Epic Rhythm:

Metrical Shapes in Greek Hexameter

Stephen A. Sansom and David Fifield

I
N THE Homeric Hymn to Hermes, the newborn god has just been
caught. When Apollo sees the skins of his cows hanging in
Hermes’ cave—sure signs of his little brother’s audacity—he
tries to bind Hermes and vent his anger (403–416). But Hermes
takes out an instrument of his own invention, the kithara, and
plays so pleasantly (ἐρατὸν κιθαρίζων 423) and so orderly (κατὰ
κόσµον 433) that Apollo relents and pledges eternal friendship
for knowledge of the instrument (436–462). Yet it is not only the
verbal descriptions of the kithara’s “lovely sound” (ἐρατὴ …
ἰωή 421) and soporific effect (ἥδυµον ὕπνον ἑλέσθαι 448–449) that
should alert us to the quality of Hermes’ song. From when Her-
mes starts to play, well into Apollo’s response (409b–441), nearly
every metrical shape is in place. Every dactyl (‒ ⏑ ⏑, as in µήτερα,
“mother” [430]), spondee (‒ ‒, such as υἱόν, “son” [430], and
the appositive group ἣ γάρ, “for she” [430]), pyrrhus (⏑ ⏑, λάχε,
“obtain” [428]), choriamb (‒ ⏑ ⏑ ‒, ἀθανάτους, “immortals”
[431]), almost every shape is in a more expected, or what O’Neill
calls “localized,” metrical position.1 It is as if Hermes “soothed”

1 E. O’Neill, Jr., “The Localization of Metrical Word-Types in the Greek

Hexameter: Homer, Hesiod, Alexandrians,” YCS 8 (1942) 102–176. The one
exception is the shape – ⏑ ⏑ – – of the appositive group at the beginning of
Hermes’ catalog of divinities, Μνηµοσύνην µέν (“Memory” 429), which is very
unexpected in its position. By expected, we mean not among the least
expected ≈ 3.30% of shapes in their positions according to a definition we
will develop in §1. Our definition of “metrical shape” is identical to that of
O’Neill’s “word-type”; for terminological distinctions see §1. The mean num-
ber of unexpected shapes per 181 metrical shapes (i.e., the number of words

—————
Greek, Roman, and Byzantine Studies 64 (2024) 350–377
ISSN 2159-3159
Article copyright held by the author(s) and made available under the
Creative Commons Attribution License
CC-BY https://creativecommons.org/licenses/by/4.0/
 EPIC RHYTHM 351

(ἐπρήϋνεν 417) Apollo with familiar rhythms, a lullaby (449),

each shape placed kata kosmon (433).2 Here, at least, the meaning
of the story can be felt through the rhythm of the poetry.
This article is a preliminary investigation into the expectancy
of metrical shapes in the rhythm of Greek epic (seventh century
BCE to fifth century CE). It argues that a new measure of
metrical regularity according to expectancy enables readers to
compare the distribution of metrical shapes and locate passages
within texts with unusual metrical tendencies, such as Hermes’
song above. Its source data are the metrical shapes, positions,
and basic statistics of the roughly half-million words and ap-
positive groups (for example a preposition and its object) from
Homer’s Iliad to Nonnus’ Dionysiaca.3 From these, it provides
new stylistic information about the distribution of metrical
shapes, especially when they occur at unexpected positions, in
texts, authors, and characters, and how they may affect the
experience of the poetry. This investigation opens possibilities
for future research into the ways in which rhythm engages

or appositive groups from δεσµά [“bonds” 409b] to δῶρον [“gift” 442]), is

7.13 in the Archaic corpus; the percentage of windows of 181 shapes that
have 0 or 1 unexpected shapes is 0.72%.
2 For the kosmos of the hymn see O. Thomas, The Homeric Hymn to Hermes

(Cambridge 2020) 48.

3 S. A. Sansom and D. Fifield, “SEDES: Metrical Position in Greek

Hexameter,” DHQ 17.2 (2023). This study thus includes the smallest two
prosodic domains that constitute complete metrical shapes: 1) the phonetic
word and clitic group (i.e. enclitic and lexeme) and 2) the appositive group
(lexical word with one or more nonlexical, e.g. particles), cf. A. Devine and
L. Stephens, The Prosody of Greek Speech (Oxford 1994) 303–349, and, most
recently, A. Blankenborg, Audible Punctuation: Performative Pause in Homeric
Prosody (Washington 2022) 71–73, and “Mark the Words: Early Music’s
Representation in Writing,” Greek and Roman Musical Studies 11 (2023) 13–14;
C. Bozzone, Homer’s Living Language: Formularity, Dialect, and Creativity in Oral-
Traditional Poetry (Cambridge 2024) 92. Because rhythmic expectancy arises
from what the audience likely had heard, prosodic apposition is a necessary
factor in the assessment of rhythmic expectations; cf. e.g. S. Hagel, “How is
Technology Useful in the Study of Ancient Music,” Greek and Roman Musical
Studies 10 (2022) 269–289, at 284. Appositive words are taken, with some
additions, from S. Hagel, “Tables Beyond O’Neill,” in F. Spaltenstein et al.
(eds.), Autour de la césure (Bern 2004) 135–215, at 137.

—————
Greek, Roman, and Byzantine Studies 64 (2024) 350–377
352 STEPHEN A. SANSOM AND DAVID FIFIELD

expectations, including the style of authors, mimesis of generic

models, rhetoric of speakers, and the kosmos of poetry.
Rhythm is an essential characteristic of epic poetry.4 The
“omnipresent,” “immutable pattern of sound” gives epic a
distinct identity and pleasure of “immersion.”5 Epic rhythm
comes out of a long tradition of oral performance, where
rhythmic units emerged from phrasal patterns and converged
into the “special language” of verbal art.6 Comparative studies

4 Meter is the “science of measuring verses,” whereas rhythm is “simple

verbal diction, based on the quantitative duration of the syllables,” B. Gentili

and L. Lomiento, Metrics and Rhythmics: History of Poetic Forms in Ancient Greece
(Pisa 2008) 27–28; cf. Arist. Rh. 1408b; S. Liebhaber, “Rhythm and Beat: Re-
evaluating Arabic Prosody in the Light of Mahri Oral Poetry,” JSS 55 (2010)
163–182, at 168.
5 E. Wilson, “I Began with Sound,” Public Books (2023) https://www.

publicbooks.org/i-began-with-sound/, accessed 5 December 2023.

