Hurrian Meter and Phonology in the Boğazköy Parables
Chelsea Sanker
Cornell University
1
Introduction
The Hurrian Parables, a Hittite-Hurrian bilingual from a collection of texts from Boğazköy
(KBo 32), have been characterized as having four stressed syllables per line (e.g. Bachvarova
2011); others have suggested that the pattern of unstressed syllables may also contribute to
the meter (Haas and Wegner 2007; Neu 1988), though variable line lengths pose a problem
for an isosyllabic meter.
1.1
DR
AF
T
I offer evidence for a syllable-counting but non-isosyllabic meter, with lines in groups of 2-3,
forming semantic units. Within these groups, the average difference in syllable count between
lines is significantly lower than it is between non-grouped lines. The similarity within groups
of lines suggests that, despite apparent differences based on the orthography, lines within
groups may have matched exactly in number of syllables. Postulating this exact match, I
offer three phonological interpretations of the orthography that build on the phonological
system standardly given for Hurrian.
1.1.1
Overview of Hurrian
Hurrian History and Linguistic Environment
Hurrian was a language spoken in northern Mesopotamia and southeastern Anatolia from at
least the late 3rd millennium BC until the 12th century BC. There is evidence for contact
between Hurrian and several language families of the Near East, most importantly Akkadian,
Sumerian (Wilhelm 2008), Hittite (Wegner 2007) and an unattested Indic language (Watkins
2006), among others as far north as the Caucasus Mountains (Diakonoff 1985) and as far
south as Egypt (Wegner 2007).
1.1.2
Hurrian Phonology
There is still uncertainty about the phonological inventory of Hurrian. Because no writing
system used for Hurrian was native, there is ambiguity in whether apparent lack of distinctions reflected an actual absence in Hurrian or only limitations of the writing system. The
influence of foreign scribes also obscured the phonological system, though in some cases they
recorded allophones whose differences were ignored by Hurrian scribes (Wilhelm 1983).
Most Hurrian texts are written in the cuneiform syllabary borrowed through Akkadian from
the Sumerian logo-syllabary. In this writing system, each symbol represented a syllable,
either a consonant followed by a vowel (CV), a vowel followed by a consonant (VC), a vowel
alone (V), or consonant-vowel-consonant (CVC). Although cuneiform was somewhat adapted
for representing Hurrian, it retained some limitations in representing syllable structure and
may have failed to develop ways of contrastively representing Hurrian sounds which were not
already part of the system. The syllabic writing system made it impossible to represent initial
or final consonant clusters, or internal clusters of more than two consonants. However, it
seems that Hurrian had actual anaptyctic vowels that would have broken up any such clusters
1
(Wilhelm 2008). There were also some Hurrian texts written in the Ugaritic consonantal
alphabet, which can supplement the information provided by texts in the cuneiform syllabary.
A suggested inventory of Hurrian consonant phonemes includes: p t k ts f š s h m n l r
ˇ
w j, though the quality of some of these phonemes remains unclear and this inventory
is
likely too small; there was probably another set of obstruents which was distinguished by a
phonetic feature not yet identified, which did not exist in Akkadian or Ugaritic and was thus
not represented in their writing systems.
T
The labiodental fricative seems to be warranted despite its absence in Semitic and the lack
of <f> sign, because it was usually written with the <wa> sign, while /w/ was more often
written with the <ú> sign. The glide /w/ occurred in internal and final positions, and only
occurred word initially in loanwords and foreign names (Wilhelm 2008). In final position,
there was variation in the sign used to represent labials: the same word might appear with
<p>, <b>, or <w> signs (Speiser 1940). It is often ambiguous whether an <i> sign ought
to be interpreted as a vowel or glide, as the spelling could be the same for both (van Soldt
2010).
DR
AF
The vowel inventory is more clear, and seems to have included a e i o u, with phonemically
contrastive length, though length distinction is not systematically marked in all texts. In
texts which write high and mid vowels distinctly, /o/ tends to be marked with <u> signs
and /u/ with <ú> signs, adapted from Akkadian, which lacked the /o/ vowel and used both
these sets of signs to mark /u/. /i/ and /e/ are also often written with distinct signs, though
texts vary in how consistently this orthographic distinction is maintained. It is unclear
whether diphthongs existed, or if there was a consonant, perhaps a glottal stop, between
orthographically adjacent vowels (Wilhelm 2008).
Some phonological processes are evident. There was vowel insertion to split up clusters made
by the /-n/ affix of the jussive or ablative and a following enclitic pronoun, and either vowel
insertion or deletion (depending on what is assumed to be the underlying form) caused by the
presence or absence of a suffix which changed the position of the stress. The /a/ of enclitic
pronouns and initial /a/ on the enclitic connective /-an/ contracted to produce a long vowel.
/a/ and /i/ were deleted between identical dentals, and the first of two vowels in contact in
certain morphological conditions was also deleted (Wilhelm 2008).
1.1.3
Hurrian Stress
Hurrian words had penultimate stress, calculated from the word-end, including suffixes but
not including clitics. The position of Hurrian stress is based on evidence from internal vowel
deletion and the spelling of penultimate vowels. If an unaccented light syllable followed a
stressed syllable, the vowel was deleted, as shown by alternations of roots with and without
suffixes. Most penultimate vowels were written plene, with a vowel both in the CV sign and a
following independent V sign. Such marking has typically been interpreted as marking length,
following the usage in cuneiform Akkadian (e.g. Huehnergard 2011), though alternative
explanations for the Akkadian plene patterns have also been proposed (e.g. Diakonoff 1991).
The patterns of plene spelling in Hurrian have been interpreted as showing that there was
lengthening of stressed short vowels.
Duration is a feature commonly associated with stress, and could easily have been an element
2
of the phonetic realization of Hurrian stress. Morphological variation shows that there was
also a tendency to lower vowels in this position, which also could have been conditioned by
stress (Wilhelm 2008). Plene spelling was not used consistently in all occurrences of a word,
though there is a clear distinction between words which sometimes have plene spelling and
those which never have it. Whether an extra vowel sign is included or not in a word which
has variable spelling may be motivated by the scribe’s desire to fit the words to the length
of the line (Neu 1996).
1.2
Poetic Traditions of the Ancient Near East
1.2.1
T
As several features of Hurrian poetic tradition seem to be borrowed from Mesopotamia
through Akkadian, including the writing system, musical notation (West 1994), and content (Bachvarova 2012), it seems plausible to imagine a similar origin for other aspects of
Hurrian poetry, such as meter, to have the same origin or to have been influenced by contact
between the languages. Other languages which came in contact with Hurrian may also have
had an influence on Hurrian metrics.
Couplets/Tercets
DR
AF
Many of the ancient Semitic languages and other languages in the region of the ancient Near
East had poetry characterized by groups of lines bound together by syntactic and semantic
similarity and patterns of parallelism with no formal meter regulating stress or syllable count
(Michałowski 1996). The syntax and metrical lines were closely aligned; line breaks never
fell in the middle of a phrase (von Soden 1981; Buccellati 1990). There was a tendency for
lines within a group to be of similar lengths, but this may result from the parallel structure,
which often produces groups within which each line has the same number of words (Watson
1984).
The lines within a group could be tied together by parallelism in form and meaning, rhyme,
or repetition (Hecker 1974; Vantisphout 1993). Sometimes metrical patterns would be maintained for a few consecutive lines, including Hebrew texts with occasional couplets of five
stressed syllables each (Watson 1984). Metrical line breaks were written to coincide with textual line breaks in some poetic texts of Akkadian, Sumerian (Güterbock 1951), and Ugaritic,
which can shed light on the types of variation that were allowed in the type of parallelism
used, when it was used, and the size of the unit within which it operated (Watson 1999).
However, the patterns of parallelism are not a reliable mark of poetry in all of these languages; in Sumerian, parallelism is not a required features of verse and is also found in prose
(Michałowski 1996; Vanstiphout 1993).
Many texts followed a couplet pattern or a mix of couplets and tercets, as in Akkadian
(Buccellati 1990), while in some traditions, larger groups of lines were more common (Watson
1984). Sometimes tercets were used to mark ends, beginnings, and climaxes within Hebrew
texts (Watson 1984). In Egyptian, tercets and other metrical irregularities were sometimes
marked by punctuating dots (Foster 1980).
Stanzas within Hittite poetry have also been proposed. Güterbock identified stanzas in the
Song of Ullikumi, based on syntactic and semantic units and the existence of horizontal
dividing lines, though the lines varied in number of syllables and stresses, even within a
3
stanza. He suggested that the inconsistency in verse form reflected a poor attempt to render
the meter of the unattested original Hurrian poem. In other Hittite texts of Hurrian origin,
such as the hymnic introduction to the Cow Story, there was also inexact strophic structure,
though in original Hittite texts this structure cannot be found, so this form may owe its
origins to Hurrian, even if it entered Hittite in a changed form (Güterbock 1951).
T
In Vedic Sanskrit as well, whose metrical tradition may have had features shared with the
Indo-Iranian language which Hurrian came in contact with, there are meters with stanzas of
2-3 verses, although these meters are based on syllable weight and not stress (Arnold 1905).
In these meters, lines were roughly isosyllabic, which is probably an archaic feature, as it is
shared in meters of several distant Indo-European languages (Campanile 1999). Although
lines did not show a greater similarity in length within couplets, lines patterned differently
within couplets than across them: hiatus was avoided more between lines of a couplet than
across couplets, as was resyllabification of coda consonants when it would result in a light
syllable (Günkel & Ryan 2013).
DR
AF
Some analyses of these texts also propose more restricted patterns of stress and syllable
weight (e.g. Sauren 1971-2), but the lack of evidence for stress and vowel length particularly
in Sumerian makes it difficult to confirm or disprove prosodic features of verse (Vanstiphout
1993). However, there is some evidence for recurring stress patterns, e.g. stress of the
penultimate syllable of lines in Hebrew, sometimes even when the word stress would not
normally fall on that syllable (West 1997).
1.2.2
Stress-counting meter
There was also a stress-based meter in Mesopotamia and Anatolia, generally characterized
by four stresses per line, but also with lines of three or five stresses, and sometimes with
even greater variation (West 1997); some attempts have been made to explain rules of stress
assignment, particularly in Hittite, to allow these lines to fit into this meter either by consistent stress rules (e.g. Durnford 1971), or optional stress rules in poetry (e.g. Melchert
1998). Several optional stress rules have been proposed to fit longer and shorter lines into
the four-stress pattern, e.g. allowing secondary stress within long words (West 1997) and
stress on heavy clitics (von Soden 1981).
It is likely that stress and length generally coincided in Akkadian, but the orthography
does not distinguish between the two and does not consistently mark either (Hecker 1974).
Akkadian stress has been described as being ultimate if the final syllable contained a long
vowel and a final consonant; otherwise on the rightmost non-light syllable; and initial in
words which meet neither of those conditions (Huehnergard 2011). These are essentially the
same stress rules that are in Classical Arabic; however, the earliest reconstructible stress
system of Arabic does not follow this system, so the similarity could not be a shared feature
inherited from Proto-Semitic (Knudsen 1980).
