With a technique devised by one of the authors (A.M.G.), more than 300 electroencephalograms have been subjected to a Fourier analysis. The technique consists in taking a strip of record as a shadowgram on film, fastening the 2 ends of the record together to form a belt, rotating the belt between a light and a photocell, and leading the signal thus obtained through an exceedingly sharp continuously variable filter to a recording galvanometer. The resulting record is a continuous plot of the amplitude vs. frequency function of the particular strip of electroencephalogram chosen for analysis. Such a plot is a Fourier transform. It shows the distrib. of energy over a wide range of frequencies and can properly be regarded as a spectrum. Since all analyses thus far have given continuous spectra, they cast doubt on the usefulness of such crude categories as o, B, and 8 waves. By comparing frequency spectra, slight differences in records from time to time in the same person or from person to person can be noted and accurately characterized. This type of analysis reveals features of the electroencephalogram which cannot otherwise be seen. It affords a rigorous and purely objective description of the data.