Figure 8 shows, for MC 21648, the lines Hβ, Hγ, Hδ, Hε, and K, for four out of the nine focus settings. The change in line depth, and the blunting of the intensity curve, are at once apparent. TABLE VIII Spectrum 2c 2d 2e 2b 2a Setting 16.2 16.4 16.6 16.8 17.0 Hβ .53 .53 .51 .49 .47 4481 .13 .13 .09 .09 .09 Hγ .60 .59 .60 .60 .63 Hδ .63 .65 .65 .66 .62 Hε .62 .62 .66 .62 .65 K .15 .14 .13 .12 .10 c. _Accuracy of microphotometer records._—As was pointed out in Harvard Bulletin 805, the width of the analyzing beam, which is not in any case greater than one-tenth of an Angstrom, is such that no smoothing effect need be considered at the line center. In a few cases the same line of the same spectrum was registered twice. The measures made upon the two tracings were always satisfactorily accordant. [Illustration: Figure 9.—Test of the consistency of spectra taken with different apertures.] Ordinates are distance from “clear film” to “line center” taken from microphotometer tracings of spectra of α Aquilae, MC 20800. Abscissae are the ratio (_l_ + _m_ for one aperture)/(_l_ + _m_ for twice the aperture). The fact that the points lie on a smooth curve indicates that the results are satisfactorily consistent. The consistency of the results given by the tracings of several spectra of the same star, when photographed with different apertures, may be examined by means of the plot shown in Figure 9. Ordinates are values of _l_ + _m_. Abscissas are values of the ratio ((_l_ + _m_) for one aperture)/((_l_ + _m_) for twice the aperture). It is evident that if the points thus derived fall on a smooth curve, the results derived from different tracings of the same spectrum will be mutually consistent. The method of interpolation described in Section 6 may therefore be used in deriving the differences of intensity between line and background. If the method of Section 6 is to be successfully applied, it is essential that the total range (“darkness” to “clear film”) shall be uniform for a single series of tracings. In general the variations in total range do not exceed three or four scale units, but, for some spectra, occasional changes of eight or ten units have occurred, generally owing to changes of voltage or room temperature. Under these circumstances, it has been thought best not to attempt to apply any correction for variations in total range, but to reject from the “selected mean” readings from spectra that gave very discordant total ranges.