This thesis evaluates the human ERG obtained with sinusoidally varying light stimulation. Eighteen test frequencies are used, varying from 2 to Hz. The test is administered to both normals and patients with retinitis pigmentosa (R.F.) The output waveforms are subjected to a Pourier analysis, and the Fourier coefficients form the data base for a cluster analysis. The clustering methods have three main goals: 1) To automatically separate the normals and the R.P.'s, based on the Pourier coefficients, 2) To find groups among the R.P.'s based on these coefficients, and 3) To investigate these groups for a relationship with physiological or clinical variables that the ophthamologist measures. Three methods of clustering are employed: 1) The K-Keans algorithm, 2) A hierarchical method, and 3) Approximation of the frequency versus phase of the first harmonic graph. Normals are grouped correctly by all methods, while all but three R.P. eyes are separated as abnormal. The greatest correlation occurs with the flash ERG. The phase approach leads to the most promising results, and further research in this area is suggested.