Sie sind auf Seite 1von 73

THE DIALECT GEOGRAPHY OF ANCIENT GREECE

By ROBERTCOLEMAN
RECURRENT
REFERENCES
BECHTEL,
F. Die Briechischela Dialekte 1-111 (Berlin 19214).
BUCK,C. D. The Breek Dialects (Chicago 1955).
PORZIQ,
W. ' SprachgeographischeUntersuchungen zu den altgriochische
Dialekte,' 1.F. 61 (1964), 147-169.
RISCH,E. 'Die Gliederung dor Griechischen Dialekte in neuer sicht,'
Mus. Helv. 12 (1955), 61-76.
RODRUIQEZ
ADRADOS,
F. La Dialectologia Oriega coma fuente para el estudio
de las migraciones indoeuropeas en Qrecia (Acta Salmaticensia 1952).
RUIPEREZ,
M. S. Sobre la prehistoria de 10s dialectos Griegos, Emerita 21
(1963), 263-66.
SCHWYZER,
E. Dialectorurn Graecarum Exenzpla Epigruphica Potiora
(Leipzig 1923).
THUMB,A. Handbuch der Giechischen Dialekte I (rev. E. Kieckers,
Heidelberg 1932), I1 (rev. A. Scherer, ib. 1958).
1

THE student of morbid dialectology has to contend with a


number of difficulties imposed by the actual data that are not
shared by his colleagues who work with living dialect material,
These are well enough known but their methodological
implications are not always sufficiently recognized, and Greek
dialect studies are often conducted as if they were dealing with
a living language.
In the first place we are severely hampered by the closedness
of the corpus. We cannot return to the field to elicit new sets
of responses in order to verify or refute our conclusions, nor
can we fill out an imprecise or incomplete picture of dialect
usage by framing and putting new questions. What we have
as our data are a fixed store of potential answers, which
determine the range of questions we can usefully ask. We are
compelled not by any doctrinaire choice on our part but
because no choice is possible to frame our analysis in such a
way that the same body of facts will yield answers to the
maximum number of questions.
Then there ia the limitation imposed by the geographical
* See for example : E. Risch, ' Altgriechische Dialektgeographie,' M Z M .
Helv. 6 (1949), esp. 20-21.

R. COLEMAN-THE DIALECT GEOGRAPHY OP ANCIENT GREECE

59

unevenness of our data. We may find that the dialectal shape


of a whole vaguely defined area has to be inferred from the
evidence of one or two localities. Where, as in Elean and
Phocian, these localities are important centres of pan-Hellenic
cults, liable to steady infiltration from other dialects, this
difficulty is especially acute. Conversely the material may be
widespread over an area but very fragmentary a t any particular
point, as in N.W. Greek or in Laconian, where before the
Hellenistic period there is scattered material from the
surrounding countryside but little enough from Sparta itself.
I n both cases there is the danger of falsification: either
through taking one locality as typical of a region and so
presenting a much more clear-cut boundary between dialect
groups than could possibly have existed, or else by piecing
together the fragments from neighbouring localities into a n
artificially unified dialect.
Given the nature of the materials and the fact that
archaeologists are unlikely to uncover large amounts of new
inscriptions, the obstacles are unavoidable. The only way in
which our methodology can allow for these difficulties is by
operating with as large a number of geographical units as
possible, If we persist in talking as if for example Dorian
were a monolithic group, then major divergences within the
group will be blurred and comparisons with non-Dorian
dialects marred in consequence. Thumb-Kieckers for instance
divided Greek into seven dialect groups : Dorian, N.W. Greek,
Boeotian, Thessalian (subdivided into E. and W.), Lesbian,
Arcadian and Ionic ; and this division, with the addition of
Linear B and Cyprian, is still adhered to explicitly by Risch
and implicitly by most other investigators, As a result we
are forced to ascribe the two reflexes of "esmi, +.I and +i!,
indiscriminately to Dorian, and even more unsatisfactorily the
three reflexes of *pant@,ndvoa m%a naiaa, indiscriminately
to N.W. Greek and Dorian. Even Buck's nineteen dialectal
divisions are still inadequate to differentiate local variants,
Thumb-Iiieckers,5s 61, 76 ; Risch, 75.
See espccially Buck's Charts I and 11.

60

TRANSACTIONS OF THE PHILOLOGICAL SOCIETY 1963

for instance as between Rhodes and COB or within Cretan.


An examination of the table of contents in Schwyzers Exempla
shows that to obtain anything like a precise picture of
geographical variation upwards of forty entries would be
needed.
The fact that material from most areas is widely scattered
in time brings us to the further problem of chronological
disparity. There are two quite separate aspects of this.
Firstly the diachronic scatter within any given dialect. The
earliest central Ionic text is a single-line Naxian inscription
from Delos dated to the seventh century (Schwyzer 757) and
the first of any substantial length is more than two centuries
later (Schwyzer 766). Prom Boeotia we have a few brief
inscriptions from the sixth century (Schwyzer 440, Buck 37,
38), but there is nothing extensive till the fourth century
(Schwyzer 467). The situation is a t least as bad elsewhere
in Greek.
Sometimes this scatter is useful, in providing evidence of
the patterns of change within the dialect, e.g. the loss of /w/
generally or of secondary intervocalic /s/ in Laconian.
Internal reconstruction could sometimes have enabled us to
estimate the earlier forms, as in the Laconian example just
quoted, but direct evidence establishes the chronology of the
change more exactly. More often however the material from
different periods does not correspond in this way. In that case
we can occasionally make reasonable inferences from structural
considerations. Thus, again in Laconian, our evidence for
vowel-contraction is very incomplete before the introduction
of the Ionic alphabet. But the falling together of /oo/ with
/o*/ attested in fifth century 76 air6 (Schwyzer 12) supports
the assumption that /ee/ similarly fell together with /c/,
and this is confirmed both by the results of compensatory
lengthening a t this period, with o6&, not 066cir (<*oudenns)
and by the fact that even after the introduction of the Ionian
alphabet 7 not E L was regularly employed for the reflexes of
contraction till the third century.
Incidentally this appeal to graphemic changes serves as a

R. COLEMAN-THE

DIALECT GEOGRAPIIY OF ANCIENT GREECE

61

reminder that in morbid linguistics as a whole the phonemic


system and the range of free variation and allophonic alternation has to be inferred from the written language. We are
completely a t the mercy of the discrepancies between writing
and speech that are common to all written languages. I n
Cyprian for instance the syllabic system of writing deprives
us of precise evidence on compensatory lengthening or vowel
contraction. The problem of opaque graphemes such as r r in
Boeotian or L/1 in Arcadian' is too well known to require
detailed discussion.
The other chronological problem arises in comparing material
of different dates from two or more dialects. No dialectologist
of modern English in his right mind would treat material from
eiyhteenth century Devon and twentieth century Buchan as
equivalent items in a single descriptive account. But in
ancient Greek the choice is not ours to make. The earliest
evidence from Pamphylian and Cyprian for instance belongs
to a period when many of the Aegean dialects were already
showing signs of contamination from the Ionic-Attic lcoinz.
I n considering the methodological implications of this let us
imagine as an extreme example two dialects A and B, where
the material from A is exclusively earlier than from B. If we
select two equivalent items x1 in A and x2 in B, both derived
from an earlier *x in the unified dialect to which A and B had
belonged, we have the formula :
(i) *x(AB) > zl(A) and > x,(B)
Now if x1 represents a stage through which x2 must have
passed in the development from *x-thus :
(ii) *x(AB) > *xl(B) > z,(B)
then the contrast between A and B with reference t o x1/x2 is
useless. But if on the contrary x, represents a stage through
which x1 must have passed in its development-thus :
(iii) *x(AB)> *",(A) > z,(A)
or if x1 and x, are not mutually derivable-thus :
(iv) *z1> 5 2 , *% > 2 1
1

Buck, 347-349, discusses these and other l0Cd alphabetic variants.


1963.
E

PHILO. TRANS.

62

TILANSACTIONF, O F THE PHILOLOGICAL SOCIETY 1063

then the contrast between z , ( A ) and a,(B) is useful, in spite


of the chronological discrepancy. To take two specific
instances of a less extreme kind. By (ii) the contrast between
6eoO in fifth century Ionic and t e - o j o in Linear B is useless,
since the Ionic form derives from *&oio ; but by (iii) the
Ionic form can be set beside the third century Thessalian
.
by (iv) we can admit the contrast of
form T O X ~ ~ O C O Again
rraiua in fourth century Lesbian with TSua in fifth century
Ionic and by (iii) the contrast of both with rrdvoa in third
century Thessalian.
This criterion depends upon the diachrony of divergences.
Siwdarities between dialects of discrepant dates cannot be
subjected to any comparable method and so cannot be
admitted in a descriptive account without qualification.
Where two dialects share a common item which is not ruled
out by chronological discrepancy, the correspondence may be
significant in one of several ways.
Firstly, it may be genealogical : evidence for the derivation
of the two dialects concerned from an earlier conimon dialect.
It is important in this connexion to see the individual item in
relation to other items characteristic of the two dialects
respectively, and also to take account of the isogloss pattern
to which the correspondence belongs within the language as a
whole. In this way we avoid the dangers on the one hand of
setting up a spurious *proto-AB on the basis of atypical
agreements between A and B and on the other of misrepresenting a genuine *proto-ABCD or the like as an exclusive
"proto-AB simply because we have not recognized that the
agreements are shared by other dialects.
Secondly agreements may be typologically significant :
evidence for independent development in a similar direction.
There are several conditions under which such similarities may
occur. For instance where the speakers of two distinct dialects
are racially identical, the physiological determinants of linguistic change will be similar over the whole area. Or again
1 Much still remains to be done in oxploring the character and range of
these genetically transmitted determinants by collaborative resoarch in

R. COLEMAN-THE

DIALECT GEOGRAPHY OF ANCIENT GREECE

63

within a homogeneous culture-pattern one can expect a priori


similar patterns of semantic and lexical change. Even morphology is susceptible to cultural factors. As Meillet and others
have pointed out, the dual survived as a living grammatical
category much later in predominantly rural dialects (where
the social pattern and closer kinship structure presumably
made it more useful) than in areas of high urbanization.
Moreover otherwise distinct dialects may cover a homogeneous linguistic substrate which produces pressures towards
convergence especially a t the phonological level. Finally even
widely divergent dialects will share not only a substantial
corpus of inherited morphological material but also a prevalent
structure, which is after all what we mean by classing them as
dialects of the same language : hence purely structural
pressures may result in similar innovations quite independently.
Thirdly the agreement may be geographically significant :
as evidence for interpenetration between the dialects concerned a t a period when they were contiguous. Although
convergence of this kind is often employed as a category of
explanation in dialectological studies, some caution is needed.
For this interpenetration depends upon a long period of
settled bidialectal contact over considerable areas of a t least
one of the dialects concerned, or over an area of it that can
plausibly be regarded as a focus for the diffusion of the
intrusive phenomena. During the immigration of Greek
speakers into the Aegaean area, which began in the early
Bronze Age and continued into the period of the " Dorian
invasions ", there must have been many opportunities for
contact between dialects that were subsequently far removed
linguistics, physiology and genetics. The ideas set out by C. D. Darlington,
The genetic component of language, Heredity I (1947), 175 ff. have already
been developed by L. F. Brosnahan, Tlae Boulacls of Lalagwge (1961).
At this stage it seems legitimate at least to distinguish in principle between
those factors in linguistic change that can be attributed directly to the
structure of the preceding language (linguistic substrate proper) and those
that can be attributed directly to the physiological processes in linguistic
behaviour that are constant in a population regardless of what language
has preceded (physiologicaldeterminants).

64

TllANSACTIONS 01 TlIB PHILOLOCilCAI, SOlIlCTY 1OW3

from each other. But these contacts may not have been
prolonged enough to produce much dialectal c0ntamination.l
On the other hand in the centuries of more stable settlement
following the Dorian invasions and the trans-Aegaean migrations intercommunication cannot have been so prolonged or
extensive in most areas as to permit more than the most superficial convergence. The pan-Hellenic literary languages-of
the Homeric sagas and the dialectally orientated genres of
lyric and iambic verse and of literary prose-could have
played some part, though only a very limited one in the
processes of diffusion. One thinks for instance of Aeolisms in
Alcman like .rraicrai, 28pLrvai, K X E V V ~ , 8am+dvEuui, which are
without parallel in epigraphic Laconian. The increase in
commercial and cultural communications and the emergence
of large religious and political federations like the Amphictyonic and Peloponnesian Leagues provided conditions
favourable to bidialectalism and convergence. The Koine
super-dialects of N.W. Greek, Sicilian Dorian and Ionic-Attic
are clear results of this.
Obviously dissimilarities can be classed under the same
three headings. Their significance may be genealogicalevidence that two dialects were not inimediately derived from
an earlier common dialect ; typological-evidence that the
racial composition or the linguistic substrate of the two areas
concerned were not homogeneous ; geographical -evidence
for prolonged separation between two dialects or the convergence of one of them with a typologically more remote
dialect or even with a foreign language.
In attempting to ascertain the significance of any given
similarity or dissimilarity in terms of the categories just
discussed, we must take account especially of the hierarchy of
1 The immigrations, including the Dorian invasion (see Ruiperez, 202),
were probably more complex and less clear-cut events than is often assumed.
J. Chadwick, T h e Greek Dialects and Greek Prehistory, G.R. (1966),
aptly reminds us (48-9) that the dialectal differencee in the second millennium
B.C. must have been much less marked than in, say, the fifth century.
This might have facilitated convergence, but i t makes i t correspondingly
harder to detect,.

it. COLEMAN--TISF.

DIALECT OEOGILAPHY OF ANCIENT GREECE

65

levels. The more highly structured parts of a language system


are less susceptible to infiltration than the less structured
parts,l and similarities a t the deeper levels therefore more
likely to be the result of genealogical relationship.
Lexicon is obviously the most susceptible level. The
pressures to borrow are primarily extra-linguistic in that the
deficiencies in the linguistic system become apparent only in
the light of new behavioural contexts. A new situation, to
which no adequate linguistic response can be made in terms
of the existing lexicon-whether by semantic extension,
compounding or analogical formation, can only be met by the
importation (via bilinguals) of lexical material specifically
identified with that situation. As between dialects it is often
impossible to distinguish independent possession of a particular
lexeme from borrowing, unless there is some phonological or
morphological characteristic which reveals the foreignness of
the loan in its new setting. It may happen too that a lexeme
borrowed from dialect A into dialect B is then lost in A , orwhat amounts to the same thing-is not attested in our sources.
This raises the whole question of the reliability of lexical
evidence in dialect studies.
Although lexicon has been the basis for a great deal of
linguistic comparativism, including the statistical study of
relationship,Z it is the least satisfactory field for such enquiries
especially with morbid material. And that not only because
of the problem of borrowing which we have just noted. For
there is a natural mortality in the lexical stock of a language,
and we do not need to follow the glottochronologists all the
way in their fanciful and arbitrary attempts t o give precision
to this process in order to recognize this neglected truth to
which they have rightly recalled our attention.
It is misleading here to point to the apparently dominant
1 See U. Weinreich, Language8 in Contact, New York (1953), 5 2.53 and 54,
for a cautious discussion of this matter.
2 See G. Herdan, The Calculus of Linguistic Observations (1962), ch. 11.9,
87 ff., and the use made there of A. S. C. Rosss PIE lexical material from
J.B. Stot. Soc. (B) 12 (1960), 39.

66

TRANSACTIONS OF THE PHILOLOGICAL SOCIETY 1063

place given to lexical matters in modern dialect surveys.'


For the eliciting of lexical material direct from native informants in live linguistic contexts automatically reveals a large
amount of data a t other levels-phonological, morphological,
etc. Moreover as the corpus of a living language is open, it is
possible to ascertain whether a particular item forms part of
the lexical stock, and if not, whether there is a semantic
equivalent for it. But we are completely a t the mercy of
morbid material, A long ritual or legal text may yield
abundant information a t other levels but hardly a typical
sample of the dialect's vocabulary.
To take a specific instance. In the bi-dialectal inscription
Schwyzer 731 we find the E. Ionic noun ~ T ~ O K ~ T ~ T ~ P L O V
rendered on the Attic portion by halmarov. The former
occurs in other Ionic texts e.g. a t Naucratis (SIG 1121) and
(in the diminutive form) in Herodotus 1-25, as well as in the
North Dorian dialect of Aegina (IG. IV.39.11, with the
expected -&PEP).
Later Ionic examples of iakurarov are
found, but 6 ~ 0 ~ p a ~ r j ie
p ~not
o v recorded at all from Attic.
Since the semantic field of the Ionic word, ' that which is
placed underneath a bowl ' is much more restricted and so
more precisely appropriate to the context here than that of
the Attic word, which means simply ' that which is placed
upon' or ' that on which something is placed ', we may
conclude that the former was indeed unknown to Attic.
However, it is only the chance availability of this bidialectal
text that enables us to infer this. Where we have no such
testimony, more caution is needed. Even where we can say
that lexeme x in dialects A B C etc. is the semantic equivalent
of lexeme y in dialects M N 0 P etc. (which is not very often),
the possibility always remains that in A B C etc. y was retained
with a different meaning. We can rarely assert that what has
not survived did not occur. This is why isoglosses based on
pairs like kkm/(+Aw
or 7 q V O S / ( ; ) K & V O S cannot be used
1 e.g. E. Dieth and H. Orton, A Queationnnirefor a Linguistic Atlas of
Pnglund, (1951).