The number of stressed syllables in a line was not always equal to the number of words.
There is evidence for some elements being consistently unstressed in Hittite; the placement
and spelling of sentential clitics in poetry and in prose suggests that they never bore their own
stress (Kloekhorst 2011). Other elements, such as postpositional phrases, do not have such
evidence in prose for bearing only phrase level stress (Melchert 1998). Lines in Akkadian,
4
Hebrew, and Ugaritic could begin with words which were not treated as part of the metrical
structure, even when they had lexical stress (Watson 1984).
Usually there was a syntactic break within a line, dividing it into two smaller units, or cola
(Hecker 1974), each of which contained two stress units (Buccellati 1990). A stress unit often
was a single word, but also could be a group of words within a prosodic unit (Buccellati
1990). In the Babylonian Theodicy, vertical strokes divide the words into compartments
within lines, seeming to indicate that lines were composed of four units. In lines with more
than four words, the words were usually grouped into semantic units (West 1997; Hecker
1974).
1.2.3
Syllable-counting meter
DR
AF
T
It has also been suggested that there was a syllable-counting meter in the earlier works of
Old Babylonian and in Sumerian, though ambiguity in the orthography hinders detailed
scansion. It has been suggested that in Sumerian, stressed syllables were often divided by
1-2 unstressed syllables and lines tended to end with a trochee, particularly in early poetry
(von Soden 1981). However, the apparent prevalence of stressed penultimate syllables may
also be the result of the frequency of penultimate word stress, particularly in verbs, which are
typically sentence-final (Hecker 1974). The small number of poems with complete couplets
that could be in this syllable-counting meter is too small to identify a consistent pattern for
the relative length of lines within a couplet (von Soden 1981).
Syllable-counting meter has also been proposed for Hittite, in particular for the Song of Neša
(e.g. Kloekhorst 2011), though such metrical analysis is hindered by unresolved questions
of stress and vowel length in Hittite (Edzard & Wilhelm 1993-7). Without reformulating
current ideas about Hittite orthography, the three lines of this poem cannot be isosyllabic,
though they only differ by one syllable (Melchert 2007). Eichner (1993) suggests that this
difference in line lengths can be resolved by allowing contraction of word final /a/ with a
following word initial /a/.
1.2.4
Poetry in Hurrian
There are few Hurrian poems which are preserved in continuous sections large enough to
make a metrical analysis of them, but some of them have clear poetic features, including
chiasmus, syntactic repetition and parallelism, alliteration (Haas & Wegner 2007; Thiel 1975),
hyperbaton, asyndeton (Haas & Wegner 2007), semantic parallelism, and repetition of the
same or similar sounds and morphemes in the same position in adjacent lines (Thiel 1975).
Some Hurrian texts included notation for musical accompaniment; however, this is not an
informative resource on meter, because the number of notes and syllables were often very
different, perhaps indicating a repeating melody within some of these texts (West 1994).
A four-stress meter, as identified in many other languages of the region, has been identified as
occurring in Hurrian, where it seems to have been extremely consistent in having lines with
exactly four stresses (Bachvarova 2012). It has also been suggested by Thiel (1975) as well as
others subsequently (e.g. Haas & Wegner 2007; Neu 1988; Wilhelm 1991) that the pattern
of unstressed syllables contributes to the meter for several Hurrian texts, with a tendency
for lines to contain a similar number of syllables. Wilhelm (1991) observed that the prayer
5
of Tadoheba (ChS I/1 41) has been divided into segments of about 17 syllables each.
ˇ
Stanzaic groups within Hurrian poetry have also been proposed. Thiel (1975) suggested
that lines form groups sharing the main semantic theme, with the length of lines following a
pattern based on order within such a group.
2
The Hurrian Parables
2.1
Poetic Features
T
The Hurrian Parables in the collection of texts1 from Boğazköy, KBo XXXII 14, published
in Neu 19962 have been observed to share stress-counting characteristics with the four-stress
meter found in Akkadian and Hittite. The Parables are a well-preserved unit that is of a
consistent style and thus ideal for analysis. The content is repeated in Hittite, but the two
versions have different syntax and style, perhaps following distinct poetic traditions. The
Hurrian version of the text contains a number of poetic features, including those demonstrated
in the following examples. Although in other meters of the ancient Near East, parallelism
was required by the meter, here it seems rather to be used as an optional stylistic feature.
DR
AF
(1) parallelism (Haas & Wegner 2007), e.g. Vs. 23-253
ku-u-le-eš
an-ti
ti-i-ib-ša-a-ri
u-la-ap-uua
ka4 -du-ul-li
“
4
kōl=e=ž
and=i tı̄p=šār=i
ol(i)=ā=ppa
kad=ol=ile
set.aside=2p.imp this=th word=col=th other=ess=2p.abs tell=dur=1s
‘Set aside this story. I will tell you another.’
a-mu-u-ma-a-ap
šal-hu-u-la
ma-ta-a-ap-pa
ka4 -du-ul-li
ˇ
amōm=ā=p(pa)
šalġ=ōl=a
mād(i)=ā=ppa
kad=ol=i=le
observe=it.imp=sbj listen=dur=2p.imp wisdom=ess=2p.abs tell=dur=1s
‘Observe; listen. I will tell you wisdom.’
(2) unusual word orders5 (Melchert 1998), e.g. Vs. 1-2, in which the absolutive nal=i
precedes the ergative fab(a)n(i)ni=ž, the inverse of the usual order.
1
There has been debate whether this collection is really a unit, based on consideration of the ordering of
the tablets, as well as debate on the most accurate term for the texts. See Wilhelm 2012 for a discussion.
2
The transliterations and translations that I use are based on Neu 1996 and Bachvarova 2011. Morpheme
glosses are based on Wilhelm 1992, Wilhelm 2008, Wegner 2007, Neu 1996, and Bachvarova 2011. A key to
abbreviations can be found in the Appendix.
3
Examples will be cited with a transliteration, showing the standard value of cuneiform signs divided by
dashes, followed a transcription with assumed phonetic shapes and morpheme boundaries. Vowels written
plene are marked with a macron in transcription, though it is possible that plene should not always be
interpreted as indicating vowel length. Sounds given in parentheses are assumed to be part of the underlying
form which are not realized in their given phonetic environment.
4
In the singular, absolutive case is indicated by a bare stem, as seen here.
5
The basic word order was ergative-absolutive-verb, with indirect objects and nouns with other relationships to the verb either following the verb or coming between the subject and object. However, topicalization
was allowed, so other orders could occur even in prose (Wilhelm 2008).
6
na-a-li i-te-[e-]i-né-eš
pa-pa-an-ni-iš me-la-ah-hu-um
ˇ
nāl=i
idej=ne=(e)ž
fab(a)n(i)ni=ž mel=ahˇh=o=m
ˇˇ
deer=th body=abl=erg6 mountain=erg bring=forth=tr.prf=3s
‘A mountain brought forth from its body a deer’
(3) hyperbaton (Haas & Wegner 2007), e.g. Vs. 10-11, in which the verb šid=ār=a
divides the participle šēdu=ili=a=ni=š from the noun nāl(i)=ni=š which it modifies.
i-ia-a-at še-e-du-i-li-ia-ni-iš
ši-ta-a-ra
na-a-al-li-iš
“
“
ijā=t(ta)
šēdu=ili=a=ni=ž
šid=ār=a
nāl(i)=ni=ž
why=1s grow=pass.prt=indv=erg curse=tr=3s deer=indv=erg
T
‘Why does the deer, grown fat, curse me?’
(4) repeated forms of a word at the end of one line and the beginning of the next (Neu
1988; Wilhelm 1991), e.g. Rs. 10-12.
ha-a-šar-ri pu-ú-zi-hu-um
ˇāžar(i)=ni pūz=iġ=ō=m
ˇ
h
ˇ
oil=loc
dip=in=tr.prf=3s
DR
AF
ta-la-ah-hu-u-um
e-ep-hé-e-ni
ˇ
ˇ
ˇ
tal=ahh=ōm
ēphē=ne
ˇˇ
ˇ
pull=forth=tr.prf=3s
oven=loc
‘He pulled (it) out from the oven. He dipped (it) in oil.’
pu-ú-zi-hu-ub
ˇ
pūz=iġ=o=b
dip=in=tr.prf=3s
ha-a-šar-ri na-ah-ha-ab ú-la-a-nu-u-um
ˇāžar(i)=ne nahhˇ=a=b
ˇ
h
ul=ān=ō=m
ˇ
ˇˇ
oil=dir
sit=intr=3s
eat=fct=tr.prf=3s
‘He dipped it in oil. He sat down, having eaten.’
Both the Hittite and the Hurrian versions seem to be in the four-stress meter, though there
are certain places in the text where it seems to shift to a different meter for curses (Bachvarova
2012). It has been proposed that there may be a meter in this text based on syllable count
(Neu 1988; Haas & Wegner 2007), and that there is a larger metrical pattern in which not
every line is the same length (Neu 1988). I will endeavor to show that this poem indeed has a
meter with both of these characteristics. Building on Bachvarova’s thorough demonstration
of the four-stress meter in this poem, I will show that there is a restriction that these stressed
syllables must be separated by one to three unstressed syllables, and furthermore, lines occur
in groups of twos and threes with equal or nearly equal length.
2.2
Metrical Lines
In these parables, there is great consistency in having four stressed syllables in each syntactic
unit, which are taken to correspond to metrical lines (Bachvarova 2012). There are 96 lines
which can be measured and grouped with confidence, some of which are repeated. Following
is an example of a typical pair of lines:
6
Suffixaufnahme, case-doubling usually is found with genitives, is found here with the ablative “body,”
based on its association with the subject “mountain” (Neu 1996).
7
(5) Vs. 31-32
S
S
S
S
[na-a]-li ma-a-an-nu-u-bur ma-[a-an-n]i tar-šu-ua-a-ni
“
nāl=i
mān=ō=vor
mānn=i
taržuvān=i
deer=th is=th=neg
is=3s
man=th
‘It is not a deer; it is a man.’
S
S
S
S
[hal-zu-uh]-la
ma-a-an-ni e-eb-re-eš
na-a-hé-la-a-i
ˇ ġl(i)=a
ˇ
hˇalz(i)=ō=
mānn=i
ēvr(i)=i=ž
nāh=e=lāi
ˇ
ˇ
district=th=lord=ess is=3s
lord=3s.poss=erg appoint=tr=ger
T
‘As a district administrator his lord appointed him.’
DR
AF
The graphic organization of the text also supports the existence of lines, as the paragraphs,
identified semantically and often also by horizontal dividing lines, always contain an integer
number of syntactically and semantically defined metrical lines. There are also winkelhaken,
punctuation marks which only occur at these metrical line breaks, including one at the end
of the first line and subsequently often at places where the line break or group of lines is
somehow unusual. The apparent metrical function of these marks has previously been noted
(e.g. Wilhelm 1991). It is also noteworthy that repeated sections generally correspond to
metrical lines, though there are a few that seem to fill half-lines. These repeated sections
will prove useful in determining syllable count, as they provide multiple lines for comparison,
allowing the equation of parts of lines.
