AbstractTo obtain the amounts of atmospheric ozone from measurements with Dobson's photo‐electric spectrophotometer, allowance has to be made for the extinction of light by scattering due to molecules and large particles. Dobson has proposed two slightly different formulae which lead to appreciably differing values on hazy days. The assumptions involved in the formulae are examined and Dobson's equation is generalized, assuming the particle scattering to vary as λ‐n. The expression for the ozone amount is split up into two terms, the first main term being that obtained on the assumption of a pure atmosphere and the second correction term denoting the haze‐scattering. On calculating the ozone amounts with different values of n, it is found that these show much greater consistency if n has a value intermediate between those assumed in Dobson's two formulae and somewhere near zero.On hazy days in winter, it is observed that (λ‐n), the differential haze‐scattering for the long wavelengths 3300A and 4450A, is positive and its magnitude increases with increase in the haziness of the sky. This is similar to what is observed in Europe. But on most hazy days in the pre‐monsoon hot season in North India, (λ–n) is found to be negative, which means that the apparent extinction of light at 3300A is less than that at 4450A.One possible explanation is that with certain sizes of particles of given refractive index, the extinction by scattering increases with increase in wavelength as was found from observations by Götz in Switzerland and by Stratton and Houghton in U.S.A. An alternative explanation is that Dobson's instrument analyzes not only the transmitted radiation from the sun but also some scattered radiation from the surrounding hazy sky. It is probable that the North Indian summer haze scatters in the forward direction an appreciable amount of radiation which is richer in shorter wavelengths than the directly transmitted sunlight.For calculating the daily values of ozone, it is proposed to assume the particle scattering to be nearly neutral and thus varying as λ−n with n=o. The correction term in the generalized formula then becomes 0.165 (δ′–δ″) and is obtained from observations on the long wavelengths.