Auceps syllabarum.

Laudator Temporis Acti sent me to Tom Keeline and Tyler Kirby, “Auceps syllabarum: A Digital Analysis of Latin Prose Rhythm” (Journal of Roman Studies 109 [2019]:161-204), which looks like a useful and well-done study; the abstract:

In this article we describe a series of computer algorithms that generate prose rhythm data for any digitised corpus of Latin texts. Using these algorithms, we present prose rhythm data for most major extant Latin prose authors from Cato the Elder through the second century ᴀ.ᴅ. Next we offer a new approach to determining the statistical significance of such data. We show that, while only some Latin authors adhere to the Ciceronian rhythmic canon, every Latin author is ‘rhythmical’ — they just choose different rhythms. Then we give answers to some particular questions based on our data and statistical approach, focusing on Cicero, Sallust, Tacitus and Pliny the Younger. In addition to providing comprehensive new data on Latin prose rhythm, presenting new results based on that data and confirming certain long-standing beliefs, we hope to make a contribution to a discussion of digital and statistical methodology in the study of Latin prose rhythm and in Classics more generally. The Supplementary Material available online (https://doi.org/10.1017/S0075435819000881) contains an appendix with tables, data and code. This appendix constitutes a static ‘version of record’ for the data presented in this article, but we expect to continue to update our code and data; updates can be found in the repository of the Classical Language Toolkit (https://github.com/cltk/cltk).

But what clinched the decision to post about it was the title; I am not Latinist enough to recognize it, but Prof. Google tells me that auceps syllabarum, literally ‘bird-catcher of syllables,’ has the transferred sense “a person who quibbles over words, argues over semantics or other technicalities; a pettifogger.” Cicero, in de Oratore 1, 55, 236 (about a third of the way down the left-hand page here), calls a lawyer “leguleius quidam cautus et acutus, praeco actionum, cantor formularum, auceps syllabarum” (J.S. Watson: “a sort of wary and acute legalist, an instructor in actions, a repeater of formulae, a catcher at syllables”), and the phrase seems to have appealed to lawyers and others, as you will see from the many uses found in a Google Books search (e.g., from Thomas Moore, Memoirs of the Life of the Right Honourable Richard Brinsley Sheridan, Vol. 1 (1825), p. 235: “a study in which more than the mere ‘auceps syllabarum’ might delight”). I will try to remember to add it to my stock of learnèd insults.

For those interested in the article by Keeline and Kirby, here is the start:

For over a century, scholars studying Latin prose rhythm have relied on the statistics generated by Theodor Zielinski’s pioneering Das Clauselgesetz in Ciceros Reden. They have also complained about his methodology and its inadequacies: Zielinski read his own Russian translations of Cicero’s speeches out loud in order to develop a feel for where sense breaks (and so clausulae) occurred in the Latin; he arbitrarily decided that the cretic was the basis for Latin prose rhythm; he did not compare his observed frequencies of clausular patterns to any expected values, thus ignoring the naturally occurring rhythms of the Latin language; he came up with dubious rules for word division and resolutions within his clausular categories. But this was path-breaking scholarship, for Zielinski had no real predecessors — and he has had no successors either. In the 115 years since Das Clauselgesetz, no scholar has had the Sitzfleisch to do what Zielinski did: he counted, by hand, 17,902 clausulae in Cicero’s speeches. He analysed his results in detail and produced elegant summary tables, all without the aid of electronic calculators. The result is an imposing and apparently authoritative monument.

The real problem with Zielinski’s analysis, however, is not its methodological basis. About his methodology Zielinski is an exemplar of openness and honesty: he lays out his assumptions and reasoning at every step of the process, and in exhaustive detail. While all of these have been questioned, no one would expect the first explorer of uncharted terrain to map it perfectly. A bigger problem is that Zielinski’s results are unverifiable and unreproducible. He seems to provide a deluge of data, but readers must trust that he has scanned and counted and tabulated correctly, for he provides comprehensive scansion for only one speech. Zielinski was a great scholar, but in most fields of scientific inquiry we do not simply accept unverifiable pronouncements. And yet it is not just Zielinski: all scholars of Latin prose rhythm who give even partial statistics have presented their varying results and varying methodologies from a black box that could not be inspected or verified. Furthermore, there looms a potentially even bigger problem: if one wanted to modify Zielinski’s methodology — disregarding some of his strictures on word division and resolution, say — it would require recounting everything from scratch.

Fortunately, computers have come to our aid. In this article, we describe a series of interrelated algorithms and modules that can produce a comprehensive analysis of the prose rhythms of a given corpus of Latin literature with a few keystrokes. This digital approach presents entirely new possibilities for the study of prose rhythm. With complete openness and transparency, we can calculate prose rhythm statistics from across the whole of extant Latin literature. Furthermore, we can be absolutely consistent in our procedures and confident in our statistics, and yet we are not bound to any one methodology.