Most apparent exceptions to the pattern of four stressed syllables per line can be resolved
by a small set of explanations. First, one must allow that sometimes lines were composed of
two or more short sentences, as in example (6). The allowance for a strong syntactic break
mid-line is in accordance with the claim, as has been made based on its occurrence in other
languages, that this meter divides lines into cola.
(6) Rs. 3-4
S
S
S
S
te-hé-eš-tab
ta-al-mu-u-ua-ab at-ta-i
a-mu-u-lu-tu-u-um
ˇ
“
teġ=ešt=a=b
talmōv=a=b
atta(i)=j
am=ōl=ud=ō=m
grow=ex=it=sbj big=it=sbj
father=3s.poss look=dur=neg=tr.prf=3s
‘He grew big. He did not look at his father.’
It is possible that long sentences could be divided over two lines, with a clausal break at the
line break, as in (7). The connective =ma, in addition to being a conjunction, could occur at
the beginning of a sentence to indicate a relationship to the actions of the previous sentence,
so such examples may not even require this allowance.
8
(7) Vs. 47-49
S
a-a-i ta-bi-ri
āi tav=iri
if
cast=ag
S
S
S
pu-ú-i
šu-u-ni si-ik-ku-ú-ul-li
fū=vi
šōn=i
sikk=ūl=le
1s.poss=gen hand=th break=dur=1s.opt
‘if only I could break my caster, his hand’7
‘and I could break his right arm inside.’
T
S
S
S
S
e-hé-ep-šu-ul-li-ma
e-ke-e-ni
pa-an-ta-ni hé-e-ra-a-ri
ˇ
ˇērār=i
eġeps=ul=le=ma
egē=ne
fandan=i h
ˇ
break=dur=1s.opt=conn within=rel.sg right=th arm=th
DR
AF
Some lines contain five words instead of four; however, many such lines do not contain five
stressed syllables, because it seems that attributive demonstratives (e.g. andi “this”), attributive quantifiers (e.g. šinzi “second, other”), and conjunctions (e.g. āi “if”) were consistently
unstressed. The behavior of clitics, which do not seem to attach to these words, is consistent
with a lack of stress.
(8) Vs. 18-19
S
S
S
an-ti
ta-a-hi ma-a-an-ni a-ar-ti-i-ta-ni
and=i tāġ=iˇ
mānn=i
ārd=i=j=dan=(n)i
this=th man=th is=3s
city=3s.poss=abl=indv
S
du-ú-ri
tūr=i
go.down=3s
‘He is this man; he ran away from his city.’
Other lines with five words actually seem to have contained five stresses, because every word
is lexical and there is no consistent appearance of two-word units which might bear only a
single stress. There are four such lines, all in pairs, both times corresponding to someone
uttering a curse. The ends of these lines are marked by winkelhaken. Thus it seems possible
that there was a distinct curse meter with five stressed syllables per line (Bachvarova 2012).
Example (9) demonstrates a pair of such lines.
(9) Vs. 56-59
S
S
S
S
S
i-ti7 -i-e
ka-a-zi te-eš-šu-u-pa-aš e-el-ki
sà-am-ma-la-aš-du-uš
id=i=e(n)
kāz=i Teššōb=až
ēlg=i
samm=al=ašt=ož
strike=tr=3s.jus cup=th Teššub=erg
coat=th tear=th=pat=opt
‘May Tesshub strike the cup; may the coating be torn off’
7
A partitive apposition construction. Although Hittite also allows this construction, it is not used in the
Hittite analog of this clause.
9
S
S
S
S
ku-ú-du
ka-a-zi pé-el-le-e-ni e-él-ki-il-la
kūd=o
kāz=i pellē=ni
ēlg=i=lla
throw=pat cup=th canal=dir coat=3s.poss=3p.abs
S
si-i-e-ni
šı̄e=ni
water=dir
‘May the cup be thrown into the canal, its coatings into the water.’
There is also one line which appears to have only three stressed syllables, as it contains only
three words. This line appears twice at the end of a section, preceded by the same line each
time.
T
This line also occurs one other time, which I will address below. It is likely that this shorter
line was permitted as a marker of the end of a stanza (Bachvarova 2012). The first two
words of this line form a syntactic unit, which might suggest that the final stress is the
one that is missing. Catalectic lines at the ends of stanzas are not uncommon in meters
cross-linguistically.
DR
AF
An alternative approach to reconciling this line with the four-stress meter would be to posit
the existence of secondary stress, such that a word could have more than one stressed syllable
counted towards the meter. However, no words in this line are unusually long, so there is
nothing to motivate an additional stressed syllable in a word within this line and not in other
lines.
(10) Rs. 3-5
S
S
S
S
a-mu-u-lu-tu-u-um
te-hé-eš-tab
ta-al-mu-u-ua-ab at-ta-i
ˇ
“
teġ=ešt=a=b
talmōv=a=b
atta(i)=j
am=ōl=ud=ō=m
be=cau=intr=3s big=intr=3s
father=3s.poss look=dur=neg=tr.prf=3s
‘He grew big; he did not look at his father.’
S
S
S
at-ta-i-bi-né-eš-ša
e-né-eš
ši-ti-la-a-i
atta=i=vi=ve=z=na
ene=ž
šid=i=lāi
father=3s.poss=gen.pl=erg=3s god=erg curse=tr=ger
‘The god of his father cursed him.’
There is also a three word phrase that seems to have only two stressed syllables, though it is
unclear which element should be unstressed. If, following Neu, one takes it as a verb followed
by direct object and modifier of the direct object, it seems most likely that the adjective was
the element without stress, though it is unclear how this might have affected the location
of stress on the noun. The phrase appears four times; in three of these appearances (two
of them identical), its position is clear based on the surrounding phrases, always as the first
half of a line:
10
(11) Vs. 3-4
u-ú-ru
te-e-lu
tap-šu-ú pa-pa-an-ni ši-ta-ri-il-lu-u-um
“
pūr=u
tēl=u
tapš=ū fab(a)n(i)ne šid=ar=ill=ō=m
begin=3s.prf quarrel=th low=th mountain curse=it=inch=tr.prf=3s
‘He began a base quarrel, and he began treating his city poorly.’
T
The other instance of this phrase has a less obvious position within a line. Following the line
in question is the third occurrence of the catalectic line discussed above, again appearing at
the end of a section. However, if the lines are thus divided, this leaves another three-stress
line preceding them, which is preceded by a winkelhaken and thus seems to be the beginning
of a section. An alternative analysis could make this section into two 5-stress lines, but they
do not form a curse like the other instances of lines with 5 stressed syllables, so such an
analysis does not seem strongly justified.
(12) Vs. 19-22
DR
AF
u-ul-ui-ne-e-ma
a-am-mi-i-ib u-um-mi-in-ni
“
ōlvi=nē=ma
ām=ı̄=b
ōmmin(i)=ne
another=indv=conn look=tr=3s land=indv
‘He was looking at another country.’
u-ú-ru
te-e-lu
tap-šu-ú a-ar-ti-i-ma
a-ma-ri-il-lu-u-um
“
pūr=u
tēl=u
tapš=u ārd=i=ma
am=ar=ill=ō=m
began=3s.prf quarrel=th low=th city=3s=conn harm=it=inch=tr.prf=3s
‘He began a base quarrel, and he began treating his city poorly.’
a-ar-ti-bi-né-eš-ša
e-ne-eš
ši-ta-la-a-i
ard=i=j=vi=ne=šša
ene=ž
šid=i=lāi
father=3s.poss=gen.pl=erg=3s.abs god=erg curse=tr=ger
’The god of his father cursed him.’
If instead of Neu’s analysis of pūr=u tēl=u tapš=u as a verb phrase ’he began a base quarrel,’
one follows Haas & Wegner (2007) in analyzing it as three asyndetic verbs, ’he stared; he
puffed himself up; he became angry,’ it would be possible to split the phrase over two lines
without compromising the strict prohibition on enjambment, and in this way supply the
missing stress of the first line in (12). This would also resolve the surprising appearance here
of the theme vowel =u, which mainly occurs in foreign names (Wilhelm 2008). However, it
would require an assumption that these three verbs could occur as a unit with two stresses
between them and also occur separately as stress-bearing units. As the analysis is unclear,
I remain ambivalent regarding this issue, and will omit these lines in the data except where
otherwise stated.
11
2.3
Couplets and Tercets
As described above, grouping of lines into couplets with occasional tercet is a standard feature
of poetry in several languages of the ancient Near East and surrounding regions, though most
of these meters are not generally analyzed as syllable-counting or stress-counting and do not
show any more similarity in length between lines within a couplet than any other lines within
the poem. In consideration of these meters, it is possible to imagine a couplet structure in
Hurrian poetry, though it will depart from them in some ways.
The existence of couplets in Hurrian meter is further supported by the observation that line
lengths in the prayer of Tadoheba (ChS I/1 41), as marked by winkelhaken, are generally
consistent, with the occasionalˇline of approximately twice this length (Wilhelm 1991), which
may suggest some confusion between the marking of lines and of couplets.
DR
AF
T
The lines within the Parables fall into groups of two and occasionally three lines, which form
a semantic and syntactic unit. Repeated and partially repeated sections of text often come
in pairs. The 5-stress curse meter also supports couplets, as it only shows up in pairs of lines.
The paragraph breaks, often accompanied by horizontal dividing lines, regularly divide up
even numbers of lines, and where they do not, there is a group of lines best treated as a tercet.
However, there is one place where a horizontal divider occurs in the middle of a couplet (Vs.
50-53), which is also in the middle of a semantically defined paragraph. The motivation for
this divider is unclear; perhaps it was drawn to aid the scribe in ending the first side of the
tablet with a strong paragraph break, without leaving blank lines at the end, as this divider
comes in the middle of the final paragraph of the first side. The divider might also have been
drawn in order to keep the lines of text horizontal, as the previous lines had begun to slant
upwards and the lines after this break are closer to horizontal. This location for a the divider
may have been chosen because the couplet being broken would likely still be recognizable as
a unit because both lines begin with the same word, tab-re-e-en6 -ni.
The winkelhaken which mark line ends were perhaps included to assist the reader in keeping
track of potentially unclear metrical line breaks (Thiel 1975), as they often occur between
instances of a word repeated at the end of one line and the beginning of another, as well
as between lines of a tercet, though not all tercets are marked this way, if they are parallel
to previous tercets. The occurrence of winkelhaken in these positions seems to support the
hypothesis that tercets existed as an acceptable but marked variant on the normal structure.
Lines within groups have a greater similarity in syllable count than is found across the whole
poem,8 which further supports the argument for such groups.9 If this poem had no couplet
structure, then the difference in syllable count between natural groups, forming semantic
and syntactic units, would not be significantly different from the difference in syllable count
between groups assigned in any other way. In calculations, pairs with the mentioned catalectic
8
The range of line lengths in the poem is 9-15 syllables, or 10-15 if the lines of Vs. 19-22 in (12), whose
line assignment is uncertain, are omitted.
