Long Diphthongs and Hiatus in Early Greek Epic Phonology and the Role of Formulaic Diction

Author(s): E. J. Bakker
Source: Mnemosyne, Fourth Series, Vol. 41, Fasc. 1/2 (1988), pp. 1-26
Published by: BRILL
Stable URL: http://www.jstor.org/stable/4431676
Accessed: 20/01/2009 08:27

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at
http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless
you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you
may use content in the JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at
http://www.jstor.org/action/showPublisher?publisherCode=bap.

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed
page of such transmission.

JSTOR is a not-for-profit organization founded in 1995 to build trusted digital archives for scholarship. We work with the
scholarly community to preserve their work and the materials they rely upon, and to build a common research platform that
promotes the discovery and use of these resources. For more information about JSTOR, please contact support@jstor.org.

BRILL is collaborating with JSTOR to digitize, preserve and extend access to Mnemosyne.

http://www.jstor.org
 Mnemosyne, Vol. XLI, Fase. 1-2 (1988)

LONG DIPHTHONGS AND HIATUS

IN EARLY GREEK EPIC
Phonology and the role of formulaic diction

BY

E.J. BARKER

1.0 Introduction

The present paper1) is concerned with one of the aspects of the

synhaphea ('phonetic continuity') in the early Greek hexameter,
viz. diphthongs and long vowels in hiatus. Hiatus may be defined
(see Lejeune 1972: 315-16) as the direct contact between a word
final and initial vowel/diphthong (or between two vowels/diph-
thongs within the word) without the occurrence of synhaphea preser-
ving phenomena, such as shortening, elision, crasis and aphaeresis.
Hiatus is thus essentially a disruption of the synhaphea; as such it
tends to be avoided in Greek poetry and cultivated prose.
It follows from the above definition of hiatus that, in the case of
hexametric poetry, hiatus after diphthongs and long vowels has to
be distinguished from
epic correption. Correption, the metrical
shortening of a diphthong or long vowel in word final position
before an initial vowel, belongs to the synhaphea preserving
devices. As such it is a consequence of hiatus, rather than a case of
hiatus itself.
The distinction of hiatus from correption, and hence the exclu-
sion of correption from discussion in the
present paper, has a
metrical consequence. We will be dealing with diphthongs and long
vowels in the arsis, rather than in the thesis of the dactylic foot2).

1) I am indebted to prof. C. J. Ruijgh and prof. R. S. P. Beekes for useful com-

ments and suggestions.
2) This usage of the terms 'arsis' and 'thesis' is modern ( = post-classical). The
meaning of the terms ('raising' and 'lowering') is applied to the voice of the poet,
which is thought to be raised at the marked ('accented') part of the foot. Original-
ly, however, (in ancient metrical theory) ??s?? and ??s?? were meant to apply to
the (rhythmic) movement of the foot of a poet or dancer. Thus, originally the
marked element ('arsis') was designated by 'thesis', the setting down of the foot;
see Allen (1973: 100).
2 LONG DIPHTHONGS AND HIATUS

Hiatus after diphthongs and long vowels means in practice 'hiatus

after diphthongs and long vowels in arsis'. Hiatus after diphthongs
and long vowels in the thesis of the foot, that is, the absence of cor-
reption, is a rare phenomenon, which has to be viewed in a wider
perspective. The coincidence of spondaic words (or word ends) with
foot end tends to be avoided in the Greek hexameter, with the
exception of the first and (naturally) the last foot3). Thus, hiatus
after diphthongs and long vowels involves word end after the arsis
of the foot, in practice the positions 3, 5, 7 and 9 of the hexameter.
Correption will not be discussed for its own sake here. It has

played, however, an important role in the history of the discussion

of diphthongs and long vowels (henceforth: DLV) in hiatus. In the
older literature4) DLV have been ordered by degree of increasing
capacity to resist correption. Thus they were divided into heavy (high
resistance) and light (low resistance)5). This purely phonological ap-
proach to hiatus and correption of DLV is inadequate. Resistance

against correption of a DLV on account of its phonological charac-

ter is an absurd notion, for resistance against correption (viz. hiatus
after a DLV standing at the end of a spondaic word or a word end-

ing on a spondee) virtually does not exist. As we shall see,

'heaviness' is proportionate to the frequency of hiatus after the
DLV in arsis. Thus, the account of the authors mentioned (n.4)
amounts to the conflation of arsis and thesis. This is unacceptable,
for it is manifestly false to speak about correption when correption
cannot possibly occur.
An entirely different approach is suggested in Athanassakis

(1970). In his account the phonological differences between the

various DLV are discarded. Athanassakis argues that phonology

3) Accordingly, hiatus after diphthongs and long vowels in thesis occurs most
often in position 2 of the hexameter (thesis of the first foot), for example ? 35
?e????', ?? 8? ??s??s?? e?? ???a??? ???????. Hiatus in other theseis yields metrically
flawed hexameters, e.g. X 199 ?? d' ?? ??e?????? d??ata? fe????ta d???e??(violation
of Meyer's First Law). See O'Neill (1942: 141), Beekes (1972), Janko (1982: 36-
7), Van Raalte (1986: 93).
4) E.g. Hoffmann (1842-8), Monro (1891), Chantraine (1948).
5) Hoffmann (1842: 53): "nec omnes eiusdem sint momenti, sed nonnullae
leviores, ut ita dicam, nonnullae ita graviores ut correptionem vix patiantur (??
enim gravius est quam ?, ?? quam ? seu ??'\ Chantraine (1948: ii, 89): "Les
diphthongues avec premi?re voyelle longue ne s'abr?gent pas volontiers".
 LONG DIPHTHONGS AND HIATUS ?

does not play a role in connection with DLV and hiatus. According
to him the only criterion on the basis of which the role of DLV as
regards the synhaphea may be assessed is metrical?, hiatus may occur
after any DLV in arsis; and any DLV in thesis may suffer correp-
tion. 'Heaviness' is disposed of as an entirely irrelevant notion.
In the present paper the behaviour of DLV in arsis is recon-
sidered with due attention to both parameters involved, metrics and
phonology. Awareness of phonology involves a fresh account of the
notion of heaviness', and awareness of metrics involves the exclu-
sion of correption and the thesis.Furthermore, a third parameter,
entirely neglected in the two above approaches, will appear essen-
tial: hiatus after DLV in arsis cannot be studied in isolation from
the characteristics and peculiarities of the kind of poetry in ques-
tion. The flexion
and juxtaposition of formulae in Greek epic dic-
tion will turn
out to be a powerful explanatory parameter bearing
on the behaviour of DLV as word end in arsis.

2.0 Preliminaries

2.1 Phonological

A typology of the DLV in question can be construed according

to the two following criteria:
a) The presence/absence of a phonological semi-vocal element
(/i/ or lui). This element is important as regards the synhaphea,
*
because it is a potential glide' between the preceding syllabic
('vocal') element of the diphthong and an initial vowel. Thus, on
the basis of this criterion a distinction can be made between
diphthongs and non-diphthongs (simple long vowels).
b) Quantity. Within the class of diphthongs a distinction
may be
made between long diphthongs (LD) and short diphthongs (SD),
on the basis of the quantity of the vowel. On the basis of this
criterion -et may be opposed to -??, -?? to -?? and -at to -a. The class
of long diphthongs partakes of both oppositions and in both cases
it is the marked term.
It is important to realize that diphthongs, and especially long
diphthongs, are diachronically unstable phonetic phenomena (Le-
jeune 1972: 225-31). They tend to be pronounced, in the course of
4 LONG DIPHTHONGS AND HIATUS

the history of the Greek language, as simple long vowels. In view

of the nature of epic diction, with its strong diachronic component,
the unstable character of diphthongs may well be a factor interfer-
ing with statistical data: a 'structural', recurrent hiatus after a
diphthong occurring in an (old) formulaic context need not be
revealing for the synchronie phonology of that diphthong.

