horizontal "RefraSlhn of the Air. 57 



but, as the fhip did not fink below the horizon, I had 

 not an opportunity of obferving at what time, and in what 

 order, the images would have vanilhcd, if the fliip had fo 

 difappeared. 



Being defirous of feeing whether the fame effeft was pro- 

 duced upon the other (hips which were vifible, I direfted my 

 telefcope to another (liip A, (Fig. 2.) whofe hulk was juft in 

 the horizon xy ; when I obferved a complete inverted image 

 B, the main-maft of which juft touched that of the fhip it- 

 felf. In this cafe, there was no fecond image as before. The 

 fhip A moving upon the horizon, B continued to move with 

 it, without any variation in its appearance. 



The next fliip which I dire6ted my telefcope to, was fo far 

 oil the other fide of the horizon xy^ as juft to prevent its hulk 

 from being feen, as is reprefented by A, (Fig. 3). And here I 

 obferved onlv an inverted image of part of the fliip, the image jk 

 of the topfail, with the maft joining that of the fliip, the 

 image x of the top a of the other maft, and the image z of 

 the end c of the bowfprit, only appearing at that time, 

 Thefe images would fuddenly appear and difappear very 

 quickly after each other ; firft appearing below, and running 

 up very rapidly, fliowing more and lefs of the mafts at dif- 

 ferent times, as they broke out; refembling, in the fwiftnefs 

 of their breaking out, the fhooting out of a beam of the 

 Aurora borealis. As the fliip was defcending on the other 

 fide of the horizon, T continued my obfervations upon it, in 

 order to difcover what changes might take plyce; when I 

 found that, as it continued to defcend, more of the image 

 gradually appeared, till at laft the image of the whole fhip 

 was completed, with their main-mafts touching each other ; 

 and, upon the fhip defcending lower, the image and the fhip 

 feparated ; but I obferved no fecond image, as in the firft 

 cafe : a fecond image, however, might probably have ap- 

 peared if the fliip had continued to defcend. 



Upon moving my telefcope along the horizon, in order to 

 examine any other fliips which might be in fight, I ob- 

 ferved, juft at the horizon xy, (in Fig. 4.) the top a of the 

 maft of a fhip ; and here an effeol was obferved which had 

 not been before difcovcred ; for there was an inverted image B, 

 Vol. \'1I. I vertical 