9
This pattern of couplet-dependent variability in line length is parallel to the pattern in Tamil epics, in
which lines have a strong tendency to match in syllable count and metrical pattern within couplets, but
show great variability between couplets (Ryan 2011). This does not imply any connection between Hurrian
and Tamil metrics, though if the Hurrian couplet meter is borrowed from an Indo-Iranian language, such a
connection could be plausible.
12
line in example (10) were omitted, because if it was catalectic, it was not meant to match the
length of other lines. Couplets which occur multiple times were only counted once, though
couplets which differ by at least one word were counted independently.
T
Natural groups were compared to those formed by pairing each line with the next line that
occurred within a different natural group; their average differences in syllable count for
natural groups is 0.88 syllables (0.95 if the lines of uncertain grouping in Vs. 19-22 are
not omitted) and for artificial groups is 1.34 syllables (1.35 if the uncertain lines are not
omitted); the difference is statistically significant (p value = 0.022).10 The result is weaker if
the artificial groups are assigned completely randomly: the average difference between length
in lines grouped in this way is 1.19 syllables (p value = 0.098 for this average being different
from the average difference between naturally grouped lines). This might suggest that having
adjacent couplets differ in length was a desirable feature of the meter, perhaps to emphasize
the couplet structure. 33% of naturally grouped lines match exactly in length, while 15-17%
of the artificially grouped lines do, depending on the method of grouping.
DR
AF
As a starting point for syllable count calculations, each written vowel or sequence of identical
vowels was treated as a single phonetically realized vowel; each sequence of three different
vowels in which the middle vowel was high was treated as two syllables, divided by a glide; and
each sequence of two adjacent vowels was treated as a monosyllabic diphthong. The first and
second of these are standard, fairly uncontroversial assumptions, but the third assumption
has been the matter of some discussion, which has not yet yielded a clear answer; it will
be discussed further later. If all sequences of two vowels are treated as being disyllabic, the
average differences in syllable count between lines within natural groups is 1.1 syllables and
between lines in artificially formed groups is 1.5 syllables.
If one assumes that in the poem as written, every couplet/tercet was internally consistent
in line length, several potential features of Hurrian phonology or spelling conventions can be
identified as being consistent with greater similarity within pairs of lines. In some places, it
is apparent which part of the line must be the location for an alternative reading of syllable
count, based on partially repeated couplets which differ in only one or two words. These
features can also reduce some of the apparent long sequences of unstressed vowels. Identifying
spelling conventions is hindered by the variability in some aspects of Hurrian spelling and the
limitations in vowel information from syllabic and consonantal writing systems, which make
testing a hypothesized spelling convention difficult. Fortunately, logograms are infrequent,
so there is rarely the additional problem of determining the underlying Hurrian word. The
one logogram that appears in the Parables is <IGI> “eye”; following Neu (1996), it has been
treated as having a Hurrian value [ši].
10
This pattern of line length and couplets cannot be found in Hurrian prose, even in formal texts with
some stylistic features that are also found in poetry. For example, in the first half of the third column of the
Mitanni letter, one of the best preserved continuous sections, lines assigned based on semantic and syntactic
divisions vary from 8-26 syllables, even allowing much flexibility in the syntactic units that can constitute a
line. The average differences between groups of such lines (4.5 syllables) is not significantly different from
the average difference between artificially grouped lines (4.1 syllables). However, phrases frequently contain
3-5 words, because syntactic units in this text tend to match up with lines on the tablet.
13
2.4
Phonological and Orthographic Features
2.4.1
Disyllabic Vowel Sequences
DR
AF
T
The first analysis to be considered is that plene writing could be used to mark a consonant,
possibly a glottal stop, because cuneiform lacked a sign for a glottal stop. Although plene is
generally interpreted as marking long vowels, as mentioned above (§1.1.3), it seems to have
additionally been used in Hittite to clarify vowel quality, mark accent on short vowels, and
to write monosyllabic words (Hoffner & Melchert 2008). In Old Babylonian and Akkadian,
word-internal glottal stops could be marked by plene or by sequences in which a vowel was
written independently instead of in a VC sign with the following consonant. The glide [j]
could also be written with <a> signs. In many cases, the same spellings used to mark these
consonants could also be used for writing vowel sequences (Huehnergard 2011). Though a
glottal is not generally included in hypothetical phonological inventories, Wilhelm suggests
that it could have been part of the Hurrian phonological inventory, which could explain why
certain vowel sequences seem to be disyllabic (Wilhelm 2008).11 Treating every orthographic
sequence of two different vowels as disyllabic does not produce strong consistency of line
lengths within couplets, which suggests a contrast between vowel sequences separated by a
glottal stop and diphthongs.
The Hurrian texts in the consonantal Ugaritic alphabet can contribute to the question of how
to treat vowel sequences, although there is a small number of Hurrian words known in both
writing systems, because Hurrian does not lend itself well to a consonantal writing system,
frequently making interpretation of these texts uncertain. Ugaritic had three aleph signs
for a glottal stop + vowel: P a P i P u (van Soldt 2010). Some words are consistently written
with these aleph signs, e.g. hP urn for [haurun] ‘heaven’ (Dietrich & Mayer 1999). In texts
ˇ
ˇ
written in the Ugaritic consonantal
alphabet,
most words spelled with <u> word internally
“
P
could also appear with the sign < u>, which could be interpreted as reflecting
a glide or a
glottal stop and a vowel (Lam 2011). However, it is often assumed that Hurrian lacked a
glottal stop, and thus that in Hurrian texts, the Ugaritic glottal stop + vowel signs simply
represented vowels (Speiser 1940).
Hurrian words which have appeared both in the Ugaritic alphabet and the cuneiform syllabary
provide evidence for interpreting these signs. For example, <tP inm> has been interpreted
¯
as a form of “man,” followed by suffixes. This word was rendered in syllabic cuneiform as
<ta-a-e> (Lam 2011), which is consistent with an analysis as [taPe] or [taPi]. The consonant
that was marked by plene in Hurrian might also have been a glide, as suggested by the name
ęauška, written as <tP utk> in the Ugaritic texts (Lam 2011), but which appears as ęawuška¯ ¯
(d ša-wu-uš-ga-an) in later Hittite (Beckman 1998).
In the Ugaritic alphabet, aleph signs are consistently used for vowel-initial Hurrian words
(Lam 2011), but this does not necessarily indicate that all or apparently vowel-initial Hurrian
words written in the cuneiform syllabary actually began with glottal stops, as there is no
clear parallel elsewhere in cuneiform for this usage. The status of word initial glottal stops in
Akkadian and its descendants is unclear; they might not have been phonologically contrastive
11
Perhaps the glottal stop was lost late during the recorded history of Hurrian. This might explain why
the <a-e> vowel sequence of the instrumental, for example, generally seems to be disyllabic (Wilhelm 2008),
though in later Hurrian it tends to be written as a single vowel (Diakonoff 1985).
14
in this position, so interpretations of word initial VC or V-VC are varied (Buccellati 1996),
while in Sumerian a glottal stop is generally not included in the phonological system at all
(Edzard 2003). Word initial glottal stops would not have a large impact on the syllable
count of the lines, though in the few places where it would be predicted to prevent elision,
it improves the matching of lines within couplets, which provides a limited point of support
for the existence of word initial glottal stops.
The following lines demonstrate the context and effect of positing glottal stops in orthographic
sequences of a plene vowel before another vowel:
(13) Vs. 31-32
T
[na-a]-li ma-a-an-nu-u-bur ma-[a-an-n]i tar-šu-ua-a-ni
“
nāl=i
mān=ō=vor
mānn=i
taržuvān=i
deer=th is=th=neg
is=3s
man=th
DR
AF
‘It is not a deer; it is a man.’
syllable count: 11
[hal-zu-u]h-la
ma-a-an-ni e-eb-re-eš
na-a-hé-la-a-i
ˇ
ˇ
ˇ
halz(i)=ō=ġl(i)=a
mānn=i
ēvr(i)=j=š
nāġ=e=lāi
ˇ
district=th=lord=ess
is=3s
lord=3s.poss=erg appoint=tr=ger
‘As a district administrator, his lord appointed him.’
Syllable count:10
Syllable count if =lai is disyllabic: 11
The most common vowel sequence with a plene vowel is <Ca-a-i>. The <a> vowel is
particularly useful because it is not complicated by the ambiguity of spellings which are used
both for high vowels and for glides, so it has been taken as the main focus of testing for the
disyllabicity of vowel sequences. In vowel sequences with plene high vowels, it is sometimes
clear that the vowel sign is to be read as a consonant, based on alternative spellings of a
word. It is sometimes unclear whether a vowel sign is best interpreted as reflecting a vowel
or a glide; the interpretation of such vowels also depends on the conditions assumed for the
high vowel diphthongization rule proposed subsequently (§2.4.3).
Table 1 shows how positing disyllabic vowel sequences improves the matching of line length
within groups of lines. For comparison, the average difference in artificially paired lines is
given; the difference in length between artificially paired lines remains consistently higher
than the difference between naturally paired lines, though disyllabicity in plene vowel sequences has a side effect of making the overall line length slightly less variable. It is likely
that vowel sequences other than <Ca-a-i> also could have been disyllabic, so a column is
included to show the results of assuming that all vowel sequences including a plene vowel are
disyllabic, but is a negligible difference between including this more general rule or not.
How to interpret vowel sequences seems to depend on the particular morpheme, as one morpheme may have vowels divided by a consonant and another may have a diphthong: =la-a-i
15
consistently scans better as disyllabic, and =ma-a-i consistently scans better as monosyllabic.