R. COLEMAN-THE DIALECT GEOGRAPHY OF ANCIENT GREECE

67

with much confidence, though of course they need not be


dismissed completely.
At the phonological level it is often difficult to decide among
the various possible interpretations of an isogloss. Occasionally
the same phonetic change in several dialects can represent
independent instances of a general tendency. For instance
three stages in the history of initial /h/ can be discerned.
The first, represented by Corinthian, Laconian, and West
Ionic among others, is the general retention of the phoneme.
The second, represented by East Ionic, Cretan, Lesbian, etc.,
shows general loss prehistorically. A third group, including
Thessalian, Boeotian, Pamphylian and probably Achaean
shows loss of initial /h/ only in the nominative singular
masculine and feminine of the definite article, and there is
evidence from Locrian and Phocian that this stage was
reached there also within the period covered by our records.
Now it is possible that diffusion through bidialectalism
explains the distribution in this third group (though Pamphylian would in any case have to be excluded). But given
the likelihood that the situation seen in the second group
began precisely with the loss of /h/ in the atonal forms
of the article and the fact that the tendency t o loss of initial
/h/ was pan-Hellenic, then all these instances of restricted
loss could be independent.
Another interesting phenomenon here is the change :
/rs/ > /rr/.
This is found in Arcadian, Phocian, Attic,
W. Ionic, Rhodian and Theran, with traces of both clusters
in Laconian, W. Locrian and Megarian. Adrados argued
that this was a Dorian change which then spread to Attic
and Arcadian. The difficulty with this is that many Dorian
dialects, e.g. Messenian, Corinthian, Argive and Coan, do not
show the change. However, it is geographically restricted
to a band acros8 central Greece (excluding the isthmus !) ;
so the occurrences can hardly be independent. Beyond
this we cannot say very much: for we have 110 way of
deciding whether the change was diffused from a focal area
As Ruiperez, 259, observed against Adrados, 55.

68

TRANSACTIONS OF THE PHII,Ol.OC:IC'AL SOCIE'PY 1963

in this band--which for a cluster-reduction of this kind seems


a priori improbable-or due to community of linguistic
substrate or of physiological determinants. This uncertainty
is one that arises again and again in morbid phonology,
whenever we pass beyond mere description t o attempt to
explain the phenomena.
Finally there is the most highly structured level, that of
morphology and syntax. Here diffusion is very unlikely.
Even the massive infiltration of English lexicon from Romance
sources in the centuries following the Norman Conquest had
very little effect on the morphological system (e.g. the establishment of -able as an adjectival suffix) and even less on the
syntax. Clearly where contiguous dialects exhibit similar
developments at this level, contact between them will reinforce
these trends. But in general we may start from the hypothesis
that similarities here are significant genealogically or typologically rather than for diffusion through geographical
contiguity.
A typologically significant example is the spread of thematic
forms from the present to the perfect, which is a feature of
Aeolic. In Ionic the phenomenon is confined to the participles
and occurs only at Chios and Smyrna. We know from the
ancient historical tradition that these two areas were originally
Aeolic-speaking and subsequently conquered by Ionians : so
that a bidialectal situation followed by linguistic substrate has
to be reckoned with here, and we can thus connect the geographically limited Ionian use of -wv, -0vros with the
generalization of thematic forms in Lesbian, which has not
only yeydvovra etc. but also ~ T T ~ G T ~ r&&cr)v,
K E ,
eta.
However the extension of thematic forms to the perfect is
not peculiar to Aeolic. Besides Phocian, where forms like
clhd+a, &~OTETEIKEV, 6e6wKo&oas could be due t o the influence
of Thessalian or Boeotian, we have examples of the spread in
Argive, Heraclean, Cretan, Rhodian and Cyprian, where it
could not be so explained, unless we substitute for diffusion
some vague prehistoric Aeolic substrate (see I1 below). The
correct explanation is probably to be seen in terms of structural

It. COLEMAN-THE

DIALECT GEOGEAPHY OY ANCIENT GREECE

69

pressures operating independently over the whole of Greek.


It is well known that the temporal orientation of the perfect
in Greek, a t least before the Hellenistic period, is towards the
present rather than the past tenses of the verbal system (to
which latter of course the pluperfect corresponded). Given
the morphological anomaly of the whole perfect paradigm and
this semantic-syntactic orientation, it is not surprising that
there should be analogical pressures from the present system.
In Lesbian and Phocian we have evidence for the wholesale
spread of thematic present forms. EIsewhere the data suggest
a partial disturbance of the inherited system, e.g. Rhodian
yEyOIvEiv, r E r i p & . K E i but 8E8&avri,
Argive hda/34KEiv but
heha/3rpctds, Boeotian Kara/3e/3&wv but &o&SdavOi.
This is
precisely the spasmodic pattern of interference between the
two tenses that we should expect to find in such circumstances.
We need not reject for Chios and Smyrna the possibility of
Lesbian in%uence,but we must see it as reinforcing a general
tendency operating in different ways a t different places
throughout the Greek language.

I1
In the light of the preceding general discussion we may now
pass on to consider certain hypotheses that have been put
forward or restated in recent years regarding Greek dialectal
relationships and to examine in detail the arguments on which
these are based.
(A) Ionic-Attic and Arcado-Cyprian were in close contact
before the Dorian invasions as parts of a sinqle dialect
complex.
This view has in one form or another been accepted by a
number of modern investigat0rs.l The supporting arguments
turn on a number of shared isoglosses :
A. Tovar, Esayo sobre la estratigrafia de 10s dialectos Griegos : I,
Emerita 12 (1944), 245ff., esp. 330-331; Porzig, 156-164; Risch, 7 0 ;
Chadwick op. cit. 42-3. L. R. Palmer, in A Companion to Homer (ed.
t A. J. B. Wace-F. H. Stubbings, London 1962), 88-91, revives the older
view of an Achaean group unjting Aeolic and Arcado-Cyprian ageinst
Ionic (see below).

70

TRANSACTIONS OF THE PITILOLOGICAL SOCIETY 1063

(I) The change */ti/ > /si/ by way of the palatalization


of the dental stop : *[ti] > "[tli] > *[tail > [si] ; e.g.
+t?povai (Arc.), 2'xoiai (Lesb.), beside & i a o u v n (Them),
~ ~ ~ C T O(Cret.).
V T L
/si/ here characterizes Linear B, Arcado-Cyprian, Ionic-Attic
and Lesbian. We know that this change was also a feature of
Anatolian phonology,' so that the Greek phenomenon may be
due to some older substrate extending over the areas occupied
by the dialects listed. This would support the view that
Ionic-Attic and Arcado-Cyprian (with Linear B) formed part
of a pre-Dorian complex and would also account for the
divergence between Lesbian and Thessalian, without recourse
to the assumption of Ionic influence upon the former.
The few examples of /sio/ for /ti./ in West Greek need not
be counted against the general retention of /ti/ in that area.
For we cannot be certain that the immediate starting point
for these was not a positional variant *[tjo], which would then
bring these adjectival forms into the orbit of the regular West
Greek shift */tj/ > /s/, seen in * n a v ~ ~>a ndvaa etc. * 7 0 7 L O S
> T ~ U U O S ,but with the retention of /i/ by morphological
analogy with /ti/. However most of the instances of /sio/
forms are proper names, geographical and religious, e.g.
KapvEiCIuiov (Messenian, cf. Kapveidras in Sicyon), Fa8wutw
(Boeot., cf. Att. r A 8 ~ u u ~ ~AtOfhca
u),
(Laconian), 'A+opSiaiius
(Pamp.), ~ U T & U l O V (Cret.), and so belong to a lexical class
that is particularly susceptible to diffusive influences, so that
they need not be West Greek a t all in origin.
A special case of diffusion is perhaps t o be found in the
variant forms of the god's name, Poseidon.
(a) 1. IloTEi8dluV--Cret., Boeot.

2. I~OTELSEV-Cor., Cret., Rhod., Aet., Phoc., Meg.


3. LlorciSoCv -Them.
4. I ~ O T O L
-Lesb?
~~V
See H.Kronaaser, Vergleichcnde Laut- und Fmmenlehre des Retldlschen
(1956), 8 72.

R. COLEMAN-THE DIALECT GEOGRAPHY OF ANCIENT GREECE

71

(b) 1. Ifouer8dov-Lin. B (po-se-da-o)


2. 1 7 0 ~ ~ ~-lit.
6 6 ~Lesb., Arg., Cor., E. Cret., Ther.,
Rhod., Coan, Ach. and Arc.
3. 17oaci8Bv -Att., Ion.
4. I7ouoi86v -Lac. (with /s/ > /h/), Arc.
Presumably the starting point is PoteidGGa, *PotoidGa, with
the change of /t/ to /s/ before a front vowel in certain dialects,
cf. nduis, which was then generalized in the proper name. The
fact that individual dialects often show more than one variant
is precisely what we should expect from diffusive influences
working across dialect boundaries. There is no justification
whatever for arguing that the s-forms are due to pre-Dorian
Aeolic substrate in West Greek, especially as only Lesbian
among the Aeolic dialects shows /s/ here ! The diffusion is
much more likely to have been by way of the pan-Hellenic
cults and the Homeric poems.
Another particular instance of the ti/si isogloss that
requires special discussion is the words for ' twenty, two
hundred ', etc. For these we have the following distribution :
( a ) ( F ) l K a n -West Greek generally (except Ach.), Pamp.,
Thess. and Boeot.
ZKOUl
-Ach. and (beside Z K a r i ) Phoc.
EilKoui
-Lesb., Arc. and Ion.-Att.
( b ) rpiatcdrioi-West Greek generally (except Ach.), Thess.,
etc.
Boeot.
rpLaKdo~oL-Lesb., Ion.-Att.
rpia Kduioi-Arc.
As Adrados rightly pointed out, the community of an
inherited feature, implying no change in the dialects concerned,
is less significant for relationship than the community of an
innovation.' Hence the distribution of Fkari (<*wTkpti,
cf. Skt. vimdatih, Lat. Ggint;) is less significant than that of
E?KOUl (< *e-wikoti). Although Risch and others have argued
for a change */?/ > /o/ in Lesbian, the examples are inconclusive, as we shall see later, A more probable factor in
The relative significance of archaism and innovation is thoroughly
discussed and illustrated by Adrados, 15-17, 30-38, 46-59.

72

TJLANXACTIONS OW THX PHILOLOGICAL SOCIETY 1983

the emergence of -0- forms is the influence of T ~ L C ~ K O V T C L


(< -*komta), etc. The same analogical spread is seen in
T P L C L K O G T ~ Sfor * T ~ L ~ K ~ U (<
T ~ S
-*kpt-to'-, cf. Skt. trimiattamah), etc. In Boeotian we find both FLKarL and FLKCLUT~S,
in Thessalian Z K ~ T L but I K O G T ~ S , with analogical -0- already
established in the ordinal. I n Lesbian alongside EZKOUL we
,
which -01shave ~ZKOLUTOS with - 0 1 - from T ~ L $ I C O L U T O Sin
(< -*om) has spread from the cardinal. The details of the
analogical extension are thus very varied.
The Arcadian data are interesting. For a t Tegea we find,
and
L Z K ~ U T ~(or
on the same inscription, the forms ~ I K O U
E)IKCLUTCL),
which suggest that analogy from the cardinal
forms had not yet penetrated the ordinal system: thus
providing the complementary situation to Thessalian.
A point that is perhaps insufficiently stressed in the discussion of these forms is the association of the -0-forms exclusively
with -01. Thus we do not find * ( F ) i K o r i or E Z K ~ U L . This
means that the doubly innovating EZKOGL is particularly
important for gencalogical grouping. Whether or not we
attribute the doublets "wzkpti and *ew6koti to proto-Greek
and interpret the subsequent situation as due to selection,l
the distribution between West and East (viz. non-West !)
Greek, with Aeolic overlapping the two, is striking.
The appearance of CZKoaL, r p t a t c d o i o e in Megarian and
Corinthian has been ascribed by Porzig (p. 164) to Aeolic
substrate a t the Isthmus. As with the -s- forms of Poseidon,
this depends on the assumption that it was the Lesbian
division of Aeolic, with /ti/ already > /si/, that provided the
prehistoric substrate: and this is assuming rather a lot,
Moreover there is evidence of -ti- forms a t the Isthmus, not
only in Argive FLKarL but also in Megarian ] ~ a r l a ~ Isn. fact
the -s- forms noted by Porzig are all late enough to be explained
in terms of influence from the Koine, thus belonging with
Coan - K ~ U L O L and probably Achaean ZKOUL, where -UL is
contrary to the general retention of /ti/ in that dialect.
For the use of this concopt and its relation to archaism and innovation
see Adrados, 19, 27, 31-39.

R. COLEMAN-THE DIALECT GEOGRAPHY OF ANCIENT GREECE

73

(2) The reflexes of */tj/, e.g. o"uor (Attic) o"ouor (Argive).


Porzig lays great emphasis on this isogloss, but a closer
inspection of it reveals a more complicated picture. We find
the following distribution of /tj/ reflexes :
/ss/-Tar., Arg., Cor., El., Phoc., Thess., Lesb., Pamp.
/tt/-Boeot.
/s/ and /ss/-C.
and E. Ion., Arc.
/s/ and /tt/-Att.,
W. Ion.

I n Cretan there are some peculiar developments, e.g. o"ooa in


Dreros, Itanos, but a t Gortys ~ { o L later
,
deBdKw, and later
still GTTOL. The graphemic conventions of Linear B and
Cyprian are opaque, though both show sibilant reflexes of
some sort.
There are clearly two distinct aspects of this isogloss. First
the phonetic one : whether the development is a stop or
fricative cluster. Secondly the phonological one : whether
or not the phoneme-cluster was split, i.e. whether /VtjV/
> /VssV/ or /VttV/ exclusively or > partly /VsV/, partly
/VssV/ or /VttV/. The split is seen clearly in Ionic-Attic :
(a) * T O - T ~ O S > T ~ U O S ,cp. *TOT-;OS > T ~ O U O S(Lesb.), T ~ T T O S
(Boeot.)
*pE-#!Los > ~ & o s ,CP. *pee-Los> ~ & T O S (Lesb.), p&OS
(Boeot.)
( b ) * + T - p > zp&ow (Ion.), <p&, (Att.)
* K p T - p V > KpE)UUwV (Ion.), Kpl!TTWV (Att.)
The phonological junctures here are clearly affected by
morphological structure, the demarcation of root from suffix
being more prominent in words of type (6) than of type (a).
Thus Attic is phonologically comparable to Ionic with the
distinction between simple and geminate, but phonetically
comparable to Boeotian in the form taken by the geminate.
It is therefore misleading to construct an isogloss solely on the
A complete account of these phenomena would also include e.g.
(Att.), n p l j ~ ~(Eretr.)
w
< - * K ~ J , ihanov (Oropus) < -*xiov, +UA&TW
(Att.) < * - K ~ J . See M. Lejeune, Traite' de Phondttique Grecque (Par is, 1955)
87 ff., Adrados, 5 6 7 , Risch, 66-7.
1

npLmw

74

TRANSACTIONS O F THE PHILOLOGICAL SOCIETY lD63

basis of &ros/&ruos, ignoring on the one side the distinction


between Attic and Boeotian, and on the other the divergence
within Ionic-Attic. The correct pattern of relationship on this
isogloss appears as :
t t & S -4

ssas4-/

tt

ss

where the top line represents locally restricted phonetic developments (whether due to diffusion or substrate) in contrast
to the general tendency towards sibilants represented in the
lower line. The left-hand vertical represents a locally restricted
phonological development (in this instance no doubt significant
genealogically) which overlaps both phonetic developments.
What emerges from all this is that while the isogloss brings
Attic and Ionic close together, their relationships outside of
this group are too complex to admit easy generalization.
(3) &opa in Ionic-Attic and Arcado-Cyprian against o"vvpa
elsewhere. This isogloss is based on inadequate factual
evidence. For Cyprian there are apparently no data. Arcadian
K ~ E C ~ V Owhich
~ O Sas, a proper name is in any case not decisive,
can be contrasted with ]wv6po in the same dialect. Moreover
the change /om/ > /urn/ is attested in Arcadian 6polocs,
urvpi'ov, as it is in Lesbian. Even in Ionic-Attic, which thus
remains as the only certain area for the survival of o"vopa,l
we have also Irr&vvpos, &v&vvp,~s.This isogloss must therefore be rejected as insignificant.
(4) Nominative plural ot in the definite article appears in
Ionic-Attic and Arcado-Cyprian as an innovation for inherited
701, which is found generally in West Greek. The innovation
is shared by Lesbian and in part by Thessalian ( o l in Pelasgiotic, roi in Thessaliotic) and Cretan (ol a t Gortys, r o i a t
Itanos). In this last dialect the analogy of the nominative
irvopa turns up in
Cretan (Dreros), Rhodian and Aetolian, but earlier
instances of bupa are attested in all three areas, so that the -0-forms may
be due to the Koine.

R. COLEMAN-THE

DIALECT GEOGRAPHY OF ANCIENT GREECE

75

singular 6, which accounts for the innovation elsewhere,


could have operated independently, though ' Mycenean '
substrate should not be discounted. With this one exception
the innovation is confined to East Greek, including Lesbian
and partially Thessalian, but not Boeotian.
(5) The athematic infinitive in -Val, - E V ~ L as
,
against - p w .
We find Etvai (Ion.-Att.), +ai (Arc.), cp. Zpp~v (Them.),
$ p ~ (Lac.)
v
; 8oGvai (Ion.-Att.), 6Gvm (Arc.), cp. 8 d p ~ (Them.,
v
Lac.). The evidence from Cyprian is thin but KvpepZvai,
8oFbai point clearly to - V a l , not - p ~ v here
,
also. In fact the
restriction of -vai and -mai to Arcado-Cyprian and Ionic-Attic
is c1ear.l
Lesbian - p ~ v in
a ~Zppwai, Gdpcvai is unique. Porzig argued
that it represents a contamination of Aeolic Z p p ~ v ,8 d p w
(cf. Thessalian) by Ionic EFvai, 8oGvar (or more strictly &ai,
GoFCvai). This is not impossible, but against it must be set
the equal possibility that -pcvai was inherited. Sanskrit
shows among many other infinitive forms both karman and
vidmcine. Porzig arbitrarily dismisses the latter as a coincidence. It is true that in Sanskrit this form in -mane can be
identified with a living paradigm type, whereas - p ~ v a i
cannot ; but the same could be said of all the Greek infinitive
forms. Moreover, it is strange that only a t this point in the
entire verbal system should Lesbian show infiltration from
Ionic.
It is better to treat - p ~ v a ias an instance of independent
selection from among the group of fossilized case forms which
provided the various Greek infinitives. These might well have
included pairs such as -*men, -*menai (cf. the Skt. forms
cited), -*wen (cf. Hitt. eswan), -*wenai (as in *8F&ai, remodelled later to give c16CvaiJ and 6oF&ai), -*sen (as in
*qkvyrurv > + & y w in Attic, +6yqv in Lesb.) and *-senai
(cf. -seni in Skt. bhzi&ni, nesdni). It is even possible to
regard Sdp.~vaias a contamination within Lesbian itself of
The isolated 8 a f v ~ rat Troezene, cp. +EU, 8dp~tv, eto.,' elsewhere in
Argive, could be an early Ionism. For a different explanation see ThumbKieckers, 122.26a.

