Ancient Greek Dialectology

in the Light of Myccnacan*

Warren C. Cowgill
YALE UNI V EIIS ITY

1. I n the first half of this century many (probably most) experts be-

lieved that therc werc four principal dialects of Ancient Greek (leaving
aside the difficult and litLie understood Pamphylian): West Greek (including Doric and Northwest Greek), Aeolic (Asiatic Aeolie, Thessalian,
Boeotian), Arcado-Cypriall, and Attic-Ionic. It was also believed that
the latter three constituted something of a unit (East Greek) against the
first, with Areado-Cyprian intermediate between Acolie and Attie-Ionie.1
A substantial minority classed Arcndo-Cyprian and Acolie together as
form s of a single dialect ("Achaean" or the like), resulting in a threefold
classification : West Greek, Achaean, Attic-lonic.1 When features considen!d typical of one dialcct were found ill local varieties of another,
this was generaliy taken as evidence for prehistoric tribal migrations and
overiayering:s ; for instance, "A colic" dative plurals ill - EllUL were taken
to indicate an Acolic substratum in Aetolia, Corinth, and so forth.
2. Closely related to problems of G reek dialectology is the problem of
the prehistory of the Homeric epic. From the time of r ick linguists had
been agreed that under the Old Ionic component of Homer's language
there is an Acolic layer, implying that Ionians learned to compose epic
from Aeolian poets, and no doubt took over much of t heir subjcct matter
from them as wcll. But. perceptive scholars li ke Antoine Meilletl and
Milman Pa rry4 had postulated a still earlier "Achaean" layer, reflecting
the language of poets of the Mycelmeall age; traces of such a layer seemed
clearest in the words peculiar to Homer and Arcadian or (especially)
Cyprian. But in HJ50 Manu uumann's book Homerischc W arier appeared, expressing strong skepticism toward this view; Leumann sug I wish t o exp ress here my gratitude to E. L. Bennetlll.nd his Nes/or, whie h
has immelUmrably s implified the task of collecting mflterial for this repor t . That I
have neve rtheless left many gaps T am painfully aware.
I So, for example, C . D. Buck , The Greek dia/ects' 7- 9 (Chicago, 1D55) .
So, for example, O. Hoffmann, Dle griechisehen Dialekle Lv ii (GOttingen, 1891).
AperiU dune hiMo;re de 10 longue grccque' liS (Paria, 1930).
Ha n'a r d Slu(Ncs in Classical Philology 43.1- 50 (1932).
77

78

ANCIENT (lREEK DIALECTOLOGY IN TilE L I(lHT

01<'

MYCENAEAN

gested that many of the Homeric words in Arcadian and Cyprian were in
fact borrowed by thc dialeds from the language of epic poctry.
3. No one, it seems, has recently questioned the validity of dialect
groups that can be labeled West Greek, Acolie, Arcado-Cyprian, and
Attic-Ionic. Discussions of Greek diaJcctology in recent yenrs have rather
ccntcred on the various possibilities of grouping (:ombinations of these
fou r in to larger entities, and on the question of whether Mycenaean
constitutes a fi ft h main gronp or belongs wit.h one or another of the already
established groups. .i\Jodem techniques of dialect geography have been
wed to provide approximate dates for innovations, and the realization
that innovations can spread aeross existing dialect boundaries has led to
soberer views of prehistoric migrations. Mycenaean evidence has provided important support for a Mycenaean (Achaean) element in the
Homeric language; indeed, the question nOw is whether an Aeolic component is needed between Achaean a nd Ionic.
4. Even before the decipherment of Mycenaean the traditional picture
was beginning to change. I n an article written ill ]9015 but not published
until 1954 ,~ Walter Porzig emphasized the agrccments between AWcJonic and Arcado-Cyprian, suggesting that these were "verschiedenarUge
Entwicklungen desselben alten Dialekts" (p. 157), which was once
spokcn in the Pcloponncse, Attica, and Boeotia (p. 164). Agreements between Arcado-Cyprian and Acolic were the result of contact, not of
original identity, as Hoffmann thought (p. 163). Acolic features in the
Peloponncsc and Crete indicated an Aeolic migration from Thcssaiy,
and it was these Acolians who created the l\'[ycenacan culture (p. 166).
5. I n 1949 Ernst Risch& used principles of dialect geography to show
that when isoglosses do not agree with tribal boundaries (e.g., the treat
ment of -ns) , the most likely inference is that thc inllovations ill question
are relatively latc, rathcr than that tribal mixture and o\'eriaying has
occurred.
G. When in 1953 Ventris and Chadwick first published the decipherment of Mycenaean/ it was immediately clear that the new dialect did
not belong with West Greek- cf. most strikingly the development of ti
to si in worus like dQS08t"i 'they will give'. But what was its relationship
to the East Greek group? Was it part of an undifferentiated ancestor to
all of them? Was it ancestral or elosely related to the ancestor of ArcadoI "Spraehgeographisehe Untersuchungen zu den altgriechise hen D ialektcn,"
IF 61.147--{lfl .
M!48. He/v . 6.9-28.
, " Ev idence for Greek dialect in the ~Iyee nnean a rchivcs," J ll S 73.84- 103.
I Mycenaean forms in italics nre direct transli te ra tions of the syllabic or iginal;
those in Greek icttel1l are phone t ic interp retations.

