n of th e asp ira te [h] to zero (see p p . 50 if . ) .

The grammat ica l t r ad i t ion d iv ides the consonants in to two

pr im ary ca tego ri e s, ήμ ίφ ω να and άφ ω να , co r r e spond ing to

cont inuants* and plos ives respect ively; thus e .g . Dionysius

T h r a x ,

Ars Gramm.,

p . 1 1 U , 'ή μ ίφ ω να μεν έστιν οκ τώ ·

^ ξ ψ λ μ ν ρ σ . . . ά φ ω ν α δε έσ τιν εννέα , β γ δ κ π τ θ φ χ ' . I n A r is to tl e,

Poetics,

1456b the la t t e r a re desc r ibed as 'hav ing con tac t '

(μετά προσβολής) l ike the former , but as not being pronounce

ab le w i tho ut a vowel . T h e a l loca t ion of

φ,

θ, χ to the category of

ά φ ω να is a fa ir indic a t io n of the i r p losive , non-fr ica t ive na tu re ,

since fr icat ives w ou ld b e classifiable wi th σ as ή μ ίφ ω να, be in g

c o n t i n u a n t s a n d so ' i n d e p e n d e n t l y p r o n o u n c e a b l e ' . T h e s a m e

alloc ation is foun d e ven at a m u ch late r da te in e .g. A rist ides

Q u i n t .

(De Mus.

i i . 11 , p . 76 W I ) , wh o fur ther speaks of the

δασέα as being pronounced 'ενδοθεν έκ φάρυγγας '—which

wo uld be a co m m end ab le desc r ip t ion of asp i rates bu t co mp le te ly

inappropr ia te to f r ica t ives , s ince these do not involve any

dif ference in g lot ta l ac t iv i ty but only in ora l aper ture .

Other c lea r ev idence comes f rom the language

itself.

W h e n

a f inal voiceless unaspirated plosive

(ΤΓ,

τ, κ), as in e.g. ουκ

or e l ided απ ' , κατ ' , s tands before an aspira ted vowel ( i .e . in i t ia l

[h] ) , i t is changed to φ, θ, χ; which can only mean that φ, θ, χ

here stand for aspirated [ph], [ th] , [kh], and not for fr icat ives.

2

In such cases a spell ing of the type καθ

5

ήμέραν, with the aspira

t ion also marked on the following vowel, is , s tr ict ly speaking,

redundant , s ince the asp i ra t ion i s t r ans fe r red to the consonant ;

i t i s a normal iz ing t rad i t ion or ig ina t ing in Byzant ine p rac t ice ,

but i s not genera l in those inscr ip t ions which otherwise indicate

th e rou gh b re at hi ng (see p . 50) , ju s t as i t is no t in co m po un ds

1

E.g.

Suppl. Artis

Dionysianae, p. 107 U.

2

T he fricative pronu nciation of a com parable junctio n of plosive +

as in

e.g. [gou9am] for Gotham, N.Y., is a spelling pronu nciation , based on the non-

junctional value of the digraph th in English (contrast [gotam] for Gotham, Notts.).

2 1 7

A V

G


34/174

C O N S O N A N T S

suc h as καθη μέριος. A sim ilar tran sfer of as pi ra tio n is fou nd in

crasis, e.g. ττ) ήμ ερα -> θήμ έρα, κα ι ό π ω ς -» χώ ττω ς (n ot e also ,

wi th in te rven ing ρ , the compound προ-όδος -> φ ρο ύδο ς: cf.

p .

41) ; but here the Byzant ine t radi t ion a lso omits the or ig inal

vowel -asp i ra t ion and marks the combina t ion by the s ign

Kopcovis, having the same shape as the apostrophe (and, in

modern pr in t ing , a s the smooth brea th ing) . In the case o f

compounds and established formulae the effects of el is ion and

cras is do not of course necessar i ly prove the aspira ted, non-

fr ica t ive nature of

φ,

θ, χ for the 5 c. B.C., but only for the period

of fo rm at ion ; bu t the con t in ua t io n of th is p ro nu nc ia t io n i s

ind ica ted by the sam e effects in the case of in d ep en d en t w ord s .

