• Introduction
• Indefinite and abstract -i
• The vowel in budī and šudī
• Interpretations of man
• Two outstanding issues
• Conclusion
• Bibliography

Persian poetic meter is described as quantitative, using syllable length as the building block for its verse patterns. However, in addition to syllable length, Persian is characterised by a stress system which is independent of its syllable length (Hosseini 2004:26). This stress system plays an important role in another feature of Persian poetry, one that runs throughout Ferdowsi’s Shahnameh – rhyme. Stress generally falls on the last syllable of the word in Persian, which would entail that each line of Persian poetry end with a stressed syllable. However, exceptions to this stress rule and other interfering phenomena, some of which are presented in 1-3 below, often make the last syllable unstressed.

1. 1. 1. The indefinite-forming suffix -ē is unstressed,
      2. The object marker rā is enclitic and therefore never stressed.
      3. Personal suffixes, which originate as enclitics, are unstressed: ketā́b-ešān ‘their book’.

It is significant then to see how rhyme behaves in Ferdowsi’s Shahnameh when these unstressed elements occupy the last position in the line.

We see in the above that indefinite endings -ē, object markers -rā and possessive endings -šān and -i man all share the property of not exhausting the rhyme: The rhyme has to extend into preceding syllables. This extended rhyme phenomenon is called radif in Persian poetry and appears there extensively. In the Shahnameh, “7652 lines out of 48601 include radif, that is 15.74% of the whole corpus” (Radmanesh 2002:90). We will maintain here that the radif phenomenon is directly related to stress placement in Persian: the rhyme begins on the last stressed syllable of the line. If any syllables follow the last stressed syllable, they must be identical. This phenomenon occurs regularly and cross-linguistically when rhyming emerges in poetic traditions of the world, as can be observed in the following English, French and Russian examples:

There was a young man of Nantucket.
Who went down a well in a bucket;

Ils se ressentent si bien vivre,
Les pauvres petits pleins de givre!

They feel themselves living so well
The poor little ones full of frost

vot begaet dvorovyj mal’čik
shalun už zamorozil pal’čik

There runs around a yard boy,
The trickster’s already got a finger frozen

The first insight we can get if we accept that radifs reveal stress placement is that a change occurred between Ferdowsi Persian abstract -i and Modern Persian abstract -i. Below, we see that in Modern Persian, the indefinite -i (which comes from indefinite Ferdowsi Persian -ē by regular sound change) is unstressed, while the abstract -i is stressed (as above, the stressed syllables are underlined):

In Ferdowsi Persian, we have seen radifs indicating that the indefinite -ē is unstressed. However, we see the same picture for abstract -i in the Shahnameh, which means that, for abstract -i, stress has undergone historical change and became word-final in Modern Persian

It is worth highlighting that verse-final instances of the -ī abstract are involved in an extended rhyme across the board. The example from 3.26 above doesn’t show a special case of badī having an extended rhyme, rather all cases of badī that we could observe follow this pattern. Beyond the multiple (41 out of 127) attestations of the badī / ezadī pair, other verse-final words paired with badī show the same extended rhyme, with representative examples given below:

The regularity with which badī occurs in extended rhymes brings out a further insight when we look at the outlying cases. The four instances below are the only ones where the badī rhyme does not conform to the pattern above:

In all these cases, we see that badī rhymes with either budī or šudī, which descend respectively from the present stems bawa- and šawa– in Middle Persian. Since the evidence discussed above indicates that badī must be stressed on its first syllable, we reason that budī or šudī must have the same stress placement on the first syllable and we conclude that, to make the rhyme work, the Middle Persian sequence -awa- had to be in a state rhymable with -a- at the time of Ferdowsi Persian.

For all observable rhyming pairs of budī or šudī, the majority of cases (101 of 108 budī rhyme pairs, 38 of 42 šudī rhyme pairs) are verses with the final verbal element being fully unstressed, which pushes the extended rhyme further left and gives budī and šudī identical corresponding pairs:

However, the outlying cases all point to the first syllable of budī and šudī rhyming with -a-. Together with the badī cases, the follwing examples exhaust all the attestations of verse-final budī and šudī rhyming with -adī in the Shahnameh.

Taking a closer look at man, we see different interactions between this personal pronoun and extended rhyme.

1.1 The rhyme onset starts on man

1.2 The rhyme onset starts on a syllable preceding man

The distribution of the rhyme onset between 1.1-type lines and 1.2-type lines is easy to account for. Instances of man that exhaust the rhyme as in 1.1 are a cases of man being sentence-initial in an enjambed quoted speech. This discourse-initial position places stress on man and allows it to exhaust the rhyme. This puts it in opposition to 1.2, where man is part of the possessive ezafe construction, which is unstressed like the possessive suffixes seen in ketā́b-ešān. There is however a sizeable amount of line-final man instances in which the pronoun receives the rhyme onset and isn’t discourse initial. For these instances our interpretation of the passage must be guided by the fact that man has to be exceptionally emphasized.

