Abstract
A method is described of refining an instrumental profile (initially derived from emission-line observations) of a high-resolution stellar spectrograph, by using profiles of the telluric oxygen bands obtained with a solar spectrometer of still higher resolution. Owing to the light thrown into the wings of the instrumental profile, the observed profiles of absorption lines in late-type stellar spectra are substantially shallower (by ~ 10 per cent of their central depressions) than the true profiles. It is not true that accurate photometry with a high-resolution spectrograph will result in a close resemblance between the spectrum being observed and the observations being made of it. It is not true that the resemblance can be appreciably improved by any increase in resolving power. Even with accurate photometry, the equivalent widths measured with the spectrograph here discussed must be 5–10 per cent less than the true values. Other fundamentally similar spectrographs are unlikely to be much better and may easily be much worse. Important improvements in the observational accuracy of stellar line profiles only seem possible by the use of double-pass spectrometers, which unfortunately have low luminous efficiency.