76

TRANSACTIONS OF THE PHILOLOQICAL SOCIETY 1003

8dpw and So Fivai, but in view of the Sanskrit parallel selection


seems a preferable explanation. The motivation for this
isolated retention of an earlier doublet form could well have
been the expediency of distinguishing the infinitive and first
person plural flexions. The possibility of homonymy here
could only arise in Aeolic, since West Greek regularly shows
- ~ E S in the first plural alongside - p ~ v in the infinitive.
Although Thessalian shows no trace of any attempt t o
differentiate in this way, this is no reason to reject the possibility that this affected the Lesbian selection.
(6) The appearance of -0- vowels in the verb ' to wish '
marks off Arcado-Cyprian and Attic-Ionic, together with
Lesbian from the rest of Greek, where -e- predominates.
Typical instances are GEIAopai (Phoc.), /3kAAopub (Thess.),
/3dMopaL (Lesb.), /3odAopai (Att.), /3dAopai (Eretrian). The
Phocian form is typical of West Greek, including (so far as
the vocalism is concerned) Boeotian. An /o*/is found in Cretan
as well as in Attic and Central and East Ionic, /o/ in W. Ionic,
Arcadian, Cyprian and perhaps Pamphylian, where the length
of the -0-in /3oAtp~vusis uncertain.
The general pattern of distribution is reminiscent of that
seen in (4))and with certain exceptions we may reckon -0forms as East Greek, -e- forms as West Greek. What makes
the isogloss less convincing for affiliation is the problem of
derivation. This arises particularly in Aeolic, where it has
been argued by some that /3dAAopai (Lesb.) is the original
form, and /3lAAopaL (Thess.) due to West Greek influence, and
by others that /3kMopaL is original Aeolic and PdMopai due
to Ionic influence.
The root is clearly *gwel. For proto-Greek we may reconstruct the following theoretical possibilities :
(a) *gweZ-o-with e-grade and thematic suffix, as in A+,
Gixopat, cf. the cognate gdlati (Skt.). No direct Greek
reflexes.
(b) *g*el-so- -with e-grade and -so- suffix, as in A&,bopar.
This could be reflected in some a t least of the attested
forms, with */els/ > /el/ or /ell/.

R. COLEMAN-THE

DIALECT GEOGRAPHY OF ANCIENT GREECE

77

*gwj-jo- -with zero-grade and j-suffix as in /3alvwJ


cf. the cognate quellan (OHG) with secondary -e-. This
would yield /3dhhw and, if we accept */i/ > /ol/ in
Lesbian, j3dMopal also.
(a) *gwol-ejeThe iterative-causative suffix with o-grade
of the stem, as in $0/3lw. This would give /3oXlw and
/3ohlopar, which, transferred to the -pl class, could be
the origin of the Pamphylian form.
(e) *gwol-o- -with o-grade and thematic suffix, as in
Xdyos, -60x0s would provide substantival forms /3dhos,
/3dXii etc., derivative verbs from which (cf. c&Gw)
would be indistinguishable from the reflexes of (a).
( f )*gwol-nE-with o-grade of the stem and -n& suffix, cf.
etc.
AoMrjs (<-*nes-) beside E&J. This would give
whence / ~ O A E & etc. and the denominative /36X~&wetc.
The only unproductive type here is, rather remarkably, (a).
Conversely the only attested forms that cannot be derived
from this scheme are /36Xopal and fldhopal. It looks therefore
as if at an early stage in Greek *gwelomaiwas affected by
analogical pressure, partly from within its own paradigm, e.g.
the perfect *gwegwola, and partly from forms with -0- in the
same semantic field, e.g. in ( d ) and (e), and so replaced by
*gwolomai. The pressure from the very productive Greek class
( f ) might have led to the creation of a verb *gwolnomaias a
doublet to *gwolomai ; alternatively the reflexes of (f)
/36Ad, /36X~dw might a t a later stage have contaminated
/3dAopal to produce /36hopal,
All this is of course very speculative and may seem unduly
complex. But the attested forms are complex and do not
admit an easy solution. It is perhaps safest t o conclude that
the distinction between e and o which characterizes broadly
West and East Greek (with Aeolic again astride the division)
is the result of selection and independent analogical extensions
operating within these respective areas.
(7) The conditional particles EL, 4 as against al. The
distribution is as follows :
&-Arc.
(c)

PHILO. TRANS.

1963.

78

TRANSACTIONS OF THE PHILOLOGICAL SOCIETY 1863

4-Cypr.

~ t 4, (in i t i v < *4 &)-Ion.-Att.


at-Dor.,

N.W. Gk., Aeol.

Here we have once again a clear division between East Greek,


with EL (the locative of *o-) and 4 (the instrumental of *o-),
and West Greek with at (the locative of *&), Aeolic being
aligned with West Greek, This selection from a cluster of
synonymous or quasi-synonymous forms is reminiscent of
Italic, with Osc. svai and Lat. sei. The original selection
within the ' Southern ' group of East Greek is represented by
Ionic-Attic, and the further specialization by Arcadian and
Cyprian respectively.
(8) The presence of a"v in Ionic-Attic and Arcadian over
against Cyprian has been used as evidence for a period of
unity among the three following the departure of the Cyprian
colonists. This seems plausible. The distribution of the
modal particles is :
&-Ion,.
Att., Arc.
KE-cypr., Lesb., Thess.
Ka-W. Gk., including Boeot.
The relation between these three is uncertain. It has recently
been suggested that K a < *kp beside KEV, and that a"v arises
from wrong division of 06 Kav; cf. E ~ K & and ~l 6'a"v in
Arcadian and the distribution of Homeric 06 KEV, O ~ Ka"v.2
The idea is ingenious ; but there are a number of difficulties.
In the first place although KEV is regular before vowels in
Lesbian and Homeric texts, the epigraphy of Lesbian,
Thessalian and Cyprian shows only KE, This is strange, if KEY
was in fact the original form. There is only one instance of
W.Gk. KUV beside KU, in Boeotian. Moreover, in literary
Dorian KU regularly shows a long vowel, which can hardly
K. Forbes, Qlotta 49 (1958), 179-82, whose argument against tho
traditional connexion of dv with Latin and Gothic an seems definitive.
See also Palmer, op. cit., 90-91.
a J. van Leeuwen, Enchiridion Dictionis Epicae (1918), 8 326, especially
p. 409.

R. COLEMAN-THE

DIALECT GEOGRAPHY OF ANCIENT GREECE

79

be written off as metrical licence. Finally the alternation of *kp and ken is apparently without parallel in Greek.
Thessalian pd beside p b is not strictly comparable, since the
two are functionally distinct, p& . . . pcl . . . regularly
corresponding to ,U&J . . . 66 in other dialects.
The comparative material adduced for ken, or rather *kern
outside Greek is not as impressive as it seems a t first sight.
The prepositions kam (Skt.) and ZB (Slavonic) are functionally
remote and phonologically suggest *kwem a t least as plausibly
as *kern. The Hittite particle kam used in conjunction with
nzi, szi and hi seems more attractive as a cognate.
However the doubtful status of the -m in Greek is still
troublesome. If we assume ke, not ken as the original form,
then we could connect this with the Latin deictic particle ce,
a flexionless pronominal stem, with k6 as the corresponding
feminine and ka, if it existed a t all1 as the reflex of k6 in
hiatus. Then we might also adduce the Hittite particles ha
and ki (e.g. in the combination ki-nu-un) and kan in nu-kan.
A deictic particle would correspond functionally to the locatives
EL and al, and KC could even be interpreted as a feminine
instrumental, parallel to the neuter 4. The addition of -n
could be in part ephelcystic, though there are other instances
of particles with doublet forms in -n, e.g. T O M ~ (Horn.)
KL
beside T O M ~ K(Cret.,
L V Lac.) and T O M ~ K(Ion.),
K vd beside
v h . It is tempting to relate the conjunction ~ a l / ~ ctol s the
same deictic root, Kai being either the feminine locative or
~6 + deictic - L , and K ~ being
S
analysed as ~d s.
None of these speculations of course disturbs Forbes's
plausible derivation of a"v from K ~ V .If it is accepted, however,
then it does mean that K C ~ Vsurvived in Arcadian right through
to the historical period. Hence a"v in itself cannot be used to
support an association of Arcadian, Ionic and Attic in prehistoric times. Instead it is the acquisition of -n, following
the selection of KU as against K E , that is significant for this
hypothesis.

1 ~d is perhaps supported by Phocian hivrr K ' & O T ~ U ~ ~ (Schwyzer


L
323B),
though one could wish for an example where the following vowel was not /&/.

80

TRANSdCTIONS OF THE PHILOLOGICAL SOCIETY 11M3

(9) &E, T ~ T Eas against &a, T ~ K U(W. Greek) &a, ZAAOTU


(Lesb.). Porzig used this isogloss plausibly to isolate ArcadoCyprian and Ionic-Attic from West Greek. &a seems to be
original in Aeolic. &a turns up in Boeotian, but there are
too many West Greek features in this dialect to justify our
separating this one from West Greek. Thessalian d ~ is
opaque, but in view of Lesbian GTU probably represents ~ T U
also.
The connexion between these three forms is obscure.
Adrados (p. 33) suggested that &a and d r contain
~
reflexes
of the labiovelar *Ico/e-. This is very plausible for GTE, more
difficult for &a. Por the latter Ionic ~ K W S ,~ K O ~ Uetc.,
,
would provide a parallel, and indeed it is tempting to regard
~ K Uand ~ K W S
as fossilized case forms of the same pronominal
root, the former a neuter plural, the latter an instrumental
with secondary additions of -s. The labio-velar hypothesis
makes both - K a and - r e relatives in origin, which is appropriate
functionally. This still leaves &a. Remembering that in
Lesbian and Thessalian the article (or deictic pronoun) is
regularly employed as a relative, we can analyse &a as
*jo
the neuter plural of the article. It would then be
functionally a precise equivalent of W. Greek &a.
The results of this section may be briefly summarized as
follows : A close association between Ionic-Attic and ArcadoCyprian is demonstrated by the exclusive isoglosses (5), (7)
and (9), and supported by (1) and (6), which include also, in
particular, Lesbian, Linear B agrees with the four on (l),
the only one of these isoglosses for which it shows any clear
evidence. (a), (3), (4) and (8) do not support the association,
though it is fair to add that they cannot be used to overthrow
it either. As we shall see, the results of factorial analysis
strengthen the hypothesis considerably.
(B) lhe divergence within Aeolic between Thessalian and
Lesbian.1
Some of the isoglosses relevant to this topic have already
been discussed in A : KaTLyViTOS (Thess.)/tcaa/yvq.ros (Lesb.)

See Porzig, 149-155 ; Risch, 70-71 ; Chadwick, op. cit., 46.

R. COLEMAN-THE DIALECT GEOGRAPHY OF ANCIENT GREECE

81

and ~'KUTL
(Thess.)/ei'Koui (Lesb.) in (I),8 d p v (Thess.)/Sdpvai
(Lesb.) in (5), /3&lopui (Thess.)//3dMopai (Lesb.) in (6). We
must now consider the other items cited in this connexion.
(1) The reflexes of *.rravqu : ~IT$vuu(Thew.), .rraiua (Lesb.),
.rrG:aa (Ion.-Att. and W. Gk.).
This pattern of distribution was used by Porzig to support
his view that the Lesbian divergences from Thessalian were due
to Ionic influence. In considering the evidence it is important
to take account of the reflexes of */entj/ and */ontj/ as well
as of */antj/.
No dialect, not even Linear B, shows -*Vnti- in any of
these contexts.
/Vns/ is probably concealed by Lin. B pa-sa and is clearly
attested in early Thessalian, in parts of Crete, and in Argive
and Arcadian. As all the other dialects must have passed
through this stage, this part of the isogloss is of limited
importance though it does serve to illustrate the conservatism
of the dialects which exhibit it.
Three distinct reflexes of this /Vns/ are attested :
( a ) /V,.s/, with simplification of the cluster and compensatory lengthening of the preceding vowel, which in the
case of */e/ and */o/ falls together with the inherited longvowel phonemes, e.g. @puua. Examples occur in Laconian,
Heraclean, Theran, Elean, Boeotian and Pamphylian (in part
at least).
(b) /V,.s/, where the resultant long vowel in the case of
original */e/ and */o/ is kept distinct from the inherited
long-vowel phonemes, e.g. ( ~ Q O V U U . Examples occur in
Corinthian, Megarian, Rhodian, Coan, N.W. Greek, Attic
and Ionic.
( c ) /Vis/, e.g. qGpoioa. This development is peculiar to
Lesbian. Its appearance in the text of Alcman is interpreted
by some as a genuine Laconism, but this is improbable in view
of the subsequent appearance of /V,.s/ in tho epigraphic
material. The only other occurrence of -Viu- epigraphically is
a t Cyrene. It is not a feature of the mother-dialect Theran,

82

TRANSACTIONS OF THE PHILOLOGICAL SOCIETY 1963

and so could be an independent development in the colony.


However there are other Lesbian features too in Cyrene (see
Thumb-Kieckers 0 146) which could be due to the participation
of a Lesbian group in the sixth century re-colonization.
The reflexes of original */ns/ in final position are comparable
in many respects to those of */ntj/, though there is a greater
range of variants :
( a ) /Vns/ unchanged, e.g. rdvs: a t Troezene (Arg.) and
Gortys (C. Cret.).
( b ) /V,-s/, e.g. &s, s;7 : Lac., Tar., Boeot., Pamp.
( c ) /V,.s/, e.g. 706s : Cor., Meg., Rhod., Aetol., Phoc.,
Att., Ion.
(d) /Vis/, e.g. 701s : Lesb., Elean.
The Elean situation can hardly be due to Lesbian influence in
view of the remoteness of the two dialects from each other and
the relatively recent character of this pattern of change in
Greek. Moreover, as */Vntj/ reflexes in Elean do not show
this form, whereas they do in Lesbian, it seems best to regard
the shift as independent in the W. Greek dialect. This is confirmed by the early orthography of Elean which shows -0s in
this context, as distinct from -os, perhaps denoting a nasal
vowel or some other intermediate stage between /om/ and
/oh/.
( e ) /Vs/, e.g. T ~ :S Thess., Arc., Coan, Theran. I n many
dialects, e.g. Arg., Cret., Rhod., and Phoc., there is evidence
for this treatment alongside one of those noted above. The
distribution of 769, T ~ V S ,T& must in origin have been determined phonologically, 76s before consonants, T ~ V S before
vowels (cf. 2s E)VS is), with subsequent generalization resulting
in the situation seen in Thessalian, etc.
Both these changes are thus pan-Hellenic. The realization
of a general trend in the language most probably occurred in
each dialect independently. Contacts with nearby dialects
might confirm a trend already in existence, but there is no
justification for assuming that the contrast between, say, the
conservative situation in Thessalian and the innovatory one
in Lesbian is due to any external influences.

R. COLEMAN-TEE

DIALECT GEOGRAPHY OF ANCIENT GREECE

83

(2) The genitive singular of the thematic declension:


The divergence of Lesbian -w from Thessalian -010 has been
attributed to the influence of Ionic, which shows -ow, viz. -0.
Here as in (1) the argument is vitiated by concentrating
attention on too restricted a sector, via. Thessalian, Lesbian,
Ionic, without taking account of the pattern of variation over
the whole Greek area.
I n the inherited genitive -*/osjo/ [s] was lost prehistorically.
-/ojo/ is already attested in Linear 13 -0-jo, and later in
Thessalian and Homeric -010. I n all of these the phonetic
value is probably [ojjo]. For all the other forms of the Greek
0- stem genitive the immediate starting point seems to have
been [ojo]. This was replaced first by [ O O ] , a stage attested
only in the -00 forms restored on metrical grounds in some
Homeric passages, and then by the contract reflexes set out
below. The diachronic relationship between [ojo] and [ojjo]
is uncertain. Loss of [s] from */osjo/ could have been accompanied by compensatory lengthening of the semivowel, and
the resultant [ojjo] later reduced to [ojo]. Alternatively loss
of [s] could have led directly to [ojo], with [j] thereafter either
lost (as in most dialects) or lengthened to [jj].
Two reflex types result from the contraction of [OO] :
(a)a vowel identical with inherited /o-/, e.g. 8 4 beside
OrGv. This development is seen in Lac., Her., Mess., Arg.,
Cret., El., W. Locr., Boeot., Lesb., Arc. and perhaps Cypr.
(b) a vowel distinct from the inherited /o./, e.g. 8eoC
beside 8cGv. Examples are found in Cor., Meg., Ther., Rhod.,
Coan, Ach., Aet., Phoc., Pamp., Att., Ion.
The pattern is very similar to that for compensatory
lengthening in (1)and (2) above and can be interpreted as the
independent realization in all dialects of a general Greek
tendency. In fact even Thessalian shows a reduction of /ojo/
within the historical period : in Pelasgiotic we find /oi/, e.g.
MvaacpaXdoL beside noX+oio (Larissa), MEVEUT~IOL
beside
@ ~ h d y p o ~(Gyrtun),
o
in Thessaliotis /o,/ as in (a)above, e.g.
ZwudvGp6, later Aapa~pclov (Kierion). For the change

84

TRANSACTIONS OF THE PHILOLOGICAL SOCIETY 1983

*/ojo/ > /o-/ in Lesbian the influence of Ionic-or


other dialect-is thus wholly redundant.

any

(3) The dative plural of -6-and -0-stems shows Thessalian


-OK,
e.g. FoiKLdraLS, Lesbian -aim, - o m , e.g.
cip+orkpami, ZAAoiui. This has been claimed as an example of
W. Greek influence in Thessalian, but in fact the distribution
of the two types is more complex than this reasoning implies :
-ais and

-s forms-Lac., Meg., El., W. Locr., Boeot. ( T O X ~ T ~ S< -*ars,


rrpo&~vs < - * O K ) , Thess., Arc., Cypr., Att.

