Prefatory note: I have slightly modified my encyclopedia entry on ancient Greek accentuation

(Gunkel, Dieter. 2014. “Accentuation.” In Encyclopedia of Ancient Greek Language and Linguistics. Vol 1: A–F,
ed. Georgios K. Giannakis et al. Leiden/Boston: Brill, 7–12.)
by replacing the transliteration required for the EAGLL with the Greek alphabet and IPA and adding some
footnotes. I hope this makes the article more useful for classicists and linguists. An offprint of the actual
publication begins on p. 12 of the pdf.

ACCENTUATION
ABSTRACT
The accents marks in modern editions of Ancient Greek texts primarily reflect the accentual system of an
educated register of the Koine of the early 2nd c. BCE. In this system, phonological, morphological, and
lexical factors conspire to associate a pitch accent with one syllable of each lexical word. The phonology of
the language permits limited contrasts in accentual position (λιθοβόλος vs. λιθόβολος = lithobólos vs.
lithóbolos) and type (ἰσθµοί vs. ἰσθµοῖ = isthmói̯ vs. isthmôi̯); in the latter case, the syllable marked with an
acute accent hosts a High tone, and that marked with a circumflex hosts a High-Low falling contour tone. In
any given form, the maximum number of phonologically licit accentual contrasts is three. Within the
bounds set by the phonology, morphological and lexical factors, e.g. the inherent accentual properties of
particular suffixes, further determine the accentuation of a word. Comparison with related Indo-European
languages, especially Vedic, shows that the Greek system developed from an earlier system that likely lacked
a contrast in accent type but permitted more positional contrasts; Greek accentuation is more dependent on
the rhythmic structure of the language.

1. THE NOTATIONAL SYSTEM

The accent marks written in modern editions of Ancient Greek texts derive from a grammatical tradition
that most likely began in Alexandria in the early 2nd c. BCE, with Aristophanes of Byzantium, to whom the
invention of the written signs is attributed, and his successor as librarian, Aristarchus of Samothrace. Given
the important functional role that accent played in the language, conveying accentual information in
writing facilitated the difficult task of reading poetic texts written in scriptio continua. For example, the
unaccented graphic sequence απονου = aponou could represent ἀπ᾽ ὄνου = ap’ ónoː 'from a donkey'1 or ἀπο

1
The apostrophe shows that a word-final short vowel has been elided (deleted), in this case /o/. Here, as often, elision avoids
vowel hiatus.

1
νοῦ = apō nôː 'from (your) senses'2 (cf. Aristoph. Nub. 1273). These lectional signs conveyed the accentuation
of the Koine spoken during that period, and to a lesser extent, the accentuation of other dialects; it is
possible that the scholars also consulted oral traditions, e.g. rhapsodic performances of the Homeric poems,
to determine the accentuation of forms that were unfamiliar to them. In the 2nd c. CE, Aelius Herodian
codified the tradition in περὶ καθολικῆς προσῳδίᾱς = perī katholikɛ̂ ːs prosɔːi̯díaːs 'On Prosody in General',
which served as the basis for later works on accentuation. In the early accented papyri, which date from the
2nd c. BCE onwards, notational conventions vary, as does the frequency with which written accents were
applied. The notational system familiar to us, where each accented word is marked with an acute,
circumflex, or grave, was first applied in minuscule manuscripts of the 9th c. CE by scribes following the
precepts of the same grammatical tradition. The early works on accentuation including Herodian's do not
survive as such, but scholia and short treatises based on them provide us with indirect access. On the
grammatical tradition, the papyri, and the manuscripts, see Probert 2006:21-52 with refs.

2. PHONETICS AND PHONOLOGY OF THE GREEK ACCENTUAL SYSTEM

Phonetically, accent may be studied from an articulatory, acoustic, and perceptual standpoint. It is clear
that the most salient perceptual correlate of ancient Greek accent — what the Greeks themselves 'heard' —
was pitch, at least until the 2nd c. BCE. Within the word, pitch peaked during syllables marked with an acute
or grave accent, and it both peaked and fell again during syllables marked with a circumflex. Phonologically,
we may say that syllables marked with acute accent hosted a High tone and those marked with a circumflex
hosted a High-Low tone, i.e. a falling contour tone. Evidence for the phonetic nature of Greek accent comes
from several sources. The words used to refer to 'accent' have to do with musical pitch, e.g. τόνος = tónos
refers to the ‘tension’ and therefore to the perceived pitch of vibrating strings, and the basic meaning of
προσῳδίᾱ = prosɔːi̯díaː is 'singing along (to music)'.3 The adjectives used to specify the three different types
of προσῳδίᾱ = prosɔːi̯díaː are ὀξεῖα = oksêi̯i̯a ‘high’ for acute, βαρεῖα = barêi̯i̯a ‘low’ for grave, and ὀξυβάρεια =
oksubárei̯i̯a ‘high-low’ for circumflex. Fragments of non-strophic musical compositions dating from as early
as the 3rd c. BCE — the Delphic hymns in particular — provide a richer source of phonetic detail. As in
vocal music traditions in a number of languages with contrastive tone (Devine and Stephens 1994:160-171),
the fragments display a relatively strict correspondence between the pitch movements of speech and the
melody of the music it is set to (Devine and Stephens 1994:172-194; Probert 2006:47-48; West 1992:199). For
example, the accented syllable of a word is set no lower than its unaccented syllables, such that the pitch

2
I have chosen the (combining macron =) IPA mid tone to convey the “grave” accent, which is an underlying high tone that has
been lowered in the (post-lexical =) phrasal phonology. See further below.
3
τόνος = tónos and προσῳδίᾱ = prosɔːi̯díaː are the source of English tone and prosody.

2
peak of a word corresponds to a local peak in the music. If a syllable bearing a circumflex is set to a two-
note melism, the first is usually higher, respecting the falling pitch contour of circumflex accent. The grave
accent, a phrasal sandhi variant of the acute, proves to be a lowered version of the acute (Devine and
Stephens 1994:180-183), and it is possible to reconstruct the accentual contour of entire words. For example,
in a proparoxytone word of five syllables (e.g. ἐκλεγόµενος = eklegómenos), pitch rose steadily over the initial
two unaccented syllables, peaked at the accented syllable, then fell steeply over the first post-accentual
syllable and less steeply over the final syllable (Devine and Stephens 1994:183-189).

ek le gó me nos4

The turning point between the post-accentual fall in pitch and the rise to the following accent coincides
with word-boundary, which certainly had a 'demarcative' function, i.e. made word boundaries audible
(Allen 1973:246); this may point to a Low word-final boundary tone in the phonological representation
(Devine and Stephens 1994:180).5 Statements by grammarians and other ancient scholars provide a further
source of information about the phonetic nature of word-level accentuation (Devine and Stephens 1994:171-
172; Probert 2003:4-7), and comparison with the accentual systems of related languages, especially Vedic and
Balto-Slavic, suggests that in Proto-Indo-European, one syllable of each accented word was realized with
high pitch (cf. Olander 2009:53-100 with refs.).
The accentuation of a word is determined by interacting phonological, morphological, and lexical
factors. The phonology plays two important roles in this system. First, it places restrictions on which
syllables can host an accent and on what type of accent (acute and/or circumflex) can be realized there. The
most important of these restrictions, the so-called 'Law of Limitation' [CrossRef], essentially sets the

4
The diagram, based on Devine and Stephens 1994:189, is not included in the encyclopedia entry.
5
In my view, both the word-final grave and the circumflex accent on penultimate syllables are the result of tonal crowding
avoidance with this low word- or clitic group-level boundary tone.

