Commentary Par II 94-105
In Aristotelian manner, Beatrice extols the virtue of experiment. In this 'thought experiment' that she proposes for Dante to perform, the two equidistant mirrors (one of which is, strictly speaking, unnecessary) represent the surface of the Moon, while the one set farther back stands in for the indented portion, where the dense part begins. The three lights reflected in the mirror will show equal brightness, if not equal size. Thus, as Beatrice clearly means to instruct Dante, the quality of the light is not affected by distance; only the quantity is.

For a discussion of this passage and the significance of mirrors in Paradiso, see Miller (Mill.1977.1). See also Tate (Tate.1961.1), offering a first 'Trinitarian' reading of the third experimentally unnecessary mirror. Boyde (Boyd.1995.1), pp. 14-16, actually performed a version of the experiment in order to test Dante's method (it passed his test). Gilson (Gils.1997.1), pp. 204-6, discusses various other similar 'experiments' described by earlier writers (Pseudo-Thomas on the Meteorologica, Chalcidius on the Timaeus, Roger Bacon, Albert the Great on the De causis), concluding that, while Albert's is the closest to Dante's, in this (as in so much else) Dante simply cannot be pinned down. And now see Turelli (Ture.2004.1), in polemic with Kleiner (Klei.1994.1), p. 105 (who believes that 'dopo il dosso' means 'directly behind your back,' thus making the experiment literally impossible). But see Landino's gloss (DDP Landino.Par.II.100-108); he simply (and understandably) assumes that the light is above the experimenter's head ('et sopra el capo tuo ti sia un lume'). In our own time, Turelli convincingly defends the experimental 'correctness' of Beatrice's verbal demonstration; Moevs (Moev.2005.1), p. 115, however, accepts Kleiner's presupposition, arguing that in the thought experiment we must realize that Dante possesses -- meaningfully, of course -- a transparent body. On the other hand, see the second version of this 'experiment' in the poem ([Par XXVIII 4-9]), where the same conditions pertain and the reader once again feels justified in making Landino's assumption.