Although there are not enough examples of each morpheme to test this distinction reliably,
the variation of spellings for each morpheme are noteworthy: the variants =ma-a-i and =ma-i
both occur, but =la-i never occurs. If the <a-i> vowel sequence is interpreted in both cases
as the instrumental, or at least historically derived from it, as Neu (1996) suggests, the difference is surprising. However, as the specific history of the development of these two forms
is unknown, I will suggest that there was perhaps a glottal stop in the sequence preceding the
instrumental in <la-a-i> /laP + ai/, but not in <ma-a-i> /ma + ai/. If these two suffixes
differ in their syllable count, this might in part explain their distribution, which is not clearly
motivated by differences in meaning, as both function as gerunds.
disyllabic Ca-a-i
0.88
0.93
1.34
1.13
disyllabic Ca-a-i
except =ma-a-i
0.83
disyllabic other
plene vowels
0.81
1.18
1.16
T
naturally
paired lines
alternatively
paired lines
unamended
DR
AF
Table 1: Average difference in number of syllables, between pairs of lines, with the assumption
of disyllabic vowel sequences
2.4.2
Elision
The next process to be discussed is elision, the elimination of word final vowels when followed
by a vowel within a line. This has parallels in the observed phonology, as elision could occur
word internally in certain morphological conditions. The exact conditions for word internal
elision have not yet been definitively identified; it may only occur with the theme vowels /a/
and /i/12 (Wilhelm 2008). The clearest results for elision between words in this text are also
for the vowels /a/ and /i/, though this may reflect these vowels’ frequency word finally. For
the sake of consistency in the absence of strong motivation for restricting elision to a further
subgroup of vowels, statistics are based on the assumption that all short vowels could elide,
though elision of /e/ has no net effect and elision of /u/ has a negative effect. For high
vowels, the effect of elision might overlap with the effect of the high vowel diphthongization
rule proposed in §2.4.3, if it could apply between words as well as within them. Elision in
Hurrian poetry has previously been suggested for IBoT II, in consideration of patterns of line
length (Thiel 1975). The following lines demonstrate the context and effect of elision:
(14) Rs. 58-59
a-a-i ši-ia-li-ri pu-ú-i
šu-u-ni
sí-ik[-k]u-ul-li
“
āi sijal=iri fū=vi
šō=ni
sikk=ul=le
if
build=ag 1s.poss=gen hand=indv break=dur=1s.opt
‘If only I could break my builder, his hand’
syllable count: 12
12
e.g. fı̄radi + ardi > fı̄radardi ‘nobility’ (Wilhelm 2008)
16
e-h[é-e]p-šu-ul-li-i-ma
i-ke-e-ni
pa-an-ta-ni hé-ra-a-ri
ˇ
ˇērār=i
eġepš=ul=le=ma
egē=ne
fandan=i h
ˇ
break=dur=1s.opt=conn within=rel.sg right=th arm=th
‘and I could break his right arm inside.’
syllable count: 14
syllable count after elision: 13
DR
AF
T
If elision was indeed at work, it did not occur across line breaks, which occurred at major
syntactic breaks. It is possible that elision also failed to occur at a line internal syntactic
break, though there is limited evidence for the outcome of vowel sequences in this environment. Lines with an internal syntactic break frequently have this break after the 5th or 6th
syllable. However, not all lines have a syntactic break, and having a word break here is not
significantly more common here than in other positions. Thus, it does not provide evidence
against cola divided by a break, as has been suggested for other varieties of this 4-stress
meter, nor does it provide very strong evidence in favor of it. The behavior of vowels at
line-internal syntactic breaks does not resolve the issue, because line internal breaks have
different effects in different poetic traditions and there is insufficient data to test it within
this text. In some Latin verse types, hiatus without elision was permitted at the metrically
required caesura, and sometimes elsewhere as well, particularly at syntactic breaks, though
it was generally avoided (Sturtevant 1916). There seems to be no tendency in the Hurrian
Parables to prevent a vowel sequence from occurring at syntactic breaks, and it seems that
strong syntactic breaks did not preclude elision, but it is possible that they could license hiatus. If mid-line syntactic could optionally block elision, the lines could be improved further;
however, I wish to be cautious about positing optional rules, because the text is too short to
identify patterns in support of rules which did not apply systematically.
Table 2 shows how applying the elision rule improves the matching of pairs of lines, as well
as how allowing optional hiatus at the mid-line syntactic break would further improve this
matching. Although the improvement made by the assumption of the elision rule is small,
applying this rule in conjunction with other rules will show a greater result, as shown in
Table 4.
unamended
actually paired 0.88
lines
alternatively
1.34
paired lines
with elision
0.86
with optional
hiatus at break
0.71
1.24
1.18
Table 2: Average difference in number of syllables, between pairs of lines, with the assumption
of elision
2.4.3
High Vowel Diphthongization
The last process to be considered is high vowel diphthongization; I suggest that high vowels
became glides when followed by another vowel word internally, reflected in the spelling by
17
a high vowel followed by a glide. In some cases it is clear from the morphology that such
spelling sequences were the outcome of high vowels followed by other vowels.
T
To verify this rule would depend on the knowledge of underlying sounds which orthography does not always mark unambiguously, because it does not make systematic distinctions
between several sounds. <u> signs were sometimes to be read as labial fricatives or stops
“
instead of glides; in some words,
the <u> signs alternate with <p> and <b> signs, which
“ with few attestations, a lack of alternation is not
are assumed to reflect fricatives. In words
sufficient reason to assume that the sign reflects a glide. Both [j] and [w] could be written
with high vowel signs <i> and <u>. In Hurrian texts, <u> signs were also sometimes to be
read as [o], but in this text they seem to consistently be written distinctly. Although in most
cases there are two distinct signs, sometimes a single <Ci> sign could also be used for both
[Ci] and [Ce]. The established readings of high vowels versus mid vowels generally match up
with places in the text where assumption of the gliding rule improves the matching of line
lengths or not, which is a point of support for the existence of this rule. If confirmed, this
rule also provides a distinct line of evidence for the reading of vowel qualities.
DR
AF
(15) Vs. 9-11
na-a-li pa-pa-an-ni-iš
ši-ta-ar-na
ku-lu-u-ru-um
nāl=i
fab(a)n(i)=ni=ž
šid=ar=na
kul=ūr=o=m
deer=th mountain=indv=erg curse=nmz=indv.pl say=ex=dur=3s
‘The mountain said curses to the deer:’
syllable count: 11
i-ia-a-at še-e-du-i-li-ia-ni-iš
ši-ta-a-ra
na-a-al-li-iš
“
“
ijā=t(ta)
šēdu=ili=a=ni=ž
šid=ār=a
nāl(i)=ni=ž
why=1s grow=pass.prt=indv=erg curse=tr=3s deer=indv=erg
‘Why does the deer, grown fat, curse me?’
syllable count: 12
syllable count if the internal <ii> is nonsyllabic: 11
“
Depending on the reading of the underlying sounds, this and the previous rules suggest that
what is written as a vowel sequence with a repeated high vowel could represent two vowels
separated by a glide, two vowels separated by a glottal stop, or a glide followed by a vowel.
The specific prediction for how the phonological rules should interact depends on how they are
conceived. I propose that the high vowel diphthongization rule applied only in places where an
underlying high vowel occurred before another vowel, rather than being a rule reducing uwV
> wV and ijV > jV. There was a reduction process like this in Babylonian poetry, in which
the vowel /i/ could become a glide before other vowels, but such sequences could also remain
disyllabic (von Soden 1981). The only clear examples which would clarify the character of this
rule are combinations of morphemes with previously established shapes. It is also possible
that an orthographic convention developed based on a phonological diphthongization process,
such that glides between a consonant and a vowel were written as if preceded by a homorganic
high vowel. This is consistent with the lack of cases where there is a sequence written with a
18
consonant followed by a glide where the glide sign is to be interpreted as a glide, except word
initially, as seen in the first word of the second line in example (15).13 I will tentatively assume
this orthographic convention. If scribes writing Hurrian were native speakers of Hittite or
were originally trained in writing Hittite, the development of such a convention is particularly
plausible, as a glide likely occurred to break up Hittite vowel sequences (Kloekhorst 2011).
The spelling could additionally have developed because of the ambiguous values of signs for
high vowels and glides.
DR
AF
T
It this rule was present, it could clarify some of the conditions in which the internal elision rule
applied and what the underlying sounds in given words were, based on the different outcome
which each combination of assumed rules and sounds would predict. Because it only would
occur with high vowels and glides, and not mid vowels or labiodental fricatives, it could
also establish the pronunciation of ambiguously spelled words or confirm pronunciations that
have been assumed for other reasons. It could alternatively be imagined as an optional rule
applied when metrically necessary, perhaps as an extension of an existing phonological rule
in the spoken language. I include the results of making this assumption, though I still wish to
maintain a distinction between the results that can be made by applying rules systematically
and those that can be obtained by applying rules optionally, because the results of the latter,
while superficially greater, are less meaningful.
unamended
actually paired 0.88
lines
alternatively
1.34
paired lines
/i/ gliding
0.90
/i/ and /u/
gliding
1.0
optional gliding
of /i/ and /u/
0.69
1.32
1.42
1.29
Table 3: Average difference in number of syllables, between pairs of lines, with the assumption
of high vowel diphthongization
Although the data is not significantly improved – and in fact the average difference between
line lengths is increased – by the application of this rule alone, the rule fares much better
when applied together with the previous rules, as will be seen in the following discussion and
Table 4.
2.4.4
Net Result
While each rule alone has a small effect, they work together to improve couplet length match
more than the sum of each one independently. The following table shows the cumulative
effect as each rule is applied. For the first three rules, I include only changes made if they are
applied without exception. The next column is based on interpretations for two individual
morphemes based on the presence or absence of a glottal stop within them: that the gerund
morpheme =mai is monosyllabic, and that the morpheme =iri ‘one who does X’ begins with
13
There are, however, places where <uV> signs occur after VC signs when the <u> is to be read as a
“ between <ua> 42%
labial stop or fricative, as in example (1),“where final sign of the morpheme =ppa varies
“ <ap>,
and <pa> 56% in the 12 times it appears in this recurring couplet. The preceding sign is always
which is why the consonant is taken to be [p], though [f] might also be imagined.
19
a glottal stop. Both appear six times, and in each case, making this assumption improves
the matching of syllable count. Assuming that the form of the agent-deriving morpheme was
/Piri/ means that when it followed a vowel, the resulting vowel sequence would be disyllabic.
Although neither vowel is doubled in the spelling of such sequences, it is noteworthy that all
cases of disyllabic Ca-a-i improving the meter are found within a morpheme, so it is possible
that plene spelling to mark glottal stops within morphemes was not used across morphemes.
In the column for optional rules, I change the calculations to treat hiatus at syntactic breaks
and high vowel diphthongization as permitted but not required.
elision
0.81
high vowel
gliding
0.74
P in specific optional
morphemes rules
0.62
0.36
1.03
1.11
1.18
T
unamended disyllabic
Ca-a-i
actually paired 0.88
0.93
lines
alternatively
1.34
1.13
paired lines
1.16
DR
AF
Table 4: Average difference in number of syllables, between pairs of lines, with the assumption
of all of the preceding rules
After applying these rules, the range in line length is unchanged (10-15), but the maximum
difference between paired lines is reduced to two syllables. If the optional and word-specific
rules are also applied, the maximum difference between paired lines is one syllable, while the
maximum difference between lines paired by shifting remains four syllables. Having optional
rules which allow variation number of syllables and placement of stress in a meter is not
uncommon (Halle & Keyser 1971), so including such rules here would not be without basis.
However, it can be unclear how to interpret such rules in written poetry; they can indicate
tolerance for variation in the meter, reflecting no phonetic change (Halle & Keyser 1971), but
they can also indicate variation in pronunciation to fit the meter (Kiparsky 1977). It is not
clear which of these possibilities would be the case for this text and whether phonological
variation in Hurrian poetry could be assumed to apply to the language generally. Furthermore, it is impossible to check such rules within a small corpus, and thus there is a risk
that they could provide apparent explanations which do not reflect actual rules of Hurrian
composition. With this caveat in mind, they are nonetheless interesting to consider, which is
why I have given them independently from the systematically applied rules, to give a better
impression of some of the processes which may have allowed pairs to match exactly. With
all of these rules, there remain a total of 15 syllable differences in 42 pairs of lines; further
rules could likely eliminate these remaining differences.
