A non-linear technique is predominantly used for the recording of transiently-evoked otoacoustic emissions (TEOAEs). The aim of this study was to compare linear and non-linear TEOAE recordings. TEOAEs were recorded in 22 normal hearing subjects to clicks from 90 to 30 dB SPL in 10 dB steps with the ILO88 system using both linear and non-linear recording techniques. The non-linear recording technique reduces stimulus artifacts for early latencies, but total elimination could not be proved. Both artifact reduction and significant differences between the two kinds of TEOAE recordings were reduced for longer latencies and lower stimulus intensities. For longer latencies (>10 ms) there was no significant difference between "linear" and "non-linear" TEOAEs. A higher signal-to-noise ratio was found for "linear" TEOAEs, resulting in better identification and a higher test-retest correlation. The linear recording technique, which includes new methods of artifact cancellation in comparison to the mainly utilized non-linear recording technique, should be used especially in hearing screening.