6 G. Nagy “Language and Meter,” in E. J. Bakker (ed.), A Companion to the

Ancient Greek Language (Malden 2010) 370–387, at 370; cf. Arist. Poet. 1449a
(ἐκβαίνοντες τῆς λεκτικῆς ἁρµονίας). The bibliography on rhythm in Greek
poetry is vast; representative arguments can be found in D. S. Raven, Greek
Metre (London 1962) 45; W. S. Allen, Accent and Rhythm. Prosodic Features of Latin
and Greek: A Study in Theory and Reconstruction (Cambridge 1972) 158; B. Pea-
body, The Winged Words (Albany 1975) 31; G. Nagy, “Metrical Convergences
and Divergences in Early Greek Poetry and Song,” Historical Philology: Greek,
Latin, and Romance 87 (1992) 151–185; Devine and Stephens, Prosody 85–156,
esp. 99–116; Gentili and Lomiento, Metrics and Rhythmics 262–266; P.
Kiparsky, “Indo-European Origins of the Greek Hexameter,” in D. Gunkel
et al. (eds.), Sprache und Metrik (Leiden 2018) 77–127; E. Bakker, “Learning the
Epic Formula,” in C. Reitz et al. (eds.), Structures of Epic Poetry I (Berlin 2019)
81–98. For rhythm and formularity see e.g. A. Lord, The Singer of Tales
(Cambridge [Mass.] 1960) 47; A. T. Edwards, “ΚΛΕΟΣ ΑΦΘΙΤΟΝ and Oral
Theory,” CQ 38 (1988) 25–30, at 29; M. P. Parry, The Making of Homeric Verse:
The Collected Papers of Milman Parry (Oxford 1971 [1923]) 427 (cf. xxv), “[The
repeated words and phrases] are like a rhythmic motif in the accompaniment
of a musical composition, strong and lovely, regularly recurring, while the
theme may change to a tone of passion or quiet, of discontent, of gladness or
grandeur.” Cf. recent work in language acquisition, e.g. S. Fujii and C. Y.
Wan, “The Role of Rhythm in Speech and Language Rehabilitation: the
SEP Hypothesis,” Frontiers in Human Neuroscience 8.777 (2014) 1–15, at 10;
A. Langus et al., “Rhythm in Language Acquisition,” Neuroscience and Bio-
behavioral Reviews 81 (2017) 158–166.
—————
Greek, Roman, and Byzantine Studies 64 (2024) 350–377
 EPIC RHYTHM 353

in oral poetics suggest that rhythmic patterns can serve literary

functions, such as to characterize dramatic passages.7 And con-
temporary theorists of transhistorical poetics have explored how
the rhythms of metrical verse “afford an organizing of temporal
experience … in the moment of reading,” coordinating aesthetic
and social regularity.8
Metrical shapes contribute to the rhythm of epic in what is
known as the “inner metric” of hexameter.9 As O’Neill has
shown, metrical shapes frequent or “localize in” only a few of all
possible metrical positions.10 Porter argues that this regularity
generates “patterns of expectancy” that poets can “distort” to
create “tension between what is expected and what is actually
spoken.”11 In other words, each word or word group has a
combination of long and short syllables, that is, a quantitative
“shape”; the regular occurrence of these shapes in positions of

7 G. Herzog, “The Music of Yugoslav Heroic Epic Folk Poetry,” Journal of

the International Folk Music Council 3 (1951) 62–64; cf. H. Saussy, The Ethnography
of Rhythm: Orality and its Technologies (New York 2016) 33.
8 C. Levine, Forms: Whole, Rhythm, Hierarchy, Network (Princeton 2015) 79; cf.

V. Barletta, Rhythm: Form and Disruption (Chicago 2020). For application to

Greek epic see M. Ward, “ΓΑΜΕΣΣΕΤΑΙ/ΓΕ ΜΑΣΣΕΤΑΙ: Homer Iliad 9.394
and the Constitutive Role of Irregularity,” JHS 141 (2021) 224–240. For
earlier “formalist” positions see H. N. Porter, “The Early Greek Hexameter,”
YCS 12 (1951) 3–63, at 7–8.
9 O’Neill, YCS 8 (1942) 105 n.2, defines the “outer metric” as the six feet

of hexameter in combinations of dactyls (‒ ⏑ ⏑) or spondees (‒ ‒) and the

“inner metric” as “purely metrical limitation on this composition,” such as
Hermann’s Bridge. For the rhythm of larger narrative and traditional units,
such as type scenes of sacrifice within a nostos (“return”) story, see J. M. Foley,
The Singer of Tales in Performance (Bloomington 1995) 175–176; for the cor-
respondences of rhythms of epic poetry and the physiological rhythms of the
human body, Parry, Homeric Verse xxxiv, of thought, lxii. For rhythm as a
product of quantity see e.g. K. Witte, “Wortrhythmus bei Homer,” RhM 70
(1915) 481–523, and Parry, The Making of Homeric Verse 93. For Platonic views
of rhythm as “a configuration of movement organized over time” see
E. Benveniste, “The Notion of ‘Rhythm’ in Its Linguistic Expression,” in
Problems in General Linguistics (Miami 1971) 281–288, at 287; cf. Barletta,
Rhythm 30.
10 Cf. Hagel, in Autour de la césure 135–215.

11 Porter, YCS 12 (1951) 8–9.

—————
Greek, Roman, and Byzantine Studies 64 (2024) 350–377
354 STEPHEN A. SANSOM AND DAVID FIFIELD

the hexameter line forms a verbal rhythm that becomes ex-

pected by an audience and offers poets rhythmic material to
manipulate.12 Musicologists such as David Huron have found
similar tension in musical rhythms, where rhythmic patterns
imprint on a listener’s mind through “statistical learning” and
guide aspects of their attention:
the evidence strongly suggests a form of statistical learning in
which listeners perceive the world in a manner that corresponds
to their past listening experience … Listeners appear to be sensi-
tive to the frequency of occurrence of various rhythmic patterns,
and their cognitive processing of rhythmic information is disposed
to interpret stimuli in terms of familiar preexisting rhythms.13
By disrupting learned patterns of rhythm and sound, un-
expected sounds can trigger a variety of affective states in the
listener, such as increased attention, or surprise and its physical
manifestations in gasps of awe, chills of frisson, and automatic
laughter.14 For readers of epic, awareness of the rhythmic char-

12 Thus when this paper uses the term rhythm it primarily means the
syllabic quantities of words and word groups that are matched, or “textset,”
to the overarching meter of the hexameter line, cf. Devine and Stephens,
Prosody 101. For “textsetting” see Bozzone, Living Language 79–80 (with bibli-
ography), and R. Jackendoff, “Rhythmic Structure in Music and Language,”
in P. Kiparsky et al. (eds.), Phonetics and Phonology I Rhythm and Meter (San Diego
1989) 36–37. For notions of verbal rhythm in relation to music, see Blanken-
borg, Greek and Roman Musical Studies 11 (2023) 7 and 10.
13 D. Huron, Sweet Anticipation: Music and the Psychology of Expectation (Cam-

bridge [Mass.] 2006) 190–194; cf. M. R. Pearce, “Statistical Learning and

Probabilistic Prediction in Music Cognition: Mechanisms of Stylistic En-
culturation,” Annals of the New York Academy of Sciences 1423 (2018) 378–395.
For rhythm in poetry and music see e.g. D. Attridge, “Rhythm,” in Princeton
Encyclopedia of Poetry and Poetics4 (Princeton 2012) 1195–1198.
14 J. A. Slaboda, “Music Structure and Emotional Response: Some

Empirical Findings,” Psychology of Music 19 (1991) 110–120, at 113, and

“Empirical Studies of Emotional Response to Music,” in M. R. Jones et al.
(eds.), Cognitive Bases of Musical Communication (Washington 1992) 33–50;
Huron, Sweet Anticipation 283. For the relation of Huron’s “statistical learning”
in music to O’Neill’s “localization” see C. W. Forestall and W. J. Scheirer,
Quantitative Intertextuality: Analyzing the Markers of Information Reuse (Cham
2019) 39. For the qualia of music, or subjective feelings resulting from sta-
tistical properties of a stimulus, see Huron 362–363.
—————
Greek, Roman, and Byzantine Studies 64 (2024) 350–377
 EPIC RHYTHM 355

acter of a passage can provide not only insight into the potential
cognitive/attentive experience of an early audience but also new
ways to approach affect in and of texts, such as Apollo’s unusual
laugh in response to the “startling” sound of Hermes’ kithara
(Hom. Hymn Hermes 420–421).15
But O’Neill’s breakthrough, while fundamental to our knowl-
edge of regularity in epos, has been limited in its usefulness for
reading texts. One problem is its measurement. Even with
counts and percentages for a given metrical shape, it is often
unclear what metrical positions are in fact unexpected for it,
since shapes may localize in various positions to various degrees;
percentages alone prove insufficient for comparison. Another is
the difficulty of its application. Although looking up the statistics
in the published charts for one metrical shape is easy enough, it
soon becomes laborious to apply the information from printed
tables of statistics to every word in a line, let alone all words in a
passage, poem, or poetic corpus. Both problems have made it
difficult to tell whether shapes, either expected or unexpected,
are characteristic of books, speakers, or of epic as a genre, and
to compare the expectancy of shapes of a passage in situ. Previous
efforts to integrate this metrical information into commentary
and interpretation have thus been limited, and there is much
that remains unknown about the basic rhythm of metrical shapes
in passages of poetry.16
In what follows, we attempt to increase the utility of metrical
shapes for reading passages by analyzing their stylistics in three
respects. After (§1) updating previous methodologies, we provide
(§2) new stylometric information about the density of un-
expected shapes in particular authors, poems, and books and