2.2 Orthographical

In all of the discussions of DLV as word end known to me, e? and

?? are without more ado treated as (short) diphthongs. This is in-
correct: e? and ?? are orthographically ambiguous. Apart from
being the notation of a short diphthong they may also represent a
long vowel: long closed ? and ?. These vowels are, diachronically,
the result of contractions (????te? < -ee6); e?e?? < -ee?; ????? <
-??) or compensatory lengthening (e?e??a <*e?e?sa; spe???
<*sp????; ?????? < ???f??7). Inscriptional evidence points out
that /-e?/ and /-??/ are not monophthongized until the fourth cen-
tury (thus coinciding with l?i and loi). Consequently, it is false to
treat themin early epic as a single phoneme8).
To observe the ambiguity of -e? and -?? is especially important
in the case of-??. As will be seen, the great majority of the instances
of-?? as word end is formed by the o-stem genitive ending singular,
and this is clearly a case of contraction9). Therefore, statistics in
which the different 'values' of -?? are not recognized are

misleading. Thus, the complete list of the DLV under study may
be given as follows:
a) long diphthongs (LD): -a, -??, -??
b) short diphthongs (SD): -a?, -e?, -??, -a?, -e?10), -??

6) Of course the contraction of Iti and hi (??ate?>???te?) is a diphthong.

7) Compensatory lengthening as the source of the 'spurious' diphthongs -e? and
-?? is not relevant here, because the product of compensatory lengthening is never
word end.
8) See Lejeune (1972: 229) and Buben?k (1981: 45-6).
9) As a matter of fact diphthongs with a labial semivocal element (l-ul) are rare
as word end. We find some particles and some vocatives (??, a?, ?e?).
10) Notice that -e? is ambiguous too: it represents both a real diphthong (?e?)
and a 'Ionic contraction' (??e?). However, it is hardly worth while to separate the
two values of-e? in the tables, as they are too infrequent to be statistically infor-
mative.
 LONG DIPHTHONGS AND HIATUS D

c) long vowels (LV): cl) open: -a, -?, -?

=
c2) closed: -e?, -?? ( ?, f)

2.3 Statistical

No investigation of the kind of phenomenon under study here

can do without statistics. I have made countings in a number of
sample texts:Iliad A, ?, ?, X (2177 lines), Odyssey a, ?, ?, ? (1998
lines), Hesiod Works and Days, Theogony (1850 lines) and the five
Homeric Hymns of greater length: Demeter, Delian Apollo, Pythian
Apollo11), Aphrodite and Hermes (together 1914 lines). On account of
the more or less equal number of lines comparison is possible be-
tween Homer, Hesiod and the Hymns, not only of the relative fre-
quency of hiatus after the DLV (expressed as a percentage, see
below), but also of the absolute frequency.

3.0 Metrics and phonology

The phonological approach to hiatus and correption may be

exemplified by a citation from Monro (1891: 355-6, cf. Chantraine
(1948: i,89)):
"The long diphthongs (as they may be called), viz. ?? and ??, are
the most capable of resisting the shortening influence of hiatus; next
to them are e? and ??, and the long vowels ? and ?: while e?, ?? and
a? are at the other end of the scale. A measure of this may be gained
by observing how often each of these terminations is long before a
vowel, and comparing the number with the total number of times
that the same termination occurs. Thus it appears that out of every
100 instances of final
??, it is long before hiatus about 23 times.
Similarly final -?? is long 19 times, -e? 6.7 times, -?? 6 times, -? 5.7
times, -? 4 times, -e? 1.8 times, -?? 1.6 times, and -a? only 1.3 times.
Thus hiatus after ?? and ?? is scarcely avoided, while after e?, ?? and
at it is very rare".
Some things are taken together here which had better be kept
apart. The frequency of hiatus after a given DLV has to be distin-

11) Seejanko (1982: 99-100) for a survey of the arguments for the partition of
To Apollo into a Delian and a Pythian part.
O LONG DIPHTHONGS AND HIATUS

guished from
the shortening (correption) of that DLV. Hiatus is
primarily an arsis affair, whereas correption exclusively belongs to
the thesis. Monro's 100 instances indiscriminately include DLV in
arsis and in thesis. As unshortened DLV in thesis before hiatus are
virtually nonexistent, a correct percentage of hiatus after DLV
should be based solely on DLV as word end in arsis. However, in-
adequate as Monro's list may be, it reveals a clear difference be-
tween LD and SD. In fact the high percentage for LD (23% and
19% resp.) will turn out to be even
higher when the thesis cases are
subtracted. The difference
as regards hiatus between LD on the one
hand and SD/LV on the other, which is still better displayed by
tables 2A-D below, constitutes the main subject of this article.
The opposite, metrical, view is advocated by Athanassakis
(1970). His conclusion amounts to a complete reversal (1970: 144):
"The only statement which can be made with certainty on the
basis of the facts of correption in Hesiod is that any diphthong or
long vowel can be found long in the arsis of a foot and short in the
thesis of a dactyl. All diphthongs and long vowels, when found
before a vowel, behave in like fashion both in the arsis and the thesis
of a foot. Where metrical quantity is concerned we cannot indeed
maintain, or for that matter, be able to determine that some
diphthongs are longer or 'heavier' than others because their first
vowel is a long rather than a short one". (Emphasis added).
Athanassakis does indeed distinguish between arsis and thesis,
but it is as yet an open question whether he is right in minimizing
the role of phonology, as his statement that the phonological char-
acter of the DLV is immaterial is made without argumentation and
is not based on relevant statistical data.
In the following pages the
matter will be reconsidered against the background of the two views
mentioned. It will be shown that any discussion of the subject is in-
complete as long as epic diction and its diachrony are not taken into
account.

4.0 Hiatus in the arsis: some statistics

Athanassakis has to clear

away one difficulty before he can for-
mulate his metrical approach to hiatus and correption: his tables,
too, display a far higher frequency of hiatus in arsis after -?, -??,
 LONG DIPHTHONGS AND HIATUS /

-?? than after any of the other vowels or diphthongs. He therefore

suggests that this is due to the metrical character of the words of which
-?, -??, -?? are the final vowels (1970: 144): "...it may happen that
the endings ??, ??, ? are found more frequently at the end of forms
the penultimate and antepenultimate of which have the metrical
value vj^ or ?, in which case the ultimate syllable cannot be
shortened (a?a??, ?e??e???, ??stef????, ???a?ap etc.). Conversely,
a?, ??, e? may happen to come more frequently at the end of words
of which the penultimate and antepenultimate syllables are of the
metrical character -u, in which case the terminali syllable has to be
shortened (examples: ???eta?, ?????e?a?, ????e?a?, ????????, ??p???
etc)". Now adducing such examples is, on the face of it, entirely
gratuitous. One could as well cite ?????e?a?, ??stef????, e????e???
etc., and, conversely, ????????, ???a???, ??p?? etc. The only possi-
ble way for Athanassakis to substantiate his suggestion would have
been simply to count the various DLV at the end of the metrical
forms he envisages (i.e. forms the last syllable of which can only oc-
cur in the arsis: ww-, ?^w-, w?ww? etc. The following table
represents the results of this counting. The left column contains the
figures of the frequency of the forms in question12), of course in
Athanassakis' corpus {W&D, Theog, Shield)', the right column con-
tains the total figures of hiatus in arsis found by Athanassakis.

