A visual photometer is described for measuring the brightness of the zenith sky during twilight. The results of measurements made with it at Bombay on a large number of clear moonless evenings during the period December 1937 to March 1938 are discussed. A red filter RG1 (Schott and Gen.) and a green filter VG1 (Schott and Gen.) were used. With the latter, it was possible to continue observations even when twilight conditions had completely ceased. The variations in brightness agree with a relation proposed by Link, namely,dI/I=β sin θ sec2 θdθ, where I is the brightness of the zenith sky, θ the angular depression of the sun below the horizon and β a constant. If the brightness is due only to primary scattering from the atmosphere illuminated by sunlight, it is possible to connect β with the temperature of the atmospheric layer responsible for the scattering. The difficulty however is to ascertain to what extent secondary scattering is important.