3
accentable domain of a word, which consists of the final three syllables, if the ultima6 is light, and the final
two, if it is heavy (Göttling 1835:21-28; Steriade 1988:273-275). Note that for the Law of Limitation, a single
word-final consonant is weightless; final syllables ending in a short vowel (-V#) and those ending in a short
vowel followed by a single consonant (-VC#) both count as light, e.g. βασίλεια = basílei̯i̯a 'queen', acc. sg.
βασίλειαν = basílei̯i̯an.7 All other syllable rhymes count as heavy.8 Within the accentable domain, an acute
accent is phonologically permissible on any syllable (with one systematic exception noted below), but the
circumflex accent is subject to further restrictions. It is phonologically licit on final syllables containing a
long vowel or diphthong (i.e. final VV-syllables), where an acute is also possible, as reflected by contrasts
such as ἰσθµοί = isthmói̯ 'isthmuses' (nom.) vs. ἰσθµοῖ = isthmôi̯ 'on the isthmus' (loc./adv.). The circumflex
also occurs on penultimate VV-syllables, where it is in complimentary distribution with the acute according
to the so-called 'sōtêra rule' [CrossRef]: the accent on a penultimate VV-syllable is realized as a circumflex if
the word-final syllable contains a short vowel, e.g. σωτῆρα = sɔːtɛ̂ ːra 'savior' (acc.), οἶκος = ôi̯kos 'house';
otherwise, it is realized as an acute, e.g. σωτήρων = sɔːtɛ́ ːrɔːn 'saviors' (gen.), οἴκοις = ói̯koi̯s 'houses' (dat.). In
other words, if the penult is accented, the phonology determines which type of accent is realized there,
meaning that phonologically, a contrast in accent type is limited to word-final VV-syllables.9 Together, these
phonological restrictions permit only the five combinations of accent location and type already recognized
by the ancient grammarians.

Oxytone (ὀξύτονος = oksútonos): acute on the ultima, e.g.

• ὀφθαλµοί = ophthalmói̯ ‘eyes'
Perispomenon (περισπώµενος = perispɔ́ ːmenos): circumflex on the ultima, e.g.
• ὀφθαλµῶν = ophthalmɔ̂ ːn ‘eyes’ (gen.)

6
The term ultima is used in classical linguistics to refer to the word-final syllable, in the same way penult (< paenultima) is used to
refer to the next-to-last syllable, etc.
7
For the purposes of all other weight-sensitive morphophonological processes as well as the quantitative meter, only syllables
ending in a short vowel (i.e. only syllables whose rhyme consists of a short vowel alone) count as Light. This motivates the claim
that one word-final consonant in ancient Greek is “extrametrical.” Note that the extrametricality may be related to the fact the
one word-final consonant is “re-syllabified rightward” phrase-internally where followed by a V-initial word.
8
Many grammars of ancient Greek state that for purposes of accentuation, final syllables containing a long vowel or a diphthong
are heavy, and all others are light. This incorrectly entails that final syllables closed by two or more consonants are light.
9
Strictly speaking, the contrast between acute and circumflex (High and High-Low) on accented, long-voweled penultimate
syllables is not purely phonological. The morphology plays a role if the final morpheme is oi̯ or ai̯ or ends in oi̯ or ai̯. See §4 below
and the EAGLL entries on the “Law of Limitation” and the “Sōtêra Rule.”

4
 Paroxytone (παροξύτονος = paroksútonos): acute on the penult, e.g.
• σωτήρων = sɔːtɛ́ ːrɔːn ‘saviors’ (gen.)
Properispomenon (προπερισπώµενος = properispɔ́ ːmenos): circumflex on the penult, e.g.
• σωτῆρα = sɔːtɛ̂ ːra ‘savior’ (acc.)
Proparoxytone (προπαροξύτονος = proparoksútonos): acute on the antepenult, e.g.
• ἥλιος = hɛ́ ːlios ‘sun’

However, the maximum number of ways any given form may be accented is three, as exemplified by the
following nonsense words.

mei̯duplɛːrɛː: mei̯duplɛːrɛ̂ ː, mei̯duplɛːrɛ́ ː, mei̯duplɛ́ ːrɛː

mei̯duplɛːros: mei̯duplɛːrós, mei̯duplɛ̂ ːros, mei̯dúplɛːros
mei̯duploros: mei̯duplorós, mei̯duplóros, mei̯dúploros

3. LEXICON, MORPHOLOGY, PHONOLOGY: RECESSIVE ACCENTUATION

Within the bounds set by the phonology, morphological and lexical factors determine the accentuation of a
given word. Thus, Greek accent is only 'free' insofar as the accentuation of a word is not determined by
phonological factors alone. This limited freedom is reflected in minimal pairs that differ only in position
and/or type of accent, e.g. κήρ = kɛ́ ːr 'doom' vs. κῆρ = kɛ̂ ːr 'heart', λιθοβόλος = lithobólos 'pelting with stones'
vs. λιθόβολος = lithóbolos 'pelted with stones'.
If every inflectional form of a word is accented as early (i.e. as far 'left') as permitted by the Law of
Limitation, that word is said to exhibit 'recessive accentuation', e.g. ἄνθρωπος = ánthrɔːpos '(hu)man', gen. sg.
ἀνθρώπου = anthrɔ́ ːpoː, dat. sg. ἀνθρώπῳ = anthrɔ́ ːpɔːi̯, acc. sg. ἄνθρωπον = ánthrɔːpon, etc. In this accentual
subtype, we observe the second role of phonology interacting with lexical and morphological factors.
Recessive accentuation is both the property of particular lexical items, such as ἄνθρωπος = ánthrɔːpos, and
the property of entire morphologically circumscribed classes of words, such as finite verbs, 3rd declension
neuter nouns, and most types of compounds, including those whose first member is a governing preposition
or verb (Kiparsky 2003; Vendryes 1945:189-196), e.g. φιλ(ο)-X = phil(o)-X 'X-loving' compounds such as
φίλοινος = phíloi̯nois 'wine-loving', φιλόσοφος = philósophos 'wisdom-loving', φιλόπαις = philópai̯s 'boy-loving',
φιλόρτυξ = philórtuks 'quail-loving', φιλοσπῆλυγξ = philospɛ̂ ːluŋks 'cave-loving', etc. In short, lexical and/or

5
morphological features determine whether a word is recessively accented; the phonology determines the
accentable domain and locates the accent 'leftmost' in that domain.
A number of facts suggest that recessive accentuation was the unmarked or default type of
accentuation in the language (cf. Probert 2006:128-144): among accented words, recessive accentuation is
more frequent than non-recessive accentuation by both type and token; comparison with Vedic and
Germanic points to a tendency within the history of Greek to innovate recessive accentuation in inherited
lexical items, e.g. páros 'formerly' vs. Vedic purás 'before' < PIE *pr̥ hxós or *pr̥ hxés; entire morphological
classes of words (noted above) are recessively accented in Greek, but no such class is associated with a non-
recessive type of accentuation; finally, in Lesbian, recessive accentuation was generalized to virtually all
accented words, arguably due to an extreme form of the tendency just noted.10
A central question regarding the Law of Limitation and recessive accentuation is whether the
accentable domain is related to the rhythmic phonology of the language in general (cf. Devine & Stephens
1994:154). In other words, can the accentable domain be equated with a rhythmic/prosodic constituent that
is also reflected in meter, word formation, and other (morpho)phonological processes? Building on Steriade
1988, recent studies suggest that the accentual domain is — or is aligned with — a unit of rhythmic
organization known as a foot (cf. Probert 2010 with refs.). The span between the accent (´) and word-end (#)
consists of two light syllables (LL), e.g. εὑρήµατα# = heu̯ rɛ́ ːmata#, a heavy syllable (H), e.g. εὑρηµάτων#
heu̯ rɛːmátɔːn#, or a heavy-light sequence (HL), e.g. εὕρηµα = héu̯ rɛːma# — in other words, ´LL# or ´H(L)#.
This span has been equated with a word-final quantity-insensitive trochaic foot (Sauzet 1989) and a
quantity-sensitive one (Golston 1990). The latter, a bimoraic rhythmic unit consisting of either two light
syllables (LL) or one heavy syllable (H) may also be reflected in word formation (Gunkel 2011),11 meter
(Golston & Riad 2000; 2005; Gunkel 2010:43-75), and constraints on minimal word size, alias word minima,
in the language (Devine and Stephens 1994:93; Golston 1991). On that analysis, in recessively accented
words, the beginning of the post-accentual fall in pitch (`) is aligned with the first mora of the word-final
foot, e.g. εὑρή(µὰτα) = heu̯ rɛ́ ː(màta) and — representing the bimoraic long vowel ω = ɔː as oo = ɔɔ —
εὑρηµά(τὸον) = heu̯ rɛːmá(tɔ̀ ɔn). For indispensable in-depth treatments of Ancient Greek rhythmic

10
Proto-Indo-European may well have have a leftmost default accent for underlyingly unaccented as well as morphologically
deaccented words (cf. Kiparsky 2010, Yates 2015). The domain for accentuation in PIE appears to have been the word. A number of
IE branches did not change the domain (much), but did generalize leftmost/initial accentuation (more or less). These include
Germanic, Italic, and Tocharian.
11
Gunkel 2011 argues that a change in word formation introduces a (lexically and morphologically restricted) form of Trochaic
Shortening into the langauge. According to mosts phonologists, Trochaic Shortening is a process that optimizes moraic trochees
at the end of the word. Trochaic Shortening in Greek thus provides additional evidence for right-aligned moraic trochees in the
language.