Table 1: Hiatus after word type (..)ww- in Hesiod

DLV End of (..)^- Hiatus

-?? 73 11
-?? 54 25
-a? 141 11
-e? 28 2
-?? 93 3
-? 167 21
-? 21 1
-?? 83 18

Athanassakis' suggestion is not borne outby the facts; for ex-

amples, -??, which is followed by hiatus more than twice as often

12) Including-(because of Athanassakis' example ???a???), although the

last syllable of this form is not strictly confined to the arsis; cf. W&D 297 ?? ?????
?????ta?, d d' a?? ??????? ????.
8 LONG DIPHTHONGS AND HIATUS

as -a?, occurs at the end of the (..)??- word-type less than half as
much. For Athanassakis' suggestion to be right, -?? had to come at
the end of the (..)^^- word-type about 300 times. Only
Athanassakis' figure for hiatus after -? (21 times) corresponds to a
high frequency of -? as word end of the (..)ww- type (167 times).
However, here the figure for hiatus turns out to be erroneous:
Athanassakis treats an initial vowel after ? (disjunctive particle) as
but in view of the current etymology of ?, ? (?^?, ?/e ?
hiatus13),
cf. Lat. -ve), initial vowel after ? had better not be treated as hiatus.
Once ? is left out of account, the figure for hiatus after -? in Hesiod
appears to be considerably than 21 (6 in W&D + Theog.)**).
lower
Table 1 is a first step towards presenting the frequency of hiatus
after the DLV as a ratio, rather than as an absolute figure15). In the
following tables the frequency of hiatus after the DLV in the corpus
mentioned above (2.3) is presented as a percentage, viz. as a ratio of
the figure of hiatus with the total number of occurrences of the
DLV as word end in arsis.Thus the frequency of hiatus after the
DLV is viewed in relation to the frequency of the DLV at the end
of the (..)ww- word-type as well as at the end of any other word-type
(-^-, ?, whose last syllable may, in case of correption, occur in
the thesis as well). Monosyllables have been left out of account, as
hiatus after monosyllables appears to be much less frequent than
after words of two or more syllables. For reasons of security initial
digamma has not been included in the figures on the basis of which
the percentage is computed. However, its figure is presented (in

13) This can be deduced from his treatment of initial vowel after ? in thesis
(1970: 145). According to Athanassakis there is hiatus in thesis in Asp. 407 a????
??ess?????? ? ????t???? ???f???. However, the non-occurrence of correption in the
case of ? in thesis before vowel (which is rather frequent in early epic poetry) has
to be explained in the same way as cases like I 406 ??e? xott Gf?a???a; digamma
is involved.
14) It might seem odd that the monosyllable ? is mentioned in a discussion of
the (..)ww- word-type, but A's point is that his total figures of hiatus in arsis might
be proportionate to the frequency of DLV at the end of the word-type in question.
15) Athanassakis gives absolute figures; his table (1970: 143) presents the total
number of hiatus (as well as correption) in Hesiod for each metrical position in
the hexameter separately (the arsis being marked with *A\ the two short elements
of the dactyl with 'be', and a spondee with *AB\ The separate treatment of the
various metrical positions has some advantages (see below), but one should use
percentages.
 LONG DIPHTHONGS AND HIATUS

brackets) after the figures of the total number of occurrences of the

DLV as word end in arsis (the first column of each table).

Table 2: percentagesof hiatus after DL V in arsis

A. Iliad ?, ?, ?, X B. Odysseya, ?, ?, ?

DLV Word end Hiatus % Hiatus DLV Word end Hiatus % Hiatus
-at 4 0 0 -a? 3 0 0
-?? 90(3) 31 34.4 -?? 70(2) 29 41.4
-?? 76(3) 28 36.8 -?? 78(2) 30 38.4
-at 106 (7) 9 8.4 -a? 103 (8) 8 7.7
-e? 81(3) 4 4.7 -e? 73(1) 2 2.7
-?? 118(11) 10 8.4 -?? 128 (8) 9 7.03
-a? -a?
-e? 5(2) 20 -e? 50
-?? -??
-a 28(1) 1 3. -a 26 1 3.8
-? 172(21) 22 12. ?? 143 (13) 21 14.6
-? 96(3) 7 7. -? 100 (5) 7 7
-?(*) 10 2 20 14 1 7,1
-??(?) 37 13 35. -??(?) 44(1) 14 31.8

C. Hesiod (W&D, Theog.) D. Hymns

DLV Word end Hiatus %Hiatus DLV Word end Hiatus %Hiatus
-a? 4 0 0 -a? 1 0 0
-?? 98(3) 10 10.2 -?? 67(4) 13 19.4
-?? 84(4) 20 23.8 -?? 64(3) 12 18.7
-a? 152 (5) 10 6.5 -a? 124(5) 8 6.4
-e? 73(1) 2 2.7 -e? 60 0 0
-?? 110(3) 3 2.7 -?? 84(5) 4 4.7
-a? -a?
-e? 1 0 -e?
-?? -??
?a 14 2 14.2 -a 12(1) 0 0
-? 210(7) 6 2.8 ?? 141 (16) 3 2.1
-? 45(1) 0 0 -? 85 (3) 4 4.7
-"(?) 1 0 0 -*(?) 8 0 0
-??(?) 100 (2) 12 12 -??(f) 57(3) 6 10.5

There is no proportionate relationship between the frequency of

the DLV as word end in arsis and the frequency of hiatus after
them, as appears from the widely divergent percentages in these
tables. The usefulness of ratios as against Athanassakis' absolute
figures is clear. For example, the Iliad table lists 13 instances of
10 LONG DIPHTHONGS AND HIATUS

hiatus after -?? ( = g) and no less than 22 after -?, and yet the
percentage of -?? is more than two times as high. Without the
percentages absolute figures reveal little or nothing.
Now what do the percentages reveal?
According to the tables
hiatus after long diphthongs (LD) is indeed, contrary to
Athanassakis' assumption, markedly more frequent than after SD
and LV, both absolutely and as a percentage. This seems to be so
in all stages of epic diction, though in Hesiod and the Hymns the
percentages of -?? and -?? are considerably lower (ranges of percen-
tages: Iliad 34.4%-36.8%; Odyssey 38A%A1A%; Hesiod 10.2%-
23.8%; Hymns 18.7%-19.4%).
On the basis of this observation the following, entirely phonology-
oriented, interpretation of the facts as presented by tables 2A-D
would seem to be tempting. The
percentages of hiatus after LD in
Homer are so high that apparently an initial vowel after -?? or -??
( = -LD V-) was not felt as hiatus'. In other words, we might argue
that an initial vowel after -?? or -?? did not, in Homer, involve an
interruption of the synhaphea. An adequate ground of explanation
of this could be that the two above mentioned (2.1) phonological
characteristics, vowel length and a semivocal element /-j/, are at
least in Homer still a reality. The lower percentages in Hesiod and
the Hymns would seem to be an index of their gradual disap-
pearance. In the case of SD and LV, on the other hand, either of
the two phonological characteristics is at any rate absent, and this
would seem to be reflected by the considerably lower percentages
for these sounds: the synhaphea, one may assume, is felt to be
broken when an initial vowel follows (-SD/-LV V-) and thus it
tends to be avoided, at least far more
strictly than after LD.
Thisinterpretation amounts, of course, to a reformulation of the
19th-century distinction between 'heavy' and 'light' DLV, entirely
discarding Athanassakis' non-phonological approach, which seems
indeed to be contradicted by tables 2A-D. However, a few ques-
=
tions are left unanswered. For example, why has -?? ( ?>) in
Homer percentages that are as high (Iliad) or nearly as high

(Odyssey) as -?? or -??? And is phonology the only, or for that mat-
ter, the best ground of explanation for the decreasing percentages
of LD? Is not in Hesiod and the Hymns, on the whole, initial vowel
after all DLV s less frequent than in Homer? To cope with these
 LONG DIPHTHONGS AND HIATUS 11

questions we must turn to the third explanatory parameter men-

tioned above, epic diction and its flexion and juxtaposition of
formulae.

