314 Mr. W. Le Conte Stevens on 



stereoscope, though vision with the unaided eyes is preferable. 

 The visual result is thus obtained without strain, and by optic 

 parallelism or divergence, either of which conditions makes 

 Brewster's theory of intersection inapplicable. By reversing 

 the inclination of the cards to the parallel visual lines, the 

 apparent convexity is changed into concavity. We have thus 

 stereoscopy obtained from a pair of perfectly similar conjugate 

 pictures, and capable of reversion at will without conscious 

 motion of the eyes or removal of the stereograph. 



This is, so far as I have been able to learn, a new 

 phenomenon in binocular vision. Brewster's unexplained 

 observation had apparently passed into oblivion on account of 

 the exceeding strain upon the eyes which it involved. The 

 curvature of the phantom wall, in a plane passing longitu- 

 dinally through the observer's body, has lately been redis- 

 covered by Professor Joseph Le Conte, and shortly afterwards 

 its curvature in all directions by myself. On working out 

 the explanation, which would not have been possible except 

 by rejecting Brewster's theory of triangulation, I at once 

 devised a simple attachment for my adjustable stereoscope, 

 capable of use however with any such instrument, if so 

 arranged that the stereograph may be rested upon a cross 

 bar, instead of sliding within a box. Let MN (fig. 2) be 

 such a cross bar, in front of the eyes whose optic centres are 

 at and 0'. A pair of extra short bars PQ and P'Q' are 

 pivoted at C and C ; , over which rest the centres of two 

 similar series of concentric circles whose horizontal diameters 

 are included between E and D, E' and D', respectively, the 

 plane of each card being perpendicular to the page. The 

 central visual lines CO and C'O' are parallel, but for obvious 

 reasons made very short in the diagram. The retinal pro- 

 jections of ECD and E'C'D' are ecd and e'c'd' respectively. 

 Because the triangles EOD and E'O'D' are oblique, their 

 medians divide the angles at and 0' unequally ; hence 

 dc>ce and d'c' <c'e f . The retinal images in the two eyes are 

 hence dissimilar; and this dissimilarity may be made so great 

 by increasing the angles NCD and MC'E', that the binocular 

 image becomes indistinct if the eyes are not made to play 

 rapidly over the picture. If the attention be momentarily 

 withdrawn from C and C to D and D r , the visual lines 

 become divergent to an extent measured by the difference of 

 the angles dOc and d'O'c'. The associated contraction of 

 the external rectus muscles which this necessitates, at once 

 produces the sensation that habitually accompanies recession 

 of the object binocularly viewed. The same is true if the 

 attention be restored to C and C, and then given to E and 