15 Cf. A. Becker, “Rhythm in a Sinuous Stanza: The Anatomy and

Acoustic Contour of the Latin Alcaic,” AJP 133 (2012) 117–152, at 141 on
the capacity of poetic lines to “delight, arrest, and impel the reader’s ear”
based on statistical deviation. For Apollo’s laughter as a response to sound
see S. Halliwell, Greek Laughter: A Study of Cultural Psychology from Homer to Early
Christianity (Cambridge 2008) 15 n.35; cf. A. Vergados, The Homeric Hymn to
Hermes (Berlin 2013) 503, and Thomas, The Homeric Hymn to Hermes 376.
16 For an example of the application of the data from Hagel’s extension of

O’Neill’s charts see Thomas, The Homeric Hymn to Hermes 145 and 179 n.94.

—————
Greek, Roman, and Byzantine Studies 64 (2024) 350–377
356 STEPHEN A. SANSOM AND DAVID FIFIELD

(§3) a brief case study of the expectancy of shapes in the Homeric

Hymn to Hermes before (§4) concluding. It is now possible to pro-
duce statistical analysis of the rhythm of metrical shapes in epic,
compare them meaningfully, and see in vivid detail passages
where metrical shapes localize or occur against expectation.
1. Metrical shapes beyond O’Neill and Hagel
The present study focuses on metrical shapes as quantitative
rhythmic units. As such, it completes O’Neill’s study of “word-
types,” i.e. metrical shapes such as ‒ ⏑ ⏑, which was based on a
limited sample of a few texts and omitted appositive groups.17
Our study diverges, however, from a more recent analysis of
localization, that of Hagel’s “word shapes.”18 Hagel measures
the localization of phonetic-metrical compounds and appositive
groups with variable quantities, breaking metrical shapes into
smaller categories of “word shapes.” For example, Hagel com-
bines O’Neill’s two shapes ‒ ‒ ⏑ and ‒ ‒ ‒ into one “metrical
shape” ‒ ‒ ⏒, whose final syllable is an anceps (may be either
long or short); then subdivides that into ten “word shapes” ac-
cording to their initial and final vowels or consonants (or double
consonants) and final quantities: for example C‒ ‒ ⏒V̆C as found
in the word κοίλῃσιν (“hollow” Il. 1.26).19 Hagel shows that word
shapes, like O’Neill’s metrical shapes, also localize and provide

17 C. J. Ruijgh, “ΜΑΚΡΑ ΤΕΛΕΙΑ ΕΤ ΜΑΚΡΑ ΑΛΟΓΟΣ: Le prolongement de

la durée d’une syllable finale dans le rythme du mot grec,” Mnemosyne 40

(1987) 313–352, treats ῥυθµός in Dionysius of Halicarnassos as word type; cf.
Blankenborg, Audible Punctuation 37–39: “in his approach, rhythm is under-
stood as the perceptible recurrence of certain rhythmical word types, like the
iamb (⏑ –)” (37).
18 Hagel, in Autour de la césure 135–215; cf. C. W. Forstall, De Homeri Ingenio:

Aspects of Oral-Formulaic Composition in the Iliad (diss. SUNY Buffalo 2014) 55–
93. For localization of phonemes see D. Bouvier, “Localisation des lettres et
des phonèmes dans l’hexamètre homérique: développement d’une intuition
refoulée de Ferdinand de Saussure,” Cahiers Ferdinand de Saussure 68 (2015)
149–169.
19 Hagel, in Autour de la césure 181–188, at 185–186. Cf. Allen, Accent and

Rhythm xiii. Hagel’s definition of “metrical shape” thus differs from O’Neill’s,
since Hagel’s shapes at times include variable quantities (anceps), which is a
higher level of abstraction than O’Neill’s.
—————
Greek, Roman, and Byzantine Studies 64 (2024) 350–377
 EPIC RHYTHM 357

valuable information on the metrical tendencies of word forms.

In contrast, the present study addresses the rhythmic expecta-
tions generated by metrical shapes alone, which localize and
generate patterns of expectancy through statistical learning in
ways that may affect the stylometry of the text.
Our study makes use of a statistical formula we call expectancy.20
The intuition is that the distribution of metrical shapes over sedes
creates an expectation for an average or “typical” number of
occurrences of a shape at any sedes, and for how much the counts
vary across sedes. Expectancy, denoted by the variable z, is the
amount by which an observed count x differs from the mean, in
units of standard deviations.21 Two appendices report ex-
pectancy x and z data for the metrical shapes of words and
appositive groups in major poems of the hexameter corpus.22
The Appendix linked to this paper is a synopsis of the distri-
bution of metrical shapes in epic, while the second appendix
(online) provides the diachronic distribution of each shape by
text.23 Expectancy allows for valid and meaningful comparisons
of shapes between texts, authors, and eras, more than
localization scores or percentages alone. Take for example the
distribution of the anapaest (⏑ ⏑ –) over sedes, shaded for
expectancy (Table 1):24

20 Cf. S. A. Sansom, “Sedes as Style in Greek Hexameter: A Computational

Approach,” TAPA 151 (2021) 439–467, at 447, and Sansom and Fifield, DHQ
17 (2023) §5.
21 The mean is a weighted average of the vector of observed counts

x1, …, xn, each weight being the count itself: μ = Σ(xi × xi)/Σ(xi). The standard
deviation is likewise weighted by the number of occurrences: σ =
√(Σ(xi × (xi − μ)2)/Σ(xi)). The expectancy of the count x is then the standard
score: z = (x − μ)/σ.
22 This corpus is the same as that of the SEDES program; cf. Sansom and

Fifield, DHQ 17 (2023) §6.

23 The online appendix, as well as the code and data, is available at

https://github.com/sasansom/epic-rhythm.
24 Statistics are based on the distribution of shapes within a text or collec-

tion of texts, such as the Iliad or Homeric Hymns. The SEDES program tabulates
sedes from the beginning of shapes, in a slight modification of O’Neill; cf.
Sansom, TAPA 151 (2021) 445.

