------+-------+-------+-------+------------------------------------------ Turns | Prob- | | Prob- | of | able | Wave- | able | screw.|errors.|length.|errors.| Remarks. ------+-------+-------+-------+------------------------------------------ 4·62 | ·0037 | 6297 |·00014 |Very bright stripe. } On 7·12 | 9 | 5569 | 2 |Brightest line of the spectrum; } a | | | | becomes noticeably fainter at } faintly | | | | appearance of the red line. } lighted 7·92 | — | 5390 | — |Extremely faint line; unreliable } ground. | | | | observation. } 8·71 | 21 | 5233 | 4 |Moderately bright. 8·95 | 49 | 5189 | 9 |This line is very bright when the red line | | | | appears at the same time, otherwise equal | | | | in brilliancy with the preceding one. 10·06 | 20 | 5004 | 3 |Very bright line. 12·33 | — | 4694 | — |}Broad band of light, somewhat less 12·59 | 22 | 4663 | — |} brilliant in the middle. 12·88 | — | 4629 | 3 |Very faint in those parts of the Aurora | | | | in which the red line appears. ------+-------+-------+-------+------------------------------------------ April 14th.—Faint Aurora; only the bright line in the green could be recognized in its spectrum. The mean of two readings gave 7·12 turns, or 5569 wave-length. I append a table of the wave-lengths of the brightest line, as exactly measured on four evenings:— 1871, February 11 5573 ” 12 5573 April 9 5569 ” 14 5569 Therefore the average result (if only half-weight is allowed to the last observation, because it only depends upon two readings) gives for the wave-length of the brightest line 5571·3, with a probable error of ·000·92. According to Ångström[21], the wave-length of this line is 5567; according to Winlock[22], on the other hand, 5570. 2. _On the Spectra of some Gases in Geissler’s Tubes, as well as on the Spectrum of the Atmospheric Air._