The gain of hearing aids is conventionally measured in a hard-walled 2-cc coupler. It has become increasingly apparent, however, that many interacting variables can affect the aid's behavior in an individual ear quite significantly. Therefore it is desirable that any departure from coupler response be known. A number of methods have been used to measure real-ear gain, some of which require expensive or nonstandard instrumentation. An alternative method is described, based on measurements of aided and unaided acoustic reflect thresholds that appears to have certain advantages over other methods. Sound-field intra-aural reflex thresholds for six one-third-octave noise bands were determined for 20 subjects using a postauricular aid under three conditions--unaided, aided/ear occluded, and aided/ear nonoccluded. Real-ear gain was defined as the difference in decibels between aided and unaided thresholds. Results showed wide individual differences in real-ear response. Average occluded real-ear gain was about 5 dB greater at 1.6 k Hz, and about 5 dB less at 3.15 k Hz, than the gain in the artificial ear. When the aid was coupled to the open ear, average real-ear response was down at all frequencies relative to the 2-cc coupler, although relative to the closed-ear response frequencies above 1.6 k Hz were emphasized. It is argued that real-ear response can be expected to vary considerably depending on the location and orientation of the microphone of the aid on the head or body. It is suggested therefore that real-ear measurements be part of every aid selection procedure, especially for persons with a reduced dynamic range. Response modifications might then be possible to enable the critical speech frequencies to be heard at a comfortable level.