-si forms-Pamph.,

Lcsb.
Both types are attested in Corinthian, Cretan (where -si belongs
mainly but not exclusively t o the central region) and in Ionic.
Ionic shows a remarkable diversity. From inscriptions
dated t o the fifth century or earlier we find, e.g. ~rjisa t Paros,
$ULV at Naxos, $pkpprliuiv a t Chios and ripais a t Erythrae.
From the same period we have in -0- stems kpoiuiv a t Miletus,
vBpo~s a t Oropus. Most of these variants are attested in
MSS. of Ionic authors as well.
Both Sanskrit and Latin show syncretism of the dative and
ablative cases in the plural, e.g. uivvebhyah, equis. I n Greek
the syncretism of ablative and genitive, found in Sanskrit only
in the (non-thematic) singular, was extended to the plural,
where the ablatival functions are taken over by the genitive
forms. The Greek dative plural is a mixture of inherited
locatival and instrumental forms, corresponding to the
functional syncretism of the case.
-01 in 0-,ci-, consonant-stems reflects the inherited locative.
For -OLOL cf. Skt. uivqu, for -auc cf. uivcisu. The Greek - L
may be due to analogy with the dative singular, but the
possibility of an inherited variant cannot be ruled out, in view
of Avestan hafs'i, tanus'i (not attested however in vocalic
stems).
-01s reflects the inherited instrumcntal -*/o.is/,
cf. Skt.
aivuih,with the regular Greek reduction of the long diphthong
in this position. -ais seems to have been formed on analogy
with -01s. The inherited &stem instrumental -*/a*bhi/,

R. COLEMAN-THE DIALECT GEOGRAPBY OF ANCIENT GREECE

85

attested in Skt. aivvdbhih (where the final -*/s/ is perhaps


secondary) is found only in the Linear B instrumental, e.g.
a-n.i-ja-pi = hzniciiphi, and in the fossilized multi-functional
-41 of Homeric.
I n Linear B dative and instrumental cases are still distinct.
Beside a-ni-ja-pi we find e-re-pa-te-jo = elephantewis, etc. as
instrumentals, and the datives te-o-i = theoi(h)i, e-qe-ta-i
= hepetd(h)i or (by analogy with o-stems already) heqetai(h)i,
which reflect the inherited locative formation. The lost intervocalic s was later restored by analogy with the consonant
stems e-ke-si-qe = enkhessi-qe, pa-si = pan(t)si, etc. Consonant stems here apparently included those with -/e. w/suffix, e.g. Ica-he-u-si = khalhwsi, which might well have
provided the starting point for the restoration of s in purely
vocalic stems. Both dative in -si, -(h)iand instrumental in -pi
also exhibit locative functions in Linear B.
It is interesting to note the greater variety of :-stem than
o-stem datives in Ionic. I n the latter paradigm the prehistoric
reduction of the long diphthongs had brought the reflexes of
inherited instrumental -*/o * is/ and locative -*/oisi/ closer
together, viz. -OK, -0iui. Phonetic change within Ionic itself
produced, in contrast, a greater disparity between the corresponding &stem forms, viz. -air, -7oi (< -*aai). The possibilities of analogical influence here are therefore more numerous.
Besides -aim, showing pressure on -*tior both from -ais and
-oiai, we find -71s and -7iai.
The presence of datives in -s and -UL side by side in Ionic
and in Cretan must represent a situation once common to the
whole Greek area, and the distribution pattern noted above
provides an admirable instance of the effects of selection.
The nearest parallel to Lesbian in this particular item is not
Ionic, where the ultimate selection over the whole region
favoured -ais and - O K , but Pamphylian. However the
pressure to avoid homophony with the accusative forms -air
and - O K ( <-*avs, -*ovs) in Lesbian cannot have been a factor
in the Pamphylian selection.
(4) The sigmatic tenses of verbs in -5w. The significance of

86

TILANSACTIONS OF THE PHILOLOOICAI, SOCIETY 1983

the distribution of this isogloss has been much discussed,


Kretschmer long ago argued that - f - was an 'Achaean',
-u- an Ionic feature. Poraig maintained that Thessalian
$a&fapkvas, 2pydfaro were due t o West Greek, Lesbian
z p d v r i a a v to Ionic influence.
The correct interpretation of
the isogloss depends once again on two considerations : the
historical origin of -5- and -a-in these forms and the pattern
of dialectal distribution.1
-5w verbs comprise two types of denominative suffix:
voiced dentals, with -5w < -*8kw, e.g. 2 h i s 2ArI8os : &l[w,
and voiced velars, with - [ w < -*y-iw, e.g. Q r a E gprayos :
Aprdlw. The unvoiced stops yielded -rrw/-auw, e.g. 2p+s :
2pkrrwJ rp&xos : r p d r r w . The geminate in these latter forms
must, originally a t least, have denoted the unvoiced equivalent
of 5 : viz. *[tj] > *[ts] > [tt]/[ss], beside *[dj] > [dz] > [dd]
in some areas /[zz].
The voiced-stops in sigmatic tenses (future and aorist)
would originally show two distinct developments : *[ds]
> *[ts] > [ss] ([tt] in Boeotian), "[gs] > [ks]. No dialect
exhibits this precise distribution. A number show both types
of reflex, but not distributed etymologically. This situation
must have been intensified, if not actually caused, by the fact
that -5w was extended to stems where there were no
supporting stop-final forms, e.g. 8apd[w, 8 i ~ d [ w ,v o p ~ c w .
Examples include : Arg. uxluas rapeve$&&
GiKdfauBai but
E'8l~au~a
; vCoan 2pydfaa%ai, 8iKauakw, Boeot. &r~t,ba$haro
&opifdpeBa, Thess. 2ppyd[uro, ~ p o v ~ [ l a ] a (late
~ i v enough, a t
Larissa, to be due to the Koine),2Arg. ~61KdUUatLVJ
rapherafd~ E V U S Ion.
,
2rieaav, T L E X B ~ V T ~Horn.
S,
;praaa, <prafa.
A third stage, the tendency to generalize one type exclusively
is seen in Attic, where -0-is generalized in -5w verbs, - f - in
-77-ones, cf. Cypr. K a T a G K v ' f U U but dpv'f..
I n w. Greek
1

See P. Kretschmer, Zur Geschichte der griechischen Dialekte,' Clotta 1

(1909), 9-59, especially 28 ff. The details are well discussed by Risch, 73.
2 The Thessalian imperfect ivedaviuoocv shows a different kind of conW
tamination : -uuw for -& ; of. U@TTW (Att.), u+d(w (Ion.), T ~ ~ T T (Att.),
xpd88w (Cret.).

B. COLEMAN-THE DIALECT GEOGRAPHY OF ANCIENT GREECE

87

(including Pamphylian) the trend was towards -I-,though


Argive and Coan, as we have just seen, preserve an earlier
stage.
While interdialectal contacts here as in other cases would
reinforce existing trends, the distribution pattern is most
plausibly explained in terms of independent development, the
result OP amalogical extension and selection. Within Aeolic
all three historically attested types are found : Boeotian like
Argive (but not N.W. Greek)shows both -&and -m-(-uu-)forms,
Thessalian like most West Greeks shows predominantly -f-,
Lesbian like Attic (but not Ionic) shows -a-.
( 5 ) 2v
acc. (Thess.), i s, cts (< *&s)
acc. (Lesb.).
Porzig (p. 150) regarded this divergence as also due to Ionic
influence. Indeed he argued that <US started in the East
Aegean, probably Ionic, and spread from there to all the
dialects where it is attested. Risch (p. SS), observing that
&, i s in Ionic-Attic agrees with Dorian against Iv ( < E'v) in
Arcado-Cyprian, used this isogloss to support the hypothesis
of a close contact between Dorian and Ionic a t the time of the
Dorian invasions. The two arguments are not mutually
incompatible, but are they a plausible interpretation of the
evidence P
The pattern of distribution falls into three sections :
(a) t'v-W. Low., Phoc., Aetol., Arc.-Cypr. (lv), Boeot.,
Thess.
(b) &s and its reflexes :
ivs-Arg., Cret.
&-Lac.,
Tar., Arg., Cop., Rhod., Coan, Lesb.,
Ion.-Att., Is-Pamp., Cret.
eIs---Cor., Coan, Lesb., Ion.-Att.

Both forms :
iv, 2s and &-Mess.
&, ds-Meg.
&, t's-Ther.
If the -s forms were diffused from Ionic, this must have
happened early enough for them to take part in the general
(c)

88

TRAKSACTIONS OF THE PHILOLOGICAL SOCIETY 1063

phonetic shift from /Vns/ in all the dialects concerned, as


Risch rightly saw, placing the innovation itself in the period
1200-900 B.C. However the wide diffusion that has to be
assumed (affecting all the dialects save those in (a) above) is
so improbable that one is prompted to seek an alternative
explanation.
The addition of -s to adverbial and prepositional forms is
well known in Greek e.g. aiWi, ad%is; 0;17w, ov"rws, Elean
&v:vrvsfor ZVVEV,Dor. +rCv beside &mas, and in prepositions
+$I, a'&s ; r p d , rpds, where the variants were semantically
differentiated (see below) and 2 ~ ,Zt, where they were not.
We may conclude therefore that the doublets Zv and Zvs were
both used originally in Greek with the accusative-though
not with the locatival dative, since no dialect shows Zvs
dative. The pattern of distribution set out above would
then result from independent selection, some dialects retaining
both forms with the accusative function (cf. ZK and (2
gen.),
others showing differentiation of function, with 2vs + acc.,
E)v dat., others again levelling out Zvs and employing Zv
with both cases.
Different structural pressures would operate in different
dialects. Thus in Ionic thc use of r p d s with the accusative in
a similar semantic function would favour the exclusive use of
ZVS, in Argive the change of 26 to 2s in certain contexts would
lend to homophony with 2s (< *Zvs), and this would favour 2v.
The selection hypothesis seems much more satisfactory here
than the assumption that an isolated morpheme was exported
from one dialect, especially as there was no functional gap for
it to fill in the other dialects. We may therefore reject the idea
of diffusion on this item before the Koine period, when the
appearance of Ionic ~ l in
s many dialects is accompanied by
the intrusion of a host of other Ionic forms.
(6) The preposition TOT/ in Thessalian corresponds to irpds
in Lesbian.
This divergence has been used on the one hand to associate
Thessalian with W. Greek against Lesbian and on the other
Lesbian with Ionic against Thessalian, cf. (4)above.

R. COLEMAN-THE

DIALECT GEOGRAPHY OF ANCIENT GREECE

89

Again we must consider both the distribution and the


derivation of the relevant forms :
( a ) m-rl-W.Gk. generally, Boeot., Thess.
(6) roi-before dentale in some W.Gk., generalized in Arg.,
Phoc. This looks like dissimilatory loss of /t/, but the
possibility of an original *PO-; cannot be ruled out.
(c) po-si-Lin. B, with /si/ < */ti/.
( d ) rids-Arc.-Cypr. Presumably < * r o u l , with the apocope
that is common in Arc., though it could represent
original *PO-s.
( e ) ~porl-in Homer. Some of the examples could conceal
Tori, e.g. /3ij G'haL n p o d v;ias but others must be
genuine, e,g. cklpovro rrpo-rl &pas.
Homeric - 7 6 here
and in no71 must reflect an old Aeolic substrate in the
dialect (we do not know a t what date Lesbian
*/ti/ > /si/). r p o d could in theory represent an
artificial contamination of nor1 and rpds, but there are
also tracea of it elsewhere. rp07' occurs once in Argive,
where i t is supported by Apollonius Dyscolus's testimony that ~ p owas
~ la Dorian form.
(f)rop.rl-beside n o d in C. Crete. The latter cannot be a
derivative of 7 ~ 0 ~ 7since
1 the change */ort/ > /ot/ is
not otherwise attested for Cretan (or for any other
dialect, e.g. in ( a ) above). r r o p ~ <
l *p$i is also phonologically improbable for this dialect. However the
metathesis of op, ap for PO, pa in Cretan is attested in
'A+opGira, araprds, KC~PTOS, so that we can derive
rrop71 from 1 ~ ~ 0in7 1( e ) .
(9) vpds-Ion., Att., Lesb. In Lesbian, where apocope
of prepositions is common, npds < * ~ p o u<~ pod is
possible. I n Attic and Ionic apocope is almost unknown,
so we cannot assume the Lesbian pattern of change,
Attic ~ p d therefore
s
< *po-s.
Metathesis of the Cretan
(h) mp7l-Pamp. TEPT&OKE.
type is attested in 'A$opGlra, Ilp~rlasbeside I l ~ p y l a s .
Pamp. m p 7 l therefore < *pre-ti,

90

TRANSACTIONS OF THE PHILOLOGICAL SOCIETY 1963

( i ) rpis-Lesb. shows the Ablaut alternant of (9) . cf. epi


and opi (Lin. R).
This welter of variants can be derived from the following
basic forms : *Po-ti and perhaps *PO-i,
*PO-s
*prO/l,-ti and *pro/e-s
To these may be added the semantically differentiated *pro
and perhaps *PO, since .rrd occurs in a number of the ~ 0 7 1 dialects, where it may not always be due to apocope. We thus
have *PO and
with some a t least of the suftixes zero, 4,
-s, -ti. How many of these were inherited it is hard to say.
For *pos there are parallels in Lith. pds = at, Italic *posti
with additional *ti (Lat. post and perhaps postid, Osc. pzist
and perhaps pzistin, Umbr. posti) ; for *poti Av. pa%, OPers.
patiy ; for *pro Lat. pro, Skt. p r a ; for *proti Skt. prati,
OCS. protiva.
There is a double isogloss here : pro/po in the root and a t
least ti/s in the suffix. It is difficult to see how the complexity
of variation and distribution can be anything but the result
of independent selection from a number of such compounds
with overlapping functions in pGk. Thessalian seems to go
along with W. Greek ; Lesbian with East Greek but with a t
least one peculiarity of its own ( r p i s ) . This means nothing
more than that in this as in so many features Aeolic does not
as a group fit neatly into one division or the other.
(7) The suffix of the adverbs .rrpdo%ev etc. (Lesb.) contrasts
with that of Thessalian a[ov^Ba. This too has been used to
establish a connexion between Lesbian and Ionic. The isogloss
need not detain us long. For 1~pdo6cwithout -v is the epigraphic form in Lesbian ; .rrpduOev, which is regular in Attic
and Ionic, also occurs in Aetolian and in Thessalian ( ~ p d u r r v !)
The relation between these two forms and .rrpduBa etc. in
Heraclean, Cretan, Phocian and Arcadian is obscure. There
is a great variety of adverbial suffixes observable in most
dialects : Thessalian has ~ ~ p d a r but
r v E'eov^Ba,Argive ZprpouBa,
+.rrpoo%a, Boeotian etrcv, &ra, Ionic 2mua, t'nm-w, &Ba,
&Btv, Arc. BV'oBev (= outside), Cypr. i;Ba. Not only does this
See Palmer, op. oit., 89.

R. COLEMAN-THE

DIALECT GEOGRAPHY O F ANCIENT GREECE

91

isogloss not divide Aeolic neatly but it fails to characterize any


of the Greek dialects.
We may now summarize this part of the discussion. I n
general it is clear that, where Thessalian and Lesbian diverge,
Lesbian is the innovating, Thessalian the conservative dialect,
e.g. A (1)) B (1) and (2) and perhaps A (6). Most of these
innovations are shared either wholly or in part by other
dialects, e.g. K a o l y q r o s , PdMopai, though in some instances
the particular development is peculiarly Lesbian, e.g. raicra.
A ( 5 )and (8))B (3), (4))(5)and (6) are all examples of selection,
and again we may distinguish those features shared by other
dialects, e.g. the dative in -OLUL and the preposition ~ V S from
,
those peculiar to Lesbian, e.g. S d p v a r , r p & . Even with Eikoai
we cannot assume Ionic influence, since the form is not
confined to these two dialects, and it is better to see this and
other items as uniting Lesbian (more closely than Thessalian)
to a loosely-knit East Greek complex, As for Thessalian itself)
the items that it shares with W. Greek are either ones originally
common to all the dialects and so inconclusive genealogically
and unlikely to be due to convergence, or else they can be
plausibly ascribed to independent selection. Nevertheless the
fact that Thessalian is more " westward-looking ')than Lesbian
remains significant.

(C) The dtffUsi0.n of Aeolic features in non-Aeolic dialects.


This argument has appeared in various though not mutually
exclusive forms. The commonest is the setting up of a prehistoric ' Central Greek ' or ' Achaean with a considerable
Aeolic content. Adrados in fact used the term ' Aeolic'
almost as a synonym for ' Achaean '.l I n addition to ' Aeolic
substrate of this kind there are also occasional instances of
))

Thumb-Iiieckers, 8 76, follow Kretschmer in setting up a Central Greek


group (Aeolic, Old Achaean) between Ionic-Att,ic and West Greek, comprising Arc.-Cypr., Boeot. and Lesb.-Them So too Palmer, op. cit., 88-91.
Adrados, M-61, etc., divides Aeolic from Arc.-Cypr., but his use of the
notion of ' Aeolic ' substrate is rightly chastised by Ruiperez, 261. Porzig,
161-168, postulates a southward migration of Pelasgians in the middle
Bronze Age.