5.0 Flexion

Whatever the role and the importance of phonology, it does not

explain the high percentages of -?? ( = ?) in tables 2A-D, as other
DLV, which are, too, marked only for length (a, ?, ?, e? ( = ?)) have
percentages that do not differ significantly from those of the short
diphthongs. The percentages of-?? are, in the Homer-tables (2A-B),
exactly as high as those of the LD -?? and -??. This is not mere
coincidence when one realizes that these three sounds have in
common their functioning as oblique
case endings of a- and o-stem
nouns16). Very often an initial vowel -??, -??, -?? is in practice
after
unavoidable, just because -??, -??, -?? is an oblique case ending. In
many of these instances the initial vowel belongs to a word that is
in some way or another related to the preceding oblique case, either
by governing it (as a preposition or preverb), or by belonging to the
same noun phrase (and hence standing in the same case). The cases
in question may be divided into five groups:
1) Prepositional expressions of a common type, mostly placed at
the end of the line:
? 402 e???st?? ??? d?f???|| (? 602 ????st?? ep? d?f???||)
? 124 ????? e?? ??d?a?e???? 11 / ? 225 ????? e? ??d???e??a? 11.
? 500 ?? ??? ?????||
W&D 599 ??t??????? ?? ?????|| etc.17)

16) -?? is never anything other than dative of o-stem nouns; -?? is most often
dative of a-stem nouns, but it may function, of course, as a subjunctive ending,
-?? ( = ?; -?? as SD in word final position does not normally occur) is always
genitive ending, except for adverbs like ????, a???? and an isolated verbal form
like d?d??.
17) In Hoekstra's notation the expressions in question are mostly P2, T2 or H2
formulae (viz. formulae filling the slot between the penthemimeres, the trochaic
caesura or the hephthemimeres resp. and the end of the line). The expressions in
question are much less frequent when they are placed beforethe main caesura (e.g.
A 30 (f 375) ??et???* (-??) ???(??) ????? (-??), with correption of the second DLV
before the trochaic caesura). Normally, prepositional expressions placed before the
main caesura (Pi or Ti formulae) contain correption of the ending of the noun,
which is placed before, rather than after, the preposition: G 233 ????? ?v ??et????,
? 36 ?a??? ?? ????dap??. This remarkable dichotomy in the localization and the
internal structure of the prepositional expressions needs further investigation.
12 LONG DIPHTHONGS AND HIATUS

2) Preverbial expressions. A noun or adjective is followed by the

verb of which it is the complement (or to whose complement it
belongs), the verb containing a preverb:
? 223 ??? ????t?et? ???p??|| (cf. W&D 27, H. Dem 286)
? 430 ?d???? ?????e ????? 11
O 24 ?est?? ?pe??set? d?f??? 11.
3) As 2), but the verb without preverb:
A 381 e??a????? ????se? (before the trochaic caesura)
A 351 p???? d? ??t?? f???? ???sat? ?e??a? ??e????
4) Hiatus between two nouns standing in the same case:
I 556 ?e?t? pa?? ???st? ??????, ?a??? ??e?p?t??? (? 242 ???t??? ?p?
???st?? ?????? ?st??s?? ??????.
? 24, ? 6, 254, ? 158 ?t?e?d?: ??a??????? (? 18 '?t?e?d?? ??a???-
*
???a) etc. (declension of Agamemnon, son of Atreus' around the
penthemimeral caesura).
5) A noun or proper name is declined within a fixed slot in the
hexameter, regardless of what follows. Example: patronymics
placed at the beginning of the line before the trithemimeres:
M 438 ???a??d^?, d?, E 457 ??de?d??, d? etc. Notice that types 1)
and 2) also yield some instances of'flexional hiatus' after -e?: A 541
??e? ?p???sf?? ???ta, I 426 (? 62) ??e? ?p?????sa?t??.
Flexional hiatus after a- and o-stem genitives and datives is a
homogeneous phenomenon. However, on account of differences in
phonological character the datives (-??, -??) will be discussed
separately from the genitive -??.

5.1 Datives: -?? and -??

The categories 1) and 2) above yield the most frequent types of

flexional hiatus; they play a central role in the discussion of the
present paper. One could ask oneself whether the prepositional ex-
pressions are original or derived from metrically equivalent expres-
sions without hiatus. Derivation may be conceived of either as flex-
ion within the same
paradigm (examples: ? 124 and ? 225 under
??? ? ?p? (?p?)
1) above, ??ss????: (-??) (?p?) ??? (???) ??ss?????
????, ? 358, ? 414, ? 345, ? 500, ? 419), or as 'flexion by analogy'
? ?p* ap???? O 275).
(????st?? (-??) ??? (?p?) d?f??? (-??) ????st^
Sometimes derivation seems plausible, for instance in the case of
 LONG DIPHTHONGS AND HIATUS 13

????? ??? pat??d? ?a??? (G 244) being a 'modification' of the much

more frequent ????? ?? pat??da ?a?a? (? 140 etc.), a typical old
'nostos-formula'. However, very often such a 'prototype' is absent
and the dative expression seems to have an existence in its own
right. Moreover, many dative prepositional expressions are of an
archaic nature. Notice, for example, the (former) presence of
digamma in ?? ??? (^)????? and the archaic prosody (initial liquid
making position) in ??? ?e????? ??p??t??.
Now formulae constitute the trademark as well as the continuity
of an epic diction. Often they represent linguistic stages much older
than the poet's own; hence they are an excellent environment for
the preservation (irrespective of actual language change) of other-
wise extinct linguistic data (lexical, morphological, phonological).
Many formulae contain prosodie irregularities which disappear in
the light of diachronic phonology. The purport of this is that the
recurrent hiatus after
-?? and -?? in the prepositional/preverbial
expressions under
1) and 2) might be indicative of a phonology of
-?? and -?? which precedes the phonology of -?? and -?? in the poet's
time. In other words, it is possible that the phonetic change of LD
(see 2.1 above) already is well on its way in Homer's time, in which
case ????st?? ??? d?f??? would be diachronically faultless as regards
the synhaphea, but not synchronically.
As for the other categories of flexional hiatus listed above, cat.5)
(declension of patronymics before the trithemimeres) is likely to be
traditional (cf. Hoekstra 1965: 34). Accordingly, hiatus occurring
here is of the same type as in the case of the prepositional idiom.
In the case of categories 3) and 4), however, formulaic connections
are less clear18). Here, without the preserving environment of an
ancient formula, the hiatus after -?? and -?? might be unavoidable
in the way any hiatus outside ancient formulae is unavoidable in the
extemporaneous composition of oral poetry. In other words, the
high percentages of hiatus of the long diphthongs in table 2A-D is
not necessarily due to the phonology of LD. They might be the con-
sequence of the structural occurrence of hiatus after -?? and -?? in

18) With the exception of the declensions of 'Agamemnon, son of Atreus'

around the penthemimeres, which is evidently just as traditional as the declension
of patronymics before the trithemimeres.
14 LONG DIPHTHONGS AND HIATUS

the common and apparently still productive prepositional formulaic

idiom, while the synchronie phonology of -?? and -?? does not war-
rant it. Before pursuing this line of reasoning, however, a word has
to be said on -?? as a genitive ending.

