The Palatine Anthology – Book 1: A Metrical and Prosodic Statistical Study

Preliminary steps and sampling strategy. Segmentation

The available automated tools for hyphenation and syllabification (the division of words into syllables and the determination of syllabic quantity) and scansion (the establishing of metrical feet)—including large-scale language models (e.g. Gemini Pro)—proved inefficient. Therefore, a manual segmentation of the edited text was performed individually for each word, considering that synapheia (cf. Bozzone, 2024, 72) loses its relevance in the context of the byzantine verse. E.g., verse (1-a) is hyphenated as in (1-b), not as in (1-c).

• (1)  a.   εἰσόκεν ἠέλιος πυριλαμπέα δίφρον ἐλαύνει                                 (AP 1.10.41)

            “as long as the Sun drives his fiery chariot”¹

        b.   εἰσ#ό#κεν ἠ#έ#λι#ος πυ#ρι#λαμ#πέ#α δίφ#ρον ἐ#λαύ#νει

        c.   *εἰσ#ό#κε#νἠ#έ#λι#ος#πυ#ρι#λαμ#πέ#α#δίφ#ρο#νἐ#λαύ#νει

In the case of elisions, since syllabic segmentation is “the lowest level in a hierarchy of prosodic prominence assignment” (Devine & Stephens, 1994, 24), resyllabification was performed (see Figure 1, cf. Devine & Stephens, 1994, 235–248). However, when caesurae intervene² or syntactic constraints occur (segmentations of syntactic phrases), they are given priority (see Figure 2).

Figure 1

Figure 2

The muta cum liquida group was divided when it lengthened the preceding short vowel and transferred to the next syllable when it did not. Subsequently, the chain of long and short syllables was divided into feet according to the verse profile.

In the process of verse segmentation, the main difficulty was the caesura, for which there is no unanimous definition among specialists (Janse, 2003; Spaltenstein & Bianchi, 2004). We work with two concepts: the metrical caesura and the functional caesura (see references above). Among the metrical caesurae indicated by the software (marked with the black symbol ‖), the functional ones were manually determined as those that represent phonological pauses and do not break syntactic phrases (marked with the red symbol ‖). E.g., in verse (2-a), the software indicates the presence of four metrical caesurae in feet 1 (fem.), 2 (masc.), and 3 (masc.+fem.) (2-b). Among these, the trihemimeral caesura is functional (2-c).

• (2)   a.  αὐτὸς ἑκών⋅ οὐ γάρ τις ἐπιχθονίων πολεμίζοι                  (AP 1.119.22)

              “[…] by his own will, for no earthly man could war […]”

           b.  αὐτὸς ‖ ἑ ¹| κών⋅ ‖ οὐ ²| γάρ ‖ τις ‖ ἐ ³| πιχθονί ⁴| ων πολε ⁵| μίζοι ⁶

           c.  αὐτὸς ἑκών⋅ ‖ οὐ γάρ τις ἐπιχθονίων πολεμίζοι

In several verses, the functional character of the caesurae is debatable. Thus, in verse (3), the feminine metrical caesura in the third foot, although it functions as a phonological pause due to inversion, is not a functional caesura (it breaks the prepositional phrase (PP)).

• (3)     ἤγαγεν [_PP ἀχράντοισιν ‖ ὑπ’ ἐννεσίῃσι] τοκῆος               (AP 1.119.26)

           “he led […] by the immaculate command of his father […]”

Data pipeline

Hyphenation and scansion were exported as three .csv files (hexameter, iambic trimeter, elegiac pentameter). These are processed by meter-specific modules that parse the hyphenated Greek text into syllables, expand the metrical scheme into per-syllable quantities, and extract the given parameters from the spreadsheet columns. The software then computes metrical and prosodic features independently of the manual extraction—bridge violations, diaereses, metrical caesurae, homodynia—and compares the results, reporting all discrepancies as warnings. Using this process, the dataset was verified and corrected across multiple iterations.

Interactive viewer

The output is a single .html file with seven overlay toggles (colors, scansion marks, feet, diaereses, caesurae, bridges, homodynia), each applicable to the selected verse or to all verses simultaneously. Functional and metrical caesurae can be displayed independently. Verses are selected by clicking or navigating with arrow keys.

