Sound and Structure in Dante's Divine Comedy

Background
I originally approached this subject through a study of modern criticism of Dante's Divine Comedy, but then attempted to work out more fully my own answers to some of the questions that the criticism raises, considering first of all the difficulties that arise in providing a general description of the poem's language and style. This is what led me into the area of humanities computing. Computers can contribute in a variety of ways to the kind of general description I have been interested in, but my particular reason for using them has been to explore a particularly interesting issue in modern criticism of the Comedy, the relationship between sound and sense.
I looked first at the distribution of alliterations and assonances in the Divine Comedy and a selection of subsequent narrative poems in Italian, then moved on to a comparative study of rhymes in Renaissance epic romances, which I have recently extended to include the Divine Comedy [4]. Computer-based analysis in these areas is made easy by the near-phonological nature of Italian spelling. In the study of alliterations and assonances, simple program routines can be written that produce a phonological transcription of the raw text, with minimal manual intervention; the analysis of rhymes can be effected electronically by simply reading off the matching characters from the end of each verse-line, again with minimal manual intervention. Unfortunately, however, for the purposes of computer analysis Italian orthography is not so conveniently informative in the other major source of sound effect in poetry, metre and rhythm. It is these that I have been mainly working recently, and that form the subject of the present paper.
Objectives and method
My present project produces (1) an electronic text of the Divine Comedy and Dante's lyric poetry marked up in terms of syllable count and accent; (2) on the basis of this, a systematic structural description of Dante's rhythmical practice, in the sense of the distribution of accentual structures and the application of syllable divisions (dialefe and diaeresis); (3) a much more powerful account of assonance and alliteration in the Comedy, since I am now able to link these features (as I could not do in my earlier studies) to accentual structure. The paper deals with problems and results relating to (1) and (2). Previous studies of rhythm in the Comedy have been selective, and/or based on only partially explicit criteria (notably those by Beccaria[1], Bertinetto[2], and Di Girolamo[3]). My project processes every line in the Comedy, and every hendecasyllable in Dante's lyric poetry, on the basis of a set of linguistic and metrical criteria that are as systematic and explicit as the nature of the subject allows.
The computer-based study of rhythm requires very substantial manual intervention; using a computer simply makes the work of scansion more accurate, less tedious and quicker to carry out than doing it by hand, and equally importantly makes the outcome far more easily to manipulate for purposes of further analysis. I have developed algorithms which correctly identify word accents in some 80-90% of cases, the remainder being dealt with by electronic reference to a marked-up list of the relevant forms; similarly, syllable divisions can be electronically identified in the great majority of cases, though manual intervention is required for the insertion of divisions that do not follow fixed rules. The computation process is therefore a complex and to a large extent an interactive one, involving the constant exercise of critical judgement.
The critical methodological issue arises over the distinction between word accents that are metrically or rhythmically relevant and those that are not, and the related decision as to which monosyllables should count as accented. In keeping with the general aims of the project, I have attempted as far as possible to devise a systematic set of rules for the inevitable exercise of judgement in this respect, with a view to incorporating in the metrical text a performance that is as explicitly rule-governed as the nature of the material allows. The rules are fairly pragmatic in character, claiming simply to produce an acceptable and, more importantly, a consistent reading of the poem. After consideration of various possible approaches to the problem, my solution retains the word accent of virtually all polysyllables, and assigns accents to most monosyllables as well, excluding some of the latter on the basis of word class, with some supplementary positional rules in a limited number of cases. The high degree of inclusiveness in this approach to accentuation is justified by the theoretical principle of the self-reflexiveness of poetry, which means that rhythmical structures should be actualized to the maximum extent, and by the related cultural fact that poetry is recited rather than spoken. Extensive use has also been made of recorded readings of the Divine Comedy.
Outcome and significance
What I have called a systematic structural description of rhythm can be looked at and used in a number of ways. It should first of all be an object of interest in its own right, by providing an exact overall view of an important aspect of Dante's poem. But it also has both an internal and an external dimension: on the one hand, the relation between sounds and other aspects of the Comedy - how certain sound structures work in the poem; on the other, comparing Dante's practice with that of his contemporaries and successors, particularly those writing in the same terza rima form (here special reference is made to Petrarch and Boccaccio). By providing extensive statistical information about sound features in the poem, the project also provides essential reference material for other scholars and critics. As it is, far too many interpretations of Dante's style are based on insufficient information about the overall incidence of these features.
Part of the outcome of the study is to confirm, more systematically and with greater precision, points that others have made before about aspects of Dante's language and style. The results overwhelmingly reinforce, for instance, traditional views as to the predominance of accentual structures involving even syllables: on my system of identifying accents, in the poem as a whole the average number of accented odd syllables in the eleven-syllable line is almost exactly one. In relation to both accentual structures and syllable divisions, applying a systematic and explicit set of criteria also permits a far more exact specification than has yet been produced of the degree of rhythmical irregularity in Dante's practice - the extent to which he uses accentual structures which, from the 16th century onwards, were not allowed - and of the degree of difference in this respect between Dante and near-contemporaries like Petrarch and Boccaccio. I can show, for instance, that while Dante's practice in respect of syllable divisions is far less regular than that of later writers, there are very specific limits to this irregularity, limits which can be systematically defined.
By defining the rhythmical norms of Dante's verse, the study also allows a systematic examination of aberrant or exceptional structures: for instance lines with an unusually high number of accents on odd syllables, or an unusually high number of irregular syllable divisions. Consideration of these aberrant features can contribute significantly to the long-standing discussions about the significance and effect of the stylistic features in question. Finally, as a by-product of the systematic analysis that the marked up electronic text makes possible, my project calls into question certain pre-electronic choices made by the most recent and most authoritative editor of Dante's poem, identifying elements of inconsistency of which he (the editor) seems not to have been aware.
References:
1. BECCARIA, G.L., 'Alliterazioni dantesche' and 'Figure dantesche' in id., L'Autonomia del significante (Turin, 1975), pp. 90-113 and pp. 114-135.