5.2 Genitives', -??, -oC and -oto

Whilethe percentages of hiatus after -?? and -?? might in princi-

ple be influenced by the diachronic component of epic diction, in
the case of -??, too, there is something to be reckoned with. Here
the percentages may in principle be influenced by the (former)
presence of the older genitive ending -oio, because any instance of
final -?? in arsis before an initial vowel is resolvable to the elided
form of the older ending: -??'19). Similarly ??e? before a vowel is in
principle resolvable to ??eG(?). -?? is irresolvable when it occurs in
the arsis and
is followed by a consonant. Now the irreplaceable
endings (-010 and -?? in arsis before a consonant) are necessarily the
forms the poet actually used; -?? in arsis before a vowel, on the
other hand, is not. But it does by no means follow that even Homer
(let alone Hesiod or a Hymn poet) pronounced the prepositional
phrase in ? 602 as ????st?:' ?p? d?f??? and that the -??-reading is due
to a rhapsode or a copyist.
Firstly, the declension of an o-stem noun ending in arsis before a
consonant in Homer clearly is possible: ? 91 p?????? ?a? d???t?t?,
E 348 p?????? ?a? d???t?t??, ? 29 p??e??? ?a? d???t?ta. Secondly,
it is highly improbable that the hiatus after -?? did not enter until
after Homer (i.e. in unequivocally
post-oral times), in view of the
normal strict avoidance of hiatus
in post-oral poetry.
It would indeed be wholly incongruent to suppose that the poet,
while the efficiency and usefulness of the (prepositional) system
presuppose the availability of irresolvable -??20), would, gratuitous-

19) -?? in thesis is resolvable to the uncontracted form: -00 before consonants
and -o* before vowels ('correption' of ??).
20) Cf. ? 327 ?e?? ?p? ?e????????sa?, ? 3 ?e?? p??a d??a f????sa and compare
with the ??e?-examples under 5) in 5.0: ? 497 ??e? p???????a? ?a????. Obviously
the prepositional system allows for the replacement of prepositions. Unfortunately
I have not found examples of the replacement of prepositions or preverbs with an
initial vowel by prepositions or preverbs with an initial consonant within the same
expression, containing an o-stem noun (e.g. *?est?? ?ate??set? d?f??? beside O 24
?est?? ?pe??set? d?f???); but cf. E 109 ?ata??se? d?f???.
 LONG DIPHTHONGS AND HIATUS 15

ly, use the (elided) older form before an incidental vowel21). Epic
diction is, on the whole, not concerned with indiscriminately
preserving old forms (let alone within a system the fluidity of which
is partly due to the correspondent younger form). Linguistic
'fossils' are not preserved for their own sake, but merely as long as
they are useful, that is, as long as the constant influx of contem-
porary linguistic data has not provided a younger form, or, when
the younger form is already available, as long as the expression (or
expression-type) containing the older form is still indispensable.
Tables 2A-D show that once the younger form -?? has intruded
into the diction, it becomes more and more frequent, at the cost of
-oio. The absolute frequency of -?? in arsis in Hesiod (table 2C) is
nearly three times as high as in Iliad ?, ?, ?, X (100 instances
against 37). This agrees well with the evidence presented by Janko
(1982: 50-4) of the o-stem genitive singulars in early epic: ir-
resolvable -?? is in Hesiod much more common than in Homer. It is
important to notice that it is this diachronic development to which
the lower percentages of hiatus after -?? in Hesiod and the Hymns are
due: the absolute frequency of -?? in arsis steadily increases, where-
as the absolute frequency of hiatus in arsis after -?? remains con-
*
stant (Iliad 13, Odyssey 14, Hesiod 12)22). This means that flexion'
as a useful approach to hiatus after -?? loses nothing of its
importance23).

5.3 Flexion and phonology

We have seen that it is important for the interpretation of tables

2A-D to realize that some of the DLV solely, or mainly, function

21) It would be still more incongruous in the case of ? 242 ???t??? ?p? ???st??
?????? ?st??s?? ?????? (cf. flexional type 4) above), since in using the older form
(??????') the poet would at the same time be discarding an equally old linguistic
feature of his diction, viz. digamma in (/)ast??s??.
22) On the basis of Janko's table 15 (1982: 54) we can work out the relative fre-
quency of resolvable and irresolvable -?? in arsis in Homer and Hesiod. Ir-
resolvable ( = before a consonant): Iliad once per 41.8 lines; Odysseyonce per 34.7
lines, and Hesiod once per 14.8 lines. Resolvable (before a vowel): Iliad once per
141.3 lines; Odysseyonce per 139.1 lines, and Hesiod once per 123.3 lines. Thus
the frequency of -?? before a vowel remains fairly constant, whereas the percentage
of hiatus decreases because of a higher frequency of -?? before a consonant.
23) Cf. W&D 328 ??? a?? d????a ?a????, W&D 735 t?f?? ?p???st?sa?ta; W&D
386 pe??p??????? ???a?t?? (? 833 pe??p???????? ???a?t???, a 16 pe??p???????
???a?t??).
16 LONG DIPHTHONGS AND HIATUS

as noun endings: the percentages of hiatus after ??, ??, ?? are influ-
enced by the phenomenon of flexion in the oral-formulaic poetry at
hand. Now is it flexion that accounts for the high percentages of
hiatus after -??, -??, -??, or the phonological character of these
sounds, or both? An answer to this question is likely to be com-
plicated by diachrony. In view of the normal and regular occur-
rence of an initial vowel after LD in the prepositional expressions,
which undoubtedly constitute a traditional formulaic idiom,
'hiatus' after LD must have once been 'legitimate', i.e. harmless
for the synhaphea, on account of the phonological properties of LD
(see 2.1 above).
But it does not follow that this was still so in Homer's time. The
question is obviously related to the
phonological considerations
presented above (3.0),for if the divergence of the percentages turns
out to be entirely, or to a large extent, due to flexion (i.e. a non-
phonological factor), then Athanassakis' anti-phonological ap-
proach to hiatus after DLV begins to make sense, at least syn-
chronically. If, on the other hand, it can be shown that the LV have
higher percentages independently of their functioning as noun end-
ings, the 19th century distinction between 'heavy' and 'light' DLV
appears to be basically right.
The question might also be put as follows: when one views -??,
-??, -?? as noun endings, the frequent hiatus after them has to be
considered unavoidable, but is it phonologically warranted? Of
course in the case of -?? the answer must be negative in any case:
this sound is phonologically characterized only for length (see 2.1
above), whereas it is only in the case of length + semivocal element that
we can speak of phonetic faultlessness at all. When -?? V- turns out
to be phonetically faultless, -?? V- is obviously derived from it only
on account of its functioning as a genitive ending in the same
paradigm, and when -?? V- is phonetically not faultless, -?? V- is
still less so. Thus, flexion is in any case the explanation of the high
percentages of -??.
To determine whether the high frequency of hiatus after LD in
Homer is primarily due to flexion or to phonology we have to know
how LD behave before an initial vowel outside contexts in which
their being an oblique case ending can be adduced as a ground of
explanation for the hiatus. For it is only outside those contexts (e.g.
 LONG DIPHTHONGS AND HIATUS 17

the prepositional idiom in the second half of the verse) that LD can
be compared with the DLV not (normally24)) functioning as ob-
lique case endings. This comparison has to take place in the light
of another major device of epic verse-making, the juxtaposition of
formulae25).

