Strabo Geography 2.1.35

[35] These arguments may be very geometrical, but they are not convincing. After having himself invented these various difficulties, he dismisses them, saying, Had [Eratosthenes] been chargeable for small distances only, he might have been excused; but since his mistakes involve thousands of stadia, we cannot pardon him, more especially since he has laid it down that at a mere distance of 400 stadia, ¹ such as that between the parallels of Athens and Rhodes, there is a sensible variation [of latitude]. But these sensible variations are not all of the same kind, the distance [involved therein] being in some instances greater, in others less; greater, when for our estimate of the climata we trust merely to the eye, or are guided by the vegetable productions and the temperature of the air; less, when we employ gnomons and dioptric instruments. Nothing is more likely than that if you measure the parallel of Athens, or that of Rhodes and Caria, by means of a gnomon, the difference resulting from so many stadia ² will be sensible. But when a geographer, in order to trace a line from west to east, 3000 stadia broad, makes use of a chain of mountains 40,000 stadia long, and also of a sea which extends still farther 30,000 stadia, and farther wishing to point out the situation of the different parts of the habitable earth relative to this line, calls some southern, others northern, and finally lays out what he calls the sections, each section consisting of divers countries, then we ought carefully to examine in what acceptation he uses his terms; in what sense he says that such a side [of any section] is the north side, and what other is the south, or east, or west side. If he does not take pains to avoid great errors, he deserves to be blamed, but should he be guilty merely of trifling inaccuracies, he should be forgiven. But here nothing shows thoroughly that Eratosthenes has committed either serious or slight errors, for on one hand what he may have said concerning such great distances, can never be verified by a geometrical test, and on the other, his accuser, while endeavouring to reason like a geometrician, does not found his arguments on any real data, but on gratuitous suppositions.