ANCIENT GREEK DIALECTOLOGY IN THE LIGHT OF MYCENAEAN

79

Cyprian, Aeolic, or the two combined? Was it part of a d ialect area

ancestral to both Arcado-Cyprian and AWc-Ionic to the exclusion of
Aeolic~ Or was it an ahen'ant dialect, with no dircct or near-dircct
descendants attested in the first millennium? The answer to t.hese questions was of course intimately tied up with the way the posL-Mycenaean
dialects were grouped together.
7. In "Evidence," Ventris and Chadwick noted (pp. 101-103) agreements of l\fycellaean with Acolic and Arcado-Cyprian, especially the
preposition tllrV (vs. Attic-Ionic and West Greek alTo) and the development (not universal) of 0 from or next to syllabic resonants, as in qetoro'four-' pemo (beside penta) 'seed'. The Mycenaean preposition po<fi recurs
in exactly this form nowhere else, but is clearly closer to Arcado-Cyprian
'lT0~ (which can be derived from it by apocope) than to the forms of any
other dialect (Attic-Jonic and Lesbian 'IT,m, Doric and Homeric 'lTpoTi,
Thessalian, Boeotian, West Greek, and Homeric lTOTI). I n 1953 Ventris
and Chadwick accepted the view that Arcado-Cyprian and Aeolic had a
common ancestor; they thought Mycenaean should be this ancestor ,
and, without going into details, suggested that the suppost.'f.i Aeolisms of
Homer were in fact Mycenaean.
8. But the next year appeared Porzig's article mcn tioned in 4 above l
setting forth reasons for grouping Arcado-Cypriall more closely with
Attic-Ionic than with Acolic,1 and then Risch's even more influcntial
article, "Die Glicderung der griechischen Dialekte in neuer Sicht," IIfus.
Helv. 12.61- 76 (1955). Risch tries to seL up a relative chronology for the
main Greek isoglosses and finds that some of the oldest link Attic-Ionic
with Arcado-Cyprian against West Greek and Aeolic. Thus (p. 64) the
former have athematic infinitives in -(E)pat (type oOfEI'a" OOVP(1'), while the
lattcr have -JL~I'(1') (type OOJ.'.EI', lib;uI'(1t). At.tic-Ionic and Arcado-Cyprian
agree in having "., from Ii and". (rather than TT or ".".) from *l(h )y in
words like EXOP"." ~Xotp". " and TWOi ; the latter phenomenon Risch convincingly explains by a dialectal assibilation of *t(h)y prior to the general Greek all'ricatioll of stop plus *v (pp. 66-(7), following Pedersen,
Festschrift lVackernagel 1l 5 (Gi:ittingen, 1923).
9. In fact, lUsch finds no early isoglosses separating Attic-Ionic from
Arcado-Cyprian or, tiS far as the evidence goes, Mycenaean. He infers
therefore for IHycenacan times a fairly uniform South Greek dialect, of
which the isolated Arcadian preserves more features than do the progressive a nd commercially active cities of Ionia (p. 70).11
f I think no one h9.s c9.Ilcd it an ancestor of Attic-Ionic to the exclusion of
Areado-Cy ptian.
10 Cf. also P . Chantraine, Rev. de phil. 28.264- 265 (1954).
11 So already Mus . lie/v. 6.25 .

80

AKCll>NT GREEK DIALECTOLOGY IN THE LIGHT OF MYCENAEAN

10. As for Arolic, Risch thinks this was originally closer to West
Greek than is generally supposed. In particular, the 1't of Boeotian and
Thessalian is not borrowed from West Greck, but is native Arolic; it is
rather the Lesbian O't that is a borrowing, from Ionic (p. 7l).lt Indeed,
Hisch sees no sure example of a difference between Arolic and West Greek
before 1200, and so (ibid.) arrives at an original dichotomy betwccn
North Greek (Aeolic and West Greek) and South Greek (Attic-Ionic,
Arcado-Cyprian, and .Mycenaean).
11. Excellent as Risch's study is, there are certain weaknesses, and
some of these were pointed out immediately by Ruiperez, Mino83. 166167 (1955) . The most important is the treatment of *r. Risch had left
this isogloss out of account, claiming (p. i2) that the data were too complex to be usable for determining eariy dialect boundaries. But Ruiperez
observes that the development to op/ po had already occurred in Mycenaean-as in qetoro fro m *k"'etwr-. This agrees with ATcado-Cyprian
and Aeolic against Attic-Ionic and West Greek, which have ap/ pu..
Neither treatment can be explained as a later development of the other:
eL, for example, Cypr. ap"Y!.Ipov and Attic op4>avor; with original *ar and
*or- preserved. The conclusion seems inescapable that already in the
second millennium the ancestors of Arcado-Cyprian and Attic-Ionic
differed in this feature, and that Mycenaean agreed. with ArcadoCyprian. Ruipercz, it should be noted, docs not dispute the essential
unity of the ancestors of Arcado-Cyprian and Attic-Tonic; he merely observes that we can recover at least one of the isoglosses running through
this area.
12. Ruiperez's other points are less cogent.. If he is right in seeing the
middle ending -TN of ArcadoCyprian and Mycenaean as an archaism
agai nst the -Ta~ of most other diaiccts,13 then, as he admits, it is possible

.0.

l~ Boeotia n and T hessalian third plurals ill

(e.g. , Boeot. ~")"'o0<8" The~ .
[Larissa) ~a .. o<ulo~.O.) support Risch's view. These clearly canno t be borrowed
from West Creek, which ha~ only '.",, and it is hardly thinkable that a WGk .
. .... would have been borrowed by continental Aeolie in time to undergo analogie
change to . ..0. before the ancestoi'8 of Boeotian and T hessalill.ll were effectively
split from euch other by the very invaders whose presence is presupposed by the
pututive borrowi ng. Homeric ",(p)oT! points in the same direc t ion; cr. Chantraine,
Grammaire hom~riq!J.e P.511 (Puris, 1957).
II Em erita 2O.S-31 (1952). Cyprian '''0< is attested in keitoi, the co rrect reading
in Schwyze r 683.6, on which see now O. Masson, Les inscriptions chypriotes sylla
bique8, no. 11 (Paris, 19(1). No te al~o the b t ago keimai, Masson 213a. This form
favors Rllipl:rez' view of the chronology of -..a, and -TO', for it is not likely that an
original uniform set -"'" -(... )", -..", -.1'''' would have been remodeled to a nonuni ,
form '>la, -("')0. '1'0< -.1'0', even with help from the nonuniform seco ndary endings

>lo.v '( ... )0 ' TO ' .1'0.

ANCIENT GREEK DIALECTOLOGY IN THE LIGHT OF MYCENAEAN

81

(but not likely) that the spread of -TClI was post-Mycenaean.'~ Ruiperez
thinks that the first millennium contrast betwccn Arcadian (Cyprian evi dence on this point is ambiguous) and Attic-Ionic in the t reatmcnt of
compensatorily lengthened E and 0 must already have existed in the
second millennium_ But I fail to sec why the Arcadian falling together
with old I) and w cannot be a post-lHycenaean inllovation, not shared by
Attic-Ionic. Perhaps I am missing something obvious here; if so, I would
welcome enlightenment fro m my fellow conferees.
13. Meanwhile, Pisani,16 who believes that Greek is an amalgam of the
speech of different groups entering Grecce from diffcrent directions,
proposed that there were originally four such groups: Dorians, Aoolians,
I onians, and Mycenaeans (p_ 10)_ Like Porzig, he assumes a pre-Doric
Aeolian invasion of the South : from the mixture of Aeolian and Mycenaean
clements come thc latcr Arcadian and Cyprian_ To be sure, he docs not
list any innovations shared by Aeolic and Arcado-Cyprian and lacking
in Mycenaean, nor have I come across any certain ones; if Risch is right
in taking Myc. tereja and lerejae as coming from HAd!! and TEAE!W' (sec
29), and if, as seems likely to me from the viewpoint of general IndoEuropean (cf. Lg. 35.5 [1959]), Greek contract presents were originally
inflected only thematically, then the athematic inflection of contract
presents would be such an innovation found in Arcado-Cyprian lind
AeoUc but lacking (or at least so far unattested) in Mycenaean_ Of more
value is Pisani's observation that the Arcado-Cyprian words that Leumann had suspected to be loans from Homer and which now t urn up in
Mycenaean are more likely to be inherited dialect forms (pp. 14-15).
14. Ventris and Chadwick's Documents in Mycenaean Greek (I956)
lists the same agreemcnts of Myccnaean with Arcado-Cyprian and Aeolic
against Attic-Ionic as "Evidcnce" did: thc prcposition chu, and 0 from
syllabic resonants (p. 74 ). To this is added the apparent athematic conjugation of tercja TEAEla 'pays, performs (?)'. With Aeolic, Mycenaean
shares adjectives of material in -HOi and -Wi, and patronymic adjectives.
,. But I am beginning to doubt that lst ag. -~, a lone wou ld have bee n able to
i nfluence n"alogiclllly 2n d ego _(")"'" 3rd "g. _ro., and 3rd pI. _~r ", without help
from some as yet unconsidered sou rce. Also, on theoretical grounds I would expect
a Proto-Indo-Europcan lilt ag. middle " -A-o-!! (not - -A -e-!!), which as far 9.l! I
now understand the phonology should gi ve Greek . -(0) ... , not -~ .... On the other
hand, influence of primary _'" hy seconda ry _0 is always possible. It might be then
that -7", for expected -"01 is a feature (whose explanation is still to be found)
of P roto -Greek, in which event the agreement of l\I ycenaean and Arcadi an in
-70< would be a significant innovation.
If "Die EntzifTerung der iigeischen Linear D Schrift und die griechisehen
D ialekte," RhM 98.1- 18 (1955).

82

ANCIENT GREEK DIALECTOLOGY IN THE LIGHT OF MYCENAEAN

With Arcado~Cypria n it shares several vocabulary items and (on the

testimony of ole P Y T o. 711 ) t he formation of temporal adverbs in -TE
(this also Attic-Ionic). Risch's and Porzig's vicw (4, 10) that the regular Acolie development of n Wtl.l:! n lcads Vent.ris and Chadwick to consider the affinities with Areado-Cyprian and Attic-Ionic more significant
than those with Acolic {po 75}. A more cogent rcason for this decision
was noted by Chantrainc in his review of Documen18, Rev. de Phil. 31.239246 (1957) : of the two l\'fyccnacan-Aeolic agreements listed by Ventris
and Chadwick, the patronymic adjectives can be archaisms, and the
adjectives of material in -1.0~ independ ent innovations (p. 241). In Grammaire homirique 1 ~.578 II. 1 he modifies this last to suggest, wit.h a referellce to Schwyzer, Gricch. Gramm. l AGan. 1, t hat the -/.Os adject.ivC5 are
also a shared inheritance.
15. At the first Mycenaean Colloquium (April, 1956) R isch reported
011 "Ia position du dialecte mycenien" (Etudes myceniennes107- 172, 249258). Although he does not insist that the a ncestors of AtticIonic a nd
Arcado-Cyprian were absolutely identical in t he second millennium
("tres proches" [p. 170J, "un groupe dialectal sensiblement un" [po 253],
" un groupe dialectal peu difMrcncie" [po 258]), his report minimizes their
differences. He still considers the developmcnt of .[ unclear, but thinks
t hat t he development to op/ po is probably an archaism and hence not
decisive for grouping .Mycenaean with Arcado-Cyprian (p. 171). Myc.
pvsi hc thinks is possibly 1fopui, melathesized (like Cret. 1fOpr( fro m
"-POT') from the 1rpouL ancestral to Atticl onic 71'pbs (ibid. and p. 256) .
'ATv 500ms an archaism (p. 172); in general, where Attic-Ionic and
ArcadoCyprian disagree, 1\fycenaean agrees with t he more archaic. All
features shared by Mycenaean with both Arcado-Cyprian and Acolic
seem archaisms, but there are innovations t hat Mycenaean shares wit h
Attic-Ionic and Arcado-Cyprian (p. 25i). Mycenaean agrees wit h AtticI onic against Arcndo-Cyprian only in archaisms {p. 258}.IS
16. As was stated above in II , I do not believe that Riseh ca n be right
in considering op fro m r an archaism, later replaced in some dialects by
ap; at most olle call say that the contrast of op and ap is not very im
portant for grouping Greek dialects. Similarly his t.rcatment of posi looks
rather desperate, and has not generally been accepted. Regarding this
word and the pair hlJ : .b6, it seems likely that Greek inherited both
proti (cf. Skt. prati) and poti (cf. I ranian pati), both apo (d. Skt. apa)
.. I find a problem he re in the idea of post-Mycenaean inllovations shared by
Arcadia.n and Cyprian. Ir we do not all ow for parallcl drift, presumably anything
shared by these two would have to have come into exis ten ce before the Dorian
conquest of the Peloponncsian coaatline, which does not Beem to leave very much
time fo r common innovation after the latest Pylian texts.

ANCIENT GREEK DIALECTOLOGY IN THE LIGHT OF MYCENAEAN

83

and *apu (cL Schwyzcr-Dcbrunner, Griech. Gramm. 2.444). Each dialect

has chosen one membcr of cach pair, and it seems rcasonable to say that
agreement in choice is in itself nO reason to group two dialects togcthcr,I7
but disagrecment is a rcason to separate them. ls 1\'[ yc. pofd and apu thus
show that Mycenaean differed from the ancestor of AUie-Ionic,Jg but
only the principle of Occam's razor impels us to see in these prepositions
evidence for close relation between Mycenaean and the ancestor of
Arcado-Cyprian.
17. I n the discussion a t the CoUoquium following Bisch's presentation,
Bcnveniste objected that Hisch's scheme rcquired a surprisingly rapid
evolution of Attic-I onic ill the centuries immediately following 1200;
would it not be more likely that characteristic Attic-Ionic innovations
had begun already in Myceancan times? (D. myc. 203.)
18. At the same Colloquium, Ruiperez (pp. 118-119) and Lejeunc
(p. 2(1) suggested that the coexistence at Pylos of perno and (in the work
of one scribe) perna for *spermt} indicated that speakers of dialects
ancestral to both Attic-Ionic and Arcado-Cyprian werc prescnt at Pylos
arowld 1200.
19. Gcorgiev, who was also present at the Colloquium, maintained a
quite diffcrent vicw,20 namely that IHycenaean spelling renccts pronunciation fairly a ccurately (D . m.yc. 173- 188, especially 175-183). If
this were so, Mycenaean would have undergone quite radical sound
changes that would make it a separate dead-end branch of Greek, not
ancestral or even nearly ancestral to any of the later recorded forms of
Greek. But as Risch pointed out (ibid. 252), and as most other scholars
agree, so aberrant a dialect is much less likely than is a grossly defective
writing system.21
1T Presumably the chllnges of :my two di alects making the same choice would
be fifty-fifty.
" l am grateful to T. Dyen fur helpillg me (0 clurify my thinking on this point,
although r am not sure that the view presented here would meet with his upproval.
Thllt agreement in choice of "ingle feutures is no renSon for grouping together becomes clearer when we consider whether or not it is likely thnt Mycenaean and
Iranian form !l. lIubgroup of Indo-Europcun ugainst Atti c and Indic.
" Unless they coexisted with as yet unattested poro8i and upo; d. the case of
I'u;, and "...10., both of which seem inherited, lind both of which are possibly at
tested in Mycenaean .
~ Proposed aillo by Merlingcn and Lur'je.
"Compllriwll wilh BOrne contemporary writing systems is iliumiuating.
Harlll noo (a Philippine language) has, as described by H. C. Conklin, lfanunooEngli8h Vocabu lary Q-10 (Berkeley, 1953), syllables of the type CV and CVC, but
the writ.ing system has only CV, 60 that syllable-final COHSOllants are regularly
left unnoted. Further, the language hilS separate phonemes rand 1, but the writing
does IlOt distinguish them before u. F. Loullijbury info rms me that Canadia.n