Of these remaining differences, there is no trend to suggest that there might have been a
pattern for either the first or second of a pair of lines to be longer. There is possibly a
pattern that tercets match in their first and third line, with the middle line one syllable
longer. Two of the four tercets show this pattern, but they are built on the same formula,
which may be responsible. One of the tercets has all three lines of the same length. The
syllable count of the lines in the remaining tercet, given in (12), is uncertain; the final line is
the catalectic one seen elsewhere in pairs. It is possible that the first line is also catalectic to
match the final line, but this would be an uncommon position within the stanza for catalexis.
20
2.5
Stress Patterns
2.5.1
Dolniks
T
Having established a likely syllable count, I will return to stress. The results of analyzing
the stress patterns may provide indications of whether hypothetical emendations of syllable
count were proper emendations to make. The stress was more regulated than just the basic
pattern of four stressed syllables per line; two adjacent stressed syllables were avoided, as
were more than three unstressed syllables between stressed syllables. This pattern is similar
to the dolnik meters found in Russian and also proposed to exist in German and English.
In these meters, the number of unstressed syllables between stressed syllables is usually 12, and there can be 0-2 unstressed syllables preceding the first stressed syllable (Gasparov
1963). In an alternative Russian form, the typical number of unstressed syllables between
stressed syllables is three. There is a tendency to have a single unstressed syllable in the final
interval between stressed syllables. Some dolnik meters allow greater flexibility in aligning
lexical stress with metrical stress positions: in German and English dolniks, the final stressed
position is allowed to be occupied by an unstressed syllable (Tarlinskaja 1992).
DR
AF
Dolniks exist as an independent form of poetry, not just loose versions of a syllabotonic meter;
they tend to avoid having a high percentage of the same pattern of stressed and unstressed
syllables. Russian dolniks avoid having lines with unstressed intervals all of the same length,
though English and German dolniks are more tolerant of such patterns (Tarlinskaja 1992).
2.5.2
A Hurrian Dolnik
The Hurrian meter can easily be described based on these dolnik rules. The main difference
is that lines consistently ended with a final unstressed syllable preceded by a stressed syllable,
because based on stress placement in Hurrian, it would be impossible to end a line with a
stressed syllable. The meter of the Hurrian Parables allowed 1-3 unstressed syllables between
stresses, though two was most common, and allowed up to two unstressed syllables preceding
the initial stressed syllable. Like in Russian dolniks, lines which only had unstressed intervals
of the same length were avoided.
It is possible to describe this stress pattern in terms of feet, but the metrical reality of foot
structure is uncertain; though feet are useful to explain periodicity, line internal structure,
and the locations of word boundaries (Hayes 1988), there are no features present in this
poem that could not also explained without foot structure. Following the common method
of describing dolniks, the metrical pattern can instead be described based on the number
of unstressed syllables occurring between stressed syllables, which furthermore obviates the
issue of how the line initial unstressed syllables should be integrated into foot structure.
Following is an example of what the stress pattern looks like in a standard couplet.
(16) Vs. 31-32
S w
w
S w
S
w w w S w
[na-a]-li ma-a-an-nu-u-bur ma-[a-an-n]i tar-šu-ua-a-ni
“
nāl=i
mān=ō=vor
mānn=i
taržuvān=i
deer=th is=th=neg
is=3s
man=th
‘It is not a deer; it is a man.’
21
w
S
w
S
w S
w
w w S w
[hal-zu-u]h-la
ma-a-an-ni e-eb-re-eš
na-a-hé-la-a-i
ˇ
ˇ
ˇ
halz(i)=ō=ġl(i)=a
mānn=i
ēvr(i)=i=ž
nāġ=e=lāi
ˇ
district=th=lord=ess is=3s
lord=3s.poss=erg appoint=tr=ger
‘As a district administrator, his lord appointed him.’
T
I looked at the patterns of how many unstressed syllables occur between the four stressed
syllables of each line, using the syllable structures resulting from the phonological and orthographic rules discussed above. The results of stress patterning seem to be consistent with
these rules; without them, there would not be a clear upper limit on how many unstressed
syllables can occur between stressed syllables. The stress patterns also are in support of the
widely accepted penultimate stress assignment in Hurrian words. I omitted the two 5-stress
couplets from calculations, because it is unclear how they ought to fit with the stress structure
of the rest of the poem, though they at least follow the same patterns for line ends.
DR
AF
Stress assignment is generally clear; it depends on whether elements following a root are
suffixes and thus part of the accentual unit, or clitics, which do not influence accentuation.
Most elements can be clearly identified either as suffixes or clitics based on position, meaning,
or behavior. In ambiguous cases, I categorized them based on the most frequent position of
plene vowels preceding that element. Based on these methods of stress assignment, I obtained
the following data.
first stress
one unstressed 17
syllable
two unstressed 49
syllables
three unstressed 7
syllables
second
stress
35
penultimate
stress
9
last
stress
73
34
40
0
4
24
0
Table 5: Number of unstressed syllables following a stress, by position
It appears that stress could shift left when a final vowel of a word was elided, based on the
line in Vs. 23-24. Here, if the stress does not shift with elision, there would be two stressed
syllables together. Even if this line were taken as evidence against elision, it would still be
a problem for the meter because it would have four unstressed syllables together. In all
other places where elision has been applied, lines would follow the stated meter regardless of
whether it caused the stress to move.
It remains unclear how to assign the two stresses in the three word phrase puru telu tapšu; if
the last word is indeed an attributive adjective which loses its stress in this phrase, it seems
that the stress of its noun is shifted right, because otherwise this would result in sequences of
more than three unstressed syllables in a row. This might suggest that the phrase is better
taken as three verbs, as discussed above.
22
All lines end in a trochaic rhythm; a tendency for this would be expected as a result of
penultimate stress in Hurrian, but it appears that clitics were avoided at the end of a line,
perhaps because of a metrical rule calling for this consistent rhythm. The final stressed
syllable is usually preceded by 2-3 unstressed syllables, which is unlikely to be accidental
given the frequency of single unstressed syllables elsewhere and the relative infrequency of
triple unstressed syllables. This calls to mind the tendency in many meters of ancient IndoEuropean languages to have invariant line ends. These fixed cadences are based on syllable
weight, not stress, but as stress and syllable length tend to align in Hurrian, the comparison
is not unreasonable. However, greater regularity at the ends of lines is a common feature of
meters cross-linguistically (Fortson 2010), so its presence is not necessarily suggestive of a
connection between these meters.
DR
AF
T
As in other dolniks, lines may have 0-2 unstressed syllables preceding the initial stressed
syllable. 46% of lines have no initial unstressed syllables, 31% of lines have one initial
unstressed syllable, and 18% of lines have two initial unstressed syllables. However, there are
four lines which seem to have more than two unstressed initial syllables. Because the number
is small, and three of the lines are very similar, it seems likely that this is not a feature of
the meter, but rather indicates a problem with assumed syllable count or stress assignment.
Three of those cases are variants of the same formula, beginning with the conjunction <a-a-i>,
as in:
(17) Vs. 47-48
www S w
a-a-i ta-bi-ri
āi tav=iri
if
cast=ag
S w
S w
pu-ú-i
šu-u-ni
fū=vi
šō=ni
1s.poss=gen hand=indv
w
S
w
sí-ik-ku-ú-ul-li
sikk=ūl=le
break=dur=1s.opt
‘If only I could break my caster, his hand’
This might call into question the disyllabic interpretation of this word, which would also make
one want to reexamine other places where a vowel sequence was assumed to be disyllabic. It
is less likely that an unstressed word would influence the stress placement of a word to its
right, when stress is calculated from the right edge. The other case is:
(18) Rs. 14-15
w w w
S w
S
w
S
w
w
w Sw
a-mu-ú-mi-ne-e-ua
ma-a-an-ni e-eb-re-eš
na-ah-hi-la-a-i
ˇ
“
amūmi=nē=va
mānn=i
ēvr(i)=i=ž
nāhhˇ=i=lāi
ˇˇ
administrator=indv=dat is=3s
lord=3s.poss=erg appoint=tr=ger
‘As administrator, his lord appointed him.’
Both of the morphemes on the root of the first word must be suffixes, unless the analysis of
this sentence is changed greatly. To reconcile this line with the dolnik restrictions, one might
suggest that the <i> is merely graphic and does not reflect a phonetically realized vowel,
because it would be much more difficult to motivate the writing of a phonetically nonexistent
23
vowel word initially or with plene spelling. If one assumes that some cases of <i> are merely
graphic, this could be used to resolve some of the remaining places where paired lines still
do not appear to match in length. This line as well as one other would also require hiatus at
the center of the line to avoid a sequence of two adjacent stressed syllables.
There is a tendency for paired lines within couplets and triplets to match in stress pattern.
A comparison with the average difference in number of weak syllables in each position within
pairs formed by shifting is given in Table 6. The differences here indicate that the matching is
unlikely to simply be the result of overall tendencies for certain sequences in certain positions
of the line. The final column gives the average of the sum of the differences from each section
of the line. It is also noteworthy that a greater percent of the natural pairs than artificially
grouped pairs match exactly in part of the line or the entire line.
following sec- following
average
ond stress
third stress total
0.53
0.61
2.65
0.81
DR
AF
0.76
T
before first following
stress
first stress
actually paired 0.95
0.53
lines
alternatively
0.92
0.65
paired lines
3.14
Table 6: Average difference in number of weak syllables, by position
It seems unlikely that all couplets and tercets matched exactly in stress pattern, because
there is much deviation, even after the application of rules to resolve the differences in
syllable count. The number of unstressed syllables preceding the first stress does not seem to
be considered in matching the stress pattern within pairs, as it shows no tendency towards
greater similarity within pairs. The greater matching in stress pattern of natural pairs reaches
significance for positions following the second stress (p = 0.037), and approaches significance
for the positions following the third stress (p = 0.08) and for the average total (p = 0.084). It
seems likely that matching in stress pattern was stylistically desirable, though not metrically
necessary. Note that the lines of the initial couplet match exactly in stress pattern.
(19) Vs. 1-3
S
na-a-li
nal=i
deer=th
w S
w
i-te-[e-]i-né-eš
itej=ne=(e)š
body=abl=erg
w S
w
pa-pa-an-ni-iš
fab(a)n(i)ni=ž
mountain=erg
w S
w
me-la-ah-hu-um
ˇ
mel=ahˇh=o=m
ˇ
ˇ
bring=forth=tr.prf=3s
‘A mountain brought forth from its body a deer.’
S
na-a-li
naal=i
deer=th
w S w
u-ul-bi-i-ni
ōlvı̄=ni
another=indv
w S w
pa-pa-an-ni
fab(a)n(i)=ni
mountain=indv
‘The deer went to another mountain.’
24
w S
w
ha-pa-a-na-ab
ˇab=ān=a=b
h
ˇ
go=fct=intr=3s
Perhaps extra care was given to produce an ideal couplet to begin the poem. This might
explain the unusual word order of the first line, in which the absolutive object comes first
in the sentence, if the ideal line calls for an initial stressed syllable. However, this word
order could also be have been motivated by parallelism with the following line, in which the
absolutive is the subject of an intransitive verb and thus naturally comes first in the sentence.