6
organization including alternative views on foot structure, cf. Allen (1973) and Devine and Stephens (1984,
1994).

4. FURTHER MORPHOLOGICAL FACTORS

Morphological features also condition the phonological status of the word-final diphthongs -οι and -αι = -oi̯
and -ai̯ in the accentual system. For both the Law of Limitation and the sōtêra rule, word-final -οι and -αι =
‑oi̯ and -ai̯ have the status of a light word-final syllable rhyme consisting of a short vowel plus a consonant
(‑VC#). This is apparent in recessively accented paradigms where, for example, nom. pl. φιλόσοφοι =
philósophoi̯ 'philosophers', βασίλειαι basílei̯i̯ai 'queens', and 2sg. aor. impt. mid. παίδευσαι = pái̯deu̯ sai̯
'educate' are proparoxytone like nom. sg. φιλόσοφος = philósophos, acc. sg. βασίλειαν = basílei̯i̯an, and 2sg. aor.
impt. act. παίδευσον = pái̯deu̯ son, which end in -VC#. It is also apparent in forms where the sōtêra rule
applies. For example, nom. pl. οἶκοι = ôi̯koi̯ 'houses', γαῖαι = gâi̯i̯ai̯ 'lands', and aor. inf. act. παιδεῦσαι =
pai̯dêu̯ sai̯ are properispomenon like nom. sg. οἶκος = ôi̯kos 'house', acc. sg. γαῖαν = gâi̯i̯an 'land', and neut.
nom./acc./voc. sg. aor. act. ptc. παιδεῦσαν = pai̯dêu̯ san, which end in -VC#. The inflectional endings -οι and
‑αι = -oi̯ and -ai̯ of the 3sg. present and aorist optative active and the locative singular — or adverbial
locative — ending -οι = -oi̯ pose morphologically conditioned exceptions. Like all other word-final long
vowels and diphthongs, they have the status of heavy -VV# rhymes in the accentual system. This is likewise
reflected in recessive paradigms where, for example, 3sg. pres. and aor. opt. act. παιδεύοι = pai̯déu̯ u̯ oi̯ and
παιδεύσαι pai̯déu̯ sai̯ are paroxytone like παιδεύω = pai̯déu̯ u̯ ɔː 'I am educating', and where the sōtêra rule fails
to apply: loc. sg. οἴκοι = ói̯koi̯ 'at home' is paroxytone like dat. sg. οἴκοι = ói̯kɔːi̯. There is no evidence that this
morphologically conditioned phonological distinction between diphthongs existed outside the system of
accentuation (Probert 2012).
The distribution of acute and circumflex accent on word-final syllables that are phonologically 'free'
to host either accent is conditioned by morphological features as well. Specifically, the distribution appears
to be based on case: nominative and accusative forms bear acute accent, genitive and dative forms bear
circumflex accent, e.g. 1st declension nom. and acc. sg. φορᾱ́ = phoráː 'carrying, bearing; load, burden', φορᾱ́ ν =
phoráːn, pl. φοραί = phorái̯, φορᾱ́ ς = phoráːs vs. gen. and dat. sg. φορᾶς = phorâːs, φορᾷ = phorâːi̯, pl. φορῶν =
phorɔ̂ ːn, φοραῖς = phorâi̯s.
Morphemes themselves have accentual properties (Kiparsky 1973, 2010, forthcoming; Probert
2006:145-148; Steriade 1988). Descriptively, there are four types in Ancient Greek. First, there are inherently
accented morphemes such as the stem ἀγρό- = agró- agró- of ἀγρός = agrós 'country' and the suffixes -άδ- =
‑ád- and -ικό- = -ikó- of words such as gen. sg. phugádos 'exile', manikós 'mad'. The last inherently accented
morpheme imposes its accent on the entire derivative, e.g. φυγάδ-ικό- = phugád-ikó- → φυγαδικός =
phugadikós 'of/for exile' (not **φυγάδικος = **phugádikos). Most inherently accented morphemes are stems
or derivational suffixes such as the -εύ- = -éu̯ - used to form nouns of occupational/ethnic appurtenance and
7
agent nouns, e.g. χαλκεύς = khalkéu̯ s 'bronzesmith' (derived from χαλκός = khalkós 'bronze'), Eὐβοιεύς =
eu̯ boi̯i̯éu̯ s 'Euboian' (from Εὔβοια = éu̯ boi̯i̯a 'Euboia'), στιγεύς = stigéu̯ s 'tattooer' (from στίζω = stízdɔː 'I
tattoo'), the -µό- = -mó- used to form deverbal nouns, e.g. βιασµός = biasmós 'violence' (from βιάζω = biázdɔː
'I (use) force'), ὁπλισµός = hoplismós 'arming' (from ὁπλίζω = hoplízdɔː 'I arm'), and the -τέο- = -téo- used to
form deontic verbal adjectives, e.g. γραπτέος = graptéos 'to be written' (from γράφω = gráphɔː 'I write'),
ὠνητέος = ɔːnɛːtéos 'to be bought' (from ὠνέοµαι = ɔːnéomai̯ 'I buy'), etc. However, there are also inherently
accented inflectional suffixes, such as the genitive plural ending -ῶν = -ɔ̂ ːn of 1st declension nouns, which
arose by contraction from -ᾱ́ -ων = -áː-ɔːn.12 Compare the accented ending in nouns such as nom. sg. πόρνη =
pórnɛː 'prostitute', gen. pl. pornɔ̂ ːn with the unaccented ending -ων = -ɔːn of 1st declension feminine
adjectives such as nom. sg. ἄλλη = állɛː 'other', gen. pl. ἄλλων = állɔːn (not **ἀλλῶν = **allɔ̂ ːn). Second, there
are pre-accenting morphemes, which differ from accented morphemes in that they induce an accent on the
preceding syllable. They include the ´-αι = ´-ai̯ used to form aorist active infinitives, e.g. τελέσαι = telésai̯
'complete', ποιῆσαι = poi̯i̯ɛ̂ːsai̯ 'do, make' (with a circumflex by the sōtêra rule) and the ´-σθαι = ´-sthai̯ used to
form perfect medio-passive infinitives, e.g. τετελέσθαι = tetelésthai̯ 'have completed'. Third, there are
inherently unaccented morphemes that adopt the accentual properties of the base form. Most inflectional
endings are of this type, e.g. the gen., dat., and acc. sg. endings -ος, -ι, and -α = -os, -i, and -a, of φυγάδος,
φυγάδι, φυγάδα = phugádos, phugádi, phugáda, and the nom., gen., dat., and acc. endings of φυγάδες, φυγάδων,
φυγάσι(ν), and φυγάδας = phugádes, phugádɔːn, phugási(n), and phugádas. The suffix -θε(ν) = -the(n) that is
used to form ablatival adverbs also has these properties; compare ἀγρόθε(ν) = agróthe(n) 'from the country'
(from ἀγρός = agrós) with ἄλλοθε(ν) = állothe(n) 'from another place' (from ἄλλος = állos 'other'). Fourth,
there are inherently unaccented suffixes that induce recessive accentuation regardless of the accentual
properties of the base. The suffix -(i̯)α = -(i̯)a has those properties,13 e.g. ἀλήθεια = alɛ́ ːthei̯i̯a 'truth' (from
ἀληθής = alɛːthɛ́ ːs 'true'), βασίλεια = basílei̯i̯a 'queen' (from βασιλεύς = basiléu̯ s 'king').