—————
Greek, Roman, and Byzantine Studies 64 (2024) 350–377
 358 STEPHEN A. SANSOM AND DAVID FIFIELD

Total
Work 1 2 2.5 3 4 4.5 5 6 6.5 7 8 8.5 9 10 10.5 11 12
(Σx)
478 1,205 1,891 440 20
Iliad ✖ (11.8%) (29.9%) (46.9%) (10.9%) (0.5%) ✖ 4,034
−1.514 −0.253 +0.936 −1.580 −2.308
322 816 1,563 240 15
Odyssey ✖ (10.9%) (27.6%) (52.9%) (8.1%) (0.5%) ✖ 2,956
−1.516 −0.561 +0.883 −1.675 −2.110
49 177 213 53 1
Hom.
✖ (9.9%) (35.9%) (43.2%) (10.8%) (0.2%) ✖ 493
Hymns
−1.904 +0.176 +0.761 −1.839 −2.684
24 68 117 32 0
Theog. ✖ (10.0%) (28.2%) (48.5%) (13.3%) (0.0%) ✖ 241
−1.616 −0.403 +0.947 −1.395 −2.277
26 58 130 25 2
WD ✖ (10.8%) (24.1%) (53.9%) (10.4%) (0.8%) ✖ 241
−1.401 −0.695 +0.894 −1.423 −1.931
18 43 63 17 1
Shield ✖ (12.7%) (30.3%) (44.4%) (12.0%) (0.7%) ✖ 142
−1.470 −0.124 +0.953 −1.524 −2.386
133 389 546 37 2
Argon. ✖ (12.0%) (35.1%) (49.3%) (3.3%) (0.2%) ✖ 1,107
−1.902 −0.224 +0.805 −2.532 −2.761
28 30 98 3 0
Callim.
✖ (17.6%) (18.9%) (61.6%) (1.9%) (0.0%) ✖ 159
Hymns
−1.254 −1.195 +0.785 −1.982 −2.069
21 130 157 42 1
Phaen. ✖ (6.0%) (37.0%) (44.7%) (12.0%) (0.3%) ✖ 351
−2.328 +0.120 +0.726 −1.856 −2.777
59 199 189 43 3
Theoc. ✖ (12.0%) (40.4%) (38.3%) (8.7%) (0.6%) ✖ 493
−1.771 +0.599 +0.430 −2.042 −2.719
184 262 452 180 6
Quint.
✖ (17.0%) (24.2%) (41.7%) (16.6%) (0.6%) ✖ 1,084
Smyrn.
−1.054 −0.416 +1.137 −1.087 −2.509
349 927 1,212 97 1
Dion. ✖ (13.5%) (35.8%) (46.9%) (3.8%) (0.0%) ✖ 2,586
−1.826 −0.073 +0.791 −2.590 −2.881
1,691 4,304 6,631 1,209 52
TOTAL ✖ (12.2%) (31.0%) (47.7%) (8.7%) (0.4%) ✖ 13,887
−1.549 −0.252 +0.903 −1.788 −2.362

TABLE 1:
⏑ ⏑ – by count (x), percentage, and expectancy z-score
(darker shades are more unexpected)

Each row of Table 1 shows expectancy computed within a

single text. Since the distribution of the shape ⏑ ⏑ – is different
in each text, the same percentage may represent different
z-scores in different texts. For example, anapestically-shaped
words occur at around the same frequency near the beginning
of the line (i.e. sedes 2) in both the Homeric Hymns and the Theogony,
around 10%. But despite this similarity in frequency, measuring

—————
Greek, Roman, and Byzantine Studies 64 (2024) 350–377
 EPIC RHYTHM 359

by standard distribution shows that the shape is more un-

expected in the Hymns (−1.904 vs. −1.616), which results in a
darker shading in the visualization. The same is true of different
metrical shapes at the same sedes. As shown in the Appendix
(in which expectancy is computed over the full corpus), although
spondees (– –) appear with less frequency than anapaests (⏑ ⏑ –)
at sedes 10 (0.0% [6/40,131] vs. 0.4% [52/13,887]), they have a
greater expectancy in that position (z = −1.697 vs. −2.362), be-
cause the distribution of spondaically-shaped words over other
sedes is generally more variable. The definition of expectancy in
terms of standard deviations shows how counts differ from what
is “typical” in a way that plain percentages do not. In addition
to tables, computed expectancies may also be visualized in the
text itself by shading each word according to the expectancy of
its metrical shape, for instance in Homeric Hymn to Hermes 13
(as above, darker shades indicate lower z-scores):25
καὶ τότ’ ἐγείνατο παῖδα πολύτροπον, αἱμυλομήτην
and then she (Maia) bore a child, many-wayed, plotting-twists.
While most shapes are expected in their given positions (within
a range of z = −0.52 to +0.61), the shape of ἐγείνατο (⏑ ‒ ⏑ ⏑) is
very unexpected (z = −5.40) at sedes 2.5; it is much more com-
mon, for example, at sedes 6.5.
Similar to the analysis of lemmata (or all morphological forms
of a word) in prior studies, the above method provides basic
information about when and where metrical shapes are ex-
pected.26 Such information is beneficial to future metrical and
linguistic study of hexameter, but also, as will be demonstrated
below, to readers of epic who are interested in the potential
stylistic effects of meter. We suggest that regularity of metrical
shapes characterizes a passage, speech, text, or author in aggre-
gate. A general impression or affective state may arise from ex-
treme rhythmic regularity, such as the impression of somnolence
or “sweetness” in the Hymn to Hermes passage discussed above; or

25 The metrical scheme would be visualized thus: 1[‒ ⏑] 2.5[⏑ ‒ ⏑ ⏑] 5[‒ ⏑]

6.5[⏑ ‒ ⏑ ⏑], 9[‒ ⏑ ⏑ ‒ ‒]. This visualization calculates expectancy according
to the Archaic corpus.
26 Sansom, TAPA 151 (2021) 439–467.

—————
Greek, Roman, and Byzantine Studies 64 (2024) 350–377
360 STEPHEN A. SANSOM AND DAVID FIFIELD

unexpected irregularity, such as general increase in attention or

other affective states found in psychological studies of musical
rhythm. The potential for the latter increases in certain texts or
individual books or poems that contain a higher density of ir-
regular shapes, a topic that is the focus of the following section.
2. Density of unexpected metrical shapes
For the purpose of analysis, we will call a shape/sedes com-
bination “unexpected” when it has an expectancy z-score less
than or equal to −2.0; that is, the number of occurrences of the
shape at that sedes is at least 2 standard deviations lower than the
weighted mean number of occurrences at any sedes. The specific
threshold is arbitrary and is only for the convenience of con-
verting the continuous quantity of expectancy into a yes/no
binary. By this definition, 3.30% (13,160/399,327) of words or
appositive groups in the epic corpus have a metrical shape that
is unexpected at the sedes at which it appears.
Unexpected shapes are unevenly distributed. For example,
4.00% (3,551/88,796) of shapes in the Iliad are unexpected (an
average of 147.96 per book, which themselves average roughly
3,700 shapes), with Book 10 having the least proportion of
unexpected shapes at 3.27% (108/3,303) and Book 2 having the
greatest at 4.68% (227/4,850). Below, we show the overall per-
centage of unexpected shapes in particular texts, as well as the
book or poem with the lowest and highest rates of unexpected
shapes (Table 2).27
We may note a general diachronic trend toward fewer un-
expected shapes. Apollonius of Rhodes, Callimachus, and Non-
nus put shapes in expected places more often than other poets.28

27 Callimachus knew Homer’s metrical tendencies but not, of course, vice

versa. For this reason, the expectancy of Archaic texts is measured relative to
the Archaic corpus only; that of Hellenistic texts from the Archaic and Hel-
lenistic corpus; and that of Imperial texts from the entire corpus.
28 This trend towards metrical regularity and the “severe style” in

