An improved method of assessing the frequency response of an in vehicle exhaust gas air/fuel ratio sensor by measuring and analyzing the sensor response to a predetermined perturbation of the fuel delivered to the engine. In a first test mode that provides both quantitative and qualitative assessments, the perturbation is achieved by applying fixed biases to the fuel pulse widths of individual engine cylinders to create a rich/lean perturbation in the exhaust gas, and by adjusting the engine throttle to gradually vary the engine speed over a test interval so that the rich/lean perturbation correspondingly varies in frequency. Since the biases are fixed, intake port wall-wetting effects are minimized. In a second test mode that provides a qualitative assessment, the perturbation is achieved by applying an alternating fuel bias multiplier to every engine cylinder, with the engine operating at a fixed speed and load setting that is of interest for diagnostic purposes. In each case, the output of the air/fuel sensor is band-pass filtered at the frequency of the fuel bias pattern to identify the sensor response, and the response is rectified and low-pass filtered to produce a D.C. measure of the response amplitude to generate a pass/fail indication. In the first test mode, the output of the air/fuel sensor can also be sampled and incrementally processed with a Fast-Fourier-Transform (FFT) technique to identify the response amplitude of the sensor at each of a plurality of frequencies, forming the basis of a Bode plot.