Signal‐to‐noise ratios at the output of an f‐m receiver are determined as functions of the input signal‐to‐noise ratio, clipping level, and i‐f and audio filter characteristics when random (fluctuation) noise accompanies the signal to be observed. Both narrow and broad band f‐m are examined. Specific calculations are made for sinusoidal frequency modulation, and it is indicated how data for this simple type of signal may be related to the more complex phenomenon of speech. The concept of the signal‐to‐noise ratio is redefined to describe more accurately the sensing of a signal in noise, since the conventional definition for f‐m proves inadequate and misleading, except for strong carriers. Extensive comparisons are made with a‐m reception employing a half‐wave linear rectifier, and it is found that for moderate and weak (<3 db) signals a‐m requires less input signal power against the same noise background than does f‐m to achieve the same desired output. This is particularly evident for weak signals. Only when the signal is strong (>10 db) is f‐m superior to a‐m; here broad band f‐m with very heavy limiting is needed. Narrow band f‐m at its best (no limiting) is comparable with or inferior to a‐m at high signal levels. Limiting proves to be detrimental in narrow‐band operation for all cases involving fluctuation noise, whereas heavy limiting is essential for successful broad‐band performance. Furthermore, the shape of the i‐f response becomes important in the latter, while it is not in the former. A large number of curves illustrating the average and mean‐square signal and noise outputs, etc., and signal‐to‐noise ratios for various conditions of operation are included.