6.0 Juxtaposition of formulae and the caesurae

Milman Parry26) was the first to refute the idea that certain
metrical and prosodie irregularities (hiatus, brevis in longo) are
conditioned by rules prescribing that in some places in the hexa-
meter these irregularities are more 'legitimate' than in others. He
showed that the irregularities occurred more often and more
naturally, in the process of oral verse-making, in places functioning
as formula-boundaries, without the irregularity as such being
legitimate in those places27). The formula-boundaries that are rele-
vant for a study of word end in arsis are the trithemimeres (position
3), the penthemimeres (P, position 5) and the hephthemimeres (H,
position 7).
A DLV standing in hiatus
at the end of a juxtaposed formula can
be any of the DLV in the list of 2.2 above; there is prin-
mentioned
cipally no reason why it should be exclusively a noun ending (and
hence a LD). Now given the fact that hiatus is particularly frequent
at 'seams' between formulae, the juxtaposition of formulae is an
ideal framework, in principle dissociated from 'flexion', within
which LD can be compared, as regards behaviour in hiatus, with
the other DLV.
The obvious method to achieve this aimis to produce percen-
tages of hiatus for the various DLV in each of the three formula-
boundaries just mentioned. The following tables, based on the
Homeric sample texts (cp. tables 2A-B) display the resulting figures.
Position 9 has been included, because comparison with a position
not (normally) functioning as a formula-boundary is potentially

24) Occasionally -? (?t?e?de?) and -?? (?????) do.

25) I will use the term 'formula' throughout, even when the formulaic (tradi-
tional) nature of the juxtaposed phrase is doubtful or arguable.
26) Parry 1971: 191-221, notably 196. I refer to the collected writings.
27) As the irregularities are in practice unavoidable, the formulaic deviations
of the normal synhaphea are, of course, regarded as 'allowed faults'.
18 LONG DIPHTHONGS AND HIATUS

revealing. For the sakeof completeness position 11 is added too,

although hiatus here does not occur. Only words of two or more
syllables have been dealt with (hence the exclusion of the first
arsis28)) and cases where initial digamma is involved are left out of
account. Thus these tables, presenting the of the
distribution
various DLV over the arseis of the line, are a specification of tables
2A-B = 'word end'; h. = 'hiatus', and %h. = 'percentage
(w.e.
hiatus'.).
Table 3: the distribution of hiatus after the DLV over the line

A. Iliad ?, ?, ?, X

pos: 11
%h w.e. h. %h. w.e. h. %h w.e. h. %h.

-?? 20 45 36 13 36.1 25 5 20 8 37.5

-?? 20 45 31 5 16.1 19 10 52.6 7 57.1
?a? 32 0 36 10 27.7 20 0 0 17 0
-e? 24 4.1 15 2 13.3 28 0 0 3 0
-ot 35 11.4 29 2 6.8 36 5 13.8 17 0
-e? 3 0 1 0 0 2 1 50
-a 5 0 5 0 0 11 1 9.1 1 0
-? 46 8.6 56 15 26.7 53 6 11.3 11 18.1
-? 28 0 29 3 10.3 20 1 5 15 20
-e? 1 100 9 1 11.1 9 1 11.1
-?? 50 10 4 40 12 3 25 1 16.6

?. Odysseya, ?, ?, ?

pos: 11

%h w.e. h. %h. w.e. h. %h w.e. h. %h.

-?? 8 62.5 28 9 32.1 21 38 10 70

-?? 13 38.4 42 15 35.7 15 40 6 50
-at 23 8.6 41 5 12.1 16 6.2 13 0
-e? 34 0 7 2 28.5 31 0 1 0
-?? 26 0 45 5 11.1 35 5.7 13 0
-e? 1 100 1 0 ? 2 0
-a 4 1 25 16 0
-? 28 14.2 43 8 18.6 56 12.5 12 8.3
-? 37 0 31 5 16.1 20 0 5 20
-e? 1 0 3 0 0 8 0
-?? 9 33.3 18 5 27.7 8 0 66.6

28) Hiatus after LD and -??, -e? in 1 (i.e. after monosyllables) is more frequent
than after the other DLV; 'flexion', again, turns out to be the explanation. Cf.
X 127 t?? ?a?????e?a?;I 189 t?? 8 ?e ????? ete?pe?; ? 432 se?*?p?te????t??.
 LONG DIPHTHONGS AND HIATUS 19

After some inspection of the data displayed it appears that the

remarkable differences in tables 2A-D between the percentages of
LD and -?? on the one hand and of SD/LV on the other are not
in every arsis equally striking. They are most prominent in pos. 9,
and least so in pos. 5 (the main caesura), poss. 3 and 7 occupying
an intermediate position. The explanation of this divergence, we
may now go on to show, lies in the amount of juxtaposition and
flexion in each of the four arseis respectively.
Position 9 does not normally function as a formula-boundary,
and it lies within the scope of the prepositional/preverbial formulaic
idiom discussed above (5.0-5.1). This sufficiently explains the com-
plete absence of hiatus after SD, as these sounds do not normally
function as oblique cases29). Hiatus after LV (-?) occasionally oc-
curs, in the case of metathesized genitives of masculine a-stem
nouns: ??????de? ??????? (A 1, 322, ? 1530).
Position 7, too, belongs to the domain of the prepositional idiom.
Accordingly, structural hiatus after LD is frequent here. But at the
same time position 7 serves as a syntactic break or formula-
boundary, in which case an incidental hiatus is by no means con-
fined to LD:
A 363 ??a?da, ?? ?e??e ????, ??a e?d??e? ??f?.
? 291 ?a???da ?????s???s? ?e??, ??d?e? de ?????d??31).
? 299 p?? da? ???? est??e ???, ? s' ??a?e de???.
Because of cases like ? 291 and ? 299, which are obviously dif-
ferent from ???? e?? ??d?a?e????, f???? ?pepe??e?' ?ta???? (? 345), the
differences between the percentages of LD on the one hand and
LV/SD on the other are less striking than in 9, as there is no com-
pelling reason why a DLV standing at the end of a colon should be
a LD. Notice that juxtapositional hiatus in 7 occurs most often after

29) With the exception of the dative of nouns in ??: I 405 ????G ivt pet???ss?.
30) The two instances of hiatus after -? in the Iliad (table 3 A) are different, con-
sider: ? 388 ?pe??????? ?f????e?, ? 175 ??e??a???? iht ?a?ta?. Middle participles
of the metrical form w-ww- often occupy the slot between ? (the trochaic caesura)
and 9. Exigencies of synhaphea are occasionally subordinate to this 'syntactic
pattern' (and at the same time localization, cf. O'Neill 1942: 146, table 20). Cf. also
irregularities of a different type, as in ? 285 ??e????e??^ ap?pe??e.
31) Cf. A 403 d? ?????e? ? ?a????s? ?e?G,??d?e? d? te p??te?.
20 LONG DIPHTHONGS AND HIATUS