Hellenistic authors, especially Callimachus, is discussed e.g. in A. S. Hollis,

Callimachus: Hecale2 (Oxford 2009) 17; F. Cairns, Hellenistic Epigram: Contexts of
Exploration (Cambridge 2016) 232; and M. Fantuzzi and R. Hunter, Tradition
and Innovation in Hellenistic Poetry (Cambridge 2004) 44 (“a series of refinements

—————
Greek, Roman, and Byzantine Studies 64 (2024) 350–377
 EPIC RHYTHM 361

Work Overall rate of Book with lowest rate Book with highest
unexpected of unexpected rate of unexpected
metrical shapes metrical shapes metrical shapes
Iliad 4.00% (3,551/88,796) 3.27% (108/3,303) Bk. 10 4.68% (227/4,850) Bk. 2
Odyssey 3.91% (2,667/68,293) 2.70% (87/3,222) Bk. 10 4.62% (159/3,444) Bk.17
Homeric Hymns 3.91% (508/12,994) 0.00% (0/88) Hymn 8 10.53% (8/76) Hymn 29
3.95% (423/10,704) 0.00% (0/24) Hymn 12
[Hymns 2–5] 0.00% (0/18) Hymn 13
0.00% (0/29) Hymn 16
0.00% (0/26) Hymn 17
0.00% (0/31) Hymn 21
0.00% (0/20) Hymn 23
Theogony 3.55% (201/5,663) - -
Works and Days 3.38% (158/4,669) - -
Shield of Heracles 3.88% (104/2,683) - -
Ap. Rh. Argon. 2.09% (668/32,027) 1.83% (179/9,761) Bk. 4 2.55% (180/7,064) Bk. 2
Callim. Hymns 1.52% (79/5,200) 0.97% (6/619) Hymn 2 1.77% (10/564) Hymn 1
Aratus Phaen. 3.27% (201/6,149) - -
Theocr. Idylls 3.12% (445/14,273) 1.17% (3/256) Idyll 20 6.35% (53/835) Idyll 15
Quintus Smyrn. 3.74% (1,836/49,063) 3.35% (123/3,677) Bk. 14 4.46% (165/3,701) Bk. 6
Fall of Troy
Nonnus Dion. 2.50% (2,742/109,517) 1.79% (33/1,840) Bk. 34 3.54% (69/1,951) Bk. 46

TABLE 2:
Density of Unexpected Metrical Shapes by Work and Book

Aratus and Theocritus, however, interrupt this tendency and

have only slightly fewer unexpected shapes than Archaic poets;
Quintus of Smyrna has nearly the same average as the Iliad. But
despite Theocritus’ closeness to Archaic figures, the density of
unexpected shapes in his poetry ranges far more widely than to
Homer; in contrast to the Iliad and Odyssey, whose difference is
only about 1.4% to 1.9% between the highest and lowest fre-

and prohibitions governing word-breaks and possible combinations means

that the Callimachean hexameter is a very strict instrument, which imposes
a marked intellectual formalism on all his poetry in this metre”).

—————
Greek, Roman, and Byzantine Studies 64 (2024) 350–377
362 STEPHEN A. SANSOM AND DAVID FIFIELD

quencies, the difference in Theocritus’ range is 5.2%. The

rhythm of Theocritus’ metrical shapes varies far more in his
oeuvre than does Homer’s.
Some individual books or poems have a surprisingly large
number of unexpected shapes, which may tell us more about
their rhythmic character. Theocritus Idyll 15, which depicts
women attending a festival of Adonis, has over twice as many
unexpected shapes as we would expect based on his corpus
(6.35% [53/835] compared to average 3.12%); this difference
exceeds those of other Hellenistic authors, e.g. Callimachus
(highest 1.77% to average 1.52%), and Homer, too (4.68% to
3.96%).29 From these figures, Idyll 15 appears to be rhythmically
distinct, at least in its density of unexpected metrical shapes. A
better understanding of the poem’s distinctive rhythm could
intervene in current scholarship, which debates its Homeric
character and its indebtedness to rhythmic prose mime, such as
that of Sophron.30 Theocritus adopts the hexameter, but his

29 The following are marked with unexpected shapes in Idyll 15 (with the
sedes, shape, and line number): ὅτι καὶ νῦν (2[⏑ ⏑ – –], 2), ὁ πάραρος (2[⏑ ⏑ –],
8), τὸν τεόν (3[– ⏑ ⏑], 11), ὡς ποθορῇ (9[– ⏑ ⏑ –], 12), τὰν πότνιαν (7[– – –], 14),
ἀργυρίω (7[– ⏑ ⏑ –], 18), ἀκούω (6.5[⏑ – –], 23), καὶ ἰδοῖσα (6[⏑ ⏑ – ⏑], 25),
ἀεργοῖς (6.5[⏑ – –], 26), πρότερον (10[⏑ ⏑ –], 29), ὕδατος (8[⏑ ⏑ –], 29), ὁκοῖα
(2.5[⏑ – ⏑], 32), καθαρῶ (10[⏑ ⏑ –], 36), ἀργυρίω (7[– ⏑ ⏑ –], 36), καὶ τὰν ψυχάν
(6[– – – –], 37), ἀπέβα τοι (6[⏑ ⏑ – –], 38), τὸν µικκόν (6[– – –], 42), τὰν αὐλείαν
(6[– – – –], 43), παρέρπων (6.5[⏑ – –], 48), ὡς ἄγριος (7[– ⏑ ⏑ –], 53), ἀνέστα
(2.5[⏑ – ⏑], 53), διαχρησεῖται (6[⏑ ⏑ – – –], 54), τὸ βρέφος (8[⏑ ⏑ –], 55), πείρᾳ
θην (6[– – –], 62), γυναῖκες (2.5[⏑ – ⏑], 64), καὶ ὡς Zεύς (6.5[⏑ – –], 64), τὰν χέρα
µοι (7[– ⏑ ⏑ –], 66), Eὐνόα (3[– ⏑ ⏑], 76), ὦ δειλά (7[– – –], 76), σοφόν τοι
(6.5[⏑ – –], 83), ὤνθρωπος (10[– – –], 83), ὃς ἁµῶν (6.5[⏑ – –], 94), κενεάν
(8[⏑ ⏑ –], 95), ἀθανάταν (7[– ⏑ ⏑ –], 106), εἰκυῖα (10[– – –], 110), ἑλένᾳ
(8[⏑ ⏑ –], 110), ὥρια (3[– ⏑ ⏑], 112), ἐν ὑγρῷ (8.5[⏑ – ⏑], 117), οἱ δέ τε (3[– ⏑ ⏑],
120), µίλατος (2[– – ⏑], 126), τὰν σαµίαν (7[– ⏑ ⏑ –], 126), πτύοντα (10[– – –],
133), καὶ ἐς νέωτ’ (6.5[⏑ – –], 143), ἀφίκῃ (8[⏑ ⏑ –], 144).
30 For the former see M. Chaldekas, “An Intersectional Approach to

