Frequency-specific electric response audiometry can be performed on difficult to test young children if the child is sedated and proper choices are made of acoustic stimuli and recording parameters, although certain compromises are necessary. A very satisfactory sedative is secobarbital, administered intramuscularly in doses related to the weight of the child. As stimuli we recommend '2-1-2' tone bursts at 500, 1 000, 2 000, and 4 000 Hz: i.e., with a rise and fall of two periods and a plateau of one period of the modulated tone. A very robust and sensitive response that is not significantly modified by the sedation and is effective for all four frequencies is the P6-SN10 of the early brainstem sequence. To record this complex favorably requires a bandpass input filter of the Butterworth type with pass-band (at -3 dB) from 50 to 1 700 Hz and rejection rates of 24 dB/octave. With this combination, polarity of stimulus is unimportant and sweep time, rate of stimulation and number of responses averaged may be selected for convenience and simplicity. A routine that requires about an hour of testing time is described and the necessary correction factors are given for estimating a child's behavioral pure-tone thresholds. We believe that our threshold estimates are generally correct within 10 dB, and are sufficiently frequency-specific for proper selection of a hearing aid.