Laudator Temporis Acti

Basil Lanneau Gildersleeve, "Brief Mention," American Journal of Philology 25.1 (1904) 104-114 (at 104):

Quite apart from any theoretical questions, the practical handling of the prepositions in any given language is a matter of exceeding complexity. Read the elaborate articles on prepositions in English that have appeared from time to time in the 'Englische Studien.' Watch the run of prepositions in England, if you are an American; in America, if you are an Englishman. 'At' and 'in ', 'in' and 'on' are as troublesome as ἐπί c. gen. and ἐπί c. dat. And yet the prepositions are inevitable, the mastery of them a gnomon of one's familiarity with the language, and since the appearance of Mommsen's memorable book, since the founding of Wölfflin's Archiv, the literature of the prepositions in Greek and Latin has become enormously swollen, and it is almost impossible to keep pace with the tide of doctoral dissertations that agitate the subject. In most of those that I have examined the work does not seem to have involved much brain-fag. The categories are taken from the ordinary manuals and all that is needed is care in counting—a homely virtue. But so is cleanliness a homely virtue, and the variations in statistics suffice to show that behind the most seductive array of decimals there may lurk a gross error. I have known an investigator, of whom I had reason to expect better things, to strike an average from the page number of the second volume, oblivious of the fact that there was a first. I have known another of greater note to get his columns interchanged. I have known—but if I go on, I may expose my own shortcomings in the simple matter of numeration and those who are curious in such matters can find my confessions elsewhere. But even if the figures are unassailable, even if the averages are so high as to make any possible error a negligible quantity, one asks: What is the result? What can be the result of statistical work with prepositions? Occasionally the usage of an author as determined by the statistics may help in a question of textual criticism, nay, even in a question of genuineness, but when it comes to prepositional usage as an index of style, the problem taxes the resources of the grammarian, of the rhetorician. Whose senses are so keen as to notice a variation of even ten per cent. in the total use of prepositions? One goes through the whole mass of statistics—and little abides except what any attentive reader might have observed without the statistics. And yet I welcome the statistics, especially those that deal with entire ranges of literature such as Lutz's work on the orators, such as the latest addition to the Schanz Beiträge, Die Präpositionen bei Herodot u. andern Historikern, von Dr. ROBERT HELBING (Würzburg, Stuber).

Id. 25.2 (1904) 225-234 (at 232):

At the same time every mistake in statistics is a demand that the work be done over again. All they that take statistics shall perish with statistics; and no one has protested more vigorously than I have against the misuse of figures in historical syntax.