Theocritus, Idyll 15,” Helios 49 (2022) 1–24, at 11; S. Reece, The Stranger’s
Welcome: Oral Theory and the Aesthetics of the Homeric Hospitality Scene (Ann Arbor
1993) 320; for the latter, S. Miles, “Performing Mime in the Idylls of Theocri-
tus: Metrical Mime, Drama, and the ‘Everyday’ in Theocritus, Idylls 2, 14,
15,” in P. Kyriakou et al. (eds.), Brill’s Companion to Theocritus (Leiden 2021)
154–175.
—————
Greek, Roman, and Byzantine Studies 64 (2024) 350–377
 EPIC RHYTHM 363

rhythm, at least of metrical shapes, is un-Homeric. Perhaps like

other metrical features, the positioning of shapes also contributes
to what Miles has described as “the audial impact of (Theocri-
tus’) works and the implications of tones and resonances with
past poetry.”31
The expectancy of shapes allows for future inquiry of this sort
not only into the style of different authors, but also how rhythm
relates to the possible style of individual books and poems. And
like books, shorter passages may also deviate from typical metri-
cal distributions. In what follows, we return to the Homeric Hymn
to Hermes to briefly examine the literary contexts of both un-
expected and overly regular rhythm of metrical shapes within
the hymn.
3. Rhythm, expectancy, and kosmos in the Homeric Hymn to Hermes
The rhythm of the Homeric Hymn to Hermes has been the subject
of some debate. In its larger structure, West claims that “(the
hymn has) no command of the even tempo appropriate to epic
storytelling.”32 But Thomas demonstrates how the poem “makes
use of complex internal patterning at every level” (39), where
corresponding passages develop themes such as charis (“divine
favor” 36) and kosmos (“intricate, pleasing structure” 39).33 When
Apollo prosecutes Hermes for stealing his cattle, he deploys “an
arsenal of rhetorical strategies” that include “rhythm, al-
literation, syntactic curtness, understatement and perhaps sar-
casm.”34 Hermes responds to this rhetorical onslaught with
rhythmic stylings of his own, ending 11 out of 17 lines with the
pattern [⏑ – –]VERB with six preceded by [– ⏑] (261–277).35 Thus
far, then, rhythm has been suggested to work on both a large
and a small scale in the organization of the poem: as thematic
responsion between longer intratextual passages, and as a rhe-

31 Miles, in Brill’s Companion to Theocritus 167.

32 M. L. West, Homeric Hymns, Homeric Apocrypha, Lives of Homer (Cambridge
[Mass.] 2003) 12.
33 Cf. Vergados, The Homeric Hymn to Hermes 125, for the larger-ring com-

position of the poem.

34 Thomas, The Homeric Hymn to Hermes 292, cf. 280.

35 Vergados, The Homeric Hymn to Hermes 24 and 423.

—————
Greek, Roman, and Byzantine Studies 64 (2024) 350–377
364 STEPHEN A. SANSOM AND DAVID FIFIELD

torical feature of interpersonal dispute.

Metrical shapes play a role in both the macro- and micro-
rhythms of the hymn. From a more distant vantage point, we see
that passages vary in the number of unexpected shapes
(z ≤ −2.0). We can view this variation as a graph of unexpected
shapes in the poem ( fig. 1):36

Figure 1: Number of unexpected metrical shapes in windows of 181

shapes in Homeric Hymn to Hermes, with pips indicating their location

There are peaks, where passages have more unexpected shapes

(max = 15), and troughs, with few to no unexpected shapes.
Such a graph thus identifies runs of lines that we could consider
arrhythmic (peaks) or, conversely, overly rhythmic (troughs) ac-
cording to the expectancy of metrical shapes. Both types—
arrhythmic and overly rhythmic passages—may tell us some-
thing about the style and meaning of their respective places in
the narrative. We suggest that this is the case for the hymn and
will consider a few of the outstanding sections.
Surprisingly, thematically related passages at times cor-
respond rhythmically as well. This correlation is especially
apparent for the theme of song. Hermes sings two songs in the
hymn. As mentioned at the beginning of this paper, Hermes’
theogonic song that pacifies Apollo near the poem’s end is overly

36 The method for counting unexpected shapes is called the sliding-window

technique, whereby the formula counts the number of unexpected shapes

within a given number of shapes, reports the count, then shifts one shape over
and repeats; cf. S. Katajamäki, “Analyzing Irregular Rhyme Sequences:
Methodological Experiments with Lauri Viita's Kuknor,” in V. Sykäri et al.
(eds.), Rhyme and Rhyming in Verbal Art, Language, and Song (Helsinki 2022) 246–
263, at 253.
—————
Greek, Roman, and Byzantine Studies 64 (2024) 350–377
 EPIC RHYTHM 365

rhythmic in its metrical shapes (409b–441). From the moment

Hermes breaks from Apollo’s bonds (δέσµα 408) well into
Apollo’s enamored response to Hermes’ song (δῶρον 442), there
is only one unexpected shape out of 181 shapes over 32 lines
(– ⏑ ⏑ – –, Μνηµοσύνην µέν 429). For context, in the Archaic
corpus only 0.72% of passages of this length have 1 or less
unexpected shapes.37 But this is not the first section with overly
expected rhythm. The first is, in fact, Hermes’ other song, sung
immediately after he invents the lyre at the poem’s beginning
(52–67). From when Hermes “improvises” on the “lovely play-
thing … in scales” (ἐξ αὐτοσχεδίης 55; τεῦξε, φέρων, ἐρατεινὸν
ἄθυρµα … κατὰ µέρος 52–53) singing of his birth in the cave (61),
to his first thoughts of something to eat (χρειῶν ἐρατίζων …
δείπουσι 67), there are no unexpected shapes (0/78).38 Scholars
have previously noted how the two songs are intricately related
and reflect Hermes’ social transition from the domestic sphere
of the first song to the social hierarchies of the gods in the
second.39 We can now see that the songs share overly expected
rhythm of metrical shapes as well.
Hermes’ second song also responds to themes and rhythm of
the ‘sacrifice’ of Apollo’s cows that follows their theft (120–154).
After Hermes butchers two of the cows, he cooks them on spits.
From when he begins to cut the meat (ταµὼν κρέα 120) to when
he burns the carcasses, tosses his sandals into the Alpheios (139),
and crawls back into his crib cradling his lyre (χέλυν ἐρατὴν ἐπ’

37 The scene before is likewise overly rhythmic in its position of metrical

shapes. After Hermes finishes his final case to Zeus, Zeus laughs, bids him
show Apollo where the cows are hidden, and he does so; from 381 (καὶ
δαίµονος) to 407 (κράτος), there is only 1 unexpected shape (2[– – ⏑] Ἑρµῆς
µέν 401, 1/139 shapes, of which only 2.70% of windows have 1 or less un-
expected shapes).
38 4.18% of passages of 78 shapes in the Archaic corpus have no un-

expected shapes.
39 S. I. Johnson, “Myth, Festival, and Poet: The Homeric Hymn to Hermes and

its Performative Context,” CP 97 (2002) 109–132, at 124, contrasts the

hymn’s structure with other instances of the myth and suggests that this
change is due to the social significance of Hermes’ raid of Apollo’s cattle. Cf.
J. S. Clay, The Politics of Olympus: Form and Meaning in the Major Homeric Hymns2
(London 2006) 140.

