1894.] Images following Visual Impressions. 137 



Experiment 2. 



For the screen with the aperture at D, tig. 1, another was substi- 

 tuted, having a horizontal slit 7 cm. long and 2 mm. wide, the image 

 of which was projected upon the screen B after reflection from the 

 rotating mirror. Thus a small spectrum was produced, which re- 

 volved parallel to itself, in a circle about 1 metre in diameter.* The 

 eyes were directed upon a fixed spot near one end of the horizontal 

 diameter of the circle. The spectrum was followed by a ghost of the 

 form rather roughly indicated in fig. 2. It extended from the orange 

 of the spectrum to the beginning of the violet, terminating somewhat 

 abruptly at the orange end, and fading away gradually at the other. 

 The image was distorted, as shown in the figure, approaching nearest 

 to the spectrum at about the middle of the green, a little on the more 

 refrangible side of the most luminous portion. The distance sepa- 

 rating the spectrum from the image increased more rapidly towards 

 the red end of the spectrum than towards the violet end, and the 

 image was widened out considerably at the violet end; but neither 

 the moving spectrum itself nor its recurrent image was so sharply 

 defined as appears in the diagram. 



It was remarkable that the whole of the recurrent image of the 

 spectrum was of a violet hue, being brightest where the distance 

 irom the spectrum was least. No trace whatever of yellow or 

 greenish-yellow could be detected at the more refrangible end, nor 

 of blue or bluish-green at the other. 



The apparent absence of any colour except violet in the. recurrent 

 image of the complete spectrum is capable of two possible explana- 

 tions. The greenish-yellow seen at one end, and the blue and bluish- 

 green seen at the other, when the spectrum colours are tested sepa- 

 rately, may be due merely to an effect of contrast, the true colour of 

 the image being in both cases a weak violet. Or, on the other 

 hand, these colours may really be present at the ends of the image 

 of the whole spectrum, being, however, of too weak an intensity 

 to be distinguishable when in proximity to the more luminous por- 

 tions of the spectrum itself and of its image. 



Two experiments were made in the hope of settling this question. 



Experiment 3. 



The slit at D was removed, and in its place was put a zinc plate 

 having two small apertures close together. A second lantern and 

 prism were set up, and two spectra were projected upon the zinc 

 plate. By the help of screens, things were so arranged that a violet 



* Helmholtz observed the after-images of a spectrum seen for an instant, but 

 failed to notice the dark interval which preceded their appearance (' Phys. Opt.,' 

 p. 376). 