And the conclusion:

Our algorithms and the data that they generate provide a powerful tool to answer questions like the ones posed above, a list which can be extended indefinitely. Because we are using computers and code, we can change assumptions or look at different texts or divide our existing texts up differently — and immediately generate refreshed data for the entirety of the corpus that we are considering. Furthermore, although it is in most cases impossible to replicate previous scholars’ methodologies with absolute precision, in broad outline we can nevertheless check their results almost instantaneously. This process of replication and verification has long been absent from studies of Latin prose rhythm. Since all our code and data are open source and publicly available, our own results can also be easily checked (and perhaps improved).

Improvements and extensions of these data may take a variety of forms. A different approach to locating clausulae, one that does not rely on punctuation, might help advance exploration of ‘internal’ clausulae, a topic which has thus far resisted rigorous analysis. More extensively marked up texts would facilitate other kinds of investigations: for example, does Cicero use different rhythms in his exordia, or narrationes or perorationes? Annotating his speeches with consistent metadata would allow for more detailed study. More sophisticated data manipulation techniques, like Principal Component Analysis, might give us other profitable ways to categorise our data beyond just ‘artistic’ and ‘non-artistic’. And this is to say nothing of further work that can be done with the data that we have already collected, like that on word division and word accent in clausulae, which would necessarily be crucial in studying the rhythms of late antique texts as the cursus begins to develop.

Of course, none of the broad brush pictures painted by statistical analysis can give insights at the level of an individual clausula in an individual sentence in an individual author’s text. Such an analysis of the details of prose rhythm in the context of a speech or a letter is eminently worthwhile and can have great explanatory power. So when Cicero describes the same event twice in almost the same words in Pro Milone, he once writes ‘respondit triduo illum aut summum quadriduo esse periturum’ (Mil. 26), but later ‘audistis … periturum Milonem triduo’ (Mil. 44). It seems likely that he wrote esse periturum in the first case because it was in clausular position (= esse uideatur), whereas in the second the infinitive came in the middle of the phrase and so he preferred simply periturum. Prose rhythm is one of the keys to unlocking the secrets of Latin word order and word choice, revealing points of emphasis and rhetorical artifice, and understanding it at the local level is essential for appreciating an author’s verbal artistry. Much of this artistry must have been put into practice subconsciously or unconsciously (see, for example, Quint., Inst. 9.4.119–20), and we remain sceptical of accounts that attempt to quantify the force of any individual clausula, but it is clear that ancient authors and ancient audiences could perceive and appreciate rhythmic prose. Today, without native speaker Sprachgefühl, we can only recover these effects by philological analysis.

While interpreting prose rhythm at the level of the sentence and clause requires close reading and analysis, at the global level, questions of prose rhythm cry out for an open-source, Big Data approach. We have offered one such approach, producing algorithms to detect and categorise the rhythms of any Latin prose text, providing comprehensive data generated by these algorithms for most of extant classical Latin prose, presenting a new statistical approach to analysing the significance of those data, and giving several examples of how to use our data and procedures to answer particular questions about authors’ propensity toward artistic rhythms. For example, we can confirm that Cicero’s letters are significantly less concerned with ‘artistic’ prose rhythm than are his speeches, but we can also show how certain letters, like the lengthy and polished Q. fr. 1.1, take particular care to be artistically rhythmical. We can with a few clicks compare the prose rhythms of the perhaps spurious Inuectiua in Ciceronem or Epistulae ad Caesarem senem with those of the undisputedly genuine Sallust: the former does not look at all Sallustian, but the latter actually does. We can compare the rhythms of speeches and narrative in authors like Sallust and Tacitus: Sallust’s rhythms never change, but Tacitus has at least two distinct rhythmic profiles (neither of which, even in the Dialogus, counts as ‘Ciceronian’). We can see almost at a glance that Trajan’s replies to Pliny’s letters in Book 10 have an entirely different rhythmic fingerprint from Pliny’s, while in the Panegyricus Pliny mirrors the rhythmic preferences that he shows in the Epistulae. It may be an exaggeration to claim that technology will revolutionise the study of Latin prose rhythm — the fundamental insights as worked out over a century ago seem to stand correct and confirmed — but it will certainly replough the entire field, offering fresh data and the possibility of countless new results. Nothing will ever make the study of Latin prose rhythm easy, but computers will certainly make it a lot easier.

Finally, I can’t resist quoting the same sentence Laudator singled out: “Latin prose rhythm sometimes looks like a species of philological witchcraft, albeit one without the seductive power of most black magic.”