92

TRANSACTIONS OF T H E PHILOLOGICAL SOCIETY 1083

more recent change in non-Aeolic areas that have been


ascribed to diffusion from an Aeolic source.
Two relevant items have already been examined : L'oa~r6Gv
in A (l),and the -8- signatic tenses of -50 verbs in B (4).
Further examples are :
(1) The close-vowel reflexes of the contraction of */ee/ and
*/oo/ which Porzig saw as the result of diffusion southwards
from Thessaly a t a relatively recent date (uncontracted forms
appear still in early Boeotian, Argive and Cretan).
As with the reflexes of */tj/ considered in A (l), it is
essential to distinguish in our pattern of distribution between
phonetic and phonological aspects of the isogloss. Phonologically two types of development are attested :
(a)*/ee/ > < /v/, */oo/ > < /w/. viz. no distinction
between the contract reflexes and the inherited long
vowels, both being represented by c, 0 (later 7, w ) , e.g.
Lac., Her., Mess., Arg., Meg., Cret., Ther., El., Boeot.,
Thess., Lesb., Arc. and probably Cypr.
4 /w/. vie. the contract
(b) */ee/ > Q: /e/,*/oo/
reflexes and inherited long vowels are kept distinct, the
former as E L , ouJ the latter as 7, w .
e.g. Cor., Rhod., Coan, Ach., Aet., Phoc., W. Locr.,
Pamp., Att., Ion.
In Thessalian both /el/ and /o./ were raised prehistorically,
e.g. SOUK KC, dvd8ecrcc. It is therefore impossible to say whether
the contract reflexes (also denoted by E L and ou) fell together
with the long vowels before or after this change occurred.
On either diachrony the Thessalian contract reflexes in E L and
ou must be kept phonologically separate from the phonetically
similar 1 and ou of other dialects, which were structurally
contrasted with 7 (or i ) and w (or 6). Hence Thessalian is
placed in (a).
In ( b ) it is hard to be certain of the precise phonetic interpretation of the digraphs. They could represent either
diphthongs resulting from end-closure, viz. *[eel > [ei],
*[oo] > [ou] ; or the ' spurious ' diphthongs, established for
Attic and Ionic, viz. [y] and [y], which could of course be

>

R. COLEMAN-TEE

DIALECT QEOQRAPHY OF ANCIENT GREECE

93

related in some instances to an original diphthongal stage, as


*[eel > *[ei] > [+,I. It is possible that (a) represents phonologically a later stage than (b) ; with the loss of distinction
between the reflexes of /ee/ and /e./, /oo/ and /o./. Now it
is worth remembering that the distinction between EL and q
as [y] and [E'], which we are familiar with in Attic and Ionic,
need not have been so marked in other dialects, where /a*/
and /e./ had not fallen together. The probable phonetic
values E L = [e.], Z or q = [e.] in these dialects would present
a more favourable context for the merging of the two
phonemes in question.
We must also take note here of the change that resulted
from the introduction of the Ionic alphabet into the areas
listed in (a). Very Boon the Ionic graphemic distinction of
a / q , ov/w began to appear in some places, e.g. Argos, Thera
and perhaps Megara (though the original situation is not
certain there), while in others, like Laconia, the previous
situation remains, 2 and 0 being replaced by q and w in both
contexts. It is of course possible that the epichoric graphemic
conventions really concealed a phonemic distinction, and that
some of the dialects classified under (a)above properly belong
to (b). More probably however the phonemic distinction was
introduced along with the Ionic spellings. This intrusion
would not be violent, since by the fourth century, when it
occurred, the Ionic long vowels /e* 9. 0-?./ (7 E L w ou) were
already becoming raised from [ c y 3' 9'1 towards [e, i. 0' u.3
and the difference between the Ionic original long vowels and
contract reflexes slightly reduced in consequence.
The raising of the contract reflexes in (b) was far too widespread to be due to Thessalian, which even vis-8-vis Central
and South Greece was for centuries too isolated and remote to
be a focus of diffusion. The fact that the dialects listed in (b)
form a band right across central Greece suggests the possibility either of diffusion, though hardly from Thessalian, or
else of substrate of one kind or the other, though again in
view of the Lesbian divergence from Thessalian unlikely to
be Aeolic.
PHILO. TRANS.

1903.

94

TRANSACTIONS 01' TlIE PHILOLOGICAL SOCIETY 1003

On the other hand the early raising of /e./ as a whole was


confined to Thessalian, Boeotian and Cyprian. The shift in
Boeotian could be due to Thessalian influence, though if 80,
it is strange that /o./ in Boeotian was not similarly affected,
since both phonemes were raised in Thessalian. The Cyprian
change must be independent. It is tempting to see the raising
of /v/ in these three (or two) areas as an independent early
realization of the general tendency to iotacism throughout
Greek. Subsequently Ionic and Attic also exhibit the shift,
and diffusion through the Hellenistic Koine of which IonicAttic was the focus no doubt hastened the development in
other regions too.
(2) The occurrence of -VV- as a reflex of */sn/ in Laconian
has been taken as an instance of pre-Dorian Aeolic substrate.
For the reflexes of */Vsn/ we have the following distribution :
(a) /Vl.n/ where for */V/ = */e/ or */o/ /Vl,/ > <
/e./ or /o,/.
Lac., Tar., Mess., Arg., Cret., Ther., Rhod., Coan, El.,
Boeot., Arc.
( b ) /V,m/ where for */V/ = */e/ or */o/ /V,,/ > <
/e*/ or /o*/.
Cor., Meg., Ach., Phoc., W. Locr., Aet., Att., Ion.
(4 /Vnn/.
Thess., Lesb.
The Laconian examples : @cij3~vvos,@a&va and Alcman's
&vva are anomalous in view of e.g. +EV < "esmen, and not
very serious ones at that, since two are proper names and the
other occurs in a literary genre which shows a number of
Aeolic features. We may therefore dismiss Aeolic substrate
in Laconian, so far as this item is concerned.
Risch (p. 71) used this isogloss (o~Adv6in his table) to
establish a contact between Ionic and Dorian a t the time of
the Dorian invasion, Given that all the dialects apart from
Thcssalian and Lesbian show a similar kind of change, (a)
and (b) above, and that Dorian dialects are evenly distributed
in both ( a ) and (b), Attic and Ionic were bound to be aligned

It. COLEMAN-THE DIALECT GEOGRAPHY OF ANCIENT GREECE

95

with some Dorians. However, as in other changes resulting in


long vowels, e.g. in B (i) and C (i), Attic and Ionic belong to
a Central Greek band, so diffusion or substrate including
N.W. Greek, North Dorian and Ionic-Attic remains a real
possibility.
( 3 ) dv for dvd occurs in Lesbian dvEleEKc etc., Thessalian
dv&IKat.v etc. beside dv&lKarv,
Arcadian liv&lvac etc.,
Cyprian 6vEleEKr. The ' Aeolic ' hypothesis here rests on
several very dubious assumptions. The first is that dv and olvd
are strict doublets. There appear to be no instances of dv or dvd
standing alone (the compounds cited above are typical) and
the only example of dvd at all is in Cyprian dva40pcil where
contamination of dv and dvd has been suggested. dv could
equally well be regarded as an Ablaut alternant of &, though
its original relation to dvd remains obscure. The second
assumption is that */$/ would give in ' Aeolic ' /on/ or /o/
instead of the usual Greek /an/ or /a/. Lejeune is rightly
sceptical about this change, for which the examples cited from
Arcadian and Lesbian, e.g. 8dKoros hacordv, are too easily
explicable by analogy from other numerals to be convincing.
It is better to avoid building one vague hypothesis upon
another, and instead to regard dv in Aeolic and ArcadoCyprian as an instance of common selection as between dv
and dvci a t a time in the history of E. Greek when both forms
were living alternatives.
(4)The apocope of prepositions has been interpreted by
Porzig as an Aeolic feature. In fact apocope of these proclitics
is widespread everywhere except in Ionic-Attic. Thus hv, K a r l
Tap, TOT are common in all other regions, with sandhi forms
like KA T ~ (Her.),
V
T ~ K U T ~ T T U(Boeot.),
~
K a pqva (Arc.) and
KdMmcv (Lesb.) ; r e p is attested in Messenian, Phocian,
Cretan, etc. Thessalian, which is most affected by apocope,
shows d ~ &,
, GT, which are almost unknown elsewhere.
In some of these we may wonder whether the apocopated
I,
forms are not in fact the older ones : e.g. &rl < k
cf. Lat. ob, Osc. zip, m p I < r e p I, cf. Latin per. But in

Lejeune, op. cit., 169.

96

TRANSACTIONS OF THE PHILOLOGICAL SOCIETY

lees

any case the phenomenon serves only to isolate Ionic-Attic


and is too pan-Hellenic otherwise to be attributed to any
particular dialect substrate .
( 6 ) The simplification of if to i s before a following consonant has also been considered to be due to Aeolic substrate.
Examples include i s TOCV beside if cipxlis (Them.), ioydvor
and even 20s i&'@v
beside i x s ' E p x o p [evG], (Boeot.),
ia6CMowes but E'fZvac (Arc.), 2s 7~08' (Cypr. : Hesychius),
i s rdhcos, i .&&/as
(Arg.), iaydvois and even is6cKaKaeS
(C. Cret.). Clearly the phenomenon entails not only the
phonetic change [eks] > [es] but also, in part a t least, an
initial selection as between if and GK. In Attic and other
dialects that retained both, the distribution was phonetically
determined: 2f before vowels, E)K before consonants. In
Thessalian, Boeotian, Arcadian, Cyprian and Central Cretan
only if (> i s ) is attested. I n Argive and Phocian i~ is also
~ a K{a?rpafd
~
[UTW]
found alongside i s : e.g. b ~ ~ p d f a uand
(Phoc.).
The reduction of [eks C] to [es C] is an expected change,
so it is highly probable that in those widely scattered dialects
where selection had eliminated E'K the change occurred
independently.
(6) The dative plural of consonant stems e.g. 7~oXl~oac
(Lesb.), K~TOCKCWECTUL (Them.), a"vSpcaac (Boeot.). -EUUC
forms are also found in Phocian, E. Locrian, Corinthian
colonies, Elean (+~ydSeoac), Cyrenaean and Pamphylian
(SiKaurtpcao~
) . Although 'EmreXiSEouc is cited from Argive,
the last three letters here are a restoration. Ot/laaai in a
metrical inscription (Schwyzer 102) may be relevant, but
L
the restoration
other Argive datives like Odovoc, V V ~ A L U make
inconclusive. Aeolic (or ' Achaean ') substrate has been
claimed for this form.
The origin within the paradigm system is still disputed.
Its motivation is presumably the phonetic disruption of the
stem final before the datival-si. e.g. #2powes, c$kp~?ociXnlSt.s,
~ X ~ T L ' U Letc. Heraclean Zvracrar represents an independent
remodelling of the dative to bring it into line with the

R. COLEMAN-THE

DIALECT GEOGRAPHY OF ANCIENT GREECE

97

remainder of the paradigm. -EUUL appears to be extracted


from the -s- stems e.g. ykvea-ul, T ~ A E U - U L(Att. etc. ykveoc,
.r&hcui), and this seems to be the source (perhaps independently
in each dialect) of Attic rdhcol for *rdAcul, Rhodian N ~ ~ E U L
and Arcadian 2u8dueac-in all of which the older -L- forms
were not in fact anomalous ! Another influential factor may
have been the relation of nominative to dative in ~ M o L ,
&loiuL, whence e.g. Bv8pes, &dpeouc, though this could only
happen in dialects with - o m , -ULUL in the -0- and -a- stem
paradigms, e.g. in Lesbian, Corinthian and Pamphylian.
Given that the anomalous forms of the dative plural were
common to all dialects and that more than one formative
factor might have produced -EUUL, the innovation could have
been independent in some areas, e.g. Pamphylian.
With regard to the rest we may ask : was the innovation
due to Aeolic substrate or diffusion 1 Now within Aeolic
Lesbian, which has - o m and -acuc datives is (like Pamphylian)
a likely place to look for the independent extension of -coal.
But Lesbian on the other side of the Aegaean seems too
remote to be a source of diffusion, unless this took place before
the migration and so formed a substrate in areas later occupied
by other dialects. Here however we encounter a difficulty.
Thessalian shows beside K U T O L K ~ V T C U U L (Larissa) 6rdpXoual
(in the same locality) and X p t p a a w (Thetonion), which
suggests that -cum is recent in this dialect. I n spite of this it
is tempting to explain the emergence of these forms in
Thessalian, Boeotian, Phocian, Locrian and Elean in terms of
diffusion of some sort, though not from Aeolic. And we have
a parallel in the later spread of -01s in consonant stems
(another independent replacement of the anomalous dative
plural : + k p o v m +pdwocs supplanting r p k p o ~ e s+kppoul) from
N.W. Greek to Boeotian, Elean and even Laconian. Our
conclusion is that -EU (u)c emerged independently in several
areas of Greek, along with -auui, - OK, to replace the older
dative case, and in N.W. Greek and Continental Aeolic was
diffused from an undefined source. There is no need in all this
to assume any pre-Dorian Aeolic substrate.

98

TRANSACTIONS OF THK PHILOLOGICAL SOCIETY 1963

(7) The athematic conjugation of verbs with vocalic suffix


+ c a X q p - b w u (Lesb.), beside KaXroplvwv, etc. (Ion.-Att.), yaopyeiprv (Thess.) beside y r w p y ~ i v
(Ion.-Att.) q5iXapL (Boeot. : grammarians) beside +ch&
(Ion.-Att.).
A number of scholars have held that where this paradigmtype occurs outside Aeolic it is due to Aeolic substrate.
Risch (p. 71) saw the community of i w etc. in Dorian and
Ionic as evidence for contact between these two groups at the
time of the Dorian invasion.
Once more we must start from the pattern of distribution.
Three groups emerge :
( a ) - l w etc.-Cor., Rhod., Att., Ion.
Laconian shows ~ K / ~ ~ V Tfrom
B S
i ~ i 3 &but
~ , only in a
literary text.
( b ) -qpc etc.-Arg., El., Pamp., Arc., Cypr.
( c ) Both -do and - q p ~types: Her., Ther., Coan, Phoc.,
Cret., Aet., to which must be added Boeot., e.g. 2iroX+ov
and dirLmr+avw&rv (with thematic -cpcv but - w - ) ,Thess.
hvX?ipcowos and Lesb. ciypcdpcvo~ (for all of which
cp. the athematic forms cited above).
Relevant here is the reverse tendency, aided by inherited
doublets like -viip/-Vuw, to extend -rw, etc., to inherited - p
paradigms, e.g. Phocian ciroKa6mrdovrrs , Cyrenean 6cSdv,
Lesbian d p v a v , Euboean T L ~ E ~ VKaBLu76V.
,
The distribution
in (a), (b), ( c ) , is clearly more complex than the neat divisions
in Rischs table suggest.
Contract verbs in -EW, -ow, -aw reflect the inherited
denominative (-*o/~-, -*Ii- + -*j6) and iterative-causative
(-*ej6) types, while their doublets in - q p J-up, -+L represent
analogical extensions from inherited root-class verbs with
long final : r & p , 816wp~,ZurEp. Several possible factors
can be discerned in this analogical extension. Firstly denominatives from the -a- stem declension like rrpdw would
originally have /a./, with a long vowel comparable to tartip.
Secondly the generalization of the long vowel in the paradigms
of inherited -pi verbs, e.g. 8iSwu6ai for 6i6ouBaL in Lesbian,
( contract verbs ), e.g.

R. COLEMAN-THE

DIALECT GEOGRAPEY OF ANCIENT GREECB

99

would lead to similarities in some dialects with contract vowel


reflexes, e.g. *{apidEuBai > {apiBuBai. Moreover patterns
resulting from contraction in one class, e.g. SovAov^v( < -*o-EV) :
8ovAodp~vos(< -*o-op~vos)might be extended to give &kiv :
+~helpevosin place of ~#tAo&p~vos.
Some of these structural pressures through internal analogy
and the like would be operative in all dialects, so that in view
of the evidence of a two-way analogy mentioned above, the
complex pattern of distribution, cutting across the usually
defined major groups and including all three stages of development, is best explained in terms of independent change in
each dialect. In any case, given the occurrence of both -Cw
and -qpi types in all the Aeolic dialects, we can hardly explain
-7pc elsewhere as an Aeolism (however we define that term).
The fact that Ionic-Attic and some Dorians agree in showing
-&,etc. is insignificant, in view of the fact that this is the
inherited type of conjugation for stems with vocalic suffix.
(8) The form Cu is used as the feminine of ECS in Boeotian,
Thessalian and Lesbian, in contrast to pia elsewhere. A form
Ids turns up in Messenian (if rdv ytdv 2vtavrdv is the right
reading) and Cretan. Porzig, rightly connecting this with la,
regarded it as an Aeolism surviving as a substrate feature
in Dorian.
Ids in both the Messenian and Cretan texts and in the
unique Homeric example Iwi . . . +ari (Iliad 6.422) seems to
have a deictic force equivalent to 2tceivos. It is probably
connected with Cyprian Iv, which is glossed by Hesychius as
airqjv, airrdv, and related to Latin im, em and the thematic
eum (< *ejom) etc., Gothic is and Lith. jis. The Greek
thematic form has zero g a d e of the root, but in view of the
position of the accent this need not trouble us. In Aeolic
l a with a change of accent and of meaning from that one
to a single one replaced the inherited pia, which was
anomalous in Greek generally in relation to the masculine and
neuter (cp. the pGk %ems, *smia, *sem). Lesbian p$ieia
beside prj8e~sp*.jj&vraises the possibility that this replacement
first occurred in the negative forms, o & h a replacing oirSipia.