DLV functioning as noun endings, this time the nominative includ-

ed (hence the relatively frequent hiatus after -??32).
In P, the main caesura, the difference between LD and LV/SD
is even less conspicuous. Admittedly, the percentages of LD are in
fact higher than those of the other DLV, but the margin is less wide
than in positions 7 and 9. Hiatus after LV and SD is not only more
frequent here than in the other positions; it is also more evenly
distributed over the various DLV. For example, hiatus after -at, -e?,
which is rare in 7 (see note 32) frequently occurs in 5. These obser-
vations are obviously connected with the fact that position 5 is
outside the scope of the prepositional idiom, and, more important-
ly, the fact that pos. 5 is the main caesura, i.e. the position of the
hexameter with the highest amount of juxtaposition of formulae.
Juxtaposition is far more important in 5 than in 7 and it is more
important than flexion in 5 itself, in contrast with 9, and, to a lesser
degree, 7. It is true that flexional hiatus does occur in 5, as is clear
from the declension of ?t?e?d?? '??a?????? around P (cf. 5.0
above)33). And occasionally a hiatus in ? is due to a preverbial ex-
pression, as in
? 312 de?p??sa? d' ??t??? e'^??ase p???a ???a.?But such cases do
not conform to a structural, formulaic pattern, as in the case of the
prepositional/-verbial idiom in the second half of the line. In the
majority of the cases the hiatus in 5 is juxtapositional; a syntactic
break is involved. The colon before the caesura (filling the Pi-part
of the line) often has a finite verb or an infinitive, rather than a
noun, in final position. So hiatus after -?, -e?, -a? frequently occurs
in P:
? 377 ?e????? f?t? d?t?, ? d' ?? ?sp?d? ?e????? d?t?.
? 142 ta?f?' ?pa?sse?, ???e?? te ? ????? ????e?
? 184 p??f???? ??????a?, ????? d? t?? ?p??? e??a?
? 366 ?e?? ?????? te??sa? ?d' a???at?? ?e?? ?????
In the case of a LD standing in hiatus in such situations, the fact

32) And hence the lower frequency (if it occurs at all) of hiatus after DLV func-
tioning as verbal endings (-a?, -e?, -?, -? (as in t??e?, ?st??e?)).
33) This flexional formulaic pattern occasionally causes hiatus after the non-
oblique -a?: ?t?e?da?, '??a?????? ?a? ?e???a?? (T 261) beside ?t?e?d??,
??a??????? ?a? ?e?e???? (? 552).
 LONG DIPHTHONGS AND HIATUS 21

that it is a noun ending is less important than its standing at the end
of a colon:
? 499 ?e? ?? ?f?a????? ? d? f? ??de?a? ??as???
In position 3 there is, again, a wide margin between the percen-
tages of LD on the one hand and of LV/SD on the other. But then
flexion is important in this position. We have noted that the slot be-
tween the beginning of the line and the trithemimeres (?w^~) is
*
frequently filled with patronymics (often serving as run-over
words'), flexion of which for the fluidity of epic dic-
is indispensable
tion: ???e?d? ?????? (? 176); ???a??d??, ??a (? 365)34) But position
*
3 may occasionally serve as a syntactic break, in cases where flex-
ion' does not seem to be the right diagnosis: ? 78 ??? ????t?, ?? ?a?.

The upshot of the foregoing survey of the various arseis of the

hexameter is that the width of the margin between the percentages
of LD and of SD/LV per arsis is inversely proportional to the
amount of juxtaposition per arsis. The greater the number of flex-
ional or prepositional cases of hiatus, the greater are the differences
between the percentages of LD and SD/LV. In position 5, where

34) In the case of ? 365 a demarcation problem arises: are we to speak here of
flexion or of juxtaposition? Similarly is 'flexion' the appropriate term for M 438
???a??d??, d? in view of ??de?d??, d?; ??de?d??, d? (E 362, 457 resp.), while cases
where a stronger syntactic break is involved (?t?e?d??* ? d' (I 332); ?t?e?da?, ?pe?
d? (? 341)) have to be treated as 'juxtaposition'? It seems to me that Parry's distinc-
tion between metrical faults caused by the flexion and adaptation of formulae and
metrical faults caused by juxtaposition of formulae (Parry 1971: 191-239 = Lesfor-
mules et la m?triqued' Hom?re) is not always useful, as there are cases where both
concepts may be applied at the same time. For example, I 341 may be analyzed
as 'flexion' on the grounds that the need to decline a patronymic ending in 3 is
stronger than the desire to preserve the synhaphea. But at the same time we may
take into account the possibility for pos. 3 to serve as a syntactic break, as in, for
example, ???pte ????, ?pe??? (A 132). Clearly, flexion and juxtaposition here are
two sides of the same coin. In other cases, too, positing a choice between the two
concepts seems to be less useful. For example, in
I 226 ???? ??????s???'??a???????? ?t?e?da? one is led to think of the juxtaposi-
tion of a Pi-expression saying 'in the tent' and a P2-expression mentioning the
owner. In view of
? 10 ?e??e????? ???s?^???as???de?? ?pp?d?????this is quite plausible. But at the
same time we may conceive of a 'system' expressing 'in/out of/to the tent of
Agamemnon':
I 178 ?????t' ?? ???s??? '??a???????? ?t?e?da?
? 9 ????? ?? ???s??? '??a???????? ?t?e?da?.
The two concepts, therefore, seem to have a considerable overlap.
22 LONG DIPHTHONGS AND HIATUS

juxtaposition is more significant than in the other positions, on

account of 5 being the main caesura, the differences between the
percentages are least significant. They become even smaller when
the instances of hiatus in ? after LD and -?? where flexion is in-
volved are subtracted from the figures of word end and hiatus in P.
For instance, out of the 13 cases of hiatus in ? after -?? in table 3A
(Iliad), 8 are due to flexion in some way or another. When these 8
cases are left out of account, there remain 5 (= 13-8) 'juxtaposi-
tionaT cases out of 28 ( = 36-8) occurrences of word end in P. Con-
sequently, the percentage of -?? on which the ultimate comparison
with the percentages (in P) of DLV that cannot possibly function
as an oblique case ending should be based is lowered accordingly:
it is not 36.1%, as in table 3A, but 17.8%. The following table
shows once more the figures and percentages of hiatus after the
DLV in ? in the Iliad, the figures of the case endings in question,
LD and -??, being modified in the way just mentioned.

Table 4: purelyjuxtapositional hiatus in ? in Iliad ?, ?, ?, X.

DLV Word end Hiatus Percentage hiat.

-rj 28 5 17.8
-? 31 5 16.1
-a? 36 10 27.7
-e? 15 2 13.3
-?? 29 2 6.8
-? 56 15 26.7
-? 29 3 10.3
-e? 9 1 11.1
-?? 9 3 33.3

Nothing is left of the clear differences between the percentages of

LD and -?? on the one hand and SD/LV on the other as displayed
by table 2A. In table 4 the percentages do not conform to any
recognizable pattern. When the indubitable cases of flexion are
excluded, and the behaviour of the DLV in P, the only position
where a substantial residue of non-flexional cases remains35) is
studied apart, it appears that there is no significant difference
between the behaviour of LD on the one hand and LV/SD on the

35) When the same is done for the figures of position 9, the percentage of all
DLV would be zero, as juxtaposition does not occur here.
 LONG DIPHTHONGS AND HIATUS 23

other as regards the synhaphea. Apparently, then, the restrictions

posed upon the choice of the initial sound (vowel or consonant) of
the P2-part of the line are no more stringent in the case of a SD or
LV than in the case of a LD36).
Now when the behaviour of LD in ? in connection with juxta-
position is not different from the behaviour of the other DLV, it
follows that the answer to the questions posed above (section 5.3)
must be that in Homer an initial vowel after a final LD is not
phonetically legitimate; there is just as much hiatus
any as after
other DLV. The higher over-all percentages of LD are due to these
sounds being oblique case endings which create as such hiatus. This
conclusion is inescapable,since otherwise the tolerance as regards
the admission after a LD standing
of a vowel at the end of a formula
or syntactic colon ending in ? would be much greater than in the
case of a SD or LV. Table 5 shows that this is not the case.