5. THE EVOLUTION OF THE GREEK ACCENTUAL SYSTEM

Examination of linguistic developments within the history of Greek and comparison with related languages,
especially Vedic, allows us to reconstruct the development of the Greek accentual system. Perhaps the most
important trend in the diachronic development of the Greek accentual system involves a trajectory from a

12
The first vowel was originally the stem-final vowel; it didn’t belong to the inflectional ending.
13
The jod (i̯), which was lost in most phonological contexts between the Mycenaean period and the earliest alphabetic attestations
of the language, still has various synchronic reflexes, e.g. the conversion of a stem-final -t- to -ss- in Ionic and to -tt- in Attic, as in
the word for honey-bee (and the source of the proper name), Ionic µέλισσα, Attic µέλιττα, both of which derive from *melit-i̯a (the
synchronic stem of the word for ‘honey’ is µελιτ- = melit-). It is not immediately obvious how to capture this in the synchronic
underlying form. I have opted for (i̯)a.

8
relatively 'free' accent system, where accent was primarily morphologically determined and phonology
played a minor role — like the Vedic system — to a less free, more phonologically constrained system.
Specific developments along this trajectory are the Law of Limitation and several prehistoric and historic
leftward accent shifts which also display sensitivity to the distribution of syllable weight, such as Wheeler's
Law [CrossRef] and Vendryes's Law, which affected Attic only. The development of the Law of Limitation in
Proto- or Common Greek was likely facilitated by the fact that the inherited morphological accent very often
happened to 'obey' the Law of Limitation before it arose (Probert 2012). For example, language learners
could analyse forms such as phérō, phéreis, phérei, phéromes/n, phérete, phéronti, etc. either as being
morphologically accented on the verbal root phér 'carry', or phonologically accented such that the accent
was aligned with a rhythmic constituent such as the word-final foot mentioned above, i.e. phé(ròo),
phé(rète), etc. An analysis of the latter sort — likely facilitated by changes in rhythmic organization and/or
its phonetic expression — produced the Law of Limitation.
In the wake of the accent shifts, speakers made sense of new weight-sensitive accentual differences
within the same word-formation type by innovating morphophonological rules for accent placement. For
example, Wheeler’s Law produced alternations such as ψῡχο-ποµπός = psuːkho-pompós 'soul-escorting' vs.
πατρο-κτόνος = patro-któnos ‘father-killing’ in a compound type which was originally oxytone, to judge by
Vedic, e.g. hasta-grābháḥ ‘hand-grasping’, bhuvana-cyaváḥ ‘world-shaking’. Speakers innovated the
following rule for the formation of these compounds: if the penult is light, accent it (τευχεσ-φόρος =
teu̯ khes-phóros, τευχο-φόρος = teu̯ kho-phóros ‘armor-wearing’); otherwise, accent the ultima (ψῡχο-ποµπός =
psuːkho-pompós). Neuter diminutives in -ιον = -ion reflect a comparable rule: they are usually paroxytone if
the antepenult is heavy and proparoxytone if it is light (Vendryes 1945:166), e.g. θηρίον = thɛːríon 'little beast'
vs. θύριον = thúrion 'little door'.
The Proto-Greek innovation of circumflex accentuation offset this trend slightly, insofar as it
introduced a new kind of accentual freedom — the contrast between acute and circumflex — that survived
in word-final VV-syllables, e.g. gen. sg. φορᾶς = phorâːs vs. nom. pl. φορᾱ́ ς = phoráːs. The circumflex in such
forms, and likely in Greek in general (Jasanoff 2004), arose via the contraction of an accented vowel with a
following unaccented vowel over which the pitch fell again, e.g. gen. sg. (PIE *bhoréh2es >) *phoráas > phorâːs
= φορᾶς, dat. sg. (PIE *bhoréh2ei̯ >) *phoráai̯ > phorâːi̯ = φορᾷ, gen. pl. (PIE *sth2tóhxo(ː)m) > *statóo(ː)m >
statɔ̂ ːn = στατῶν 'placed, standing'; compare the Rigvedic and Avestan metrical evidence for uncontracted
genetive plural forms in *-aā̆ m (cf. Kümmel forthcoming). Speakers apparently found morphological case to

9
be the best predictor of accent type, with the result that circumflex accentuation was analogically extended
to forms that originally had acute accentuation, e.g. PIE dat. sg. *sth2tóːi̯ >> statɔ̂ ːi̯ = στατῷ.14
Regarding the accentual properties of morphemes, Vedic has correlates for the four Greek types
sketched out above, as well as a fifth type of underlyingly accented morpheme that either imposes its accent
on the derivative (cf. Greek -ικό- = -ikó-) or adopts the accent of the base (cf. Greek -θε(ν) = -the(n), which is
unaccented), depending on the accentual properties of the base (cf. 'recessive accented' morphemes in
Kiparsky 2010, forthcoming, with refs.). The Vedic suffix -(m)āná-, the cognate of the Greek medio-passive
participial suffix -meno-, has those properties. Compare suffix-accented śaśamānáḥ 'having labored' with
root-accented yájamānaḥ 'sacrificing'. Which language innovated in this case is a subject for future
investigation.
The complex interplay of phonological, morphological, and lexical factors, the robust attestation of
the language, and a tradition of scholarship on the subject that has its roots in the 2nd c. BCE make Ancient
Greek accentuation a unique subject for constructing and testing linguistic theories as well as for
reconstructing the accentual system of Proto-Indo-European.

REFERENCES
Allen, W. S. 1973. Accent and rhythm. Cambridge.
Devine, A. M. and L. D. Stephens. 1984. Language and metre: Resolution, Porson's Bridge, and their prosodic
basis. Chico.
Devine, A. M. and L. D. Stephens. 1994. The prosody of Greek speech. Oxford.
Golston, C. 1990. "Floating H (and L*) tones in Ancient Greek." In J. Myers and P. E. Perez, eds., Proceedings
of the Arizona Phonology Conference, vol. 3, 66-82. Tucson.
Golston, C. 1991. "Minimal word, minimal affix." In T. Scherer, ed., Proceedings of the North East Linguistic
Society 21, 95-110. Amherst.
Golston, C. and T. Riad. 2000. "The phonology of Classical Greek meter", Linguistics 38.1:1-69.
Golston, C. and T. Riad. 2005. "The phonology of Greek lyric meter", Journal of Linguistics 41:77- 115.
Göttling, K. W. 1835. Allgemeine Lehre vom Accent der griechischen Sprache. Jena.

14
This generalization remained active in the grammar and overrode the phonologically regular vowel contraction products in
nominative and accusative forms that arose after the loss of intervocalic s and i̯, e.g. *pei̯tʰói̯a > *pei̯tʰóa >> pei̯tʰɔ́ ː = πειθώ
‘persuasion’ (fem. acc.).