—————
Greek, Roman, and Byzantine Studies 64 (2024) 350–377
366 STEPHEN A. SANSOM AND DAVID FIFIELD

ἀριστερὰ χειρὸς ἐέργων 153), there are only two unexpected

shapes out of 189 (only 2.13% of windows of this length have
two or less shapes in the Archaic corpus). A rhythmic pattern
emerges between this scene and his previous activity of crafting
and playing the lyre. Hermes crafts the lyre out of the turtle (33–
52a, somewhat rhythmically unexpected [8/104, 5.2%, of win-
dows have that many or more unexpected shapes]) then plays
the lyre (52b–67a, overly rhythmic [0/78, 4.18%]); Hermes
craftily (δολίης … τέχνης 76) steals Apollo’s cattle (105–120a,
rhythmically unexpected [8/90, 2.43%, with 8 or more un-
expected shapes]), then carefully cooks two in proper fashion
(120b–154a, overly rhythmic [2/189, 2.13%]). The final song of
Hermes recalls these two passages of cultural performance
(music, food preparation) in its themes and diction.40 In addition
to these shared themes and language, the passages are likewise
marked by overly expected rhythm.
Even more striking is the correlation between the expectancy
of metrical shapes and claims to kosmos (“orderliness”). This
correlation plays out in two passages. We have mentioned the
overly expected rhythm of Hermes’ second song and the
narrator’s accompanying commentary on its kosmos: “(Hermes)
pronounced all things in order, playing the lyre under his arm”
(πάντ᾿ ἐνέπων κατὰ κόσµον, ὑπωλένιον κιθαρίζων 433).41 Long
before this, though, Apollo demands that Hermes reveal the
whereabouts of his cows; this demand is accompanied by threats
to commit acts “not according to order” (οὐ κατὰ κόσµον 255),
such as throwing Hermes into Tartaros (256). Thomas has noted
how Apollo begins his threats with “alliteration of µ-, the mono-
syllables creating a pronounced sixth-foot caesura, the curt
syntax … (and) explosive enjambement.”42 We can also see that

40 The hides which Hermes attached to the wall trigger Apollo’s anger and

the second song (403–417); the first song, the cows and cooked meat, and the
second song are each presented as a geras or “honorable gift” (γέραιρε 60, νῶτα
γεράσµια 122, γέρας 129, ἐγέραιρεν 429, ἐγέραιρε 432, cf. γέρας 291, 573).
41 In his response to Apollo, Hermes himself also proclaims that the lyre

“knows how to speak well and in good order” (καλὰ καὶ εὖ κατὰ κόσµον ἐπι-
στάµενος ἀγορεύειν 479).
42 Thomas, The Homeric Hymn to Hermes 280.

—————
Greek, Roman, and Byzantine Studies 64 (2024) 350–377
 EPIC RHYTHM 367

Apollo’s opening lines contain an arresting amount of un-

expected metrical shapes:
ὦ παῖ, ὃς ἐν λίκνῳ κατάκειαι, μήνυέ μοι βοῦς
θᾶσσον· ἐπεὶ τάχα νῶϊ διοισόμεθ’ οὐ κατὰ κόσμον. 255
Child, who lies in a crib, confess to me about the cows
quickly· since soon we shall disagree not in good order.
Three shapes, 6[⏑ ⏑ – ⏑] (κατάκειαι), 9[– ⏑ ⏑ –] (µήνυέ µοι), and
2.5[⏑ – ⏑ ⏑] (ἐπεὶ τάχα), are unexpected (only 0.025% of windows

of 5 shapes have 3 unexpected shapes in the Archaic corpus),

while the passage as a whole (from Λητοΐδης 253 to µητρός 267
[78 shapes]) has 7 out of 78 (3.11% of windows of 78 shapes have
7 or more unexpected shapes in the corpus).43 The hymnist con-
centrates unexpected shapes in Apollo’s speech at the same time
that Apollo threatens to act οὐ κατὰ κόσµον. In other words,
Apollo performs the breach of kosmos in the irregular rhythm of
his speech.
This is not the only correspondence between notions of kosmos
and metrical rhythm in Greek literature. Later, in the early
Hellenistic period (4th/3rd cent. BCE), the poet Simias of Rhodes
makes the association explicit in his so-called technopaegnion or
“picture-poem” shaped like an egg (fr.26.7–10 Powell):
Τὸ µὲν θεῶν ἐριβόας Ἑρµᾶς ἔκιξε κᾶρυξ
φῦλ’ ἐς βροτῶν, ὑπὸ φίλας ἑλὼν πτέροισι µατρός,
ἄνωγε δ’ ἐκ µέτρου µονοβάµονος µέγαν πάροιθ’ ἀέξειν
ἀριθµὸν εἰς ἄκραν δεκάδ’ ἰχνίων, κόσµον νέµοντα ῥυθµῶν, 10
Hermes, loud-voiced herald of the gods, took it (the egg) from
under its dear mother’s wings and gave it to the world, command-
ing me to increase gradually the number of feet from the original
one foot to ten at the end, while maintaining proper order in the rhythm,44

43 Several of the unexpected shapes in his speech occur within the com-
parably unexpected metrical position of their words, for example κραταιῷ
(“strong” 265), whose lemma is z = −2.0 (3/15 instances in the Archaic
corpus) at this position, and ζόφον (“dark” 257), which has an expectancy of
z = −2.16 (3/17) as a word in the same form and z = −2.71 (3/25) as a
lemma.
44 Transl. N. Hopkinson, Theocritus Moschus Bion (Cambridge [Mass.] 2015)

569.

—————
Greek, Roman, and Byzantine Studies 64 (2024) 350–377
368 STEPHEN A. SANSOM AND DAVID FIFIELD

Plato, too, associates kosmos with rhythmos, perhaps indebted to

the earlier musicological work of Damon (Resp. 400A):
βίου ῥυθµοὺς ἰδεῖν κοσµίου τε καὶ ἀνδρείου τίνες εἰσίν· οὓς
ἰδόντα τὸν πόδα τῷ τοῦ τοιούτου λόγῳ ἀναγκάζειν ἕπεσθαι καὶ
τὸ µέλος, ἀλλὰ µὴ λόγον ποδί τε καὶ µέλει.
(We must) see which are the rhythms of an orderly and manly life. In
view of this the metrical foot and the melody must follow the
verbal expression and not the expression follow the meter and the
melody.45
Thomas suggests that the poet of the Homeric hymn may have
been aware of similar associations and metaphors of ἴχνος
(“tracks, feet”) as related to sense (cf. Plato Parm. 128C) and
rhythm (Simias fr.26) and fashioned the order of words in the
meter to reflect the meaning of the poetry, such as when Hermes
drives the cattle backward, “the front to the back and the back
to the front” (τὰς προσθεν ὄπισθεν, / τὰς δ᾽ ὄπιθεν πρόσθεν 77–78).46
We could add to this example the unexpected position of 3ἴχνια
(“tracks, footsteps”) by Apollo in 351 when describing Hermes’
deceitfulness (δολοφροσύνην 361), a word that is out of place both
according to its metrical shape (dactylic words are rarely in this
position either in the Archaic or overall corpus, cf. Appendix) as
well as its word form (z = −2.11, 1/9, in the Archaic corpus) and
45 Transl. C. Emlyn-Jones and W. Preddy, Plato VI (Cambridge [Mass.]
2013) 277. With reference to Damon (400B): “we shall also consult with
Damon as to which rhythmical movements are appropriate to illiberality and
insolence, or madness and other forms of vice, and which ones are left for
their opposites,” µετὰ Δάµωνος βουλευσόµεθα, τίνες τε ἀνελευθερίας καὶ
ὕβρεως ἢ µανίας καὶ ἄλλης κακίας πρέπουσαι βάσεις, καὶ τίνας τοῖς ἐναντίοις
λειπτέον ῥυθµούς; cf. 424C and Laches 200A–B. See E. Ermolaeva, “The Fig-
ure Poem Egg by Simias of Rhodes (AP 15, 27) and Metrical Terminology,”
Philologia Classica 12 (2017) 122–129, at 127, and N. A. Almazova, “Daktylus
und Enhoplios in Damons Rhythmuslehre,” Hyperboreus 22 (2016) 94–126, at
95. The musicologist Aristides Quintilianus (3rd cent. CE) likewise associates
kosmos with rhythmos in musical rhythm (3.25.44–47): “therefore, a cosmos of
matter is the motion of the elements, while a cosmos of the soul is the melody
of the vowels. And indeed, also of rhythms,” διὸ κόσµος µὲν ὕλης ἡ ἐκείνων
κίνησις, κόσµος δὲ ψυχῆς †ἢνουσος† µελῳδία. καὶ µὴν καὶ τῶν ῥυθµῶν (transl.
adapted from T. J. Mathiesen, Aristides Quintilianus: On Music in Three Books
[New Haven 1983] 201).
46 Thomas, The Homeric Hymn to Hermes 191 (ad 75–78).

