The occurrence of spikes in Brewer UV spectra is studied. Use is made of continuous measurement data over several years, comprising more than 90,000 spectra, from one single-monochromator and two double-monochromator Brewers. It is shown that the double monochromators, especially, may suffer from more than 200 spikes per ~5000 annual spectra. The spikes are not always randomly distributed over the wavelength range. The single monochromator is found to have an annual average of only 36 spikes above 300 nm, but it is noted that there were a significant number of spikes at shorter wavelengths, indicating possible bias in the stray light correction unless taken into consideration. The error caused by noncorrected spikes varies greatly from case to case. In an intensive study of 150 spectra measured during one summer week, the effect of one moderate-size spike was found to be more than 5% on a DNA action dose rate and close to 1% on a DNA action daily dose. When high accuracy of in situ UV measurements is required, our results suggest a need to remove spikes from the spectra. A simple statistical approach is employed. The method is applicable to any single- or double-monochromator Brewer spectroradiometer. However, under rapidly changing cloudiness it can be difficult to distinguish between noise spikes and the variation in irradiance due to changes in the state of the sky. Our data show that ancillary radiation measurements may be necessary to interpret the data correctly.