10
Gunkel, D. 2010. Studies in Greek and Vedic prosody, morphology, and meter. Ph.D. diss., University of
California, Los Angeles.
Gunkel, D. 2011. "The emergence of foot structure as a factor in the formation of Greek verbal nouns in -
µα(τ)-", Münchener Studien zur Sprachwissenschaft 65:77-103.
Jasanoff, J. 2004. "Acute vs. circumflex: Some notes on PIE and post-PIE prosodic phonology." In A. Hyllested
et al., eds., Per Aspera ad Asteriscos. Studia Indogermanica in honorem Jens Elmegård Rasmussen, 247-256.
Innsbruck.
Kiparsky, P. 1973. "The inflectional accent in Indo-European", Language 49:794-849.
Kiparsky, P. 2010. "Compositional vs. paradigmatic approaches to accent and ablaut." In S. W. Jamison, H. C.
Melchert and B. Vine, eds., Proceedings of the 21st Annual UCLA Indo-European Conference. Bremen.
Kiparsky, P. Forthcoming. "Accent and Ablaut."
Kümmel, M. Forthcoming. "Der Genitiv Plural im Indoiranischen", IF.
Olander, T. 2009. Balto-Slavic accentual mobility. Berlin/New York.
Probert, P. 2006. Ancient Greek accentuation: synchronic patterns, frequency effects, and prehistory. Oxford.
Probert, P. 2010. "Ancient Greek accentuation in Generative Phonology and Optimality Theory", Language
and Linguistics Compass 4:1-26.
Probert, P. 2012. "Origins of the Greek law of limitation." In P. Probert and A. Willi, eds., Laws and rules in
Indo-European, 163-181. Oxford.
Sauzet, P. 1989. "L'accent du grec ancien et les relations entre structure métrique et représentation
autosegmentale", Langages 24:81-111.
Steriade, D. 1988. "Greek accent: a case for preserving structure", Linguistic Inquiry 19:271-314.
Vendryes, J. 1945. Traité d'accentuation grecque. Paris.
West, M. L. 1992. Ancient Greek Music. Oxford.
Yates, Anthony. 2015. “Anatolian default accentuation and its diachronic consequences.” Indo-European
Linguistics 3:145–187.

