© 2025. All rights reserved.
Site created with
Hakyll.
Modified theme
lanyon-hakyll
based on
Lanyon.
© 2025. All rights reserved.
Site created with
Hakyll.
Modified theme
lanyon-hakyll
based on
Lanyon.
Dualismen i vor nyeste Philosophie (Copenhagen: den Gyldendalske Boghandel (F. Hegel), 1866; 76 pp., printed by I. Cohens Bogtrykkeri) is Georg Brandes’s first book, published when he was twenty-four. It is a polemic against the doctrine that faith and knowledge are absolutely heterogeneous — that they are principles of such different kinds that they cannot contradict one another and can be held together in one consciousness without discord. The doctrine is Rasmus Nielsen’s, and the book is Brandes’s intervention in the controversy that Nielsen’s Grundideernes Logik and Philosophisk Propædeutik had set running and that Brøchner was fighting on the other front.
Brandes’s argument is that the doctrine cannot be held. If faith and knowledge are utterances of one and the same being, they cannot be utterances of absolutely heterogeneous principles; the heterogeneity that is supposed to keep them from colliding is precisely what would make a single consciousness impossible. What the doctrine actually produces, he argues, is not peace but a division of the person — one man in his science and another in his life — and the third chapter presses the point that a principle defended as beyond reason turns out, under examination, to be against it.
Complete: printed pp. 1–76, comprising the half-title, title and Indhold (pp. 1–5) and the four numbered chapters:
The book closes with an unheaded Efterskrift on p. 76, noting that a piece by Nielsen which appeared while the book was in press — the reply to Zeuthen — gives Brandes no occasion to add to or correct what he had written.
Twenty-two footnotes carry the citations, mostly to Nielsen’s Grundideernes Logik and Philosophisk Propædeutik and to Heegaard’s Indledning til Ethiken, with one to Kierkegaard’s Afsluttende Efterskrift.
This is a diplomatic transcription. Nineteenth-century orthography stands as printed, printer’s errors are transcribed as printed and logged in a comment at the site, and nothing is normalised.
Three things are worth knowing about how it was made, because they bear on how far it can be trusted.
It was made from two scans of two different copies. The copy-text is the Google Books scan of the British Museum copy (600 dpi bitonal), and the Royal Danish Library’s scan is kept throughout as a second, independent machine reading. Every batch was diffed against both before it was spliced. The reason is specific: the copy-text is JBIG2-compressed, and JBIG2 is the one scheme that can substitute a whole glyph rather than blur it — but a substitution invented by a compressor cannot also appear in a photograph of a different copy. Nine disagreements between the two were adjudicated at the image during set-up, and no such substitution was found.
The check was tested before it was trusted. Faults of the three classes
that matter were planted in a finished batch and the diffs re-run. A dropped
ikke and a substituted word were caught by both witnesses. A modernised
archaic form was caught by neither — because both machine readings modernise
that word themselves. That is the method’s real limit, and rather than assume
it small, the affected class was audited at the image: the book’s mixed
orthography (Deel beside Del, heelt beside helt, naturligviis on p. 59
and nowhere else) is recorded occurrence by occurrence in the transcription’s
header, and is not normalised in either direction.
The error rate was measured, not claimed. Every batch had reported no corrections, which is exactly the pattern that ought to invite suspicion, so 14 of the 70 body pages — a fifth of the book, drawn by a seeded random generator so the sample could not be chosen after the fact — were read at the image against the finished file by readers who had not transcribed it and were told to assume it wrong. They found no discrepancies: no dropped words, no modernised forms, no silently corrected printer’s errors, no punctuation or emphasis faults. The measured rate is 0.00 errors per page, with a 95% upper bound near 0.21.
What remains unmeasured is stated in the header rather than left to inference. Fifty-six pages have been checked by the two machine witnesses and by their own transcriber, but not by a second reader. Emphasis inside footnotes is the weakest class in the edition: the book’s only emphasis device is letterspacing, neither text layer records it at all, and the mechanical detector that finds it is unreliable in the smaller footnote type — it missed a four-word letterspaced run in the note to p. 21, which was caught by eye.