Stimulated by recent observations described in a companion paper, we have revisited existing theories of the Farley‐Buneman instability throughout the altitude range 90 to 125 km. We have assumed that the irregularities detected by radars at a given altitude are dominated by structures moving at the threshold speed in a direction associated with maximum linear growth rate conditions. We included recent nonisothermal electron and ion theories, which can modify threshold speeds by considerable amounts. We included altitude‐dependent models of ion and electron temperature and of the ion motion in the phase velocity calculations. Our treatment of the instability explains why some spectra are slow (Doppler shifts typically 200 m/s) and narrow, while others are fast (1500 m/s or close to the E × B) and narrow. These narrow spectra have all the characteristics of what has been labeled as “Type III” and “Type IV” in the past. Our calculations also offer an explanation for the observation of a strong asymmetry in the number of events with positive Doppler shifts near the nominal ion‐acoustic speed and those with negative Doppler shifts of the same magnitude.