11
 abstract nouns 7
of abstractness (and predicativeness) of their Lehmann, Christian. 2010. “Roots, stems and word classes”.
singular, as (14) shows. In: Parts of speech. Empirical and theoretical advances, ed.
by Umberto Ansaldo, Jan Don and Roland Pfau, 43–64.
Amsterdam – Philadelphia.
(14) (hē) akróāsis ‘hearing’; pl. ‘thing listened to, Lehmann, Winfred P. 1958. “On earlier stages of the Indo-
lectures’ European nominal inflection”, Language 34, 2:179–202.
Luraghi, Silvia. 2009. “The origin of the feminine gender
in PIE. An old problem in a new perspective”. In: Gram-
Secondly, ANs do in general not occur in com- matical change in Indo-European languages, ed. by
pounds. Yet this feature probably depends on the Vit Bubeník, John Hewson and Sarah Rose, 3–14. Amster-
degree of the verbal origin (i.e., ­predicativeness) dam – Philadelphia.
Lyons, John. 1977. Semantics 2. Cambridge.
of such nouns rather than their abstract seman- Rico, Christophe. 2002. “Le suffixe -mos/-mós dans la recher-
tics (cf. Lazzeroni 2010 and Civilleri 2012). For che linguistique, de Ferdinand de Saussure à nos jours”,
instance, -sis nominals (which are the most Emerita 70, 2:305–320.
verbal DNs; cf. Civilleri 2012) are never com- Schmidt, Johannes. 1889. Die Pluralbildungen der indoger-
manischen Neutra. Weimar.
pounded. However, the corresponding -(s)ía Simone, Raffaele. 2003. “Masdar, ‘ismu al-marrati et la fron-
nominals undergo composition according to the tière verbe/nom”. In: Estudios ofrecidos al profesor J. Bus-
pattern shown in (15). tos de Tovar, ed. by José Luis Girón Alconchel, 901–918.
Madrid.
Simone, Raffaele and Anna Pompei. 2007. “Traits verbaux
(15) (hē) héxis (= hek-sis) ‘having, possession’; dans les formes nominalisées du verbe”, Faits de Langue
compound: pleon-exía ‘advantage’. 30:43–58.
Vogel, Petra M. 2000. “Nominal abstracts and gender in
Bibliography Modern German. A ‘quantitative’ approach towards the
Bader, Françoise. 1974. Suffixes grecs en -m-: recherches com- function of gender”. In: Gender in grammar and cognition:
paratives sur l’hétéroclisie nominale. Paris. I. Approaches to gender, II. Manifestations of gender, ed. by
Benveniste, Émile. 1948. Noms d’agent et noms d’action en Barbara Unterbeck, Matti Rissannen, Terttu Nevalainen
indo-européen. Paris. and Mirja Saari, 461–493. Berlin – New York.
Bolelli, Tristano. 1953. “Origine e sviluppo delle formazioni Wodtko, Dagmar S. 2005. “Nomen und Nominalisierung im
greche in men/mon”, ASNP 22:5–74. indogermanischen Lexikon”, IF 110:41–85.
Brugmann, Karl and Berthold Delbrück. 1889. Grundriss
der vergleichenden Grammatik der indogermanischen Germana Olga Civilleri
Sprachen II. Strassburg.
Bruno, Carla. 2000. “Fra nome e verbo: osservazioni sulla
sintassi dei nomi in -sis”, SSL 38:129–167.
Chantraine, Pierre. [1933] 1979. La formation des noms en
Accentuation
grec ancien. Paris.
Civilleri, Germana O. 2012. Nomi deverbali nel continuum 1. The Notational System
nome-verbo: il caso del greco antico. Munich.
Croft, William. 1991. Syntactic categories and grammati- The accent marks written in modern editions
cal relations: the cognitive organization of information.
­Chicago. of Ancient Greek texts derive from a grammati-
Flaux, Nelly, Michel Glatigny and Didier Samain. 1996. Les cal tradition that most likely began in Alexan-
noms abstraits. Histoire et théories. Villeneuve d’Ascq. dria in the early 2nd c. BCE, with Aristophanes
Gagnepain, Jean. 1959. Les noms grecs en -os et en -a. Contri-
bution à l’étude du genre en indo-européen. Paris.
of ­Byzantium, to whom the invention of the
Hamp, Eric P. 1982–1983. “Indo-European substantives in written signs is attributed, and his successor as
*-mó- and *-ma-”, ZVS 96:171–177. ­librarian, Aristarchus of Samothrace (→ Philo-
Holt, Jens. 1940. Les noms d’action en -sis (-tis). Aarhus. logical-Grammatical Tradition in Ancient Lin-
Krasuchin, Konstantin G. 2003. “Quantitativer und qualita-
tiver Ablaut in der Nominalderivation”. In: Indogermanis-
guistics; Metrics (métron), Ancient Theories of).
ches Nomen. Derivation, Flexion und Ablaut, Akten der Given the important functional role that accent
Arbeitstatung der Indogermanischen Gesellschaft vom played in the language, conveying accentual
19. bis 22. September 2001 in Freiburg, ed. by Eva Tichy, information in writing facilitated the difficult
Dagmar S. Wodtko, Britta Irslinger, 83–104. Bremen.
Kuryłowicz, Jerzy. 1956. L’apophonie en indo-européen. task of reading poetic texts written in scriptio
Wrocław. continua. For example, the unaccented graphic
Lallot, Jean. 2008. “De Platon aux grammairiens: regards sequence aponou could represent ap’ ónou ‘from
grecs sur la structure des mots non simples”. In: Regards a donkey’ or, apò noû ‘from (your) senses’ (cf.
croisés sur les mots non simples, ed. by Barbara Kaltz,
51–63. Lyon. Aristoph. Nub. 1273), or apónou ‘without toil
Lazzeroni, Romano. 2010. “Nomi d’agente e composizione: (gen. sg.)’. These lectional signs conveyed the
una scala di nominalità”. In: La morfologia del greco tra accentuation of the Koine spoken during that
tipologia e diacronia, ed. by Ignazio Putzu, Giulio Paulis, period, and to a lesser extent, the accentuation
Gian Franco Nieddu, Pierluigi Cuzzolin, 256–265. Milan.
8 accentuation
of other dialects; it is possible that the schol- of phonetic detail. As in vocal music traditions
ars also consulted oral traditions, e.g. rhapsodic in a number of languages with contrastive tone
performances of the Homeric poems, to deter- (Devine and Stephens 1994:160–171), the frag-
mine the accentuation of forms that were ments display a relatively strict correspondence
unfamiliar to them. In the 2nd c. CE, Aelius between the pitch movements of speech and
Herodian codified the tradition in perì katholikês the melody of the music it is set to (Devine and
prosōidíās ‘On Prosody in General’, which served Stephens 1994:172–194; Probert 2006:47–48; West
as the basis for later works on accentuation. 1992:199). For example, the accented syllable of
In the early accented papyri, which date from a word is set no lower than its unaccented syl-
the 2nd c. BCE onwards, ­notational conventions lables, such that the pitch peak of a word corre-
vary, as does the frequency with which writ- sponds to a local peak in the music. If a syllable
ten accents were applied. The notational system bearing a circumflex is set to a two-note melism,
familiar to us, where each accented word is the first is usually higher, respecting the falling
marked with an acute, circumflex, or grave, was pitch contour of circumflex accent. The grave
first applied in minuscule manuscripts of the accent, a phrasal → sandhi variant of the acute,
9th c. CE by scribes following the precepts of proves to be a lowered version of the acute that
the same grammatical tradition. The early works nevertheless represents the pitch peak within
on accentuation including Herodian’s do not the word (Devine and Stephens 1994:180–183),
survive as such, but scholia and short treatises and it is possible to reconstruct the accentual
based on them provide us with indirect access. contour of entire words. For example, in a prop-
On the grammatical tradition, the papyri, and aroxytone word of five syllables (e.g. pres. mid.
the manuscripts, see Probert (2006:21–52) and ptc. eklegómenos ‘picked’), pitch rose steadily
references therein. over the initial two unaccented syllables, peaked
at the accented syllable, then fell steeply over
2. Phonetics and Phonology of the first post-accentual syllable and less steeply
the Greek Accentual System over the final syllable (Devine and Stephens
1994:183–189). The turning point between the
Phonetically, accent may be studied from an post-accentual fall in pitch and the rise to the
articulatory, acoustic, and perceptual standpoint. following accent coincides with word-boundary,
It is clear that the most salient perceptual corre- which certainly had a ‘demarcative’ function, i.e.,
late of ancient Greek accent – what the Greeks made word boundaries audible (Allen 1973:246);
themselves ‘heard’ – was → pitch, at least until this may point to a Low word-final boundary
the 2nd c. BCE. Within the word, pitch peaked tone in the phonological representation (Devine
during → syllables marked with an acute or grave and Stephens 1994:180). Statements by grammar-
accent, and it both peaked and fell again dur- ians and other ancient scholars provide a further
ing syllables marked with a circumflex. Phono- source of information about the phonetic nature
logically, we may say that syllables marked with of word-level accentuation (Devine and Stephens
acute accent hosted a High tone, and those with a 1994:171–172; Probert 2003:4–7), and comparison
circumflex hosted a High-Low tone, i.e., a falling with the accentual systems of related languages,
contour tone. Evidence for the phonetic nature especially Vedic and Balto-Slavic, suggests that
of Greek accent comes from several sources. The in Proto-Indo-European, one syllable of each
words used to refer to ‘accent’ have to do with accented word was realized with high pitch (cf.
musical pitch, e.g. tónos refers to the ‘tension’ Olander 2009:53–100 with refs.).
and therefore to the perceived pitch of vibrating The accentuation of a word is determined
strings, and the basic meaning of prosōidíā is by interacting phonological, morphological,
‘singing along (to music)’. The adjectives used to and lexical factors. The → phonology plays two
specify the three different types of prosōidíā are important roles in this system. First, it places
oxeîa ‘high’ for acute, bareîa ‘low’ for grave, and restrictions on which syllables can host an
oxubáreia ‘high-low’ for circumflex. Fragments accent and on what type of accent (acute and/
of non-strophic musical compositions dating or circumflex) can be realized there. The most
from as early as the 3rd c. BCE – the Delphic important of these restrictions, the so-called
