Reference for the claim that only 1% of ancient literature survives

People sometimes make arguments from what our surviving collection of classical texts do NOT contain.  I tend to reply by pointing out that only 1% of classical texts survive, which makes such a procedure very risky.  This figure comes from a statement by N. G. Wilson on the Archimedes Palimpsest Project web page, although the site has changed and I couldn’t find it just now. 

When I met N. G. Wilson by accident at the Oxford Patristics Conference some years ago I asked him about this, and he said that the figure came from Pietro Bembo, the renaissance scholar.

The discussion on CLASSICS-L of the same issue has now produced a quotation with a modern academic reference.  I reproduce the post (by Atticus Cox) here:

In partial answer to Jeffrey B. Gibson’s original question — “What percentage/How much of pre-second century CE literature is lost to us and how has this figure, whatever it may be, been determined?”

Rudolf Blum in his Kallimachos : The Alexandrian library and the origins of bibliography (Wisconsin, 1991) [= transl. by Hans H. Wellisch of BLUM, R.: Kallimachos und die Literaturverzeichnung bei den Griechen : Untersuchungen zur Geschichte der Biobibliographie (Frankfurt, 1977)] states as follows (p.8):

Literary criticism was made difficult for the ancient philologists (similar to what their modern colleagues experience when they deal with medieval literature) because among Greek authors as well as among medieval ones there are so many namesakes (Apollonios, Alexandros, etc.).  Thus, for example, we find that Diogenes Laertios (3rd cent. A.D.) lists for 29 of the 82 philosophers with whom he deals in his work on the lives and opinions of famous philosophers (the most recent are from the end of the 2nd century A.D.) several known namesakes, most of whom had also been authors.[n.31]  Quite a few bearers of the same name were active in the same field, were compatriots and contemporaries.[n.32]  It also happened frequently that an author had the same name as his father, equally well-known as an author.  In such cases the customary procedure for the identification of persons — complementing the personal name by an indication of the father’s name (in the genitive) and the place of birth or domicile (in adjectival form) — was not sufficient.  One had to find further biographical details and to add them.

The large number of authors with the same name was a corollary of the large amount of Greek literature, the sheer bulk of which alone would have been enough to keep the ancient biobibliographers busy.  The small nation of the Greeks was immensely productive in art and scholarship.  Although it is impossible to ascertain the total number of all works written by Greek authors, there were certainly many more than those that have been preserved or are merely known to have existed.  For example, we have no adequate idea of the multitude of works which Kallimachos listed in the 120 books of his Pinakes.  Of the Greek literature created before 250 B.C. we have only a small, even though very valuable, part.  We do not even have the complete works of those authors who were included in the lists of classics compiled by the Alexandrian philologists.  Of all the works of pagan Greek literature perhaps only one percent has come down to us.[n.34]  All others were in part already forgotten by the third century A.D., in part they perished later, either because they were not deemed worthy to be copied when a new book form, the bound book (codex), supplanted the traditional scroll in the fourth century A.D.,[n.35] or because they belonged to ‘undesirable literature’ in the opinion of certain Christian groups.

In n.34 Blum explains that his figures are based on the counts of Hans Gerstinger [= GERSTINGER, H.: Bestand und Überlieferung der Literaturwerke des griechisch-römischen Altertums (Graz, 1948)] —

n.34: “According to Gerstinger (1948) p.10, about 2000 Greek authors were known by name before the discovery of papyri.  But the complete works of only 136 (6.8%) and fragments of another 127 (6.3%) were preserved.  Gerstinger counted, however, only authors whose names were known, not works known by their titles.  The numerical relation between these and the works that are preserved wholly or partially would certainly even be much worse.  Whether a count of known titles would serve any purpose remains to be seen.  The main sources would be the biobibliographic articles in the Suda, but even the authority on which it is based, the epitomator of the Onomatologos by Hesychios of Miletos (6th century A.D.), no longer listed many authors which e.g. Diogenes Laertios (3rd century A.D.) had still named in his work.”

The other notes are as follows:
n.31: “He lists on average 5-6 homonyms, in one case 14 (Herakleides), in two cases 20 each (Demetrios and Theodoros).”
n.32: “E.g. in the fifth century B.C. there were two Attic tragic poets by the name of Euripides other than the famous one.”
n.33: “Despite the more precise Roman system of naming persons (/praenomen/, /nomen gentile/, /cognomen/) there were many homonyms, although relatively few among authors, because there were fewer of them than in Greece.”
n.35: “Widmann (1967) columns 586-603 [= WIDMANN, H. : ‘Herstellung und Vertrieb des Buches in der griechisch-römischen Welt’ in /Archiv für Geschichte des Buchwesens/ 8 (1967) 545-640].