100

TRANSACTIONS OB THE PEILOLOGICAL SOCIETY 1983

The fact that an inherited pronominal type survived in


Messenian and Cretan and perhaps elsewhere too-who
knows ?-with an athematic doublet form in Cyprian, and
was adapted to a new function in Aeolic is surely no reason
for assuming Aeolic substrate in other areas of Greek.
(9) The appearance of byplw as the semantic equivalent of
alp& in various parts of Greece is also put down by some to
Aeolic substrate.
The distribution is :
(a) &yp&-Lin. B, El., Boeot., Lesb.
( b ) aEplw-Tar., Cor., Arg., Cret., Rhod., Ther., W. Locr.
( c ) Both verbs-Coan,
Attic-Ionic, where &yp& is used
(with alpt'w) by Archilochus, and occurs in the compound [wyp&, Lac., Cret. Aet. &yvdw may also belong
here, though its meaning is uncertain.
The derivation of +piw is clear. As a denominative of
dypds it is ultimately related, like &yv& and Ionic iyivdw,
to gyu. alp& is more difficult. If it really is related to Skt.
sdrati, then we can set up for pGk the possibilities *s$
> *ai'pw and *sore$ > *dplw. alp& could then be a
contamination of &yp& or *6plw with "azpw ; but the whole
question is obscure. There is no reason to believe that the
situation represented by (c) above was not an ancient one.
dypdw and alplo could well have formed part of the same
lexicon, with an area of overlapping function which led to
independent selection in many dialects.
In any case the evidence is very incomplete. We do not
know whether Thessalian had alp&, hypiw, both or neither ;
nor whether Boeotian and Lesbian had alp& as well as Zypkw.
Where the evidence is so scrappy, as it often is in morbid
lexicon, we must be cautious.
(10) The same caution is needed over the p~rtl/rr~Gd
isogloss. The view that mSd is due to Aeolic substrate is
widely held.
The distribution of the two prepositions is as follows :
(a) p ~ ~ 6 - wLocr.,
.
El., Phoc., Arg., Cor., Ion., Att. It is

B. COLEMAN-THE DIALECT GEOGRAPHY OF ANClENT GREECE

101

widespread in the Koine period, but partly perhaps as


the re,sult of diffusion from Ionic-Attic.
(b) dd-Lac., Arg., Cret., Ther., Cor. (Corcyra), Boeot.,
Lesb., Arc,
perd is clearly inherited (of. Gothic mip) and re6d innovatory ; but there is no reason t o assume that the innovation
must be Aeolic in origin. Indeed both p r d and nr6d could
have belonged to pGk.1 Linear B shows both and the situation
attested at a later date is easily explicable in terms of selection.
To summarize : the isoglosses on which the hypothesis of
Aeolic substrate depends are all capable of quite different
interpretation. In some, e.g. C (3), the item concerned seems
East Greek rather than Aeolic in particular ; in others, e.g.
B (a), C (a), ( 5 ) , (7) the wide pattern of distribution is best
explained by independent change ; in others again, e.g. C (9),
(lo), the actual Aeolic evidence is slender ; while fmally in
C (8) what is in fact revealed is an apparently exclusive
Aeolic development. The concept of Aeolic substrate can thus
be rejected, and with it the evidence for a pre-Dorian ' Old
Achaean ' or ' Pelasgian ' with marked Aeolic characteristics.
As for diffusion a t a later date we saw the possibility in C (1)
that the Boeotian raising of /e./ might be due to Thessalian
and in C (6) that -EUUL also might be diffused, though probably
not from Thessalian.
(D) A t the time .f the Dorian invasions lo& and Dorian
were for a while in close
Some of the relevant isoglosses have already been examined :
E)v and C'VS in R ( 5 ) , aeAh&va, aehdvv6 etc. in C (a), -&o/-qpt
in C (8). To these we may add two more :
(1) The change */T/ > /ar/ or /ra/, */Q/ > /an/ or /a/
has been held to mark off Ionic-Attic and W. Greek from the
dialects showing o-vowels in both these sets of reflexes. The
examples here are difficult, and the following discussion will
be confined to the detail of a few of them only.
pdms (Arc.), p i m a (Cret.), pdu4a (Lesb.), pdurroS~(Thess.), may well
contain initial * ~ c T -(beside p d ) .
Risch, 71-72, 74 ; Chadwick, op. cit., 4344.

102

TILANSACTIONS OF THE PHILOLOGICAL SOCIETY 1963

(a) U T ~ U T ~ US ,T ~ ~ T O SSkt.
.
stpkih confirms the probability
that /ro/ here does reflect */?/ and not the Ablaut */ro/.
a ~ p a r d sor its compounds occurs in Arg., Cret., Thess., Att.,
Ion. ; U T ~ ~ T OinS Boeot., Lesb. The data are woefully incomplete, since we have nothing from Arcado-Cyprian and a very
inadequate representation from Dorian. The split in Aeolic is
significant however, suggesting a more complex pattern of
reflexes than Risch allowed. For it begs the question to assign
Thessalian -pa- to West Greek influence.
( b ) The numeral ' four, fourth ' :
r4~upros (Att., Ion.) T E T U ~ T ~ U(Coan)
S
T C T ~ ~ T U (Arc.).
U
Two forms of the cardinal root are reflected in Greek:
(i) *kzuetw6r-(cf. Skt. cutv&ralb)> srluupes (Horn.) with [sur]
< "[tjur] < *[twur] < "[twor] ; and ~ k o p c s(Dor.) with T C T by analogy with T E T ~ ~ K L S T&CLpTOS
,
etc. ; (ii) *k%wr> T ~ T T U ~ E (Att.)
S
with [ettar] < "[etw;] and ~ 4 u ( o ) u p ~ s
(Lesb.) with [ssur] < *[tjur] < *[tur] (cf. Skt. acc. pl. cutzirah).
The inherited ordinal form is *kwetTt6s with *tr for *tw,r. The
difficulty here is that evidence is very meagre not only for the
ordinal itself but for ordinal and cardinal together in the same
dialect. The possibility of analogical influence from one to
the other is considerable and the -op- of, for instance, T E T ~ ~ T U U
could well derive not from /g'but by analogy from -op- in
the cardinal. In this respect Cnidian T C T O ~ E S beside Coan
r ~ r u p ~ 2is
v svaluable in showing the ' regular ' pattern of /or/
in the cardinal, */;/ in the ordinal. But again the overall
picture is too fragmentary to permit decisive argument.
Examples of the former in Att., Ion.,
( c ) ypu+-, ypo+.
Arc., of the latter in Her., Arg., Cor., Cret., Mel., Phoc., Leah.
Both are attested in Elean. The root is *gerbh- and cognates
are found in OHG kerbun, OCS ir6bS. Hence in ypdqh
/ra/ < */r/. But ypo+ is opaque. For apart from rare verb
forms like ypomds (Lesb.), most of the occurrences are in
nouns, where the - 0 - could be Ablaut, e.g. &ypo+os (C. Cret.),
u u v y p o ~ o v(Phoc.).'
A large number of such pairs could be listed (see Adrados,
1

cf. thc difference of gradation,e/zero, in KPCTOS/KP&TOS,

Bipoos/B&puos

etc.

R. COLEMAN-THE

DIALECT GEOGRAPHY OF ANCIENT GREECE

103

p. 44). For some the material is too restricted, e.g. ppoxtk


which in contrast to ppax6s characterizes Lesbian and
Boeotian ; but the word. is too rare to be of help beyond that.
For others the testimony is ambiguous, e.g. E)@opKds in
Arcadian could reflect */?./, but it equally well could be
Nevertheless some
influenced by the inherited Z@opa.
general pattern does emerge from the chaos: in Arcadian
and Lesbian /ro/ and /or/ predominate, in Dorian and
Ionic-Attic /.a/ and /ar/.
The chronology is uncertain. to-pe-zuin Linear B is aligned
with Arcadian. But if the change in Lesbian and Arcadian
had also occurred before the Dorian invasion, did Ionic still
retain /r/, as Rischs hypothesis would seem to require ? The
possibility that the changes in Arcadian and Lesbian, Ionic
and Dorian respectively were connected remains an open one,
but independent change, perhaps determined by physiological
or linguistic substrate, should not be excluded.
( d ) Evidence for /o/ < */?/ or */?/ is very thin indeed.
Risch cited Homeric 6rarpos beside Ionic ZAoxos, but the
former compound is obscure. The only direct evidence from
s Lesb.,
Lesbian and Arcadian is all in the numerals : 8 d ~ o r o in
Arc., Lin. B (de-ko-to) for *dekqa-to- ; but Arc. %KO contrasts
with Lesb. 6 d K a (declined), and Cyprian has SKaroL ; both
&TOS
and &&a occur in Lesbian, ~ K O T ~inV Arcadian. The
possibility of analogy in so restricted a cIass of occurrence
renders these examples inconclusive.
All the more so, since in other contexts when we should
expect to find reflexes of */GI only av or a is attested. In both
Lesbian and Arcadian, e.g. Lesb. /lacrlAqs, &~T&JK=,
Arcadian A p f a , dvdtkz. It looks then as if av, a are to be
reckoned the normal Greek reflexes, in which case the
agreement of Dorian and Ionic here is insignificant.l
(2) The reflexes of the labio-velar before front vowels.
Risch (p. 71) argued that the contrast between &pa- etc. in
1 A divergent treatment of medial and final */Q/ is possible. &~vrd&ov etc.
in Cyr., Her., Phoc., against 7640s etc. in Att., Ion., looks like /o/ < * / g / ,
but the dialectal distribution here still does not fit Rischs hypothesis.

104

TRANSACTIONS OF THE PIIILOLOUICAL SOCIETY 1063

Ionic, Attic and Dorian on the one hand and &po- etc. in
Aeolic on the other is due to contact between Ionic-Attic and
Dorian. His earlier discussion (p. 69) of Arcadian seems to
imply that he regards &p- etc. there as similarly due to
Dorian contact. Linear B of course preserved /kw/.
Now the overall tendency in Greek is to palatalization, of a
very distinctive kind, the change from *Icwisto 71s proceeding
by way of initial *[kw] > *[kW']> *[kj] > *[c] > "[ti] > [t].
There are one or two aberrant developments: for instance
in Thessalian and Boeotian where *[kji] > [ki], e.g. K i s
(Phalanna) S i E - K i (Larissa), (but "[kwc] > [pel usually :
~ i p m &-Tapes).
,
In parts of Arcadia we find evidence for
development to an affricate, viz. *[tl] > [ts], e.g. am, nis
(Mantinea), ;(is (Kleitor), T ( E T ~ U K ~ T L U C(Tegea). But the
usual Greek development is also attested in Arcadian, e.g.
&is, ChJTECOdTW, T E T ~ ~ T U U
(also Tegea), &S (Orchomenus).
In fact only Aeolic and Cyprian show any exception to the
general tendency, and, as we have noted, even Thessalian had
traces of the early stages of palatalization before /i/ in
contrast to the normal change to labials in that dialect. For
Cyprian the evidence is scanty and inconsistent : pe-i-se-i
points to labial treatment, unless p - is here analogical, and
o-pi-si-si-ke (= *;+is KE) could derive from *& m s KE by metathesis. 4 KE) uis on the other hand is reminiscent of the
development in certain parts of Arcadia.
Where the general pattern is one of palatalization, it seems
arbitrary to assign the shift to a single dialect group and then
assume diffusion.
To summarize then: the evidence for contact between
West Greek and Ionic seems negligible. The alternate forms
in B ( 6 ) turn up too often together in the same dialect, and
the pattern is better explained in terms of selection. C (2) and
D (2) serve rather to set Aeolic apart from the other dialects
than anything else, D (1) brings Lesbian and Arcadian
together but leaves us uncertain about other dialects outside
of Ionic-Attic. Finally in C (8) independent change could
account for the pattern, and the fact that Ionic and Dorian

R. COLEMAN-THE

DIALECT GEOGRAPHY OF ANCIENT GREECE

105

agree in retaining an inherited feature is insignificant. One


last point: given the population movements and general
unsettlement at the time of the Dorian invasions, it is unlikely
that any period of contact between West Greek and Ionic
speakers would have been sufficiently prolonged to bring about
the kind of infiltration that Risch envisaged.
The analysis of the four hypotheses in turn has left only one
of them (A) unscathed, though even there the evidence has
turned out less weighty than has been generally supposed.
As we shall presently see, there is something in both the
Thessalian-W. Greek and Lesbian-Ionic hypothesis of (B) and
the Ionic-West Greek one in (D), though the facts are more
complex and the relationships more subtle than the above
formulations implied, and both hypotheses need to be
substantially modified.

I11
Throughout the discussion of individual isoglosses in the
preceding section three general principles have emerged :
(1) That any given item must be seen in the structural
context of its own dialect before an isogloss including it is
drawn, in order to establish precisely what the point of
comparison is.l
(2) That the pattern of distribution and variation relevant
to each item must be examined for the whole linguistic area
and not merely for a narrow range of dialects in which the
investigator may be temporarily interested. Only in this way
is it possible to discriminate between the normal and the
aberrant phenomenon, the frequent and the rare or unique,
the geographically restricted and the widespread or spasmodic.
(3) That the diachrony underlying the situation represented
by the isogloss must be considered in detail, in order to
See U. Weinreich, Is a structural dialectology possible ? Linguistics
Today (ed. Martinet-Weinreich, New York 1956), 268 ff., especially 2 7 2 4 .
The importance of structural considerations is well illustrated by Ruiperezs
discussion, 259-60, of Att.-Ion. -?I&-, -w.

106

TRANSACTIONS OF THE lHI1,OLOUICAL SOCIXTY 1 W 3

determine the possible genetic relationships among the items


on the isogloss and to distinguish between cases of retention,
innovation and selection.
But even if we bear these principles in mind, the exclusive
attention to individual isoglosses or small groups of isoglosses
is still too limited methodologically and its conclusions
uncertain. For it is easy enough to bring forward a set of
isoglosses uniting almost any two dialects, and no less easy to
produce other sets dividing them and uniting each in turn to
any one of a number of quite different dialects. What is
required is a whole list of isoglosses, comprising a sufficiently
large and varied range of items to characterize each dialect
both in itself and in relation to all other dialects. Against the
pattern provided by this list each particular isogloss can then
be properly considered and assessed.
How many items do we need to establish this comprehensive
picture ? Rischs table (p. 75) contained twenty-one items,
nine of which may be classed as phonological, eight morphological and four lexical. Rucks table lists no less than sixtyeight, The classification here is harder to ascertain : some
; others
are clearly phonological, e.g. 818w~i, @pu, ~ k p m
morpho~ogical, e.g. FlKaTi, &uav, nd8raoL ; others again
,
lexical, c.g. ~ j j v o s ,rL, m 8 8 ; and a few, like ab1s ~ a 6n-d
+ dative, syntactic. But where do we place lap&, yivvpai,
GKOTOS
for instance ? In spite of this uncertainty it is still
fair to say that the majority of Bucks items are phonological
or morphological, which in the light of what was said in
section I seems the correct bias. Obviously Bucks list is in
principle more useful than Rischs, since the larger number
we include, the better chance we have of precision both in
characterizing and comparing the dialects.
It is essential also to have a recording technique that yields
maximum information, and a t this point Bucks table is
seriously defective. For he merely indicated for each item the
dialects in which it occurred, without differentiating in the
other dialects between those which showed a comparable but
discrepant form of the item (negative occurrence) and those

N MONO

El.

TIIE

GREEKDIALECTS.~

Phoe. Low. Aet, Boeot. These. Lesb. Arc. Cypr. Pamp. Ion.

Att.

L.B.

-77

.64

*66

.67

-61

.45

.39

*40

*20

*45

.34

.34

*56

-80

*75

.67

.44

*f30

.46

.37

*35

.24

.GO

a31

a34

-37

*88

.72

.64

.77

.70

a46

.34

-32

.23

.45

.38

.35

044

*74

.61

.56

*68

.58

.45

.38

*41

.26

.23

*29

-31

.50

*61

.72

*77

*77

.52

.39

*37

-33

*13

*45

.41

.42

*52

-70

.SO

.86

-84

~ 5 2 -44

-29

-44

-18

-47

-40

a41

a45

.52

.58

-46

a57

.49

.45

*35

.43

.32

*50

a30

.28

.36

.71

-77

363

-67

a51

a49

-32

*35

.17

-50

*36

*37

a33

*62

.66

.63

.68

.51

-37

.35

-28

.14

*42

-37

a42

.53

.64

.70

.69

.74

*50

*45

,49

*33

a23

a 3 1

.41

.35

44

-67

*83

.78

.84

.54

.44

.26

*37

*19

.57

.39

-39

-25

.73

*67

.65

.69

a53

.46

.36

-28

a47

.34

.27

*35

.73 1.00

.83

*84

.54

.51

.31

-36

.22

.48

-39

.33

.30

.67

.83 1.00

*91

.52

*44

*32

.39

-17

a34

.47

.49

-41

.65

-84

*91 1.00

*47

.46

.60

*27

.24

-53

.45

*48

-31

1.00

-69

.54

352

a47 1.00

-53

-51

.44

.4B

.68

.58

*35

.38

.35

*53

*22

-23

.36

1-00

*56

-43

-43

*38

.32

.33

.46

.31

.32

*60

*28

.35

.56 1.00

a38

*41

a29

+50

-40

-36

*36

.39

.27

.55

.38

.43

-38 1-00

-63

.36

-47

-44

a57
*56

.28

.22

*17

*24

.35

.43

.41

*63 1.00

-34

a39

*37

*47

.48

.34

*53

.53

.38

.29

*36

.34 1.00

a32

-32

-27

-34

*39

.47

*46

*22

.32

.,50

.47

-39

*32 1.00

*SO

a63

.27

.33

.49

-48

.23

.33

.40

-44

*37

.32

-80 1.00

.35

.30

.41

.31

.36

.28

.55

.57

.66

*27

.63

.6l

.61
1.00

iibits x items and dialects B and C respectively z - b, z - c items comparable t o A , then

-.

3ffieient of A and R is given by the formula A and of A and G by the formula q


x-b
x - c
[Facing p . 107

CORRELATIONCOEFFICIENTS
Lac.

Her. Mess. Arg.

Lac.

1.00

-88

.78

.77

Her.

a88

1.00

.93

.79

Meg. Cret. Ther. Rhod. cos. Ach.


etc.
etc.
*74 .SO
.67
*78 a 7 3
.69
*60
.69
.76
.70
-73 .59
.79
a74

Mess.

.78

*93 1.00

.81

.71

.74

*78

-75

*70

a77

*65

Ag.

.77

.79

.81

1.00

.73

-75

a77

.72

-74

-72

-91

Car.

.74

*69

.71

*73 1.00

-87

.60

a68

-75

*68

*67

Meg.

80

.76

*74

,75

.87

1-00

*69

*71

-81

*65

.69

Cret.
-67
.70
Ther. etc. ~ 7 8 .79

-78

*77

.60

-69 1.00

-71

-69

-62

*46

*75

-72

.68

-71

-71 1.00

-71

-80

*65

Rhod.

Car.

.74

.70

.74

.75

.81

.69

-71 1.00

.76

.63

Cos.etc. .69

*73

.77

.72

-68

.65

a62

.SO

*76 1.00

.63

Ach.

*60

*59

-65

.91

a67

.69

.46

-65

.63

a63 1.00

El.

.77

*80

.88

74

.61

.70

-52

-71

.62

a64

Phoc.

a64

.75

.72

.61

.72

*80

-58

-77

*66

*70

.83

Locr.

-66

-67

-64

.56

.77

.86

.46

*63

.63

.69

*78

.73

a67

Aet.

-67

.45

*77

.68

.77

-84

-57

.67

.68

*74

a84

Boeot.

a61

*60

a70

.58

.52

.52

-49

*51

a61

.50

*54

Thess.