There are other instances of unusual word order in the poem, but they generally do not seem
to be motivated by the meter.
2.6
The Possibility of Other Examples of This Meter
DR
AF
T
The parallel Hittite parables written alongside the Hurrian text does not seem to share the
Hurrian dolnik meter. Although this text can be explained as following the four-stress meter,
the lines are too long to follow the Hurrian dolnik restriction on the number of adjacent
unstressed syllables and very widely in length, with lengths ranging 9-24 syllables. While
the lines of the Hurrian Parables have an average length of 11.9 syllables and a standard
deviation of 1 syllable, the Hittite Parables have an average length of 15.4 syllables and a
standard deviation of 3.4 syllables.14 Hittite stress is often discussed in the context of Hittite
meter, but work on this topic is mainly focused on which words contain stress and not on
the position of stress within words (e.g. Melchert 2007; Kloekhorst 2011); plene spelling in
Hittite is not a reliable source of information for stress, as it occurs inconsistently and often
marks vowel length instead of stress (Hoffner & Melchert 2008).
Furthermore, there is no evidence to suggest couplets or tercets in the Hittite Parables; the
greatest difference in length between semantically grouped lines is 8 syllables and there is no
significant difference between the average difference in number of syllables in semantically
natural groups (3.4 syllables) and in groups made by pairing each line with a line in the
following natural group, (3.4 syllables), or in randomly assigned groups (3.7 syllables).
However, it might be possible to find this meter elsewhere in Hittite. Generally, it has been
assumed that such a meter should be the same in every line, as in Classical epic traditions
(McNeill 1963), so attempts to describe Hittite meter have been centered on finding features that are consistent across entire poems. The flexibility within a consistent stress-based
framework of this meter could possibly be suited to explaining the metrical system of texts
which are not isosyllabic but have more consistency in line length than would be predicted
in a meter indifferent to syllable count.
The Song of Ullikummi may share some features of the meter of the Hurrian Parables. It
seems to be structured in stanzas, which align with syntactic units and are separated by
horizontal dividing lines, as has been observed by Güterbock (1951). Within these stanzas,
the lines are of a significantly more consistent length than lines paired form different stanzas,
with an average difference of 2 syllables and 3.4 syllables, respectively (p = 0.024), though
the difference in similarity of syllable count between lines paired in couplets and lines paired
14
For comparison, a selection from the Mitanni letter, a Hurrian example of stylized prose, has an average
length of 16.2 syllables and a standard deviation of 4.3 syllables. In Job 1-2, an example of a couplet/tercet
meter without isosyllabism, lines defined syntactically have an average length of 21.5 syllables and a standard
deviation of 5.6 syllables; the average difference between lines within a syntactically and semantically couplet
or tercet is 4.5 syllables, while the average difference between artificially paired lines is 7.3 syllables. Of
course, syllable counts in Hebrew texts are unclear because many vowels were not indicated orthographically.
25
randomly within a stanza is not statistically significant. In addition to the lack of evidence
for couplets, the meter of this text is more variable than in the Hurrian Parables: even when
the outlying short lines which introduce speeches are omitted, the average line length is 13
syllables and the standard deviation is 3 syllables.
T
The Ritual of Iriya shows more evidence for couplet/tercet form; it is naturally broken into
lines based on syntactic structure, as demonstrated in the section of the encounter with the
mountains (Melchert 2007), though the lines are not isosyllabic. There is some semantic
motivation for grouping lines, though damaged areas and unknown vocabulary make some
of the divisions unclear. The result of comparing the differences in line lengths within these
sets of lines and in randomly paired lines is significant (p = 0.0097). However, this text
does not seem to be consistent with the Hurrian dolnik meter, because if Hittite stress really
does correspond with plene spellings, two adjacent stressed syllables are permitted. The line
lengths are also more variable than in the Hurrian Parables: the average length is 13 syllables
and the standard deviation is 3.1 syllables.
DR
AF
Texts from several other languages of this region are also consistent with a dolnik meter
of the type described above, in Akkadian and its descendants as well as other languages.
Some works which have been described as being in the four-stress meter might be amenable
to analysis using this dolnik meter, though the uncertain position of stress in Akkadian,
as in Hittite, means that it would be difficult to distinguish between a dolnik meter and a
stress-based meter.
The Babylonian Theodicy, as presented by Landsberger (1936) contains metrical lines with
slightly more variable length than lines in the Hurrian Parables, which are graphically divided
into stanzas, but are also consistent with a lower-level couplet structure. Using 118 lines
which are complete enough to make a syllable count and which belong to a semantically
natural group, lengths range from 8-18 syllables, with an average length of 12 syllables and
a standard deviation of 1.7 syllables. A small number of lines contain only three words,
which are problematic in a dolnik analysis just as they are in a stress-counting analyses.
The average difference between lines within semantically natural groups of two or three is
1.2 syllables; the average difference between pairs of lines assigned randomly within a stanza
is 1.8 syllables; and the average difference between lines paired randomly across the whole
poem is 2.1 syllables. The difference between natural groups and random pairs is significant
(p = 0.00075), as is the difference between natural groups and nonadjacent lines within a
stanza (p = 0.0096). It is possible that adjacent lines have greater similarity due to extensive
repetition and parallelism, which warrants further investigation, particularly in consideration
of the potential relationship between parallelism-based couplet meter and syllable-counting
couplet meter.
The Hattic text in KUB XXVIII shows a pattern which might fit into the dolnik scheme,
though not with couplets:15 Eichner (1993) offers a possible analysis of a four-footed syllablecounting meter. If Hattic stress is correlated with plene spelling, as is suggested by Watkins
(2006), no line has two adjacent stressed syllables nor more than four stressed syllables,
though there are sequences of more than three unstressed syllables in a row. However, plene
spelling in Hattic may not have been used exclusively to mark stress. Though there are several
15
Given in Laroche 1947.
26
logograms whose underlying Hattic value is unknown and the text is short, the metrically
interpretable lines are of similar lengths, and the consistency is even further improved if the
short lines that occur at the ends of sections are omitted from calculations, resulting in an
average line length of 11.7 syllables and a standard deviation of 1.6 syllables.
3
Conclusions
DR
AF
T
I have presented evidence that the Hurrian Boğazköy Parables, KBo XXXII 14 are in a
meter characterized by couplets and tercets that match in length, unified by a dolnik meter
in which each line has four stressed syllables, separated by 1-3 unstressed syllables, with
one unstressed syllable at the line end and 0-2 unstressed syllables at the beginning. The
poem’s apparent obedience to this meter is improved by positing several phonological and
orthographic features: elision, high vowel gliding, and disyllabicity of <Ca-a-i> sequences.
Based on metrical evidence for phonological and orthographic rules such as these, it is possible
to provide an alternative line of evidence in the investigation of Hurrian phonology, supporting
existing theories and producing new ones. Further research on Hurrian prose may establish
whether the phonological rules that seem to be reflected here were metrically licensed or
standard features of spoken Hurrian.
The identification of this meter warrants further investigation into whether it has been
recorded in other texts, in Hurrian or in other languages. Although most Hurrian texts
are short, it may be more possible to identify them as metrical if one approaches them with
a known meter to test. Furthermore, while the parallel Hittite translation of the Parables
does not appear to be in the same meter, it may be possible to find this meter in other
Hittite texts, if it really was a meter that existed in ancient Anatolia. A number of texts,
discussed above, are strong candidates for being in this meter. If such texts could be identified, this could prove useful in confirming theories about stress placement and other aspects
of phonology in Hittite or other languages which use the meter.
Establishing a meter for this text adds to the picture of the poetic tradition of the ancient Near
East. It is noteworthy that a final trochaic pattern is a natural result of Hurrian phonology,
though in many other languages of the ancient Near East which have a tendency for final
trochees in metrical lines, the stress system is uncertain or not characterized by penultimate
stress. It seems possible that the tendency for trochaic line ends in other poetic traditions
may have been a result of borrowing this meter from Hurrian. Without chronological evidence
for this meter’s language of origin, the regularity of the meter in Hurrian compared to the
less restrictive four-stress meter found in other languages might is also consistent with a
Hurrian origin of the meter, later adopted in a simplified form in other languages, as has
been suggested by Güterbock (1951).
27
References
[1] Arnold, Vernon E. Vedic Metre in its Historical Development. Cambridge: Cambridge Univ.
Press, 1905.
[2] Bachvarova, Mary. “The Meter of Hurrian Narrative Song.” Altorientalische Forschungen 38, no.
2 (2011): 285-308.
[3] Beckman, Gary. “Ištar of Ninevah Reconsidered.” Journal of Cuneiform Studies 50 (1998): 1-10.
[4] Buccellati, Giorgio. “On Poetry – Theirs and Ours.” In Lingering Over Words: Studies in Ancient
Near Eastern Literature in Honor of William L. Moran. Harvard Semitic Studies 37, ed. Tzvi
Abusch, John Huehnergard, and Piotr Steinkeller. Pp. 105-34. Atlanta: Scholars Press, 1990.
[5] ——. A Structural Grammar of Babylonian. Wiesbaden: Harrassowitz Verlag, 1996.
T
[6] Campanile, Enrico. “Indogermanische Metrik und Altirische Metrik.” In Saggi di Linguistica
Comparativa e Recostruzione Culturale, ed. Maria Patrizia Bologna, Filippo Motta, and Chatia
Orlandi. Pp. 127-143. Pisa: Istituti Editoriali e Poligrafici Internazionali, 1999.
DR
AF
[7] Diakonoff, Igor M. “Hurro-Urartian Borrowings in Old Armenian.” Journal of the American
Oriental Society 105, no. 4 (1985): 597-603.
[8] ——. “On Plene-Spelling in Akkadian.” In Semitic Studies in Honor of Wolf Leslau on the Occasion of his Eighty-Fifth Birthday, vol. 1, ed. Alan S. Kaye. Pp. 295-6. Wiesbaden: Harrassowitz,
1991.
[9] Dietrich, Manfried and Walter Mayer. “The Hurrian and Hittite Texts.” In Handbook of Ugaritic
Studies, ed. Wilfred G.E. Watson and Nicolas Wyatt. Pp. 58-75. Leiden: Brill, 1999.
[10] ——. Der hurritische Brief des Dušratta von Mı̄ttānni an Amenhotep III. Alter Orient und
ˇ
Altes Testament 382. Münster: Ugarit-Verlag, 2010.
[11] Edzard, Dietz Otto. Sumerian Grammar. Leiden: Brill, 2003.
[12] —— and Gernot Wilhelm. “Metrik.’ In Reallexikon der Assyriologie und Vorderasiatischen
Archäologie, ed. Dietz Otto Edzard. Pp. 148-150. Berlin: Walter de Gruyter, 1993-1997.
[13] Eichner, Heiner. “Probleme von Vers und Metrum in epichorischer Dichtung Altkleinasiens.”
In Die epigraphische und altertumskundliche Erforschung Kleinasiens: Akten des Symposiums
vom 23 bis. 25 Oktober 1990. Hundert Jahre Kleinasiatische Kommission der Österreichische
Akademie der Wissenschaften 236, ed. Gerhard Dobesch and Georg Rehrenböck. Pp. 97-169.
