Inspired by an earlier question about Yoda's word order in Latin, I would like to know how common different word orders were in classical Latin. To be more specific, I would like statistical information.

For example, an analysis of a single work of literature (preferably prose) for the number of instances of the different possible orders of subject, object and verb (SOV, SVO, OSV, OVS, VSO, VOS) would be great. I leave the exact choice of texts analyzed open, but please indicate what the source material is in your answer. I am not aware of any such research, but I assume it must be out there. I also assume that there is a sufficiently annotated corpus to make this kind of search possible, but I lack the aptitude to run a check myself.

So, my question is: How common are different word orders (permutations of SOV) in classical Latin? I am looking for a numerical answer based on analyzing some amount of extant literature.

I don't expect anyone here to take Caesar's De bello Gallico (or anything else) and analyze every sentence by hand. I expect that there are easier ways nowadays, or that maybe someone has published the results of such tedious work.

(If you know good text corpora, please suggest them in the dedicated question!)

1 Answer 1


I was taught Latin prose composition in a way that is now almost forgotten. There were many tricks of the trade to be acquired, including such basic rules as correctly sequencing tenses; most of these could be found in the primers and study guides — as they can be still — and one of these was how to change emphasis by altering word order, which apparently underlies your question. (In the end, however, it was considered that the only way to write with good style was to develop one's own — and that could be done only by reading widely, in just the same way as one's style of English could be developed.)

Your query seems to suggest that you are looking for rules of word-order that could be applied to prose composition. Given the vast quantity and variety of source material, that is formidably difficult. I can readily understand your wish for statistical information but, even if examples could be gathered, how would we know that they represent accurately the population from which they were drawn? And what might that population be? We might think that in Caesar's 'diaries' (rejecting those books of which his own authorship is doubted), we would be analysing just a single person's style: but there are quite a few passages which are very obviously not his own and have probably been interpolated from elsewhere. How much of such material was included with Caesar's approval? Where should we draw the line for acceptance?

Then there is the probability that even a single writer's style depends on the circumstances under which he is writing. For the works of Cicero we can be comfortably, if not completely sure that he is genuinely the author, but familiarity gained by reading selections of his works reveals that his style (like that of most people) depends on the subject and the mood of the writer. For example, the measured periods of the forensic speeches are not at all the same thing as his personal correspondence. Do we know that Tiro's copies accurately reflect his master's originals — or was he allowed, or even instructed, to make changes for stylistic effect? Perhaps; but how much 'correction' has been applied by copyists and grammarians in the process of preserving his works in a 'consistent' style? And that, of course, applies not only to Cicero's, but to all texts that survive to the present day; and not only to the syntax to which your query applies, but to the accidence.

Returning to your question, how would you expect the statistical data be presented? Maybe you have in mind a way in which the six permutations of [SOV] can be related to purpose, style, etc. for some particular set of authors, by some kind of multivariate analysis. That would indeed be a formidable, indeed virtually impossible, task if it were done in the way of nineteenth-century grammarians — but indeed, how very interesting! I think it would be a big surprise, but let's hope that someone can come up with a sound analysis.

  • 4
    Although this is very insightful, it merely challenges the OP's premise instead of attempting to give an answer. On the contrary, I think a statistical analysis is possible and useful (not, of course, as a guide for our own composition), and the objections brought up here (sample size, bad data, presentation) are readily answered with some knowledge of statistics.
    – brianpck
    May 16, 2017 at 11:52
  • 1
    As one useful example, consider my unsupported statement in another question that OSV is rare in Latin: I would love to have empirical data for or against this, even if it has little bearing on whether I should use that order in a particular instance.
    – brianpck
    May 16, 2017 at 11:54
  • While I appreciate this as s comment, it does not even begin to answer the question. I'm sorry, but I have no option but to vote down. I am interested in word order statistics out of curiosity. I would certainly not deduce any hard rules from the information. The result may indeed depend greatly on the source material chosen, but I would like to see anything since I honestly have no idea what the statistics could look like at all. Making such an analysis is not such a formidable task if one happens to have a suitably structured corpus. Technology can do wonders.
    – Joonas Ilmavirta
    May 16, 2017 at 16:45
  • @ Joonas Ilmavirta I think that's a fair enough judgment. I intended my answer to be helpful, but obviously misjudged the intention of your question. I can only add that, as it read to me, you seemed to be looking for statistical evidence on the occurrence of the [SOV] order with a deeper purpose in mind, while I wasn't sure myself that there's any real value in acquiring the data without relating it to something else — as I was trying (rather obscurely?) to indicate.
    – Tom Cotton
    May 16, 2017 at 19:00

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.