Re: 3rd declension stems

From: Carl W. Conrad (
Date: Tue Dec 30 1997 - 17:54:09 EST

At 5:32 AM -0600 12/30/97, clayton stirling bartholomew wrote:
>I seldom fuss about things like this but occasionally when I see a noun from
>the third declension iota, upsilon or sigma stems I look it up in Smyth to
>review how these work.
>The rule that James Efird (p. 80, A Grammar for NT Greek, Abingdon 1990) gives
>for determining the third declension stem seems to lack something. He says
>"The stem is determined by the removal of the -OS from the genitive singular
>form." E.V.N. Goetchius also states that the stem of a third declension noun
>can be "deduced" from the genitive singular (The Language of NT, #147 remark
>#3). However the process of deduction is not spelled out very clearly.
>I don't see how this works with the Upsilon, Iota or Sigma stems. The genitive
>singular of BASILEUS is BASILEWS, the stem is BASILEU- which cannot be
>"derived" from the genitive singular unless you throw in some extra rules
>about vowel contraction.
>Can anyone state a complete set of lucid rules for determining the stem of a
>third declension noun which will clarify all cases including Upsilon, Iota or
>Sigma stems?
>This may be a tall order but I suspect that I am not the only person that gets
>confused about this.

I'm not sure that this can be done with any simple economical formula, in
view of the fact that more than one kind of sound-change is involved and
that dialectal variants cloud the picture. The old rule that you can
determine the stem from the genitive case form by removing the -OS is too
simple precisely because it will not tell you what the original evanescent
consonant was--and what do you do with genitives in -EWS from POLIS and
BASILEUS? I rather think that what one must do is learn the declension of
certain paradigmatic nouns for types, such as:

(1) nouns in -EUS (mostly agent nouns, I think) have diphthongal -EU-
before the nominative sg. ending -S, elsewhere -HF- (eta-digamma); in the
genitive sg., this yielded a form -HFOS, dat. sg. -HFI, acc. sg. -HFA; but
after loss of the digamma between the vowels, quantitative metathesis set
in so that the H became E and the vowel in the termination became long (-WS
gen., long -I dat., long -A acc.); the plural is more complicated;
quantitative metathesis again accounts for the genitive -EWN (from original
-HFWN) and the acc. -EAS (long A--from original -HFAS, where the A was
short); the dative is built on the diphthongal stem -EU- like the nom. sg.:
BASILEUSI(N); the nom. was -HFES or -HES in Homer, but later seems to have
assimilated to the EI/I type declension where -EIS is the ending for both
nom. and acc. pl.

(2) nouns in -IS (e.g. POLIS, but quite numerous process nouns in -SIS
also, such as PRAXIS, POIHSIS, FQISIS, KTL.: these have stems alternating
between the vocalic I (iota) before a consonantal ending (nom. -IS, acc.
-IN) and the diphthongal EY (epsilon + consonantal iota or Y) before a
vocalic ending; in the genitive and dative singular I think what happened
is that -EY- lost the -Y- and the -E- became -H- by compensatory
lengthening, and then the same vocalic metathesis occurred as in the case
of the -EUS nouns, so that gen. -EYOS --> -HOS --> -EWS and dat. -EYI -->
HYI (short I) --> -EI (long I). I don't know how the plural best explains
itself for these nouns; probably the -E- regularizes itself by analogy.
It's worth noting however that the Ionic form of nouns of this sort (e.g.
in Herodotus) shows the simple -I- throughout: POLIS, POLIOS, POLI,

(3) nouns in -ES stems. To the best of my recollection there are primarily
neuter nouns of this sort; they show -OS ablaut of -ES in the nom.,
elsewhere the -S- of the stem evanesces between consonants and there is
contraction of the -E- and the terminal vowel (gen. -ESOS --> -EOS -->
-OUS; dat. -ESI --> -EI; acc. -ESA --> -EA (but A is long for some reason
as vowel metathesis); nom. pl. -ESES --> -EES --> -EIS, etc., etc. But
there's an interesting combination of this type in proper names in -KLHS
(SWKRATHS, PERIKLHS, hHRAKLHS) from the root KLEF/KLU ('renown'); the stem
of these is originally -KLEFES, wherein we have TWO evanescent consonants,
the digamma and the sigma. In the nom. sg. we have -KLEFHS, where the
ending -ES lengthens into -HS, but the digamma evanesces, leaving -KLEHS,
which in turn contracts into -KLHS--but in the voc. that -ES never
lengthens, so we have -KLEFES --> KLEES --> -KLEIS (W PERIKLEIS!).

(4) There are simpler -US/UFOS stems such as NHDUS ('belly', 'womb'); these
add consonantal -S and -N directly to the vowel in nom. and acc., while in
the gen. and dative the -F- evanesced and left no trace, so that we can
pretend that the stem of nouns of this sort is simply -U-.

(5) And then there are a few other such nouns in -W like PEIQW
("persuasion," "seduction"--PEIQW is one of the handmaidens of Aphrodite in
myth and cult); here we are dealing with a stem alternating betwee W and
OF, or perhaps it would be more accurate to say that in the nominative -OF
lengthens to -WF but the digamma won't stand at the end of the word; the
other declensional forms are understand from contraction of -OS, -I, and -A
with the stem -O(F) to yield -OUS, -OI, and -W (contraction of -OA after
loss of intervocalic F). There are a few other words like this, e.g. HXW
('echo') hHRWS ('hero'), etc.

Perhaps that's sufficient to explain why it's difficult to formulate a
simple and coherent rule for these several stems; to understand them one
really needs to know some basic phonology of Greek and explain the
phenomena found in the attested forms backwards in terms of phonological
principles. There are, in short, some good historical reasons why we cannot
simply formulate mathematical principles of noun declension and deduce all
declensional forms in terms of a simple formula.

Carl W. Conrad
Department of Classics/Washington University
One Brookings Drive/St. Louis, MO, USA 63130/(314) 935-4018
Home: 7222 Colgate Ave./St. Louis, MO 63130/(314) 726-5649 OR

This archive was generated by hypermail 2.1.4 : Sat Apr 20 2002 - 15:38:42 EDT