Vienna: Österreichische Akademie der Wissenschaften, philosophisch-historische Klasse, 1993.
[14] Fortson, Benjamin W., IV. Indo-European Language and Culture: An Introduction. 2nd ed.
West Sussex: John Wiley & Sons Ltd., 2010.
[15] Foster, John L. “The Ancient Egyptian Genre of Narrative Verse.” Journal of Near Eastern
Studies 39, no. 2 (1980): 89-117.
[16] Gasparov, Mikhail L. “Statistical Investigation of Russian Dolnik Trimeter,” trans. D. Lieberman. Theory of Probability and its Applications 8 (1963), 96-102.
[17] Green, Jay, Sr. The Interlinear Bible Hebrew/English. 4th ed. Grand Rapids: Baker Book House,
1983.
[18] Gunkel, Dieter and Kevin M. Ryan. “Pāda Cohesion and Copulation in the Rigvedic Dimeter.”
28
Paper presented at the Kolloquium Sprache und Metrik in Synchronie und Diachronie, LMU
München, München, 2-5 September 2013.
[19] Güterbock, Hans Gustav. “The Song of Ullikummi: Revised Text of the Hittite Version of a
Hurrian Myth.” Journal of Cuneiform Studies 5, no. 4 (1951): 135-161.
[20] Haas, Volkert and Ilse Wegner. “Beispiele poetischer Techniken im hurritischen Schrifttum.”
Studi Micenei ed Egeo-Anatolici 49 (2007), 339-354.
[21] Halle, Morris and Samuel Jay Keyser. English Stress: Its Form, Its Growth, and Its Role in
Verse. New York: Harper and Row Publishers, 1971.
[22] Hayes, Bruce. “Metrics and Phonological Theory.” In Linguistics: The Cambridge Survey,ăed.
Frederick Newmeyer. Pp. 220-249. Cambridge: Cambridge Univ. Press, 1988.
T
[23] Hecker, Karl. “Untersuchungen zur akkadischen Epik.” In Veröffentlichungen zur Kultur und
Geschichte des Alten Orients. Alter Orient und Altes Testament 8, ed. Kurt Bergerhof, Manfried
Dietrich, and Oswald Loretz. Pp. 113-35. Kevelaer: Butzon und Bercker; Neukirchen-Vluyn
Verlag, 1974.
DR
AF
[24] Hoffner, Harry A., and Harold Craig Melchert. A Grammar of the Hittite Language. Winona
Lake: Eisenbrauns, 2008.
[25] Huehnergard, John. A Grammar of Akkadian. 3rd ed. Winona Lake: Eisenbrauns, 2011.
[26] Kiparsky, Paul. “The Rhythmic Structure of English Verse.” Linguistic Inquiry 8, no. 2 (1977):
189-247.
[27] Kloekhorst, Alwin. “Accentuation and Poetic Meter in Hittite.” In Hethitische Literatur Überlieferungsprozesse, Textstrukturen, Ausdrucksformen und Nachwirken: Akten des Symposiums
vom 18. bis. 20. Februar 2010 in Bonn. Alter Orient und Altes Testament 391, ed. Manfred
Hutter and Sylvia Hutter-Braunsar. Pp. 157-176. Münster: Ugarit-Verlag, 2011.
[28] Knudsen, Ebbe Egede. “Stress in Akkadian.” Journal of Cuneiform Studies 32, no. 1 (1980):
3-16.
[29] Lam, Joseph. “A Reassessment of the Alphabetic Hurrian text 1.004 (KTU 1.42): A Ritual
Anointing of Deities?” Journal of Ancient Near Eastern Studies 11, no. 2 (2011), 148-169.
[30] Landsberger, Benno. “Die babylonische Theodizee.” Zeitschrift für Assyriologie 43 (1936): 3276.
[31] Laroche, Emmanuel. “Hattic Deities and their Epithets.” Journal of Cuneiform Studies 1, no.
3 (1947), 187-216.
[32] ——. Glossaire de la Langue Hourrite. Paris: Éditions Klincksieck, 1980.
[33] McNeill, Ian. “The Metre of the Hittite Epic.” Anatolian Studies 13 (1963), 237 -242.
[34] Melchert, H. Craig. “Poetic Meter and Phrasal Stress in Hittite.” In Mír Curad: Studies in
Honor of Calvert Watkins, ed. Jay Jasanoff, H. Craig Melchert, and Lisi Oliver. Pp. 483-94.
Innsbruck: Institut für Sprachwissenschaft, 1998.
[35] ——. “New Light on Hittite Verse and Meter?” In Proceedings of the Eleventh Annual UCLA
Indo-European Conference, Los Angeles, November 3-4, 2006. Journal of Indo-European Studies
Monograph 53, ed. Karlene Jones-Bley, Martin E. Huld, Angela Della Volpe, and Miriam Robbins
Dexter. Pp. 117-128. Washington DC: Institute for the Study of Man, 2007.
29
[36] Michałowski, Piotr. “Ancient Poetics.” In Mesopotamian Poetic Language: Sumerian and Akkadian, ed. Marianna E. Vogelzang and Herman L. J. Vanstiphout. Groningen: Styx Publications,
1996.
[37] Neu, Erich. “Varia Hurritica. Sprachliche Beobachtungen an der hurritisch-hethitischen Bilingue
aus Hattuša.” In Documentum Asiae Minoris Antiquae: Festschrift für Heinrich Otten zum 75.
Geburtstag, ed. Erich Neu and Christel Rüster. Pp. 235-54. Wiesbaden: Otto Harrassowitz, 1988.
[38] ——. Das hurritische Epos der Freilassung I: Untersuchungen zu einem hurritisch-hethitischen
Textensemble aus Hattuša. Wiesbaden: Harrassowitz Verlag, 1996.
ˇ
[39] Ryan, Kevin M. “Gradient Syllable Weight and Weight Universals in Quantitative Metrics.”
Phonology 28 (2011): 413-454.
[40] Sauren, Herbert. “Beispiele sumerischer Poesie.” Jaarbericht van het Voor-Aziatisch-Egyptisch
Genootschap 22 (1971-2): 255-306.
T
[41] Speiser, Edgar A. “Phonetic Method in Hurrian Orthography.” Language 16, no. 4 (1940):
319-340.
DR
AF
[42] Sturtevant, Edgar H. “Elision and Hiatus in Latin Prose and Verse.” The Classical Journal 12,
no. 1 (1916): 34-43.
[43] Tarlinskaja, Marina. “Metrical Typology: English, German, and Russian Dolnik Verse.” Comparative Literature 44, no. 1 (1992), 1-21.
[44] Thiel, Hans-Jochen. “Formale Textanalyse, Metrik, Glossenkeil und Textkritik.” In Das hurritologische Archiv (Corpus der hurri(ti)schen Sprachdenkmäler) des Altorientalischen Seminars
der Freien Universität Berlin, ed. Volkert Haas and Hans-Jochen Thiel. Pp. 240-264. Berlin:
Universitätsdruck, 1975.
[45] van Soldt, Wilfred H. “The Adaptation of the Cuneiform Script to Foreign Languages.” In The
Idea of Writing: Play and Complexity, ed. Alexander J. de Voogt and Irving L. Finkel. Pp.
117-27. Leiden: Brill, 2010.
[46] von Soden, Wolfram. “Untersuchungen zur babylonischen Metrik, Teil I.” Zeitschrift für Assyriologie und Vorderasiatische Archäologie 71, no. 2 (1981): 161-204.
[47] Vanstiphout, Herman L. J. “ ‘Verse Language’ in Standard Sumerian Literature.” In Verse in
Ancient Near Eastern Prose, ed. Johannes C. de Moor and Wilfred G. E. Watson. Pp. 305-329.
Kevelaer: Verlag Butzon & Bercker, 1993.
[48] Watkins, Calvert. “An Indo-European Linguistic Area and Its Characteristics: Ancient Anatolia. Areal Diffusion as a Challenge to the Comparative Method?” In Areal Diffusion and Genetic
Inheritance: Problems in Comparative Linguistics, ed. Alexandra Y. Aikhenvald and R. M. W.
Dixon. Pp. 44-63. Oxford: Oxford Univ. Press, 2006.
[49] Watson, Wilfred G. E.. Classical Hebrew Poetry. Sheffield: JSOT Press, 1984
[50] ——. “Ugaritic Poetry.” In Handbook of Ugaritic Studies, ed. Wilfred G. E. Watson and Nicolas
Wyatt. Pp. 165-192. Leiden: Brill, 1999.
[51] Wegner, Ilse. Einführung in die hurritisch Sprache. 2nd ed. Wiesbaden: Harrassowitz Verlag,
2007.
[52] West, Martin L. “The Babylonian Musical Notation and the Hurrian Melodic Texts.” Music &
Letters 75, no. 2 (1994): 161-179.
30
[53] ——. “Akkadian Poetry: Metre and Performance.” Iraq 59 (1997): 175-187.
[54] Wilhelm, Gernot. “Reconstructing the Phonology of Dead Languages.” In Writing in Focus, ed.
Florian Coulmas and Konrad Ehlich. Pp. 157-66. Berlin: Walter de Gruyter & Co., 1983.
[55] ——. “Zur hurritischen Gebetsliteratur.” In Ernten, was man sät: Festschrift für Klaus Koch
zu seinem 65. Geburtstag, ed. Dwight R. Daniels, Uwe Gleßmer, and Martin Rösel. Pp. 37-47.
Neukirchen-Vluyn: Neukirchener Verlag, 1991.
[56] ——. “Hurritische Lexikographie und Grammatik: Die hurritisch-hethitische Bilingue aus
Boğazköy.” Orientalia 61, no. 2 (1992): 122-141.
[57] ——. “Hurrian.” In The Ancient Languages of Asia Minor, ed. Roger D. Woodard. Pp. 81-104.
Cambridge: Cambridge Univ. Press, 2008.
DR
AF
T
[58] ——. “Ein Konflikt zwischen König und Ältestenversammlung in Ebla.” In Organization, Representation, and Symbols of Power in the Ancient Near East: Proceedings of the 54th Recontre
Assyriologique Internationale at Würzburg 20-25 July 2008, ed. Gernot Wilhelm. Pp. 155-166.
Winona Lake: Eisenbrauns, 2012.
Appendix: Key to Abbreviations
abl
abs
ag
cau
col
conn
dat
dir
dur
erg
ess
ex
fct
gen
ger
imp
inch
indv
intr
it
jus
loc
neg
nmz
opt
ablative
absolutive
agent
causative
collective
connective
dative
directive
duration
ergative
essive
extension of uncertain meaning
factitive
genitive
gerund
imperative
inchoative
individualizer
intransitive
iterative
jussive
locative
negation
nominalizer
optative/voluntative
31
passive
patientive
plural
possessive
perfect
participle
relator
subject pronoun
singular
theme vowel
transitive
DR
AF
T
pass
pat
pl
poss
prf
prt
rel
sbj
sg
th
tr
32