84

ANCIENT GREEK DIALECTOLOGY IN THE LIGH T OF MYCENAEAN

ZO. Hugo Mlihleslein conlributed to the Colloquium a paper emphasizing the agreements of l\"lycenaean with Aeolic (111. myc. 93- 97). The
one exclusively shared innovation he found between the two was a
tendency to reduce -CiV- and -CeV- to -(C)CV-, as exemplified by Myc.
kuruso 'golden' for *XPOOtOi, suza 'fig tree' for (1UKtll . Aeol. ap'Yuppov 'of
silver' (p. 96). To explain the treatments of TL a nd *Iy, where :Mycenucan
appears to share a non-Acolie innovation with Arcado-Cyprian and
Attic-Ionic, Mlihlestcin returned to the old view that Lesbian (1t, not
Thessaliun and Boeotian n, is the native Aeolic development of n;
and for *ty he took Thcssalian and Lesbian T/xJ(JO~, and so on, to reflect
the same early assibilatioll as l\"Iyccnaean, Attic-Ionic, and ArcadoCyprian, but without the specifically Arcado-Cyprian and Attic-Ionic
shortening of -uu- to -u-. This is ingenious, but fails to account for the
-TT- of Boeotian ill forllls likc chrOTTIl.t:! Milhlestein docs not coillment ou
the contrast between Mye. ole and usb. ISnt, Boeot. 1fOKo..
21. In "Achiiisch, Jonisch und Mykenisch," IF 62.240 (1956), the views
of Porzig mentioned above in 4 were discussed and criticized by F.
Hodriguez Adrados, who showed that Areado-Cyprian agreements with
Attic-Ionic do not. always affect. the whole area en bloc, as Porzig had
asserted (e.g., Attic- Ionic and Arcadian share a.~, but Cyprian has >;t),
while those between Arcado-Cyprian and Aeolic sometimes do (e.g., t in
aorist and future of tw-vcrbs). This suggests for Arcado-Cypriall (of
which Adrados believes i\'Iycenacan is all archaic phase) a position betwccn Attic-Jonic and Acolic. Simlarly the West Greek features of
Boeotian and Thessalian rest not so much on late borrowings as on early
proximity. Adrados thus arranges the Greek dialects into a chain of five
members (p. 245): Doric, Bocotian-Thessalian, Aeolic (i.e., the language
of Lesbos and the neighboring coast), Arcado-Cyprian, Attic-Ionic.
22. I n 1057 Antonio Tovar, "Nochillais Ionier lind Achaeer illl Lichte
Eskimos commonly use the Cree syllabary to write Eskimo, resulting in the same
n(!glect of syllable-final consonants; and the Cherokee syllabary, which we know
was invented by a Cherokee for writing Cherokee, genemlly ignores not only
le ngth and accent but also h and glottal stop everywhere except in syllable initial,
so that each sign typicnlly hns half II. dozen values. In II. fe w instancea finer distinctions are indicated. Three of these involve signs distinguishing di, de, da from
syllables with h-clusters Ihi, the, IIw, which is perhaps of some interest for the
special d-seri es of Linear B.
OJ Two possibilities seem open to rescue M(lhlestein's view . Either Boeotinn
. .,..,.- forms were borrowed from West Greek before the affricate resulting from -k y
and (restored) t ty had become West Greek -<><T- , Boeotian -H -; or, better, - ty in
East Greek first becam(! -t~, lind this tllen developed lik(! illh(! rited - t.t to give
T heSSlIlian and Lesbian -<T<T- (li ke w~.b<U!"n), Attic-Ionic and Arcadian -</ - (like
wu~ .. <> .. ), but Boeotian _ TT_ (like ..".. ,;,IT.,....,.o). r..Iiihlesteill 's view that -ty flrst
became sy (p. 94) seelIl5 highly unlikely.

ANC IENT GREEK DIALECTOLOGY IN TilE LIGH T OF MYCENAEA N

85

der Lincar-B-Tafcln," ?!I NHM H~ XAPIN 2.188- 193, noted that Mycenacan
caused ~riOU8 trouble for Krclschmer's view that the Greeks came to
Greece in three waves, first Ionians, then Achaeans, and finally Dorians.
Accepting this three-wave view and the obvious agreements bctween
Mycenaean and Arcado-Cyprian, Tovar assigned lldycenaean to the
second wave, which leads to the paradoxical situation that the earliest
attested Greek does not belong to the first wave of invaders. Instead of
abandoning or modifying Kretschmer's scheme, Tovar is concerned with
showing that a wellcharacterized Ionic existed already in the second
millennium, and so is left with the problem of explaining what the 10nians
were doing while their Achaean cousins were ruling the Peloponnese and
Crete.
23. By this time 1dyccnaean studies were well enough established for
Mycenaean data to be WlOO in working on problems of Homcric language.
I n L'elbnent achten dans la langue bpique (Assen, 1957) C. J. lluijgh defended against Leumanll the presence of an Achaean (iVrycenaean)
layer in the epic language, using as part of his evidence lexical agreements
between Mycenaean, Homer, anh Arcado-Cyprian . He put (p. 13) l\"iycenaean and Arcado-Cyprian together as n. separate main dialect of
Greek, intermediate between Aeolie and Attic-Ionic- essentially the
pre-lOW majority view mentioned in 1.
24. ru proof that Achaean was different fro m Attic-Ionic, Ruijgh used
the by-now familiar difference ill the treatment of *r. To show that it
was diffcren L from Aeolic he used its lack of dative plurals in -tuu~
(pp. 14- 17). '1'0 be sure, his rather involved proof that the Acolic dative
plurals in - E<1<1~ existed alrcady before the Dorian expansion has not met
with universal acceptance--wit.ncss, for example, Szemerenyi. JHS
79.192 (1959), Nuchelmans, Mnemosyne 4:13. 158 (1900). I think Ruijgh
is probably right in his dating, but only because it is unlikely that features shared by Lesbian and West Aeolic would date from much aftcr
the West Greek expansion (the agreement between Boeotia n and Thessalian is less cogent, since t<1<1t is shared by the inten'ening Phocian and
East Locrian). But I do not think he has succeeded in proving that the innovation has to be that old. I agree (along with Wackernngel, Schwyzer,
etc.) that the creation of -EUUt seems to require the existcnce of -OtU, in
the o -stems,~ and hence must antedate the genemlization of -OtS in
Boeotia n and Thessalian . But locative - OIUt (replacing -Ott, the probable
reading of Myc. -oi) a nd instrumental -O IS (Myc. -0) are both inherited,
and Mycenaean still differentiates the two in usage, so that the gcnernlization of -01$ in Thessalian and Boeotian Illay not have occurred until
long after the Dorian migrations; indeed, Lesbian - OIUI shows that thc
II This would probably be enough, without positillg -.,,'" ill a-stCIIIS.