2. BERTINETTO, P.M., Ritmo e modelli ritmici. Analisi computazionale delle funzioni periodiche nella versificazione dantesca (Turin, 1973).

3. DI GIROLAMO, C., Teoria e prassi della versificazione (Bologna, 1976).

4. ROBEY, DAVID, 'The Stylistic Analysis of the Divine Comedy in Modern Literary Scholarship' in C. Grayson (ed.), The World of Dante (Oxford,1980), pp. 198-237

5. 'Language and Style in the Divine Comedy', Romance Studies, V (1984/5),pp. 112-127.

6. 'Literary theory and critical practice in Italy: formalist, structuralist and semiotic approaches to the Divine Comedy', Comparative Criticism, VII (1985), pp. 73-103.

7. 'Sound and sense in the Divine Comedy', Literary and Linguistic Computing, II, 2 (1987), pp. 108-115

8. 'Alliterations in Dante, Petrarch, and Tasso: a computer analysis', in P.R.J. Hainsworth, V. Lucchesi, E.C.M. Roaf, D..J.B.Robey, J.R. Woodhouse (eds), The Languages of Literature in Renaissance Italy (Oxford,1988), pp. 169-189
9. 'New directions in Dante criticism?', Comparative Criticism, XI (1989),
pp. 263-270.

10. 'Rhymes in the Renaissance Epic: A Computer Analysis of Pulci, Boiardo, Ariosto and Tasso', Romance Studies, XVII (Winter 1990), pp. 97-111

11. 'Analysing Italian Renaissance Poetry: The Oxford Text Searching System', Literary and Linguistic Computing, V, 4 (1990), pp. 310-313

12. 'Dante and Modern American Criticism: Post-structuralism', Annali d'Italianistica, VIII (1990), pp. 116-131

13. 'Scanning Dante's Divine Comedy: a computer-based approach', Literary and Linguistic Computing, VIII, 2 (1993), pp. 81-84

14. 'The Fiore and the Comedy: some computerized comparisons', in Z. Baranski and P. Boyde (eds), The 'Fiore' in Context. Dante, France, Tuscany (University of Notre Dame, 1997), pp. 109-131