.45

*46

*46

.45

-39

*44

*45

a49

a37

.45

a44

Lesb.

.39

*37

.34

.38

.37

.29

-35

a32

*35

-49

-26

Arc.

.40

*35

*32

.41

.33

.44

-43

.35

.28

a33

.37

Cypr.

.20

*24

.23

*26

.13

*18

-32

*17

.14

-23

e19

Pamnp.

-45

-50

.45

-23

-45

.47

.50

-50

.42

.31

a54

Ion.

a34

-31

-38

.29

.41

.40

.30

.36

*37

.41

*39

Att.

.34

*34

*35

a31

*42

*41

.28

.37

.42

.35

-39

L.B.

.56

.37

.44

.50

-52

.45

a36

.33

.53

*44

.25

The figures here are arrived a t by the following method : Where dialect A ex1
if p of these z - b and q of these z - c are identical with A , the correlation cot

R. COLEMAN-THE DIALECT GIEOGEAPHY OF ANCIENT GRXECE

107

where no comparable data were attested a t all (zero occurrence). Moreover within the class of negative occurrence we
can subdivide still further. If five dialects A , B, C , D, E
exhibit for the reflex of *x in *p-ABCDE the distribution
pattern x1x 2 x3 x1 and -, we may start by excluding E from
this isogloss, but we then have to establish the classification
among the remaining four. Now B and C stand over against
A and D,as showing negative occurrence of xl. Similarly
A , C and D stand over against B, as showing negative occurrence of x2. This means that for every dialect we have an
invariable three-value system ; positive-, negative- and zerooccurrence, into which every item can be placed. On the
other hand for every item we have a variable-value system,
depending on the number of distinct forms comprised by the
isogloss (in the example just given we should have four values,
corresponding to x1 x 2 x3 and -).
The table opposite summarizes the results of a factorial
analysis carried out in terms of the classification just proposed. The following fifty-one items were selected and the
pattern of variation is indicated for each isogloss in turn by
a single example :
1. */a./ :
> (i) /av/ ZuTEpi (Dor.)
(ii) /e*/ ~ U T 7 (Ion.)
p
2. */e,/ /o./ :
= (i) [e'] [O'] K a T h $ d 6 O V T O S , 6EGv (Lesb.)
(ii) [ y ] [o.] &YELP, Mwudwv (Boeot.)
(iii)
[?'I dvk68ri~ca~ O U K E(Thew.)
3. */ee/ /oo/ :
(i) > < /e-/ /o./ T ~ E v <apiGvn
,
(Meg.)
(ii)
4/e*/ /o*/ & r o L ~ i ,Sdpou (Cor.)
4. */eo/:
> (i) /eo/ F ~ T E O(Her.)
S
(ii) /io/ E'Ko'upiov (Cret.)
(iii) /eu/ ZTOLECVTU(Phoc.)
(iv) /o./ or /?*/ ykvous (Att.)
[ 9 7 ]

108

TRANSACTIONS OF THE: PHILOLOQICAL SOCIETY 1963

5. */ae/:

> (i) /w/ ; u ~ K & (Att.)


(ii) /e./ &'K; (Lac.)
6. */e.o/:
> (i) /e,o/ flaaihqos (Lesb.)
(ii) /eo/ flaaiXkos (Ion.)
(iii) /eo*/ flaoih4ws (Att.)
7. */en/:
> (i) /en/ E)u (Boeot.)
(ii) /in/ Iu (Arc.)
8. */ho/:
> (i) /ho/ beside /h-/ ho, h l m o i s (Lac. G)
(ii) /o/ beside /h./ d, ~ O ~ K O(W.
V
Locr.)
(iii) /o/ beside /-/ d K a d m a p c u (C. Cret.)
9. */kw/ before high front vowels :
> (i) /k/ K i s (Pelasg. Thess.)
(ii) /ts/(?) nis (Arc. Mantinea)
(iii) /t/ 71s (Arc. Tegea)
(iv) /p/ m b a c (Boeot.)
10. */ti/:
> (i) /ti/ SlSonc (Arg.)
(ii) /thi/ Kah4oovB~(Boeot.)
(iii) /si/ +poicrc (Lesb.)
11. */tu/ :
> (i) /tu/ ~ ~ ~ L T ~ E K(Cret.)
T O U
(ii) /su/ 7jpcu6s (Ion.)
12. */tj/:
> (i) /ss/ T ~ W O S (Leab.), /tt/ T ~ T T O S(Boeot.)
(ii) /as/ and /s/ Zpluuw, T ~ U O S(Ion.)
/tt/ and /s/ +mw, &OS (Att.)
13. */Vntj/:
> (i) /Vns/ +ouaa (Arg.)
(ii) /V,*s/ where /V,*/ = /e./ /o-/ & waa (Ther.)
(iii) /V,.s/ where /V,./
/e-/ /o./ Zxovua (Cor.)
(iv) /Vis/ E'Koiaa (Cyr.)

R. COLEMAN-THE DIALECT GEOGRAPHY OF ANCIENT GREECE

14. */Vns/ in final syllables :

> (i) /Vns/ 6cdvs (Arg.)


(ii) /V,x/ xpo&ws (Lac.)
(iii) /V,.s/ xpo&ovs (Att.)
(iv) /Vis/ vo'pocs (Lesb.)
(v) /VS/ m p a v y d s (Arc.)
15. */Vsm/:
> (i) /Vmm/ Zppc (Thess.)
(ii) /V,-m/ 7jpI (Lac.)
(iii) /V,*m/ a l p ' (Phoc.)
16. */Vnw/:
> (i) / V n / ~ V V ~ K (Lesb.)
CZ
(ii) /V,.n/ KUFVOS (C. Cret.)
(iii) /V,m/ &VOS (Ion.-Chios)
(iv) /Vn/ ,$.&os (Ion.-Eub.)
17. */rs/:
> (i) /rs/ $ p q v (C. and E. Ion.)
(ii) /rr/ $ppyv (W. Ion.)
18. */eks/ before consonants :
> (i) /eks/ e-xe-pa-si-n.e = 2[#Iaaev (Cypr.)
(ii) /ek/ (doublet replacement) 2K/3auts (Att.)
(iii) /es/ 2s TO; &ydvot (Arc.)
19. The name of Apollo :
(i) ' A x d M w v (Cor.)
(ii) 'A~riAAwv(C. Cret.)
(iii) "A~Aovv
(Thess.)
20. The lexeme ' sacred ' : *isaro- :
(i) lap& (Arg.)
(ii) tcpdr (Att.)
(iii) Tpos (Lesb.)
21. The numeral ' four ' :
(i) T ~ T O ~ E(Lac.)
S
(ii) &Tapas (Att.)
(iii) ~ & ( a ) v p c s (Leab.)
(iv) ~ia~cppcr
(Ion.)
PHILO. TRANS.

1963.

109

110

TRANSACTIONS OF THE PRILOLOOICAL SOCIETY 1963

22. The numerals ' twenty, two hundred ', etc. :


(i) ( a ) F i K u T c (El.)
( b ) TITEVTUKUTiUS (COan)
(ii) ( a ) E ~ O U (Lesb.)
L
( b ) T/JLUKdOLOi (Ion.)
23. masculine a^- stems :
(i) ( a ) -a' nom. sg. K u X u (Boeot.)
( b ) -a^s gen. sg. n p O K h & S (Cor. in Acarn.)
(ii) ( a ) -6s nom. sg. ' A ~ x ~(Dor.)
T~s
(b) -60gen. sg. 'Op&60 (Theas.)
24. dat. pl. of -6-stems :
(i) FOCKZTU%(W. Locr.)
(ii) 6p+o&puiac (Lesb.)
(iii) +TLV (Ion. Naxos)
(iv) Nup&jmiv (Ion, Thasos)
25. dat. pl. of consonant stems :
(i) T&L (Ion.)
(ii) n-dvmuai (Lesb.)
(iii) ndv7ois (Aet.)
26. Oblique cases of ' Zeus ' :
(i) AiFds (Arg.)
(ii) Zqvds (C. Cret.)
27. The adjective ' first ' :
(i) ( a ) Tp&ros (Boeot.)
( b ) .IrpdmrTos (Cret.)
S
(ii) ( a ) T ~ ~ T O(Lesb.)
(b) TfldTCC?TOS (Ion.)
28. nom. pl. of the definite article :
(i) ol (Lesb.)
(ii) 701 (Boeot.)
29. The pronominal compound u3Toaavrds :
with (i) 1st part undeclined (Cret.)
(ii) both parts declined (Phoc.)
30. The deictic pronoun :
(i) 6 8 (Lesb.)
~
(ii) 0"vu(Cypr.)

R. COLEMAN-THE

DIALECT GEOGRAPHY OF ANCIENT GREECE

(iii) &a (Arc.)


(iv) [&I rouvv~ouv(Thess.)
(v) [ d - l ] T U (El.)
~
31. The prepositions :
(i) r o d (Lac.)
(ii) .rrds (Arc.)
(iii) r o p d (Cret.)
(iv) m p r l (Pamp.)
(v) npds (Ion.)
(vi) r p & (Lesb.)
(vii) pod (Horn.)
32. The prepositions :
(i) pcrd (Boeot.)
(ii) m 6 d (Phoc.)
33. The prepositions :
accus. (Thess.)
(i) E)Y
(ii) 2vs accus. (Arg.)
34. The conditional particle :
(i) el (Arc.)
(ii) 7i (CYPT.)
(iii) ul (Lac.)
35. The hypothetical particle :
(i) KU (Boeot.)
(ii) KE (Lesb.)
(iii) Gv (Att.)
36. The temporal conjunction :
(i) GT~E(Ion.)
(ii) Gru (Lesb.)
(iii) GKU (Boeot.)
37. The temporal adverbs :
(i) .rrpdu& (Cor.)
(ii) .rrpdu&v (Att.)
(iii) .rrpduOu (Cret.)
38. The verb ' to wish ' :
(i) 6clXop.a~(Phoc.)
(ii) ,9dXopa~(Cret.)
(iii) ,9dXopur (W. Ion.)

111

112

TRANSACTIONS OF THE PHILOLOGICAL SOCIETY 1063

39. The paradigms of ' contract ' verbs :


(i) thematic K U ~ O V ~ & U (Att.)
(ii) athematic K U X ~ ~ E I U W (Lesb.)
U
40. The sigmatic tenses of verbs in -50 :
(i) &y@a.ro (Thess.)
(ii) &JdUTlUUU (Lesb.)
41. The sigmatic future formation :
(i) &uoCpui (Meg.)
(ii) 2'uopui (Att.)
42. The stem of the perfect paradigm :
(i) athematic ylyoua (Att.)
(ii) thematic ycydvei (Rhod.)
43. The form of the 1st person plural :
(i) -men 6nl80pcu (Boeot.)
(ii) -me8 ~ T E S O ~ K U ~ (Phoc.)
LES
44. The form of the 3rd person medio-passive :
(i) -tni A+WUTCU (Ion.)
(ii) -toi cbqbAiy6u~oi(Arc.)
45. The form of the 3rd person of the athematic secondary
paradigm :
(i) 2'tb (Arg.)
(ii) 6v40euu (Boeot.)
(iii) dvl8euau (Ion.)
46. The form of the 3rd plural imperative :
(i) qkpdwu (Lac.)
(ii) q!qdwwu (Cret.)
(iii) t$EIpowou (Lesb.)
47. The thematic infinitive :
(i) { X E V (Her.)
(ii) 2 ' p u (Cor.)
(iii) +lpcpev (Boeot.)
48. The athematic infinitive :
(i) 8 d p u (Meg.)
(ii) Gdpeiu (Rhod.)
(iii) G d p e u a ~(Lesb.)
(iv) SoGuui (Ion.)

R. COLEMAN-THE DIALECT GEOGRAPHY OF ANCIENT GREECE

113

49. The relative pronoun :


(i) &, 0"v (Att.)
(ii) d, T ~ (Phoc.)
V
50. The remoter deictic pronoun :
(i) T+OS (Her.)
(ii) K E ~ V O S(Cret.)
(iii) EIKE~VOS (Att.)
51. The verb ' to take ' :
(i) alp& (Att.)
(ii) Ayp& (Cor.)
(iii) 6yv& (Aet.)

These items were plotted against twenty-four dialect-units.


The number of positive occurrences (as defined on pp. 106-7)
in the resultant item-dialect table was as follows :
37 Rhodian, Coan (and other
51 Ionic, Attic
50 Cretan
minor Dorian islands)
48 Lesbian, Arcadian
36 West Locrian
47 Argive, Boeotian
35 Elean, Cyprian
46 Corinthian, Phocian
31 Aetolian
45 Laconian
27 Achaean
43 Heraclean, Melian-Theran- 23 Linear B
Cyrenean
22 Pamphylian
41 Thessalian
17 Messenian
40 Megarian
Several other dialects were included in the original list, but
their yield on the items chosen was too low to be statistically
useful. Indeed this same objection might be made against the
last four on the list actually adopted, especially to Pamphylian
and Messenian. Por Linear B the number is of course reduced
by a number of special factors. Firstly the phonological
opaqueness of the script a t many points (cf. Cyprian) ;
secondly the restricted character of the material, which leaves
huge gaps in the morphology ; and lastly the application of the
chronological principle laid down on pp. 61-2. Por in a number
of instances (e.g. the labio-velar series) Linear B reveals a

114

TRANSACTIONS OF THE PIIILOLOGICAL SOCIETY 1063

situation which must be assumed for the prehistory of all or


most of the later dialects.
The results of the analysis, it must be emphasized, are only
tentative. There are several ways in which they could be
given greater accuracy and precision, For instance by
operating with a larger number of dialect units. Although
the addition of, say, Acarnanian or East Locrian would not
be helpful, the further subdivision of several groups certainly
would be. There are numerous instances where for a given item
a dialect stands on more than one isogloss. Sometimes this
situation is due to the structural context of the item, but more
often the variants are distributed locally within the dialect
area. In calculating the correlation coefficients these indeterminacies (which differ only in degree from those noted on
pp. 59-60) had to be allowed for in a somewhat arbitrary
manner, but they would be largely eradicated by working with
a higher number of units. Obvious examples would be the subdivision of Arcadian, Cretan, Thessalian, Ionic and the
complexes of Melian-Theran-Cyrenaean and Coan-Cnidian-etc.
Secondly the list of items could be greatly extended.
A glance through Buck reveals potential isogloss data on
almost every page. Some of these features, like the change :
/lth/ > /nth/ or the use of 6ird with the dative case, are
geographically restricted, but important none the less as
marking off the dialects concerned from all the others. For
items like the morphology of the personal pronouns or the
local adverbs data from a widespread area are available.
Finally the actual material used in the survey has been
taken largely from the standard compilations of Bechtel,
Thumb and Buck. The epigraphic material itself has for the
purpose of this paper been consulted only where the three
authorities mentioned disagreed in their testimony or when
there seemed reason to doubt the interpretations of the data
which they offered (both of which contingencies, it must be
sadly observed, were remarkably frequent !).
It is likely that a thorough re-examination of the entire
corpus, dialect by dialect, would uncover not only a few

R. COLEMAN-THE DIALECT GEOGRAPHY OF ANCIENT QREECE

115

hitherto neglected data but more precise evidence for the


relative chronology and local distribution of many of the
phenomena concerned already but inexactly recorded.
I n all three respects the results could be improved, and
though it is doubtful whether the overall picture would be
radically altered, the details, which on so many points are of
major importance, would emerge more decisively. With these
important provisos we may proceed to a brief discussion of
some of the results themselves.
For each dialect there is a spectrum of isogrades,l which
corresponds to the intuitive feeling that many investigators
have had that dialect classification is not so much a matter of
either/or but of more/less. Within each spectrum we may
say that the dialects showing higher correlation are more
closely related t o the end dialect than those of lower
correlation. This relationship need not of course be genealogical, though in many cases it will probably be so. To confirm
the intimacy of the relationship we may employ the test of
reciprocality : viz. whether A stands in relation to B as B to A .
I n this way we can say that members of the pairs : HeracleanMessenian, Corinthian-Megarian, Theran-Coan, ArgiveAchaean, Locrian-Aetolian, Attic-Ionic and Arcadian-Cyprian
are closer to each other than those in the pairs: LesbianThessalian, Thessalian-Boeotian, Boeotian-Messenian, EleanMessenian, Cretan-Messenian,2 Pamphylian-Achaean, Rhodian-Megarian, Phocian-Aetolian.
Where the median coefficient is relatively low, the probability is that the dialect concerned has had a longer independent
existence than one with a high median. The Arcadian, Cyprian,
Lesbian, Attic and Ionic spectrums show medians under -40,
those of Heraclean, Messenian, Argive, Megarian, Corinthian
and Theran -67 or above. It is significant that all of this
1 The term is slightly adapted from D. W. Reed and J. L. Spicer,
Correlation methods of comparing idolects in a transitional area, Lang.

28 (1952), 348-60.
* The remarkable position of Messenian is no doubt due to its low
statistical yield, which gives positive occurrences largely on pan-DoriaQ
items.

116

TRANSACTIONS OF THE PHILOLOQICAL SOCIETY 1963

latter group are Dorian dialects and it is generally true that,


with the exception of Pamphylian and Cretan which have
medians of .45 and .52, all the unquestionably West Greek
dialects show a higher median than the rest.
Furthermore the same coefficient will clearly vary in significance as between one dialect spectrum and another. For
example Achaeans ~ 6 in
3 relation to Rhodinn represents the
median coefficient for the Rhodian spectrum, whereas Arcadians ~ 6 3in relation to Cyprian represents the highest
correlation in the Cyprian spectrum. Conversely Eleans 28
in the Cyprian spectrum is above the median for the latter
dialect, while Attics .28 in relation to Cretan is the lowest
coefficient of all in the Cretan spectrum.
Where the median is low, any particular high correlation is
especially significant for inter-relationship, especially if there
is a sharp drop to the next grade in the spectrum. Thus for
Ionic, with a median of .39, the Attic correlation of ~ 8 0is
especially striking, since the next in order are Linear B with
-63 and Lesbian with ~ 5 0 . Lesbian itself, with a median of
*37, shows a more concentrated spectrum, starting with
Thessalian whose correlation is -56, Linear B on .55, Ionic
on .50 and Coan on -49. Conversely where the median is
high, an especially low correlation is striking testimony to
divergence. Thus in the Dorian group enumerated above,
with medians of more than 67, it is noteworthy that Cyprian
and to a lesser degree Arcadian and Lesbian show very low
correlations.
Within the major groups we may distinguish kernel and
peripheral dialects. For instance within Dorian the top end
of the spectrums of both Rhodian and Cretan are occupied
exclusively by Dorian dialects: so we may label these two
as kernel Dorian. Corinthian and Megarian a t the other
extreme within this group show Locrian and Aetolian and to
a lesser extent Achaean and Phocian high up in their respective
spectrums, so that we may label these as peripheral Dorian.
Factorial analysis thus brings out the diversity within what is,
with relation to other dialects, a fairly homogeneous group.