86

ANCIENT GREEK DIALECTOLOGY l.N' TUE L IGHT OF MYCENAEAN

longer form was still flourishing when Aeolian colonis~s crossed the
Aegean, and lasted there down to t he relatively late time when the
change of *-ns to -,~ made -O "1't indispensable to keep dative and accusative apart. It follows that the only secure terminus ante quem for t he
creation of -E\1(H is the pGint at which clTective communication between
Asiatic and Western Aeolic ceased . Better proofs that Mycenaean is not
ancestral to Aeolic are the treatment of n and the adverb o/e. 24
25. Whatever one may think of the details, the existence of an
"Achaean" layer in Homer's language seems by now scarcely disputable.
But the attempt of Klaus Strunk to do away with an Acolic layer altogether, in Die sogenannten Aeolismen der homenschen Sprache (1957), is,
in my opinion, going too far, and has not convinced many linguists; cf.
the criticisms of Hamp, Glolla 38.194- 198 (1960), and Ruijgh, Mnenwsyne
4;14.213- 215 (1961).
26. In his 1957 review of Documents (Rev. de phil. 31), Chantraine
called attention to the seemingly inconsistent treatment of contract
verbs in Mycenaean (p. 240). Oil the olle hand, toroqejmru:no T/101I"fOjJfVOS
(py Eq 213. 1) is clearly thematic, as in Attic-Ionic; but tereja (PY Eh
940. 1, te[.]ja 495.1) seems to belong to an otherwise unknown *n;\.wl.",
with athematic inflection, as in Arcado-Cyprian and Aeolic. However,
the interprctation of neither form is established beyond doubt (for tereja
sec below 29), and even if we grant that both arc correctly understood,
thematic inflection of iw-presents is certainly inherited, and athematic
inflection of 6...,-prcsents may possibly be so (but cf. 13), so that both
forms may be archaisms, and .M ycenaean would preserve an earlier
stage from which Attic-Ionic and Arcado-Cyprian had diverged in
opposite directions. In the third printing of the first volume of his Grammaire homi:rique, Chantrainc has a new conclusion (pp. 495-513) dating
from 1957, in which he presents a good discussion of the prehistoric dialect situution, agreeing substantially with Risch except for not linking
Acolic with W cst Greek against the other dialect.s .
27. In 1958 the view that ~fycenaean was most closely related to
Aeolic (cf. 20) received another champion in Carlo Gallavotti, "n
carattere eolico del greco miceneo," Rivista di filologia 86.113- 133. GalJavott i bases his case partly on the apparent development of labiovelars
to labials before e in words like pereqota., opepa beside qereqotao, oqcqa, but
mainly on an interpretation of TU2 and T07. as ppo./ ;\.;\.a and ppo/;\.Xo, with a
specifically Acolic development of prehistoric consonant clusters (pp.
I< Ruijgh's article of 1958, "Les dati fs pluriels da ns les dialeetes grees et Ill.
position du mye~nien," Mn em08!1ne 4; 11.97- ll6, makes some useful observations,
but does not sufficiently reekon with the possibility that _au and -0''''' both existed
in the Aeolic of the second millennium.

ANCIENT GRJ::EK DIALECTOLOGY IN THE LI GHT OF MYCENAEAN

87

118-128). The development of 1"t to tTl, lacking in Boeotian and Thessalian, he takes to be an innovation of the Mycenaean variety of Aeolie
(p. 131).

28. Gallavotti's arguments havo not been generally accepted. His

view of Til:! and rO! is by no means the only possible OIle, and already at
t he second Mycenaean Colloquium (Pavia, September, 1958) Chadwick
remarked (Athenaeum NS 36.303) that Mycenaean words in which pc
alternated wit h qc regularly contained another q. Later HcubcckZ5 developed this into a regular sound law for one strain2e of Mycenaean:
labiovelar . . . labiovelar is dissimilated to labial . .. labiovelar (p. 255).
Heubeck expressly rejects Gallavotti's connection of this phenomenon
with the Aeolic change of $k-c to n whether another labiovelar was
present or not (p. 261).
29. At the same Colloquium Risch proposed to explain Mycenaean
tereja by a sound law that *eije beca me dja in Mycenaean (p. 311 ; more
fully MU8. Helv. 16.226 (1959]). 'l'ereia would thus be for nA~lH, t he
regular and well attested denominati ve to the noun stem nA~u-, and not a
derivationally monstrous and inflectionally somewhat dubious nAda.
Similarly its infinitive lerejae (PY E b 940, Ep 617.1, 4, fragmentary Eb
495) will be a regular thematic $TfAf[W', not a morphologically dubious
athematic infinitive in -tv. The Arcado-Cyprian athematic inflection of
contract verbs can then be altoget.her a post-!\'fycenaean innovation, not
shared by Attic-Ionic. The difficulty, of course, is that Risch attributes
to Mycenaean a sound change somewhat unlikely in itself and not shared
by any later dialect, so that we would have to see here a development of
the type envisioned by Oeorgiev ( 19), with its attendant difficulties.
30. This was not the only specific Myccnacan innovation that Risch
was fi nding. In BSL 53.96- 102 (1957- 1958) he pointed to apparent accusative plurals in -u, an innovation not foun d in Arcado-Cyprian or
Attic-Ionic, and so probably unrelated to similar forms appearing later
in West Greek, Lesbian, and Koine. v I n Anlhropo8 53.160 n. 40 (1958) he
mentioned Mycenaean instances of 0 from . 1] and .~ next to labilals, for
example, anowoio 'al'{lvaTO~', enewopeza 't""fn1l"~~a ' , perno (beside perna)
'udp}.Ia', kowo (PY Un 718.4) ' "Wa~'. Since other Greek dialects seem to
have 0 from $1] and 711 only US the resnlt of anthology (cf. 36), :rd ycenaean would here have another unshared innovation . It would follow ,
roo, that the coexistence of perno and pema at Pylos would indicate not
the presence of an Arcado-Cyprianlike dialect opposed to an AtticOf "Myk. pe-re-qo-1w," IP 65.252-262 (1960) .
.. I have t aken thill term from Stimson, Lg. 38.378 (1962) .
., Bnt Scherer, Hdb. der gr . Dial . 2.343 (Heidelberg, 1!)59) ta kes the Mycenaean
forma II.!I nominativcB.