Fur ther indicat ions for an ear ly per iod are provided by

Grassmann 's Law (see p . 13) , which proves that a t the t ime of

i ts operat ion the relat ionship between the values of e .g. θ and τ

was the same as tha t be tween [h] and zero , i .e . presence and

absence of aspira t ion. The law appl ies par t icular ly c lear ly to

verba l redupl ica t ion . Redupl ica t ive sy l lab les normal ly repea t

the in i t i a l conso nant o f the roo t— e.g . πέ-πω -κ α ; bu t if the

root- ini t ial is φ, 0 or χ, the reduplicat ive ini t ial is π, τ or κ—

e.g. πέ-φ ευγ-α , τί-Θ η-μι, κέ-χυ-μ αι. T h e im p o rt a n t po in t he re is

that the redupl ica t ive in i t ia l i s a plosive, w hich wo uld no t be

expected if the root- ini t ial were a fr icat ive (roots beginning with

σ , which

is

a fr icat ive, form their reduplicat ive syllables with

initial σ, [h] or zero: e.g. σέ-σηρ-α, ΐ-στη-μι, ε-σταλ-μαι).

Evide nce fo r the con t inua t ion of the a sp i ra ted p los ive p ron un c ia

t ion int o th e 5 c . B.C. an d la ter is pro vid ed b y occas ional ne w

recurrences of this type of dissimilat ion, as revealed by inscrip-

t ion al spell ings— e.g. 4 c . αρκεθεω ροσ bes ide αρ χεθεω ροσ .

Similar indications are given by occasional Assimilat ions such as

lat e 5 c. Λ εχον for έχον, w ith exte nsio n of asp ira tio n to th e

ini t ia l .

1

Fur ther ev idence comes f rom the p rocedure o f ' express ive

1

It does not affect the significance of such evidence that spellings of this type

may indicate not so much phonetic assimilation (cf. p. 25, n. 4) as an analysis of

aspiration as applying to a sequence rather than to individual sounds ( theoretical

discussions by Z. S. Harris, Language, 20 (194 4), p p . 181 if.; A llen, BSOAS, 13

(1951) , pp . 939 ff.; Η . Μ . Hoenigswald, Phonetica, 11 (1964), p. 212).

18


35/174

A S P I R A T E D P L O S I V E S

do ub l in g ' of conso nants (as in e .g . ' f am i l ia r ' ά ττα , ' h y p o -

corist ic

5

Δ ικκώ , ' im i t a t i v e

5

π ο π π ύ ^ ω ) . F o r w h e n th e d o u b le d

con son an t i s φ , θ or χ , th e resul t in g form shows π φ , τθ, κχ—

e.g. ά πφ ϋς, τίτθη , κα κχά ^ω . Su ch a spel ling indica tes th a t the

lengthening of these consonants consis ted in a s top* e lement

(IT,

T , K ) , which would no t be appropr ia te i f the o r ig ina l sound

w ere a fr icative b u t entirely so if i t w ere a p losiv e: thu s [ph, t h,

kh]

->■

[pph, t th , k kh] . H er e ag ain , howe ver , th e proof only

refers to the t ime a t which the doubl ing took place , and in

many cases th is must have been long before the 5 c .

B.C.

Similar

evidence is provided by the apocopated forms of preposi t ions ,

as in Horn, κ ά π φ άλα ρα , w he re the ass imila t ion of the final

consonant to the fol lowing in i t ia l produces a s top.

When in Att ic the nasal ν was followed by the fr icat ive σ, the

nasal was generally lost or assimilated to the fr icat ive—thus e.g.

σ υν + σιτεΐν -> σ υσ σ ιτεΐν, σ υν + στέλλειν -> συσ τέλλειν. In sc r ip

t ions show that th is was not s imply an ancient fea ture inher i ted

in co m po un ds, s ince they a lso ap ply i t a t the ju n ct io n of sep ara te