7.0 Phonology and diachrony in Epic diction

We are now in a position to cope with the dilemma posed by

Athanassakis' (1970) metrical approach to hiatus and the 19th cen-
tury phonological distinction and 'light'
between
'heavy' DLV.
Earlier Athanassakis' suggestion to account for the higher frequen-
cy of hiatus in arsis after LD37) was proved invalid (see section 4.0
above, table 1). But now we have found that all DLV in hiatus in
P, when a syntactic break is involved, behave in the same way. It
follows, then, that Athanassakis was basically right in saying (1970:
144): "Where metrical quantity is concerned we cannot indeed
maintain, or for that matter, be able to determine that some

36) It might seem that -a is an exception (see tables 3A-B), but this vowel does
not belong to the proper language of the poet (except in dual forms like ?t?e?da
and imperatives of the type t??a). It is mainly used (and preserved) in ancient fixed
formulae with an impeccable synhaphea (?e? ?e????e??? "???,? ??a???p??
?????) and, thus, does not normally play a role in contemporary devices as flexion
and juxtaposition. In the two instances of hiatus after -? in the Homeric sample
texts, both before a syntactic break, ?e? is involved, dissociated from its formulaic
connections:
? 391 s?? s??, p?t?a ?e?, ?te ??? p??f?ass' ?pa??????
? 401 ???a s? t?? ?' ?????sa, ?e?, ?pe??sa? des???
37) And for Athanassakis, erroneously, -? (see 4.0 above).
24 LONG DIPHTHONGS AND HIATUS

diphthongs are longer or 'heavier' than others because their first

vowel is a long rather than a short one"38), though he arrives at this
judgment in an uncritical and impressionistic way. At the same
time, however, his statement (ibid.) that "all diphthongs and long
vowels, when found before a vowel, behave in like fashion" is
misleading, to say the least. The 'fashions' of the DLV differ con-
siderably, depending on the function they have as word end in the
language.
Athanassakis' position is obviously a reaction against the older
distinction between heavy and light DLV, which was, as we have
seen (section 3.0), apart from its usefulness as such, applied in the
wrong way in that it was introduced on the assumption that some
DLV suffer correption more easily than others. Correption, and the
alleged resistance against it could easily be shown to be totally
unrelated (section 1.0). Athanassakis, in rejecting the distinction,
seems to imply, though he is not clear on this point, that 'heaviness'
of a DLV is in any case an illusion. It seems to be more in accord-
ance with the facts when we say that 'heaviness' of a DLV is at least
from Homer onwards not a sufficient safeguard for the admission of
an initial vowel after it, without heaviness as such being an illusion.
In other words, the phonological distinction does not synchronically
play a role in Homer, but diachronically it certainly does.
Various features of the Homeric diction, very probably belong-
ing to the traditional epic style, provide enough evidence to state
that originally an initial vowel after a LD in word-final position in
arsis was phonetically legitimate, the -t functioning as a glide
(Schwyzer (1939: 400), see also section 2.1 above). I mention,
again, the declension of patronymics before the trithemimeres and
the prepositional/preverbial expressions at the end of the line,
many of which (-??, -?? ???) seem to have an existence in their own
right39), as well as an expression such as ??at? t?? dte (? 359, 471).
Prepositional expressions of the type discussed are very frequent
in Homer. It is to these expressions, testifying to the former

38) Athanassakis here has both arsis and thesis in view. The question whether
he is right with regard to the thesis need not be answered here.
39) For example, in addition to section 5.0, the expression ??ss????? (-??) ?vi
(?p?) ??? (????) is an organic part of the elaborate system of noun-epithet-
preposition formulae for ships, cf. Parry (1971: 111).
 LONG DIPHTHONGS AND HIATUS 25

heaviness of LD, that the high percentages of LD in tables 2A-B are

primarily due, not to the contemporary heaviness of LD. The rele-
vance of these expressions for the facts dealt with in this article
gains in importance when we realize that they remain after Homer
equally frequent, while hiatus as such tends to be more and more
avoided (see tables 2C-D dealing with Hesiod and the Hymns). It
follows that after Homer the high percentages of hiatus after LD are
to an even greater extent due to flexion, i.e. a non-phonological fac-
tor. Hiatus after LD in Hesiod as an absolute figure is less frequent
(and, accordingly, has lower percentages) than in Homer just
because hiatus in ? is less frequent. The figures of hiatus after LD
in position 7 and especially in 9 remain equally high. This can be
gauged from the following table, by which we conclude. It displays
the distribution of hiatus after LD in arsis in Homer (see table
3A-B) as compared with the parallel figures of Hesiod:

Table 5: absolutefigures of hiatus after LD

position:
Iliad: -?? 9 13 5 3
-?? 9 5 10 4
Odyssey: -?? 5 9 8 7
-?? 5 15 6 3
Hesiod: -?? 1 2 3 4
-?? 5 3 6 10

2352 ?? Leiderdorp, Lokhorst 25

REFERENCES

Allen, W. S. (1973) Accent and Rhythm. Prosodiefeatures of Latin and Greek:a study in
theoryand reconstruction,Cambridge: UP.
Athanassakis, A. (1970) Hiatus, word end, and correption in Hesiod, Hermes 98,
129-145.
Beekes, R. S. P. (1972) On the Structureof the GreekHexameter, Glotta 50, 1-10.
Buben?k, V. (1981) The Phonological Interpretationof Ancient Greek: A Pandialectal
Analysis, Toronto: University of Toronto Press.
Chantraine, P. (1948) Grammairehom?rique,Paris: Klincksieck.
Hoekstra, A. (1965) Homeric modificationsofformulaic prototypes.Studies in the develop-
ment of Greekepic diction, Amsterdam-London: North-Holland.
Hoffmann, K. A. J. (1842-8) QuaestionesHomericae, Hannover.
Janko, R. (1982) Homer, Hesiod and the Hymns. Diachronie developmentin epic diction,
Cambridge: C.U.P.
26 LONG DIPHTHONGS AND HIATUS

Lejeune, M. (1972) Phon?tique historique du myc?nien et du grec ancien, Paris:

Klincksieck.
Monro, D. B. (21891) A Grammarof the Homeric Dialect, Oxford: Clarendon Press.
O'Neill, E. G. Jr. (1942) The localization of metricalword-typesin the Greekhexameter.
Homer, Hesiod and the Alexandrians, YCS 8, 105-177.
Parry, M. (1971) The making of Homeric verse, collectedpapers, edited by A. Parry,
Oxford: UP.
Raalte, M. van (1986), Rhythm and Metre. Towards a SystematicDescription of Greek
Stichic Verse, Assen: Van Gorcum.
Schwyzer, E. (1939) GriechischeGrammatik I, M?nchen: Beck.