—————
Greek, Roman, and Byzantine Studies 64 (2024) 350–377
 EPIC RHYTHM 369

lemma (z = −2.74, 3/10, in the Archaic corpus). While it is be-

yond the purview of this article to argue for direct knowledge of
such semantics of kosmos or related terms by the hymnic poet,
nevertheless we now know something new. In the hymn, there is
a positive correlation between kosmos and the expectancy of
metrical shapes.
More thematic associations emerge from exceptionally
rhythmic and arrhythmic passages. Hermes’ overly rhythmic
performances are associated with desire (ἐρατεινόν 52, ἐρατήν
153, ἐρατή 421, ἵµερος 422, ἐρατόν 423, ἐρατή 426, ἔρος 434), gifts
(γέρας 60, 122, 129, 429, 432, cf. 291, 573), sleepiness (449),
sweetness (γλυκύς 422), and laughter (Ζεὺς δὲ µέγ᾿ ἐξεγέλασσεν
389, γέλασσε δὲ Φοῖβος Ἀπόλλων 420). But we find more negative
associations in arrhythmic passages. When Apollo threatens
Hermes οὐ κατὰ κόσµον, his future deeds emphasize darkness
(λαβὼν ἐς Τάρταρον, ἐς ζόφον, οὐδέ … ἐς φάος, ὑπὸ γαίῃ 256–258,
cf. ἐς φάος 402) and helplessness (ἀµήχανον 257). Later, Apollo
describes the theft to Zeus in perhaps the most arrhythmic pas-
sage in the hymn (343–363, 12/107 unexpected shapes [0.035%
of such windows in Archaic epic]), again emphasizing darkness
(κόνις … µέλαινα 345, µελαίνηι νυκτὶ ἐοικώς 358, κατὰ ζόφον 359)
and helplessness (ἀµήχανος 346; cf. ἀµήχανος 434 and µηχανῶτα
436, with Thomas ad loc.), as well as monstrosity (πέλωρα 342 and
πέλωρ᾿ 349) and quick, skilled movement (διαπυρπαλάµησεν …
τὸ µὲν ἔνθα, τὸ δ᾿ ἔνθα 357). This is not to say that these themes
are intrinsically or even pervasively associated with rhythmicity
in epic elsewhere, only that they are so in the hymn.
The above survey of the expectancy of metrical shapes in the
Homeric Hymn to Hermes proceeded on two levels. A macro-scale
analysis of unexpected shapes identified passages with noticeably
few or exceptionally concentrated amounts of unexpected
shapes. Closer inspection of these passages revealed unantici-
pated correlations. Both of Hermes’ songs, as well as his cooking
of the cows, occur in overly rhythmic passages with few un-
expected shapes. Conversely, the most arrhythmic passages
occur in speeches by Apollo and at times accompanied threats
and other formal irregularities such as alliteration and atypical

—————
Greek, Roman, and Byzantine Studies 64 (2024) 350–377
370 STEPHEN A. SANSOM AND DAVID FIFIELD

caesurae.47 But most surprising of the findings are the asso-

ciations between rhythm and notions of kosmos. Descriptions of
“orderly” (κατὰ κόσµον) performance had the fewest unexpected
shapes, while threats of “unseemly” (οὐ κατὰ κόσµον) conflict had
a concentration of unexpected shapes. Granted, there are limits
to the current data and methodology. The edition of texts is
paramount, and the text of Homeric Hymn to Hermes is rather
thorny, for instance in the corruption of line 346 and resulting
difference in words, word order, metrical shapes, and rhythmic
data.48 Hagel’s list of appositive groups, which is the basis of the
current analysis, also needs labor-intensive refining and ex-
pansion.49 Future phonological analysis could and should have
the option to incorporate larger groups, such as minor phrases
and formulas, major phrases and cola; such reanalysis would
provide new statistics relevant under different phonological con-
ditions.50 That said, this paper advances the state of the art in
important ways, both in methodology and in analysis, and un-
covers new potential for future investigation.
4. Conclusion: the rhythm of metrical shapes
The above analysis defines and defends a central claim: when
metrical shapes are measured by expectancy, we can better
identify the rhythm of authors, poems, characters, and passages
in ways that can inform their style and interpretation. It does so
by incorporating the metrical shapes of words and appositive
groups and by extending previous analysis of metrical shapes

47 These include the curious allotment of the “Bee Maidens” to Hermes at

the poem’s end (ἐθέλουσιν 561 through Μαιάδος 574, 9/90 [0.87% of win-
dows]).
48 Compare the editions of the manuscripts and Allen (αὐτὸς δ᾽ οὗτος †ὅδ’

ἐκτὸς† ἀµήχανος, οὔτ' ἄρα ποσσίν) with Evelyn-White (old Loeb)/Perseus/

SEDES (αὐτὸς δ’ ἐκτὸς ὁδοῦ, τις ἀµήχανος, οὔτ’ ἄρα ποσσίν) and West (new
Loeb) (αὐτὸς δ᾿ οὔθ᾿ ὁδοῦ ἐκτὸς ἀµήχανος, οὔτ᾿ ἄρα ποσσίν).
49 Hagel, in Autour de la césure 137, does not, for example, include the ἐκτός

as found in line 346; future work must also clarify on a case-by-case basis
whether candidates for apposition in prepositional phrases function as pre-
/post-positions or are adverbial (and perhaps words in a larger appositional
unit, such as the minor or major phrase).
50 For speech rate and apposition cf. Bozzone, Living Language 91.

—————
Greek, Roman, and Byzantine Studies 64 (2024) 350–377
 EPIC RHYTHM 371

with a measure of expectancy. Such a measure allows for the

more accurate and useful comparison of distribution by standard
deviations, such as in locating passages that are rhythmically
unusual by their density of unexpected shapes. We see its utility
especially in extreme cases, such as in the uncommon density of
unexpected shapes in the urban mime of Theocritus Idyll 15.
But it is the unforeseen correlations between themes and the
rhythm of shapes in the Homeric Hymn to Hermes that suggest that
the method and data have a more pervasive application to the
study of Greek epic. By mapping rhythmic analysis to narrative,
this case study reveals associations between song, kosmos, and
other themes with the expectancy of metrical shapes. Such a
methodology that combines computational exploration and
closer reading could open new avenues of inquiry. Do themes
have rhythmic expectations? To what extent do social, rhetori-
cal, or affective goals condition the articulation of hexameter’s
inner metric? While this study is preliminary, the goal of this
article is to make the method and data available for future adap-
tation and exploration and provide brief proofs of concept for
the stylistic importance of metrical shapes. Such importance
becomes visible through exploration of the data within the text
in a process that brings readers closer to reconstructing the
potential experience of the poems by audiences attuned to the
rhythms of Greek epic.51

June, 2024 Florida State University

sasansom@fsu.edu
Denver, Colorado
david@bamsoftware.com

APPENDIX
(pp.372–377)

51 We would like to thank the anonymous reviewer for their insight and
feedback as well as audiences at Western University and the University of
California, Santa Barbara.

—————
Greek, Roman, and Byzantine Studies 64 (2024) 350–377