88

ANCIENT GREEK DIALECTOLOGY IN THE LIGH T OF MYCENAEAN

Ioniclike dialect, as Ruiperez nnd Lejeune thought in 1956 (18), but

rather Ii specifically "i'l'lycenaean" v-dialect opposed to an a-dialect
which could, as far as this one feature is concerned, be ancestral to any
of the later-known dialects of Greek.
31. As remarked in note 16, I find some difficulty in supposing that
the shared 1l0n-?l'Iycenaean innovations of Arcadian and Cyprian all
originated during a time of Arcadian-Cyprian contact after 1200. The
situation is made even more complicated by Kathleen Forbes' article Oil
"The relations of the particle ,h with ~E(II), ~(1, ~(1~," Glotla 37.179-182
(1958). :Miss Forbes plausibly suggests that nil, peculiar to Attic-Ionic
and Arcadian (l\fyccnaean unfortunately gives no evidence), was an innovation for KOoIl (preserved in Arcadian ('~(1~), an ablaut variant of the
KE(P) of Cyprian and Aeolic. This "points to a period of unity" between
the ancestors of Attic-Ionic and Arcadian "after the emigration of
Achaeans to Cyprus" (p. 182). Since a feature shared by Attic-Ionic and
Arcadian but foreign to Doric can hardly have originated after t he in trusion of Dorians into the lands separating Arcadia and Attica, it is
very likely that the innovation &... is already Mycenaean in date. Yet
such a characteristic Arcado-Cyprian innovation as '" for Ell can hardly
have got its start after the Doriarl invasion either.28
It appears then that the beginnings of t he differentiation of AtticIonic, Arcadian, and Cyprian can be traced back at least to the late days
of the Mycenaean empire. So long as we lack Lincar B evidence to the
contrary, we can suppose that 0.11 fo r ~t was a widespread mainland
feature, lacking in the colonial area of Cyprus. Arcado-Cyprian innovalions not found in Lincar B texts evidently wcre limited to styles or
social levels that did not find their way into the palacc records of t he
chief rulers.
32. Anton Scherer's revision of the second volume of Thumb's Handbuch der griechischen Dialekle (Heidelberg, 1959) presents a full sketch
of l\'rycellllcan and takes its evidence into account in discussing the
intcrrelationships of the other non-West Greek dialeets_ Of l\'fycenaean,
Scherer says that it "schcint ... der gemeinsamen Vorstufe des Arkadischen und Kyprischcn nahezustehen" (p. 326). Similarly, Ebbe ViIborg's A len/alive grammar of MycenaeUlt Greek (Gotcborg, 1960) provides a good summary of the !'dycenllean features that had already been
II Indeed Lejeune, Jftmoire" de philQlogie mycellie.me 169 (Paris, 1958), and
H eubeek, G/olta 39.161, have suggested that examples of this change are found in
the Linear B tablets; but none of thei r examples are sure. Hamp. G/olta 38.200
(1960), adopts the view of Sc hulze, QuaC8tione8 epicae 323 n. 3 (Giitersloh, 1892),
that Hom. "',VVTO. is a development of ,.. "TO., and suggests very plausibly that
the word is one inherited by the E pie Ill.n gull ge from Mycenaean times in 8. specifically "Achaean" shape .

ANCIENT GREEK DIALECTOLOGY IN THE LIGHT OF MYCENAEAN

89

noted as diagnostic for dialect grouping (pp. 19- 23). Vilborg's conclusion is that if Risch is right in thinking that Attic-Ionic' and ArcadoCyprian were :l " rather uniform dialect group in Mycenaean times,"
Mycenaean would belong to that group; otherwise, "there is no serious
objection to ... regarding [it] as ... specifically Arcado-Cyprian" (p. 22).
33. 1\'lorc independent ideas arc to be found in Alfred Heubeck's "Zur
dialektologischen Einordnung des J'l'lykcnischen," GlQtta 39. 159-172
(1960-1961). Heubcck accepts Risch's position that before 1200 the
ancestors of At tic-Ionic and Arcado-Cyprian were quite close to each
other, and Mycenaean was close to both (p. 1(0). Dut he contends that
Mycenaean has a number of specific innovations that show that it is not
precisely the ancestor of any latcr dialect. One is the development of
-thiV- to -.siV-, as in kori.sijo 'Corinthian'; AWe e.xamples of this development like IIpo,6a}"l,nos 'from Probalinthos' arc, hc thinks, borrowed
from Mycenaean (p. 164). But surely Ventris and Chadwick are right
(Documents 73) in seeing llpo,6a}"lO"IOS as a regular Attic-I onic form, and
Koplv8wf, and so on, in Attic, as analogie or borrowed,~ so that this feature
is not unique to Mycenaean.
34. More cogent is the dissimilation of labiovelars already mentioned
in 28. The apparent syncope in -CiV- and -CeV- sequences (which
Miihlcstein [20J and GalJavotti [27] had used as evidence for grouping
with Aeolic) in words like kazoo for *KadclES, J,.:uruso for XP~o"IOS (pp. 167168), and t he apparent alternation of ke and ze in words like aketirija,
azetirija, are both somewhat unsure because of our uncertainty about the
precise range of values of the signs za ze zo and because it is possible that
Mycenaean scribes were more prone to write allegro form s than were the
writers of later centuries. Heubcck's final point (p. 169) is the development of " 1} to 0 in words like pe71l0 and amo, already noted as uniquely
Mycenaean by Risch in 1958 (30).
From these four specific Mycenaean features Heubcck, if I understand
him correctly, infers (p. 171) that it was not Dorians but the ancestors
of thc later Arcadians and Cyprians who destroyed the :Mycenaean
palaces. I strongly doubt that the number and imporlance of wellestablished differences between Mycenaean and Areado-Cyprian are
enough to lend any significant support to such a view, which as far as I
know is also quite unsupported by either archaoology or Greek legend.
35. I n 1960 Porzig returned to the problems of Greek dialectology,1O
and not surprisingly fOUIl d occasioll to modify some of the views in
II So also Risch, MWi. llelv. 14.71 n. 23 (1957); II. prior i, one would expect that
if "thy p!uticipates ill the change of "Ill to <T. "thi would at ICll..'It some of the time
participate in the notoriously spotty change of "Ii to <T"
10 I n his review of Thumb-Scherer, GnomQl1 32.58.)-596 (1960).

