We have already repeatedly had occasion to notice how very different the system of our space-sensations--our physiological space, if we may use the expression--is from geometrical (by which is here meant Euclidean) space. This is true, not only as regards visual space, but also as regards the blind man's tactual space in comparison with geometrical space. Geometrical space is of the same nature everywhere and in all directions, it is unlimited and (in Riemann's sense) infinite. Visual space is bounded and finite, and, what is more, its extension is different in different directions, as a glance at the flattened "vault of heaven" teaches us. Bodies shrink when they are removed to a distance; when they are brought near they are englarged: in these features visual space resembles many constructions of the metageometricians rather than Euclidean space. The difference between "above" and "below," between "before" and "behind," and also, strictly speaking, between "right and left," is common to tactual space and visual space. In geometrical space there are no such
differences. (p. 181f)
By "physiological space," presumably Mach means that which is customarily called "perceptual space" (or "phenomenal space") today, including but not being confined to visual sensations. His use of the term "visual space" to denote just the visual component of perceptual space supports my previous contention that this term has long been used for the space of visual sensations (at least in science, if not philosophy). Rather than relating some of the characteristics of VS to projective geometry, Mach relates them to metageometry (i.e., what today is called non-Euclidean geometry). Other characteristics that he notes (e.g., left and right) as departing from Euclidean space are now considered to be topological properties of space.
In a previous chapter Mach notes that "optical space" [a term he uses synonymously with "visual space"] represents geometrical space (Euclid's space) in a sort of relievo-perspective." (p. 169) There is a wealth of other observations and points about visual/perceptual space in Mach's book, and I highly recommend it for that reason, and as ostensibly being the first to propose that VS is non-Euclidean, though not apparently on the basis of curvature, but on the basis of it being like a perspective projection.