hymns in particular – provide a richer source ‘→ Law of Limitation’, essentially sets the accent-
 accentuation 9
able domain of a word, which consists of the
meiduplērē: meiduplērê meiduplērḗ meiduplḗrē
final three syllables if the ultima is light, and
meiduplēros: meiduplērós meiduplêros meidúplēros
the final two if it is heavy (Göttling 1835:21–28;
meiduploros: meiduplorós meiduplóros meidúploros
Steriade 1988:273–275). Note that for the Law
of Limitation, a single word-final consonant is
weightless (is not associated with a → mora); 3. Lexicon, Morphology, Phonology:
final syllables ending in a short → vowel (-V#) Recessive Accentuation
and those ending in a short vowel followed
by a single → consonant (-VC#) both count as Within the bounds set by the phonology, mor-
light (monomoraic), e.g. basíleia ‘queen’, acc. phological and lexical factors determine the
sg. basíleian. All other syllable rhymes count accentuation of a given word (→ Classical Greek
as heavy (bimoraic) (→ Syllable Weight). Within Morphology (Survey); → Greek Lexicon, Struc-
the accentable domain, an acute accent is ture and Origin of). Thus, Greek accent is only
­phonologically permissible on any syllable (with ‘free’ insofar as the accentuation of a word is
one systematic exception noted below), but the not determined by phonological factors alone.
circumflex accent is subject to further restric- This limited freedom is reflected in minimal
tions. It is phonologically licit on final syllables pairs that differ only in position and/or type of
containing a long vowel or → diphthong (i.e., accent, e.g. kḗr ‘doom’ vs. kêr ‘heart’, lithobólos
final VV-syllables), where an acute is also pos- ‘pelting with stones’ vs. lithóbolos ‘pelted with
sible, as reflected by contrasts such as isthmoí stones’.
‘isthmuses’ (nom.) vs. isthmoî ‘on the isthmus’ If every inflectional form of a word is accented
(loc./adv.). The circumflex also occurs on penul- as early (i.e., as far ‘left’) as permitted by the
timate VV-syllables, where it is in complemen‑ Law of Limitation, that word is said to exhibit
tary distribution with the acute according to ‘recessive accentuation’, e.g. ánthrōpos ‘(hu)
the so-called ‘sōtêra rule’: the accent on a pen- man’, gen. sg. anthrṓpou, dat. sg. anthrṓpōi, acc.
ultimate VV-syllable is realized as a circumflex sg. ánthrōpon, etc. In this accentual subtype,
if the word-final syllable contains a short vowel, we observe the second role of phonology inter-
e.g. sōtêra ‘savior’ (acc.), oîkos ‘house’; otherwise, acting with lexical and morphological factors.
it is realized as an acute, e.g. sōtḗrōn ‘saviors’ Recessive accentuation is both the property of
(gen.), oíkois ‘houses’ (dat.). In other words, if the particular lexical items, such as ánthrōpos, and
penult is accented, the phonology determines the property of entire morphologically circum-
which type of accent is realized there, meaning scribed classes of words, such as finite verbs,
that phonologically, a contrast in accent type 3rd-declension neuter nouns, and most types of
is limited to word-final VV-syllables. Together, → compounds, including those whose first mem-
these phonological restrictions permit only the ber is a governing preposition or verb (Kipar-
five combinations of accent location and type sky 2003; Vendryes 1945:189–196), e.g. phil(o)-X
already recognized by the ancient grammarians: ‘X-loving’ compounds such as phíloinos ‘wine-
loving’, philósophos ‘wisdom-loving’, philópais
– Oxytone (oxútonos): acute on the ultima, e.g. ‘boy-loving’, philórtux ‘quail-loving’, philospêlunx
ophthalmoí ‘eyes’ ‘cave-loving’, etc. In short, lexical and/or mor-
– Perispomenon (perispṓmenos): circumflex on phological features determine whether a word is
the ultima, e.g. ophthalmôn ‘eyes’ (gen.) recessively accented; the phonology determines
– Paroxytone (paroxútonos): acute on the the accentable domain and locates the accent
penult, e.g. sōtḗrōn ‘saviors’ (gen.) ‘leftmost’ in that domain.
– Properispomenon (properispṓmenos): circum- A number of facts suggest that recessive
flex on the penult, e.g. sōtêra ‘savior’ (acc.) accentuation was the unmarked or default type
– Proparoxytone (proparoxútonos): acute on of accentuation in the language (cf. Probert
the antepenult, e.g. hḗlios ‘sun’ 2006:128–144): among accented words, recessive
accentuation is more frequent than ­non-recessive
However, the maximum number of ways any accentuation by both type and token; compari-
given form may be accented is three, as exempli- son with Vedic and Germanic points to a ten-
fied by the following nonsense words: dency within the history of Greek to innovate
recessive accentuation in inherited lexical items,
10 accentuation
e.g. páros ‘formerly’ vs. Vedic purás ‘before’ < PIE consonant (-VC#). This is apparent in recessively
*pṛhₓós or *pṛhₓés; entire morphological classes accented paradigms where, for example, nom.
of words (noted above) are recessively accented pl. philósophoi ‘philosophers’, basíleiai ‘queens’,
in Greek, but no such class is associated with a and 2 sg. aor. imp. mid. paídeusai ‘educate’ are
non-recessive type of accentuation; finally, in proparoxytone like nom. sg. philósophos, acc.
Lesbian, recessive accentuation was generalized sg. basíleian, and 2 sg. aor. imp. act. paídeuson,
to virtually all accented words, arguably due to which end in -VC#. It is also apparent in forms
an extreme form of the tendency just noted. where the sōtêra rule applies. For example, nom.
A central question regarding the Law of Limi- pl. oîkoi ‘houses’, gaîai ‘lands’, and aor. inf. act.
tation and recessive accentuation is whether the paideûsai are properispomenon like nom. sg.
accentable domain is related to the rhythmic oîkos ‘house’, acc. sg. gaîan ‘land’, and neut.
phonology of the language in general (cf. Devine nom./acc./voc. sg. aor. act. ptc. paideûsan, which
and Stephens 1994:154). In other words, can the end in -VC#. The inflectional endings -oi and -ai
accentable domain be equated with a rhyth- of the 3 sg. present and aorist optative active
mic/prosodic constituent that is also reflected and the locative singular – or adverbial locative
in meter, word formation, and other (morpho) – ending -oi pose morphologically conditioned
phonological processes? Building on Steriade exceptions. Like all other word-final long vowels
(1988), recent studies suggest that the accentual and diphthongs, they have the status of heavy
domain is – or is aligned with – a unit of rhyth- -VV# rhymes in the accentual system. This is
mic organization known as a → foot (cf. Probert likewise reflected in recessive paradigms where,
2010 with refs.). The span between the accent (´) for example, 3 sg. pres. and aor. opt. act. paid-
and word-end (#) consists of two light syllables eúoi and paideúsai are paroxytone like paideúō
(LL), e.g. heurḗmata#, a heavy syllable (H), e.g. ‘I am educating’, and where the sōtêra rule fails
heurēmátōn#, or a heavy-light sequence (HL), to apply: loc. sg. oíkoi ‘at home’ is paroxytone
e.g. heúrēma# – in other words, ´LL# or ´H(L)#. like dat. sg. oíkōi. There is no evidence that this
This span has been equated with a word-final morphologically conditioned phonological dis-
quantity-insensitive trochaic foot (Sauzet 1989) tinction between diphthongs existed outside the
and a quantity-sensitive one (Golston 1990). The system of accentuation (Probert 2012).
latter, a bimoraic rhythmic unit consisting of The distribution of acute and circumflex
either two light syllables (LL) or one heavy syl- accents on word-final syllables that are pho-
lable (H) may also be reflected in Greek word nologically ‘free’ to host either accent is con-
formation (Gunkel 2011), meter (Golston & Riad ditioned by morphological features as well.
2000; 2005; Gunkel 2010:43–75), and constraints Specifically, the distribution appears to be based
on minimal word size, alias word minima on case: nominative and accusative forms bear
(Devine and Stephens 1994:93; Golston 1991). On an acute accent, genitive and dative forms bear
that analysis, in recessively accented words, the circumflex accents, e.g. 1st declension nom. and
beginning of the post-accentual fall in pitch (`) is acc. sg. phorā́ ‘carrying, bearing; load, burden’,
aligned with the first mora of the word-final foot, phorā́n, pl. phoraí, phorā́s vs. gen. and dat. sg.
e.g. heurḗ(màta) and – representing the bimo- phorâs, phorâi, pl. phorôn, phorâis.
raic long vowel ō as oo – heurēmá(tòon). For Morphemes themselves have accentual
indispensable in-depth treatments of Ancient properties (Kiparsky 1973, 2010, forthcoming;
Greek rhythmic organization including alterna- Probert 2006:145–148; Steriade 1988). Descrip-
tive views on foot structure, cf. Allen (1973) and tively, there are four types in Ancient Greek.
Devine and Stephens (1984, 1994). First, there are inherently accented morphemes
such as the stem agró- of agrós ‘farm, country’
4. Further Morphological Factors and the suffixes -ád- and -ikó- of words such as
gen. sg. phugádos ‘exile’, manikós ‘mad’. The
Morphological features also condition the pho- last inherently accented morpheme imposes its
nological status of the word-final diphthongs accent on the entire derivative, e.g. phugád-ikó-s
-oi and -ai in the accentual system. For both the phugadikós ‘of/for exile’ (not *phugádikos). Most
Law of Limitation and the sōtêra rule, word-final inherently accented morphemes are stems or
-oi and -ai have the status of a light word-final derivational suffixes (→ Derivational Morphol-
syllable rhyme consisting of a short vowel plus a ogy) such as the -eú- used to form nouns of
 accentuation 11
occupational/ethnic appurtenance and agent system involves a trajectory from a relatively
nouns, e.g. khalkeús ‘bronzesmith’ (derived from ‘free’ accentual system, where accent was primar-
