Abstract The quadratic photoelectromagnetic (P.E.M.) effect observed by Kikoin and Noskov in cuprite has been observed in germanium and silicon. The effect occurs when the magnetic field B is rotated an angle φ about an axis lying in the illuminated face and perpendicular to the primitive direction of B. The linear effect is produced by the component Bcos φ of B, while the quadratic effect is produced by the action of Bsin φ on the electric current generated by Bcos φ. From this elementary argument the quadratic P.E.M. effect should be proportional to B 2 sin φ cos φ. A part of the effect is produced by the action of a magnetic field on a current produced by the action of a magnetic field on the diffusion current of the injected carriers. Calculating this term, one finds the same type of average of the collision time as appears in the magnetoresistance coefficients. Consequently, the effect becomes anisotropic and, instead of being proportional to B 2 sin 2φ, is proportional to B 2[sin(2φ + Δ) + g], where A and g are functions of the constants of the semiconductor and the orientation of the sample. For certain simple orientations (such as the crystallographic axes) Δ = 0 and g = 0. The ratio of the quadratic to the linear P.E.M. fields is, for small light intensities, a function only of the Hall angles, the magnetoresistance coefficients of electrons and holes, the impurity concentration, and the magnetic field. Agreement between theory and experiment is satisfactory.