R. COLEJIAN-THE DIALECT GEOGRAPHY OF ANCIENT GREECE

117

Where two dialects stand on the same isograde within a


spectrum we may proceed in one of several ways. For instance
)
Aetolian and Ionic both show the same correlation ( ~ 5 0with
Lesbian. By the test of reciprocality we observe that Lesbian
( -50) shows a higher correlation than Aetolian ( .45) with
Ionic ; so we conclude that Lesbian and Ionic are closer than
Aetolian and Ionic. In other cases reciprocality does not
entirely resolve the question. Thus Heraclean and Argive
both share the same isograde ( * 74) in relation to Rhodian :
but the difference in correlation between Argive and Rhodian
(*79 and -74 respectively) in relation to Heraclean is not
wholly decisive, given that there is a less concentrated scatter
around the Heraclean median ( 67) than around the Ionic
one ( -39). Here it would be necessary to examine the isoglosscomposition of the isograde. This serves to remind us that
factorial analysis like so many applications of statistical
methods to linguistic material blurs the relative value of
the different item. Where a large number of items are involved and the distribution is fairly clear-cut, this defect
is perhaps not important.1 But in instances where the
number of items is small and the median coefticient is low,
it is imperative t o examine carefully the individual isogloss
concerned. Conversely when we are concentrating on individual isoglosses, it is essential to see them against the overall
isograde picture. When two dialects show a high correlation
any given disagreement is correspondingly more important.
Where the correlation is low agreements similarly are
especially deserving of attention.
So far we have been discussing individual isogrades, i.e.
particular points in the spectrum at which one or more dialects
may stand, It is clear however that bundling patterns are no
less important. Thus Laconian, Heraclean, Messenian and
Elean form a fairly close bundle within a number of different
spectrums, and even more strikingly Corinthian, Megarian,
1 See P. IviO, ' The Structure of Dialectal Differentiation,' Linr~uistic
Essays (Word, 1962), 33-53, especially 48 and 50 for the difficulty of
assigning values to isoglosses in statistical studies.

118

TRANSACTIONS OF TRE PHILOLOGICAL SOCIETY 1983

Locrian and Aetolian. A good deal more exploration and


analysis of this kind of situation would almost certainly yield
profitablc results in terms of group relationships.
Finally a few observations concerning the position of
particular dialects in the area as a whole, bearing in mind
especially the four topics examined in terms of individual
isoglosses in I1 above.
(1) The orierttation of Achaean and Elean withi% W . Greek.
From the factorial analysis there appears no justification for
classing these two together. Achaeans nearest correlates are
Argive, Aetolian and Phocian, and this association is confirmed
by the test of reciprocality. We may therefore conclude that
Achaean is a bridge dialect between Dorian and N.W. Greek.
Whether it occupied this position from the start of its
independent history or acquired it by bidialectal contamination between the two major areas a t a later date remains an
open question.
In contrast Eleans highest correlates are all Dorian, and
indeed, as we have already noted, it belongs to a bundle
comprising also Messenian, Heraclean and Laconian, with
Phocian as the nearest of the N.W. Greek group. In fact the
pattern of the Elean spectrum differs little from those of
Messenian or Laconian and justifies the dialects attribution
to middle Dorian, inside the periphery defined by Corinthian
and Megarian.
(2) Boeotian, usually ClassiJied as an Aeolic dialect injiltrated
from West Greek. The closest correlates are the familiar Dorian
bundle of Messenian ( * 70), Elean ( 69), Laconian ( *61) and
Heraclean ( .60), followed by Thessalian ( -58). Lesbian ( -35)
appears below the median (a52) in the remoter half of the
spectrum. The test of reciprocality with reference to the four
Dorian dialects reveals Boeotian distinctly closer than
Thessalian or Lesbian, and both the order and the scattering
of the three Aeolic dialects here is typical of all the Dorian
1 Bechtel and Thumb-Kieckers place these two together at the end of
their survey of Dorian, in Schwyzers Exempla they are classed together
with N.W. Greek, and Buck places Elean firmly within N.W. Greek.

R. COLEMAN-THE

DIALECT GEOGRAPHY OF ANCIENT GREECE

119

spectrums. The noteworthy feature is that within West Greek


it is a Dorian group that shows greatest affinity with Boeotian.
The only spectrum in which the Aeolic trio bundle
together is Thessalian, whose closest correlates are Boeotian
(-58) and Lesbian (a56). Here the N.W. Greek dialects are
relatively high in the spectrum, with Phocian ( .51) standing
between Elean (.53) and Theran (.49), then Aetolian (-46)on
the same isograde as Heraclean and Messenian, with West
Locrian (-44) in the middle of a Dorian complex including
Laconian ( .45) a t the median and Achaean ( -44). The Lesbian
spectrum presents yet another pattern. Thessalian ( 56) is
followed by Linear B (.55), Aetolian and Ionic (.50), with
N.W. Greek and Dorian widely scattered therafter, from Coan
( 49) and Elean ( * 46) through to Megarian ( 29) and Achaean
(.26) in last place of all. The position of Boeotian here ( ~ 3 5 )
below the median (e37) offers a particularly striking contrast
to Thessalian.
Whereas the high degree of reciprocality between Thessalian
and Lesbian justifies the assumption that they shared in an
earlier unity (common Aeolic), the relationship of Boeotian to
this group is far from clear. If it is Aeolic, then the degree of
Dorian infiltration suggests a long period of bidialectal contact
and contamination with a Dorian group which in historical
times was geographically distinct from it. The alternative is to
regard Boeotian as a bridge dialect from the outset. Boeotians
position vis-bvis Aeolic and Dorian is clearly similar to that
observed in (1) for Achaean via-A-vis Dorian and N.W. Greek.
(3) The relationship of Lesbian to Ionic (cf. IIB above),
As we have just seen, Ionic (with Aetolian) has a relatively
high coefficient on the Lesbian spectrum and this is confirmed
for Ionic by the test of reciprocality, since the nearest correlates of Ionic are Attic ( -80),Linear B ( * 63) and then Lesbian
( *50). It is worth noting here that the Lesbian-Ionic relation
is much closer in both directions than the Lesbian-Attic one.
This could support the argument that Lesbian was for a
period in its prehistory influenced by Ionic, but it may point
to a different conclusion.

120

TRANSACTIONS OF T H E PRILOLOOICAL SOCIETY 1083

We have already seen that within a relatively homogeneous


group like Dorian there can still be considerable diversity,
ranging from kernel to periphery. And we have noted too the
possibility of bridge dialects lying between the major groups,
as perhaps in the case of Achaean and Boeotian. The complex
situation discernible in the historical period for many of the
dialects may well go back to the earliest times of Greek settlement, and the picture of prehistoric Greek as neatly divided
into discrete monolithic units, common Dorian, common
Aeolic, etc., would be as unreal as it is for later ages. Within
every more-or-less defined group (and it is worth remembering
that the further back in time we go the less marked the
divergences must have been) there were bound to be some
members more closely akin than others to members of outside
groups. The fact that Lesbian is closer than Thessalian to
Ionic or that Ionic is closer than Attic to Lesbian is perhaps
nothing more than an illustration of the natural diversity
within broader patterns of unity.
(4) Arcado-Cyprion, and lonic-Attic (cf. IIA above). Each
of these pairs exhibits a close-knit internal unity, but the
relations between the two pairs are complex. The Arcadian
spectrum shows the order: Cyprian (.63), Linear B (.57),
then after a clear gap Ionic (.47), Attic and Megarian (.44),
Thessalian and Cretan ( -43) with Lesbian several places
further away ( * 38). The order for Cyprian is Arcadian ( * 63),
Linear B (.66), then after a similar gap Thessalian (.43),
Lesbian (.41), Ionic ( ~ 3 9 and
)
Attic (-37). This last set of
figures justifies us in regarding Cyprian as kernel within a
loosely knit East Greek area, excluding West Greek but
including Aeolic. We may note further that for Cyprian the
Aeolic pair are closer than Attic-Ionic, whereas the order is
reversed for Arcadian ; also that within Ionic-Attic it is Ionic
and not Attic that shows the higher correlation with both
Arcadian and Cyprian. Moreover within Aeolic Thessalian
shows the higher correlation with both Arcadian and Cyprian
(.43 and ~ 4 3against Lesbian -38 and -41) but the order is
reversed in the Ionic-Attic spectrums, with Lesbian * 50 and

R . COLEMAN-THE DIALECT GEOGRAPHY OF ANCIENT GREECE

121

40 against Thessalian * 32 and 33. The test of reciprocality


shows that it is Ionic ( -50) rather than Attic ( .40) that has
the cIoser affinities with Lesbian (as with Arcadian and
Cyprian). The Lesbian spectrum shows for the four dialects
concerned the order Ionic (*50),Cyprian (.4l),Attic (.40),
Arcadian (.38)-all above the median of ~37-and the
Thessalian spectrum Arcadian and Cyprian ( 43), Attic ( 33),
Ionic ( 32)---all below the median of * 44. The Aeolic correlations
bring out very clearly the greater remoteness of Thessalian.
Indeed Thessalian poses a difficult problem. Its relatively
high position in the Arcadian and especially Cyprian
spectrums, where it stands appreciably higher than Lesbian
(though both are in the top half of the spectrum) is in marked
contrast to its very low position in the Ionic, Attic and
Linear B ones, where it stands (-32, .33, *28 respectively)
not only well below Lesbian (-50, -40, -55) but also below the
median in each spectrum (e39, -37, ~44).There is a good case
therefore for associating Thessalian with Arcado-Cyprian,
Lesbian with Ionic-Attic (cf. (3)). On the other hand within
the Thessalian spectrum the only East Greek dialect above the
median ( .44) is Lesbian ( * 56), whereas in the Lesbian spectrum
aZE the East Greek dialects stand above the median (a37).
This westward orientation of Thessalian is borne out by the
W. Greek spectrums, where (except for Coan and Aetolian)
it is always higher than Lesbian, and frequently (e.g. Elean,
Phocian, Melian, Heraclean) closer than any of the East Greek
dialects. It looks as if Thessalian must, like Boeotian, be
classed as a bridge between East and West, though, unlike
Boeotian, on the Eastern side of the bridge.
We may conclude that within the loose complex of East
Greek dialects there is considerable diversity, ranging from
the kernel Cyprian to the peripheral Lesbian and the bridge
Thessalian. Within this complex Ionic shows a close relationship not only with Attic but with Arcadian, Linear B and
Lesbian. It is relevant to (2) to observe that Boeotian stands
apart from the other two Aeolic dialects in all this, though
not so far apart as to divorce it entirely from them.

122

TRANSACTIONS OF THE PHILOLOGICAL SOCIETY 1963

(5) The position of Linear B. Its closest correlates are


Ionic (.63), Attic (.61), Arcadian (.57), Laconian and Cyprian
( *56), with Lesbian ( -55) very much closer than Thessalian
( -28). The attribution of Linear B to the complex of ArcadoCyprian and Attic-Ionic can be more precisely specified by
the test of reciprocality, since in all four spectrums it stands
between the two groups, and closer to Ionic than Arcadian.
We thus have a sequence for East Greek of Cyprian and
Arcadian, Linear B, Ionic-Attic, and more remote from the
kernel Lesbian and finally Thessalian. I n view of the wider
spacings of the higher correlates of all these spectrums there
is a case for subdividing East Greek into South and North
(Aeolic), which links up with the conclusions in (3) and (4)
above, and for assuming a longer period of independence for
all the East Greek dialects.
(6) Dorian contacts with Ionic-Attic (cf. IID). One of the
interesting features of the data examined in (4) and (5) has
been the intrusion of Dorian dialects among the East Greek
ones. Only Cyprian offers any exception to this. I n the
Arcadian spectrum Megarian shares an isograde with Attic
(-44)) Cretan with Thessalian (.43), and Argive ('41) and
Laconian ( e40) appear next in order. In the Linear B spectrum
Laconian and Cyprian share an isograde ( -56) and Lesbian
(.55) is followed by Rhodian (-53), Corinthian (.52), Argive
( 50) and Megarian ( * 45). In the Ionic spectrum Locrian and
Arcadian share an isograde ( 47) followed by Aetolian ( * 45)
and then a whole cluster of West Greek including Corinthian
(-41) and Megarian (m40). I n the Attic spectrum Linear B
('61) and Arcadian (-44) are separated by Locrian (a49) and
Aetolian (*48),while Corinthian (-42), Rhodian (-42) and
Megarian ( . 4 l ) all precede Lesbian (.40).
The recurrence in high positions of Megarian and Corinthian
generally, of Laconian and Argive in the Arcadian and Linear B
spectrums and of Locrian and Aetolian in the Ionic-Attic ones
is noteworthy. We have already seen that Megarian and
Corinthian are peripheral to Dorian, and the result just
obtained could well be evidence for the presence of these two
9

R. COLEMAN-THE DIALECT GEOGRAPHY OF ANCIENT GREECE

123

West Greek dialects 1 in the area of East Greek speech before


the Dorian invasions, either in a position remote from Cpyrian
or at a time when Cyprian was already detached by migration.
The correlations of Laconian and Argive with Linear B and
Arcadian could be the result of bidialectal contamination or
linguistic substrate, given the relative geographical positioning
of the four dialects within the historical period. Pinally the
Locrian and Aetolian correlations with Attic and to a slightly
lesser extent Ionic could be used to suggest a context for
interaction of the kind Risch proposed.
More probably however we should interpret this relation as
showing that Locrian and Aetolian were from the start more
closely related among West Greek dialects to the East Greek
group. The relationship itself is confirmed by the test of
reciprocality, since in the Locrian spectrum Attic (-49) and
Ionic ( * 47) are the closest of this East Greek group, and in the
Aetolian one the order is Lesbian ( .50), Attic ( .48), Thessalian
( * 46) and Ionic ( * 45). The fact that all these coefficients are
below the median (a63 for Locrian, .65 for Aetolian) indicates
that the relationship between the two N.W. Greek dialects
and Attic-Ionic belongs to remote antiquity. This would be
consistent with the assumption of geographical proximity
between these two dialect areas of West and East Greek for
sonic time in the Bronze Age, though we need not argue for
any diffusion between them.
(7)The position of Pamphylian. The placing of this
fascinating dialect has been very much disputed.2 The
distribution within its spectrum is exceptionally complex in
detail but the overall pattern is clear : all the closer correlates
are within West Greek, starting with Achaean ( .54), Aetolian
(*53),Heraclean, Cretan and Theran (-50). Only Boeotian
( e53) disturbs this sequence which continues otherwise
1 Here as elsewhere the labels Megarian, etc., are used in preference to
e.g. p o t o Megarian, etc., to refer to the earlier dialects to which they
correspond.
a Buck, Schwyzer and Thumb-Scherer assign it to the Arcado-Cyprian
group, Bechtel to S.E. Dorian, along with Rhodian and Cretan.

124

TRANSACTIONS OF TEE PHILOLOGICAL SOCIETY 1063

unbroken through the median -45, which is shared by


Laconian, Messenian and Corinthian, to Rhodian ( ~ 4 2 ) . The
complexity is well illustrated by the fact that two of the last
four positions are also occupied by West Greek dialects, Coan
( -31) and Argive ( *23)! We may assign Pamphylian therefore to West Greek. This result is confirmed by the East
Greek spectrums. In relation to Cyprian it is the highest of
the West Greek dialects, coming immediately after Boeotian
( -35). For all the others it is remote in the spectrum, even as
compared with other West Greek dialects.
On the other hand Pamphylian often stands very low in the
West Greek spectrums-particularly in relation to Argive,
Coan and Locrinn, though it is usually higher than any of the
East Greek dialects, and this is especially clear in relation to
Cretan, Achaean and Aetolian. I n fact the dialect with
which it shows the highest correlation is Boeotian, where it
stands in ninth place in the spectrum with a coefficient (-63)
just above the median ( .52). This seems significant in view of
the reciprocal correlation of Boeotian within the Pamphylian
spectrum, and suggests that Pamphylian must be regarded
with Boeotian as forming a bridge between East and West
Greek. The relatively high correlation with Achaean and
Aetolian, means that, unlike Boeotian, where the West Greek
link is with Dorian, Pamphylians links are with N.W. Greek.
Moreover as there is no East Greek correlation for Pamphylian
comparable t o the Achaean-Aetolian one, Pamphylian must
be placed towards the West Greek side of the bridge, in
contrast both to Boeotian, whose orientation is as much
towards Thessalian as towards Dorian, and to Thessalian,
whose orientation is to Lesbian and East Greek. The complexity of the spectrum and the low median both point to a
remote date of separation.
The above results, like the factorial analysis itself, represent
only n beginning. Some of them have confirmed earlier
intuitions or results arrived a t by quite different procedures,
and in so doing they can claim to have provided a quantitative
dimension to these arguments. I n addition certain new

R . COLEMAN-THE

PHILO. TRANS.

125

DIALECT GEOGRAPHY OF ANCIENT GREECE

1963.

126

TRANSACTIONS O F THE PHILOLOGICAL SOCIETY 1063

orientations and the possibility of new patterns of relationship


have been thrown up by the analysis. Much remains to be
done. From an improvement in the actual analysis of the data
on the lines proposed earlier (pp. 114-5) and the application
of more elaborate modes of inference, both historical and
descriptive, we may look forward to a number of interesting
new discoveries in the field of Greek dialect ge0graphy.l
The author wishes to acknowlcdge the help given him by Professor W. S.
Allen, who made a number of criticisms and suggestions on points of detail
discussed in this paper.

Das könnte Ihnen auch gefallen