By this method of frequent reference to the principal air lines the measures are not sensibly affected by the errors which might have been introduced from the shifting of the lines in absolute position in consequence of alterations either in the width of the slit, in the place and direction of the discharge before the slit, or in the apparatus from variations of temperature, flexure, or other causes.

The usual place of the electrodes was about 7 inch from the slit, though occasionally they were brought nearer to the slit. When they are placed in such close proximity, the sparks charge the spectroscope by induction; but the inconvenience of sparks striking from the eyepiece to the observer may be prevented by placing the hand upon the apparatus, or putting the latter into metallic communication with the earth.

The spectrum of comparison was received by reflection from a prism placed in the usual manner over one-half of the slit. As the spectrum of the discharge between points of platinum, when these are not too close, is, with the exception of two or three easily recognised lines, a pure air spectrum, this was usually employed as a convenient spectrum of comparison for distinguishing those lines in the compound spectrum which were due to the particular metal employed as electrodes. The measures, however, of all the lines, including those of the air spectrum itself, were invariably taken from the light received into the instrument directly, and in no case has the position of a line been obtained by measures of it taken in the spectrum of the light reflected into the slit by the prism.

The measures of all the lines were taken more than once; and when any discordance was observed between the different sets, the lines were again observed. The spectra of most of the metals were re-measured at different times of observing. In the measurement of the solar lines for their co-ordination with the standard air spectrum, the observations were repeated on several different occasions during the progress of the experiments. The line G of the solar table is the one so marked by Kirchhoff.1 When no change in the instrument could be detected, the measures came out very closely accordant, for the most part iden1 Untersuchungen ii. d. Sonnenspectrum, 2 Theil, Taf. iii. Berlin, 1863.

tical. The discordances due to small alterations in the instrument itself were never greater than 5 or 6 of the units of measurement in the whole arc of 4,955 units. As the apparatus remained in one place free from all apparent derangement, these alterations are probably due to changes of temperature. The method employed to eliminate these discordances has been described.

Throughout the whole of the bright portion of the spectrum the probable error of the measures of the narrow and welldefined lines does not, I believe, exceed one unit of the scale.

NOTES TO THE TABLES.

Upon a re-examination of the Tables I found that it frequently occurred that lines of two or more metals were denoted by the same number. It appeared probable that these lines having a common number were not coincident, but only approximated in position within the limits of one unit of the scale employed; and besides, there might be small errors of observation. I therefore selected about fifty of these groups of lines denoted by common numbers, and compared the lines of each group, the one with the other, by a simultaneous observation of the different metals to which they belong. Some of the lines were found to be too faint and ill-defined to admit of being more accurately determined in position relatively to each other.

The following lines appear with my instrument to be coincident: Zn, As 909

Na, Pb 1000

Na, Ba 1005

Te, N 1366

O, As 1737
Cr, N 2336

Of a much larger number of groups, the lines were, by careful scrutiny, observed to differ in position by very small quantities, corresponding for the most part to fractional parts of the unit of measurement adopted in the Tables. These are:

The character of the lines is indicated as follows:

A line sharply defined at the edges, and narrow when the slit is narrow, s.

A band of light, defined as a line, but remaining even with a narrow slit, nebulous at the edges, n.
A haze of light irresolvable into lines, h.

Double band, too close for measurement, d.

The comparative intensity of the lines is indicated by the smaller figures placed in the position of exponents against the numbers in the Tables....The scale extends from 1 to 10, and includes fractional parts of unity to represent the very faint lines.