90

ANCIENT GREEK DLUECTOLOGY IN THE LIGHT OF MYCENAEAN

his predecipherment article of 1954. In particular, he gave up the idea

of an Aeolic invasion of the South (pp. 504-596).
36. In 1961 Risch's views continued to be attacked, this time by
Ruijgh, in " I.e traitement des sonantes voyelles dans les dialectes
grccs et Ill. position du mYl.eniell," lIlnemosyne 4:14.193-216. R uijgh
first establishes that ill Arcado-Cyprian and Acolic the regular t reatment
of r and 1is op, po and oX, (Xo) (pp. 194-198), but that for . 1} and .1]1
t hese dialects, like others, regularly have a. Exceptions occur oilly ill
numerals, for example, Arc. 6t1(0(TOi ) , EKor6p, Lesh. bOrOij and the aberrant
development is shared by Attic-I OIlic in words like ~;:I(OU l and al(ll(OuWI.
Ruijgh explains all these credibly as anaiogic, eventually to 6&0 and t he
decades in -~OPTa (p. 200)j likewise for Epic 6- from S 11 1- he has a credible
analogic explanation (p. 201).11
37. iHyttnacan clearly treats *r like Arcado-Cyprian and Aeolic.
But for *1J and *1]1 there seem to be two developments. a OCCUI'S, for
instance, in the negative prefix a-, t he accusative ending -a, and the
parHciple apeasa from apes1}tya (p. 202) . But 0 is found in the numeral
enewo- and in neuter n-stems like amo and anowoto 'allOt!o.To,,'. The
occurrence of ekamateqe fx.u aTUI("~ and ekamapi lXJ.I-lln on tablets inscribed by the same hand that wrote encwo and amolewija refutes
(p. 203) Risch's view (30) that the development to 0 is phonologically
l'Ollditioned by a neighboring labial. Ruijgh's solution is to take enewo
as analogic, like Arc. 30:0,1: and to explain the 0 of n-stems as starting in
rl n-stems like tv.ncpop, a>.d.paToi: leveling to aX~lcpoP, a>'dcpoTof and
aXu.pllp, aAd.pIlTof; hence also U1rEP/!O, U1fEPJM)TOi beside U'If'EP/!Il, f1'1f'fP/!IlTOS
(p. 205). This solution is ingenious but not very convincing. There is
no obvious reason why the few and unproductive nouns in *-r, -1}i08
should exert an influence on the common and productive type in *-m1},
+m1}tos, especially when that influence consists in the creation of new
doublets, rather t han the favoring of one or another inherited form or
thc leveling of some anomaly; and it is just as difficult for me to believe
t hat one idiolect would have analogically int roduced 0 in apJM)T- but not
in fXPIlT- as to believe that one idiolect would combine forms taken
from two or more different phonologic strains. The l'.,.rycenaean trea.tment of syllabic nasals remains problematic; but for dialeclological

ANCIENT GREEK DIALECTor.oG Y IN THE LtGHT OF MYCENAEAN

91

purposes, we can say that. Mycenaean clearly shows something different

from any later known Greek dialect, regardless of whether with Risch
we view the development to 0 as a conditioned sound change, or with
Ruijgh see in iL the result of various analogies.
38. Ruijgh furthcr (p. 207) tries to show that t he early differences
between Acolie and West Greek arc greater than Risch supposes. In
this he is probably right, but the lack of texts from the ancestors of
these dialects in the second millennium makes it impossible to say for
sure when the characteristic differences between them began developing.
Thus it is conceivable (albeit vcry unlikely) that in the North PIE
and clusters of 8 with resonant existed unchanged mueh longer than in
the South, and had assumed their characterist.ic West Greek and Aeolie
shapes barely in time to predate the population movemcnts at the end
of the Mycenaean period. For the dating of the Aeolic dative plural
in - to"CH, see 24.
39. In the South, Ruijgh emphasizes the old diffcrences bctwccn
Attic-Ionic and Arcado-Cyprian, and the agreements of IHycenaean
with the latter, against Risch's view of relative homogeneity of all throe.
Of the points raised by Ruijgh, the most cogent is the agreement of
Mycenaean posi with Arcado-Cypriun 1I"O~ against Attic-Ionic 1I"pOf
( 15, 16). The differing trea tment in Attic-Ionic and Arcadian of
lengthened t and 0 I have discussed in 12; I still do not see how it can
be proved to be a particularly old feature. Ruijgh tries to show that the
Attic-Ionic and Arcado-Cyprian difference in the $tenscs of tw-verbsU
must be old, because the Arcado-Cyprian distribution of
and -qrecurs in Argolie and so must go back to a pre-Dorian substratum,
But why cannot t he agreement between Arcadian and Argolic be the
result of borrowing between neighboring dialects after t he Dorian
invasion? l\fore to the point is the agrccmcnt (as fur as the evidence
goes) between Arcadian und Cyprian: it seems quite unlikely that the
and
two dialects would have hit on exactly the same distribution of
-q- by chance. The most reasonable intcrpretation of the presently
available evidence is that aorists (if 1101. also futures) to r",-prcscnts had
been created already in the second millennium, and that one group of
southern speakers, those whose language was to develop into Arcadian
and Cyprian, had already given them the shape -0"- after velars, +
elsewhere (cf. 2 1).M

*,

u Arc.

M,~&u"", ... ,

:r"P'7';'~""",

Cypr. kale&keuwase, oruxl1

v~.

Att. Mb ....... ,

tE.j7"",,, .

.. He nce abo Ruijgh seema right in tracing H omeric {orUlll "'ith f to Achaean,
since Homeric lauguage observes the AreadoCyprian rul e to the extent of never
uaing ~ if either of the t.wo precedill8 syllablc9 contains a velar. Yet PelMgiotic
Thessalian, which is relatively little inl1uenced by West Greek, hll.8 lI'<>4><t"""~'"

92

ANCIEKT GREEK DIA LECTO LOGY IN THE LIGHT OF MYCENAEAN

In the matter of t he contract verbs, Ruijgh (p. 21 I) is skeptical of

R isch's derivation of l\I yc. tercja from nXdet ; but, paradoxically, if
Ruijgh is right in considering lCTcja an athematic form (which I very
strongly doubt), its inflection is possibly (but d. 13) an inherited
archaism, of no diagnost.ic value for dialectology, while if R isch is
right, lereja would contain an innovation shared by neither Attic-Ionic
nor Arcado-Cyprian.
40. The apparent agreement of Myc. pel~ with Arc. u40m against
Attic-Ionic uq,iu, (p. 212) can also be all archaism, so long as t he prehistory of this pronoun remains unclear. The view that Mycenaean
shares with Arcado-Cyprian the use of dative rather than genitive with
prepositions meaning 'from' (ibid.) has been dealt a heavy blow by
H ouseholder's article, "pa-ro and Mycenaean cases," GloUn 38.1- 10
(1960).38
41. l\'!ore recent discussions that I have sccn do not seem to introduce
any important new viewpoints. I n his article of 1956, " The Greek dialects
and Greek pre-history," Greece and Rome 25.38- 50 (which I failed to
treat at its proper place in this survey), Chadwick suggested (p. 44)
that the innovations in which AtLic-Ionic agrees with West Greek against
Arcado-Cyprian and l\,[ycenaean can be explained by supposing "that
round about 1000 D. C. a dialect oC t he Areadinn type came for a period
under Doric influence; but this soon censed .... ":17 T his explanation
does not sccm adequnte to account for all the agrccments involved (cf.
the skepticism of Ruijgh, Mnemo8yne 4: 1 L 106 n. 4). The distribution of
tV~ for tv with accusative, which may be a post-Mycenaea n innovation,
could perhaps be explained by Chadwick's view (although I don't see
why it couldn't equally well have started in a variet.y of South Greek
and spread from there to the super- and ad-jncent varieties of West
Greek). But other agreements of Attic-Jonic and West Greek are almost
and fp")'"E<tTO (Scherer 70, Beclltel 1.190) . The latt er, to be sure, vio lates the
rule against f after vel afllj bu t who can say that ill the Thcuaiilln of 900 B.C.
it was not still observed? It seems theoretically possible to attribu te the H omeric
t- form~ to Aeolic Btill, but eve n so the pll.ltern must be old, perhaps P roto -Greek:
Attic-I onic with Lesbian snd the bulk of Boeotian would have leveled in one
direction, West Greek (except Argolic) in the othe r. Some relevan t Mycenaean
forms would be very welcome here .
.. Head .,.</1..
with plene writing of a monosyllable, as in qoo for ")'''",(.j,?
Or 0'<1>. (,)< wi l h lhe same *-si as in -oi from *-oisi? Cr. Documents 87, 189.
" l1 ouseholder concludes that para often, perhaps always, is used in a locative
sense; if it ever means ' from', the case following it can be ablative, not yet ~yn
cretized with either dative or genitive (9).
11 Similarly in " The prehisto ry of the Greek language" (Cambridge Ancienl
Hi,loryl 2:39.16 [1063]).