Examples of database usage

The database may be used in studies on the history of versification. The Palatine corpus – Book 1 reflects the transition from quantitative versification to versification based on natural stress. It combines archaic metrical and prosodic features (Homeric and Callimachean, such as the relatively high frequency of masculine caesurae and a greater number of spondees than in the Nonnian verse) with characteristics defining late, Nonnian-style poetry (e.g., a smaller number of metrical patterns in hexameter) and late byzantine versification (the iambic dodecasyllable verse). The verse structure is more distinctly spondaic in the Christian poems and, consequently, more rhythmically varied than in the Nonnian corpus. The epigrams average .89 spondees per line, compared to .75 in Nonnian verses (van Raalte, 1986, 32). Of the 32 possible rhythmic patterns, the hexameter verses of the first Palatine book feature only ten with a frequency greater than 1% and nine others with a frequency below 1% (Figure 3). By comparison, Homer uses 14 metrical patterns with a frequency over 1%, Callimachus 11, and Nonnus 7 (for these authors, see Ludwich, 1885, 317–322; van Raalte, 1986, 28–78; Schoisswohl & Papakitsos, 2020, 170).

Figure 3

The database is useful for studying the caesura. The most common type of caesura in the Greek hexameter occurs in the third foot (masculine and feminine) (for the distribution of caesurae, see Agosti, 2004, 64–73), which in the Palatine corpus appears metrically in over 98% of the verses, but functionally in only 67%. The nearly exhaustive presence of the metrical caesura in the third foot betrays a Nonnian versification pattern. Its functional occurrence in only two-thirds of the corpus gives way, in the remaining verses, to the bucolic diaeresis, which divides the verse according to the Alexandrian pattern (60.4% of verses have a metrical bucolic diaeresis, with this being segmentally realized in 23.5% of the verses). As illustrated in Figure 4, the Palatine verse is functionally divided by the third-foot caesura or the bucolic diaeresis into two cola and fragmented by other metrical caesurae (occasionally two in the same foot⁴). These metrical caesurae divide the verse into short phonological units, sometimes the length of a single word. Corroborating this observation with lexical analysis reveals the puzzle-like nature of the verses, in which epithets of Homeric origin and words derived or composed according to the same model fit into preformed metrical patterns that are easily incorporated into verse by authors who struggled with syllabic quantities.

Figure 4

Statistical analysis provides insight into premodern versification strategies. A relevant feature in this regard is the high proportion of masculine caesurae (penthemimeral) relative to feminine (trochaic) ones. While in the hexameter poems the feminine caesura is predominant, in the elegiac distich the caesura of the hexameter is predominantly masculine, as can be seen in Figure 5. The masculine division creates a first hemistich of the hexameter which is caesurated in the same way as the following pentameter. To make the effect clear, the masculine caesurae in the two lines are often highlighted by assonance, heralding the appearance of rhyme as a structuring element of poetic discourse, as in (4). The same phenomenon is manifested in AP 1.19.1–4, 1.70, 1.82, 1.88, 1.94.1–2, passim.

• (4)     Τηλόθεν οὐχ ὑδάτων ⁞  μνηστεύετο πότνα Ῥεβέκκα,

•                  νύμφης ἐξ ἐθνῶν  ⁞  οὕνεκέν ἐστι τύπος.                                      (AP 1.70)

           “The lady Rebecca was wooed not far from the water, because she is a figure of a gentile bride.”

Figure 5

Understanding the phonological profile of Greek during the 5th–10th centuries. The study of syllable quantity and stress preceding caesurae provides insight with respect to the merely perceptively phonetic character—if not distinctively phonological—of vowel quantity. In both the iambic and pentameter corpora, the syllable to the left of the caesura tends to be long and unstressed. The relatively large number of masculine caesurae in the hexameter corpus is consistent with this phenomenon. This fact proves that, although vowel length had disappeared from the Greek language phonologically (Allen, 1973; Devine & Stephens, 1982, 1994), it was still perceptible phonetically and prosodically. Thus, the long syllable before the caesura was felt like a “delay” emphasizing the caesura. The avoidance of the natural accent, which has become a stress accent, contributes to the realization of the caesura as a syntactically and rhythmically significant “silence” (Călin, 2023, 75) or “musical hold” (Bassett, 1919, 344).