khalkós ‘bronze’), Euboieús ‘Euboean’ (from ily morphologically determined and phonology
Eúboia ‘Euboea’), stigeús ‘tattooer’ (from stízō played a minor role – like the Vedic system –
‘I tattoo’), the -mó- used to form deverbal nouns, to a less free, more phonologically constrained
e.g. biasmós ‘violence’ (from biázō ‘I (use) force’), system. Specific developments along this trajec-
hoplismós ‘arming’ (from hoplízō ‘I arm’), and the tory are the Law of Limitation and several pre-
-téo- used to form deontic verbal adjectives, e.g. historic and historic leftward accent shifts which
graptéos ‘to be written’ (from gráphō ‘I write’), also display sensitivity to the distribution of syl-
ōnētéos ‘to be bought’ (from ōnéomai ‘I buy’), lable weight, such as → Wheeler’s Law, which
etc. However, there are also inherently accented was apparently pan-dialectal, and → Vendryes’
inflectional suffixes, such as the genitive plu- Law, which affected → Attic only. The develop-
ral ending -ôn of 1st declension nouns, which ment of the Law of Limitation in Proto- or Com-
arose by ­contraction from ā́-ōn. Compare the mon Greek was likely facilitated by the fact that
accented ending in nouns such as nom. sg. pórnē the inherited morphological accent very often
‘prostitute’, gen. pl. pornôn with the unaccented happened to ‘obey’ the Law of Limitation before
ending -ōn of 1st declension feminine adjectives it arose (Probert 2012). For example, language
such as nom. sg. állē ‘other’, gen. pl. állōn (not learners could analyze forms such as phérō,
*allôn). Second, there are pre-accenting mor- phéreis, phérei, phéromes/n, phérete, phéronti,
phemes, which differ from accented morphemes etc. either as being morphologically accented
in that they induce an accent on the preced- on the verbal root phér ‘carry’, or phonologically
ing syllable. They include the ´-ai used to form accented, such that the accent was aligned with
aorist active infinitives, e.g. telésai ‘complete’, a rhythmic constituent such as the word-final
poiêsai ‘do, make’ (with a circumflex by the foot mentioned above, i.e., phé(ròo), phé(rète),
sōtêra rule) and the ´-sthai used to form perfect etc. An analysis of the latter sort – likely facili-
medio-passive infinitives, e.g. tetelésthai ‘have tated by changes in rhythmic organization and/
completed’. Third, there are inherently unac- or its phonetic expression – produced the Law
cented morphemes that adopt the accentual of Limitation.
properties of the base form. Most inflectional In the wake of the accent shifts, speakers
endings are of this type, e.g. the gen., dat., and made sense of new weight-sensitive accentual
acc. sg. endings -os, -i, and -a, of phugádos, differences within the same word-formation
phugádi, phugáda, and the nom., gen., dat., and type by innovating morphophonological rules for
acc. endings of phugádes, phugádōn, phugási(n), accent placement. For example, Wheeler’s Law
and phugádas. The suffix -the(n) that is used produced alternations such as psūkho‑pompós
to form ablatival adverbs also has these prop- ‘soul‑escorting’ vs. patro‑któnos ‘father-killing’
erties; compare agróthe(n) ‘from the country’ in a compound type which was originally oxy-
(from agrós) with állothe(n) ‘from another place’ tone, to judge by Vedic, e.g. hasta‑grābháḥ
(from állos ‘other’). Fourth, there are inherently ­‘hand-grasping’, bhuvana‑cyaváḥ ­‘world-shaking’.
unaccented suffixes that induce recessive accen- Speakers innovated the following rule for the
tuation regardless of the accentual properties of formation of these compounds: if the penult is
the base. The suffix -(i̯)a has those properties, light, accent it (teukhes‑phóros, teukho-phóros
e.g. alḗtheia ‘truth’ (from alēthḗs ‘true’), basíleia ‘armor-wearing’); otherwise, accent the ultima
‘queen’ (from basileús ‘king’). (psūkho‑pompós). Neuter diminutives in -ion
reflect a comparable rule: they are usually par-
5. The Evolution of the Greek oxytone if the antepenult is heavy and prop-
Accentual System aroxytone if it is light (Vendryes 1945:166), e.g.
thēríon ‘little beast’ vs. thúrion ‘little door’.
Examination of linguistic developments within The Proto-Greek innovation of circumflex
the history of Greek and comparison with related accentuation offset this trend slightly, insofar as
languages, especially Vedic, allows us to recon- it introduced a new kind of accentual freedom –
struct the development of the Greek accentual the contrast between acute and circumflex – in
system. Perhaps the most important trend in the word-final VV-syllables, e.g. gen. sg. phorâs vs.
diachronic development of the Greek accentual acc. pl. phorā́s. The circumflex in such forms,
12 accentuation
and likely in Greek in general ( Jasanoff 2004), Gunkel, Dieter. 2010. Studies in Greek and Vedic prosody,
arose via the contraction of an accented vowel morphology, and meter. Ph.D. diss., University of Califor-
nia, Los Angeles.
with a following unaccented vowel over which ——. 2011. “The emergence of foot structure as a factor in the
the pitch fell again, e.g. gen. sg. (PIE *bʰoréh₂es >) formation of Greek verbal nouns in -μα(τ)-”, Münchener
*pʰoráas > phorâs, dat. sg. (PIE *bʰoréh₂ei̯ >) Studien zur Sprachwissenschaft 65:77–103.
*pʰoráai̯ > phorâi, gen. pl. (PIE *sth₂tóhₓom) > Jasanoff, Jay. 2004. “Acute vs. circumflex: some notes on
PIE and post-PIE prosodic phonology”. In: Per aspera
*statóom > statôn ‘placed, standing’; compare ad asteriscos. Studia Indogermanica in honorem Jens
the Rigvedic and Avestan metrical evidence for Elmegård Rasmussen, ed. by A. Hyllested et al., 247–256.
uncontracted genitive plural forms in *-aā̆m Innsbruck.
Kiparsky, Paul. 1973. “The inflectional accent in Indo-
(cf. Kümmel forthcoming). Speakers appar- European”, Language 49:794–849.
ently found morphological case to be the best ——. 2010. “Compositional vs. paradigmatic approaches to
­predictor of accent type, with the result that cir- accent and ablaut”. In: Proceedings of the 21st Annual
cumflex accentuation was analogically extended UCLA Indo-European Conference, ed. by Stephanie W.
Jamison, H. Craig Melchert, and Brent Vine, 137–181.
to forms that originally had acute accentuation, Bremen.
e.g. PIE dat. sg. *sth₂tṓi̯ >> statôi (→ Analogy). Kiparsky, Paul. Forthcoming. “Accent and ablaut”.
Regarding the accentual properties of mor- Kümmel, Martin. Forthcoming. “Der Genitiv Plural im
phemes, Vedic has correlates for the four Greek Indoiranischen”, Indogermanische Forschungen.
Olander, Thomas. 2009. Balto-Slavic accentual mobility.
types sketched out above, as well as a fifth type Berlin – New York.
of underlyingly accented morpheme that either Probert, Philomen. 2006. Ancient Greek accentuation: syn-
imposes its accent on the derivative, like Greek chronic patterns, frequency effects, and prehistory. Oxford.
-ikó-, or adopts the accent of the base, like Greek ——. 2010. “Ancient Greek accentuation in Generative Pho-
nology and Optimality Theory”, Language and Linguistics
-the(n), depending on the accentual properties Compass 4:1–26.
of the base (cf. ‘recessive accented’ morphemes ——. 2012. “Origins of the Greek law of limitation”. In: Laws
in Kiparsky 2010, forthcoming, with refs.). The and rules in Indo-European, ed. by P. Probert and A. Willi,
Vedic suffix -(m)āná-, the cognate of the Greek 163–181. Oxford.
Sauzet, Paul. 1989. “L’accent du grec ancien et les relations
medio-passive participial suffix -meno-, has those entre structure métrique et représentation autosegmen-
properties. Compare suffix-accented śaśamānáḥ tale”, Langages 24:81–111.
‘having labored’ with root-accented yájamānaḥ Steriade, Donca. 1988. “Greek accent: a case for preserving
structure”, Linguistic Inquiry 19:271–314.
‘sacrificing’. Which language innovated in this Vendryes, Joseph. 1945. Traité d’accentuation grecque. Paris.
case is a subject for future investigation. West, M. L. 1992. Ancient Greek music. Oxford.
The complex interplay of phonological,
morphological, and lexical factors, the robust Dieter Gunkel
attestation of the language, and a tradition of
scholarship on the subject that has its roots in
the 2nd c. BCE make Ancient Greek accentuation Accommodation
a unique subject for constructing and testing lin-
guistic theories as well as for reconstructing the Language accommodation (LA) is the ten-
accentual system of Proto-Indo-European. dency of a speaker to modify linguistic behavior
­according to interlocutor characteristics. “Lan-
Bibliography guage Accommodation Theory” was developed
Allen, W. S. 1973. Accent and rhythm. Cambridge. in the 1970s (Giles 1979) in the area of social
Devine, A. M. and Laurence D. Stephens. 1984. Language psychology and is based on the assumption that
and metre: resolution, Porson’s Bridge, and their prosodic
basis. Chico, CA.
speakers are motivated to adjust their speech
——. 1994. The prosody of Greek speech. Oxford. style, or accommodate it, to express their atti-
Golston, Chris. 1990. “Floating H (and L*) tones in Ancient tude to others. The motivation for accommoda-
Greek”. In: Proceedings of the Arizona Phonology Conference, tion lies in the (unconscious?) desire of speakers
vol. 3, ed. by J. Myers and P. E. Perez, 66–82. Tucson, AZ.
——. 1991. “Minimal word, minimal affix”. In: Proceedings of to associate themselves with (positive LA), or
the North East Linguistic Society, vol. 21, ed. by T. Scherer, keep themselves apart from (negative LA), given
95–110. Amherst, MA. social groups. In order to have one of the two
Golston, Chris and Tomas Riad. 2000. “The phonology of typologies of LA, speakers should possess dif-
Classical Greek meter”, Linguistics 38.1:1–69.
——. 2005. “The phonology of Greek lyric meter”, Journal of ferent languages and/or social and regional dia-
Linguistics 41:77–115. lects → (Dialects, Classification of). In order to
Göttling, Karl W. 1835. Allgemeine Lehre vom Accent der predict instances of LA, a model based on the
griechischen Sprache. Jena.