wo rds— e.g . 5 c . B.C. εσ σ α νίδι, ε στελει ( = εν σ τή λ η ) . T h i s ,

however , does

not

oc cu r before φ , θ, χ, b u t th e ν is eith er

retained or changed in type ( to μ, y before φ, χ: cf . p. 31) in

th e sam e wa y as before a n u na sp ira ted plos ive : thus e .g. τη μ

φ υλη ν (3 76 B.C.) as τεμ π ο λιν (4 16 ), Λ ιερογ χρεμ α τον (410 ) as

Toy κήρυκα (353)· Th is t re a tm en t con t ras t s wi th th a t o f m od er n

Greek, where before the now

fricative

φ, θ, χ a final ν is lost in

the sam e w ay as be fore σ an d o ther co n t inu an ts— e.g . ac e . s ing ,

το φ ίλο as το σ ο υ γιά a n d

unlike

e .g . τον π α τέρα ( = [ tomba-

t


36/174

C O N S O N A N T S

further pp. iooif.). It is, therefore, highly significant that the

same op tion exists in the case of φ , θ, χ + liquid , as e.g. Sophocles,

O.C., 354-5, .. Κ αδμείω ν λάθρα | ά τουδ' έχρή σ θη .. . Th e same

is also true of voiceless plosives with nasals, and here again the

option also exists in the case of a form such as σταθμός, w hereas

it does not where a fricative (σ) is followed by a nasal as in

e.g. κόσμος.

The evidence thus seems conclusive that in 5 c. Attic φ, θ, χ

represented plosives (as π, τ, κ) and NOT fricatives (as σ, or as

φ, θ, χ in modern Greek).

The continuation of the plosive pronunciation into a later

period is shown by the fact that Latin renders Greek φ at first

as a simple p, later as ph (e.g. Pilipus, Philippus), bu t never in

classical Latin times as/, which would have been appropriate

for a fricative pronunciation. The fact, on the other hand, that

e.g. Latin

Fabius is rend ered in Greek as Φ αβιος is no counter-

indication even for the period of such transcriptions; for Greek

ha d no other way in which to represent the L a ti n /, and in such

circumstances it would be quite normal to represent it by the

symbol for the nearest available sound in Greek, even though

this were still a plosive [ph]. Fo r althoug h fricatives and aspir

ates are not identical, they are phonetically (and often historic

ally) related

—in

fact the ancient Indian phoneticians apply the

same term

1

both to the air-stream of the fricatives and to the

aspirated release of the plosives. There is an exact parallel to

this in modern times, when unsophisticated speakers of an

Indian language like Hindi borrow English words containing

an f; for, havin g no fricative [f] in their own speech, they

substitute for it the aspirated plosive—thus e.g. English

film

is

rendered by

pkilam.

It was presumably in such a context that

Cicero ridiculed a Greek witness who could not pronounce the

first consonant of the name

Fundanius

(jQuintilian, i. 4. 14).

How ever, there is no doubt tha t, as mode rn Greek shows, the

aspirated plosives did eventually change to fricatives. Evidence

is sometimes quoted which would suggest that the beginnings of

1

usman

lit. hea t, steam, vap ou r , glossed in this use as

vdyu

wind ; cf. Allen,

p. 26.

2 0


37/174

A S P I R A T E D P L O S I V E S

such a c ha ng e could be t ra ce d to the 2 c . B.C. As m en t ion ed

above , the Greek grammar ians genera l ly agree in a l loca t ing

φ , θ, χ to the sa m e cate go ry of ά φ ω να as π , τ, κ, β, δ, γ, aind no t

to the ca teg ory of ήμ ίφ ω να (as σ ) . Sextus Em pir ic us , how ever ,

(Adv. Gramm.

=

Math. I

102) ment ions tha t ' some people

3

classify φ , θ, χ w ith th e ήμ ίφ ω να ; he is himself w rit in g in th e

2 c . A.D., b u t Diogenes L aer t ius (v ii . 57) seems to a t t r ib ut e a

system of only six ά φ ω να ( π , τ, κ, β, δ, γ) to the Stoic D ioge nes

Babylonius of the 2 c . B.C., thereby implying a classif icat ion of

φ , θ, χ as ή μ ίφ ω να. Bu t oth er e vide nce is ag ain st so early a

development , and the c lass i f ica t ion may s imply be a Stoic

aber ra t ion . I t i s t rue tha t P la to in the Cratylus ( t427 A) classes

φ w i th σ i n a c a t ego ry o f ' π νευμ α τώ δη ' ; bu t he is he re ma in ly

concer

W. Sidney Allen-Vox Graeca_ a Guide to the Pronunciation of Classical Greek -Cambridge University...

Documents

Transcript of W. Sidney Allen-Vox Graeca_ a Guide to the Pronunciation of Classical Greek -Cambridge University...