Elegant and hefty, the 2022 edition of the TLG Canon of Greek Authors and Works is a tribute to an astonishing monument. As readers of Classical Review know, the current TLG covers almost all texts extant from ancient Greek literature (in a sense that excludes epigraphy and documentary papyri) and a good portion of Byzantine texts; it now begins to cover early modern Greek (works later than 1800 are not yet contemplated). The brief discursive introduction to the volume alludes to the foresight of a few founding figures who went beyond the narrow programme of a nineteenth-century lexicon to create, instead, a modern technological database, with M. McDonald and D. Packard deserving most of the credit. It is interesting that both combined philanthropy with scholarship, each heir to a fortune built on modern technologies.
Just as the lexicon took the leap and became a database, so the Canon is on the verge of taking a leap to become an encyclopaedia. Originally conceived as no more than a metatext for the database, the Canon stood out, already in its preceding iterations, as one of the best surveys we have for Greek literature. However, the fact that this encyclopaedic role was not originally envisaged gives rise to certain limitations.
The canon is derived from the database of texts. The definition of an author is therefore dependent on past editorial choices. The TLG did not have the resources – and did not prioritise – to make its own search for fragments. Instead, as the TLG absorbed past edited texts, it also absorbed past collections of fragments. So, for instance, Archytas of Tarentum found a solution to a problem equivalent to finding a cubic root. Like everything else by Archytas, the original work in which this solution was presented is now lost, but the solution itself was preserved by Eutocius in his commentary to Archimedes’ Sphere and Cylinder II (Heiberg 1915, 84.12–88.2). Since Diels identified Archytas as a Presocratic philosopher (incidentally, doubtful on both counts), this solution by Archytas became entrenched as a ‘fragment’ and so is one of the texts justifying the TLG Canon's heading ‘Archytas Phil. (0620, *002 fragmenta)’. Contrast this with Dionysodorus, a Hellenistic author who produced a solution to another problem, equivalent to a maximum for a certain other cubic equation, transmitted through the very same commentary by Eutocius (Heiberg 1915, 152.27–160.2). Since there is no Diels-like collection of mathematical fragments, this Dionysodorus is not listed by the Canon.
In general, the current Canon seems to have a better coverage of belletristic authors than of technical ones. The original decision to create a database instead of a lexicon was a twentieth-century leap from the nineteenth century to the twenty-first century. But the database itself collects the fruits of nineteenth-century scholarship and is constrained by its interests. The philological origins of the TLG confine it, further, linguistically. Consider now the mathematician Diocles, attested, in Greek, similarly to Dionysodorus – through citations in Eutocius’ commentary (he, too, offered a solution to the same problem solved by Dionysodorus). I am not sure why, but this author did obtain an entry in the Canon: ‘Diocles Math.’ (1317), noted for his fragments. I suspect the reason could be that the editors of the Canon were made aware of this Diocles because he is, in fact, extant. While known in Greek only in fragmentary form, a significant treatise is preserved in Arabic translation (Toomer 1976). The entry on Diocles, however, does not refer to this extant Arabic treatise, and how could it? The Canon, as currently conceived, only refers to the works in the Greek language database that it was designed to support. Similarly, the entry on the astronomer Ptolemy does not refer to his Optics – a work of extraordinary significance for the history of science, now extant (in part) only as a Latin translation from an Arabic translation. All of this, once again, makes sense for a lexicon: Arabic or Latin do not shed light on the meaning of Greek words. But as the TLG comes to be seen as a repository for all the literature produced within the scope of Greek culture, broadly defined in time and space, the choice to avoid, in the Canon, works now extant only in other languages, becomes less obvious.
Why is it useful to have a repository of all Greek literature? For past scholars, the purpose was traditionally hermeneutic: a full survey, digitally accessed, made it possible to find parallels to a given passage. The TLG gains a different meaning within the context of Digital Humanities. There are questions we may ask at the level of the corpus. What are the spatial, chronological or generic contours of ancient literature? The TLG is a useful tool for such questions, and it would be a service to the discipline if the TLG – and the Canon – developed in this direction, becoming ever more useful for the study of the corpus as a whole. A simple, if somewhat labour-intensive project, would be the completion of the Canon based not only on edited works (with the biases that come with them) but also on other reference works.
The Canon would also benefit from a more nuanced approach to statistics. It is a little difficult to use the chronological information of the TLG because of its extremely prudent chronological definitions. Even when more precise dates are known, chronologies are given in terms of centuries only. Further, it provides little information to be told, for instance, that the Homerica (author number 1802) derive from anywhere between the fifth century bce to the tenth century ce. Would it not be better to tell us something about the distribution of the text through time – the rough percentage associated with each century? This is a case of distributing an aggregate between its components. Or take a certain ‘Menecrates Comic.’ (author number 0523). Perhaps an author of New Comedy, known through a couple of titles and one brief fragment, it is indeed difficult to date him. The TLG simply states ‘post 5 b.c.’, which is prudent. But surely we can say that a New Comedy author is most likely from the third–second centuries bce, the third somewhat more likely than the second, based on our knowledge of the overall chronological distribution of comic authors? Would such more precise probabilities not be of help in large-scale searches?
It will be difficult to include such statistical and probabilistic statements in book form – but they could easily be included in digital form. With current tools, there are certain probabilistic purposes for which the book form is currently preferable. I would encourage scholars to invest more effort in the random reading of ancient texts and authors. That is: if one reads not the cherry-picked examples one is already familiar with, but instead forces oneself to read through, say, a hundred randomly generated passages as defined by given constraints, one is sure to come up with new observations and generalisations. Right now, I use the printed Canon for this kind of Sors Vergiliana (I usually rely on page numbers for this purpose), since the TLG itself – to my knowledge – does not have a device for randomly generating passages. It would be good if the TLG would create tools for such applications.
Readers would sense that my own research leads me to questions concerning the general properties of the ancient literary corpus. I would love to see more tools developed in this direction, and I cannot resist using the opportunity to ask for them. But that we can even think in such terms is thanks only to the foresight of the makers of the TLG. And so, finally, this review concludes, as it should, on a note of astonished gratitude.
