In Experiment 1 four groups of rats received 30 light-shock pairings using footshock intensities of either .2, .4, .8, or 1.6 mA. One day later all rats were tested for startle by presenting tones in the presence or absence of the light CS. Potentiated startle (the difference between startle on light-tone vs tone-alone trials) was nonmonotonically related to the shock intensity used in training, with the greatest potentiation at intermediate shock levels. Experiment 3 demonstrated a similar relationship when backshocks instead of footshocks were used. In Experiment 2 rats were trained with either a moderate or high shock and then given an extended extinction-test session 1 day later. The moderate-shock group showed a gradual decline in potentiated startle over extinction. The high-shock group showed a nonmonotonic extinction curve where potentiation progressively increased toward the middle of extinction and dissipated thereafter. The results suggest that acoustic startle bears an inverted U-shaped relationship to fear and are discussed in relation to other studies concerned with this issue.