Absolute intensities have been measured over the whole twilight at an angle of elevation of 20° in an attempt to derive quantitative data on dust concentrations in the mesosphere. With the aid of model computations based on single scattering we have been able to separate the effects of cloud and haze at the sunset point, ozone absorption, the airglow, multiple scattering and mesospheric dust in a series of 93 twilight recordings. The effect of ozone absorption on green, orange and red wavelengths is readily interpretable in terms of simple atmospheric models, and can be used as a remarkably accurate measure of the total ozone amount. Profiles of turbidity are derived up to 65 km and indicate significant quantities of dust at every level. A maximum concentration is commonly observed near 20 km, and corresponds quantitatively to that found by direct sampling by Junge and his colleagues. Above 30 km the dust mixing ratio is constant at about 25 per cent of its value at 15 km. This concentration is very small compared with particle concentrations in a noctilucent cloud, but it is consistent with the amounts of matter inferred from the sodium resonant emission in the twilight glow. A seasonal maximum in winter months is also consistent with the sodium twilight glow. Negative results include: definite evidence against the existence of the 100–150 km haze layer invoked to account for some aspects of lunar eclipses; lack of correlation with meteor shower activity, speaking against Bowen's theory of rainfall anomalies; computations indicating that thermal and photochemical effects of the dust are probably small.