Monday, May 6, 2013

The Computer Reads Aristotle's Protrepticus: Harold Tarrant's approach to ancient source criticism through computational linguistics

Professor Harold Tarrant (Conjoint Prof. Harold Tarrant at The University of Newcastle, Australia) has in a number of recent papers (2010, 2011a, 2011b, 2012a, 2012b, 2013) researched various aspects of ancient philosophical texts by means of powerful statistical methods of computational linguistics developed and employed at the University of Newcastle's Centre for Literary and Linguistic Computing.

Recently, Professor Tarrant has begun applying these methods to "the detection of layers of borrowed material in later Greek philosophy". When he told me in Cambridge last March that he was applying these methods to the source criticism of Porphyry and Proclus, I was interested to hear what he thought about the application of these methods to the Protrepticus, since our reconstruction of Aristotle's Protrepticus is based in large part on the detection of borrowed material in later Greek philosophy (e.g. in Iamblichus' Protrepticus and On the Common Mathematics). In less than a month since our conversation in Cambridge, Harold has produced a report, entitled: "The Computer Reads Aristotle: Iamblichus, Protr. VI-XII and On the Common Mathematics XXI-XXVII". He has graciously granted me permission to post it to our website at:

(By the way, the paper is very nice to read in electronic format, since it contains colored figures. Please do not reproduce or quote the paper without Professor Tarrant's permission.)

The statistical methods involve multivariate analysis, in particular principal component analysis and cluster analysis, in order to determine the proximity of a set of texts on the basis of their use of common, everyday terms (as opposed to technical philosophical vocabulary). Although such methods are in themselves fairly elementary, they yield highly useful and suggestive results, and yet are rarely seen applied in the field of Greek Philosophy. Perhaps this is because, although computers in principle make the mathematical-logical aspect of the work fairly easy to carry out, the research still requires competence not only in ancient Greek (or at least the ability to reduce Greek verbs, nouns, pronouns, and adjectives into single forms), but also field-specific knowledge (in order to select significant versus insignificant function-words from lists of common words). Further, in the case of source criticism, one must not only be familiar with the hypothetical source text (e.g. Numenius of Apamea) but also the cover text (e.g. Porphyry's Cave of the Nymphs, which may have used Numenius as a source). (I draw this example from an unpublished manuscript of Prof. Tarrant's.) This is an area requiring intense interdisciplinary work and collaboration between mathematicians, logicians, linguists, computer scientists, and, last but not least, classicists and philosophers. Thus I view Harold Tarrant as a true renaissance man in his pioneering application of these methods to ancient philosophy.

Over the next several weeks, I plan to discuss parts of Professor Tarrant's paper in greater detail, because not only do the results arguably confirm (by a completely independent method) some of our hypotheses about the relationship between Iamblichus' cover text and the source text of Aristotle's Protrepticus (such as the authentication of Aristotle as the source of Protr. VI-XI + DCMS XXVI), but they can also help us interpret the source or sources of DCMS XXI-XXV and XXVII.

But the article is fascinating even apart from the specific issues raised by the application of its method to the Protrepticus: this methodology seems to me to be very important for the future of source criticism and thus for Classical Studies and the History of Philosophy in general. The Protrepticus might serve as a model of the power of this method.


Tarrant, H. (unpublished manuscript) 'The Computer Reads Aristotle: Iamblichus, Protr. VI-XII and On the Common Mathematics XXI-XXVII. Available online at:

Tarrant, H. (2013) ‘Narrative and Dramatic Presentation in Republic III’, in N. Notomi & L. Brisson eds. Dialogues on Plato's Politeia (Republic): Selected Papers from the Ninth Symposium Platonicum, Sankt Augustin: Academia, 309-313.

Tarrant, H. (2012) ‘Appendix 2: Report of the Working Vocabulary of the Doubtful Dialogues’ (with T. Roberts), in Marguerite Johnson and Harold Tarrant (eds), Alcibiades and the Socratic Lover-Educator, London: Bristol Classical Press, 223-236.

Tarrant, H. (2012) ‘The Origins and Shape of Plato’s Six-Book Republic’, Antichthon 46, 52-78.

Tarrant, H. (2011) ‘A Six-book version of Plato's Republic: Same Text Divided Differently, or Early Version?’, ASCS 32 Selected Proceedings: Refereed papers from the 32nd Annual Conference of the Australasian Society for Classical Studies, Auckland, NZ. Available online at: 

Tarrant, H. (2011) ‘The Mythical Voice in the Timaeus-Critias: Stylometric Indicators’ (with E.E. Benitez, and T. Roberts), Ancient Philosophy 31, 95-120.

Tarrant (2010). ‘The Theaetetus as a Narrative Dialogue?’, in N. O’Sullivan (ed.) ASCS 31 Proceedings, 2010: Available online at: