Conductive hearing impairment differs from sensorineural impairment. It results in a simple attenuation of sound. Hearing aid selection procedures designed for sensorineural impairment are not appropriate. Research has shown that subjects with conductive loss do not use sufficient gain to return the sensation level of signals to normal. This may be due, at least in part, to a "head room" problem. That is, the maximum output of the aid is too low relative to the gain, causing frequent saturation. A study of the speech spectra preferred by normal hearing subjects revealed that the normal spectrum is preferred when listening conditions are easy. When listening in difficult conditions preference was shown for a high frequency boost. This is understandable as the high frequency components of the speech would have been the first to fall below threshold. A study of people with conductive loss confirmed that a frequency response selected according to the normalisation principle is preferred in easy listening conditions. The advantage was however described as slight. In cases of conductive loss, gain should be provided at each frequency to equal the loss at that frequency. However, a volume control should be available so that the overall gain can be reduced. Also, the ability to alter the frequency response to provide a high frequency emphasis is desirable. Maximum output should be increased above the 140 dB SPL commonly seen in head worn aids. For mixed losses the required gain and frequency responses of the two components should be determined separately and added.