The pronoun man can also be involved in extended rhyme when it isn’t line-final. Below we look at instances of the string …man ast, and its different treatments within rhyming patterns:

2.1 The rhyme onset starts on ast

2.2 The rhyme onset starts on man

2.3 The rhyme onset starts on neither man nor ast, but extends further into the line

Like with the discourse-initial man from example 1.1, the distribution of these different patterns can be explained with reference to sentential emphasis, which drives stress-placement.

The most common pattern, 2.3, overwhelmingly features man ast as a possessor and copula in an ezafe construction. The identity of man ast in both lines indicates that the construction is prosodically deficient, with stress placed on preceding material, which is also where the rhyme onset occurs. We conclude that the meaning of man ast in all such occurrences is a neutral an unemphatic one.

Pattern 2.2, in contrast, shows that rhyme does not extend further than man, which indicates that man must have received sentential stress and carries the meaning of an emphatic possessor. For the lines in 12d.2439, for instance, we must interpret them as saying “killing him is (specifically) my hardship”.

Pattern 2.1 is more problematic, because putting stress and therefore emphasis on the copula isn’t as likely as on the possessor pronoun. In the lines 35.1955, it is difficult to justify emphasis on the copula for the sentence “All my water and treasure are hidden”

Nevertheless, the nature of the rhyme in 35.1955 gives one no choice but to conclude that ast in this sentence receives emphatic stress. It remains the task of future research to determine the implications of this emphatic stress for the interpretation of the passage.

If we keep to the position that extended rhymes indicate stress placement in Shahnameh Persian, two issues are observed for which we do not have a clear explanation:

1. Diachronic issues involving the noun xirad
2. Stress placement variations in compound verbs

1. Diachronic issues involving the noun xirad

The word xirad ‘widom’ can often be found sentence-finally in the Shahnameh, and is therefore bound to rhyming rules. However, one observes patterns that make xirad exceptional to the above-mentioned final-syllable stress rule.

In every instance of xirad, we observe a perfect rhyme on –rad, which is counter to the rhyming pattern everywhere else in the Shahnameh. To avoid this aberrant perfect rhyme, we argue that the rhyming must begin on the preceding syllable, and that –rad follows as a radif. Two issues arise for this analysis:

• The preceding syllabic nuclei do not match to create a rhyme
  • Since the largest number of rhyming pairs for xirad, as in the examples above, have -a- as the nucleus of the preceding syllable (baragzarad, barad), no rhyming can occur. We argue that at the time of the Shahnameh’s composition, the first syllable of xirad had a vowel that allowed identity with -a- in its rhyming counterparts, the common ancestor of the -i- used in our source text and of the –e– in Modern Persian xerad. The sound changes that produced xirad that we see in the modern languages make impossible the rhyme that was available in Ferdowsi’s time.   
• Stress placement on the first syllable of xirad is unexpected from a diachronic point of view
  • The etymological reconstructions for xirad are PIE *krét-us and PIr *krátuš, meaning that the first syllable of xirad is a later innovation. If this syllable appeared as a result of epenthesis that broke the word-initial kr- cluster, it is difficult to see why stress would shift to an epenthetic vowel in a language that overwhelmingly stresses a word’s final syllable. For this, we can offer no explanation at this time.

2. Stress placement variations in compound verbs

Line-final compound verbs in the Shahnameh show rhyme patterns that can go very far leftwards of the line end.

Based on the standard description of Persian compound verbs, this is expected, as the nominal part carries the meaning while the verbal part contributes tense and agreement information. Given this breakdown, it is not surprising that the last stress of the line would fall on –run in andarun âmadand.

However, this isn’t always the case, as examples exist where the rhyme is exhausted on the verbal part of the compound verb.

The distribution between āmadand that transfers its stress to the nominal part of the compound and āmadand that doesn’t is unclear.

Two instances were particularly problematic in that, not only did the stressed syllable not fall on the emphasizeable nominal part of the compound verb, it also didn’t fall on the root of the verbal part, but rather on its historically encliticised and unstressed ending. The evidence, discussed above, that the -u- in šudand must rhyme with -a- solves this issue.

As we have seen, extended rhyme reveals to us some aspects of Ferdowsi Persian. By leveraging it, we can gain insight into historical change that the language has undergone and we can refine our text interpretations based on the phrasal stress that extended rhyme indicates to us. Interesting questions arise regarding compound verbs in Ferdowsi Persian, as extended rhyme reveals different locations of stress placement in them, but no clear rule accounting for the differences can yet be identified.

Mubashshir Akbarzod, ed. 2007. Abylkhosim Firdavsi. Shokhnoma. Dushanbe: Adib

Radmanesh, M. 2002. Radif in Shahnameh. Literary Text Research 5 (15). 90-120.

Hosseini, S. A. 2014. The Phonology and Phonetics of Prosodic Prominence in Persian. [PhD thesis]. Tokyo University of Foreign Studies.