ANCIENT GREEK DIALECTOLOGY IN THE LIGHT OF MYCEKAEAN

93

certainly older: these are t he choice of <iTO rather than <i1l'&, and the
development of (I. rather than 0 next to syllabic liquids.
42. So far, one innovation of clearly Mycenaean date had been
established for the whole South (Mycenaean, Arcado-Cyprian, AtlicIonic) : the assibilation of *t(h)i to O'~ and of *t(h)y to 0'(0'). There
appear to be also somc choices common to the wholc South: adverbs in
-TE, and perhaps athematic infinitives in _EP(l.~.38 If l\'Iycenaean material
were not SO limited, this list could p robably be extended.
43. But it seems equally clear that the South was already somewhat
differentiated in the second millellnium, despite the near uniformity of
Linear B texts from Pylos, Mycenae, and 1(n088Os. I t is indeed hardly
thinkable, even supposing t he language brought to Greece at the beginning of the second millennium was quite homogeneous, that it could
have gone on being spoken there for nearly a thousand years without
developing considerable local and social differences. The Mycenaean
empire may have led to the creatioll and spread of a koine, but clearly
this koine, if it existed, did 1I0t entirely obliterate internal differences.
Mycenaean shares some choices with Arcado-Cyprian against AtticIonic: op fro m r, 11'00'1, 0.1/'&. It is not certain that Mycenaean shares any
innovations with Arcado-Cyprinn against Attic-Ionic archaisms. If my
doubtll about R uiperez's view of t he primary middle verb endings
(note 14) are justified, t he ending -TOI for -1'Q.~ would be one such innovation. The interpretatioll of lereja and lercjae as forms of an athematic
verb stem TE;\.Eto.- is far too insecure to be used as evidence that Mycenaean shares athematic inflection of contract verbs with ArcadoCyprian. As far as I see, Mycenaean does not share a single innovation
or choice wit h Attic-IOllic against Arcado-Cyprian. 3~
44. Mycenaean seems to have a few innovations unknown elsewhere,
indicating that the language of thc tablets cannot be a direct ancestor
even of Arcado-Cyprian, so t hat we M.ve evidcnce for a minimum of
three varieties of speech in the Mycenaean empire (four, if we take into
account the evidcnce mentioned in 31 that t he dialect of Cyprus had
already begun divcrging from that of the homeland). Specific Myccnacau
innovations include the vocalization of pemo nnd similar words (37); the
apparent dissimilation of labiovebrs (28); possibly elja from cije (29)
and accusative plurals in -H (30). The apparent syncope of -1- and - ~-

94

ANCIENT GREEK DIALEt"l'OLOGY IN THE LtGHT OF MYCENAEAN

before vowels (34), while recurring in Aeolic, is not normal in AtticIonic and Arcado-Cyprian, and so probably belongs here.
45. For Homeric, both Achaean (Myccnacan) and Acolic (Thcssalian?) layers seem firmly established.
46. The one point I would like to elaborate a little further is the
puzzling position of Ionic in the second millennium. As I have just said,
it is apparent that "Ionic" and "Arcado-Cyprian" speech forms coexisted in the ~-'lycenaean empire. But what was their distribution?
It would be rather simple to suppose that "Ionic" was the speech of
the upper classes, who remained in control in Attica and elsewhere went
into exile, eventually ending up in the Cyclades and Asia Minor, while
"Arcado.Cyprian" was the speech of the lower classes, which became
standard in Arcadia aud Cyprus, in the former owing to the disappearance of the old upper class, in the latter owing to events about
which we have no information. But this feature is hard to square with
the "Arcado-Cyprian" features of Linear B and the Achaean layer of
Homer's language. Granted that accountants may have written a
different dialect from that which their employers spoke, it seems incredible that poets would have used forms (e.g., aorisls and futures in
-~-) that would hu.vc seemed substandard in the cars of the audience on
whom their livelihood depended, and equally unlikely t hat poems on
the heroic deeds of Mycenaean princes would have bccn composed and
preserved by their former serfs.
An alternative would be to imagine a geographic distribution, with
"Ionic" spoken perhaps in Attica (and elsewhere?), and "Achaean" in
at least Pylos, Mycenae, and Knossos. But, leaving aside the minor
detail that there is no direct evidence o.t all for any such distribution,
this would result i.n the anomalous situation that Achaean and Acolic,
which share presumably diagnostic features, would be cut otT from each
other by Ionic-unless indeed we imagine t hat Boeotia was "Achaean,"
too, in which event Attic-Ionic would be cut otT from West Greek, and
the old agreements between these two would have to be explained as
owing to chance. The situation can be diagrammed as fo llows:

WEST GREEK

ACHAEAN

AEOLIC

I ONIC

ANcrENT GREEK DIALECTOLOGY [N THE LIGHT Q}' MYCENAEAN

95

Each dialect shares significant features with its horizontal neighbor

and the dialect diagonally opposite, but few or none with its vertical
neighbor. Things would be much simpler if wc could posit a prehistoric
distribution like this (or with the same configuration, but reversed or

rotat-ed):

WEST GREEK

ION IC

AEOLIC

ACHAEAN

Now each dialect shares exclusive features with both its neighbors,
but not with the dialect in the opposite corner. To be sure, I do not
know of the faintest shred of I10nlinguistic evidence for such a prchistoric
arrangement of the Greek dialccts. But it is perhaps not a complete
waste of time to speculate that such an arrangemcnt may have existed,
perhaps in Greece in the (early?) second millennium, perhaps even
earlier and outside of Greece.
(Participant' in the disCtlssion jQ/lowin() the wnferenCll presentation of the firs l
of Ihi8 paper: Watkins, Winter, Puhlltil, Halllp, Elllenea!l, Col/inder.)

~er8ion