The purpose of this study was to estimate the hearing levels, at the mid-frequencies, of 233 ears with sensorineural hearing loss by classifying the corresponding transiently evoked otoacoustic emissions (TEOAE) recordings into three threshold groups. A classification algorithm was based upon a discriminant analysis of fast Fourier transform (FFT) data, evoked by non-linear click stimuli of 80 (+/-2) dB SPL per ear. To validate the efficiency of the proposed methodology TEOAE recordings were initially grouped by mean hearing level values of the 1 kHz and 2 kHz octaves into three threshold ranges according to two strategy schemes: in the first, TEOAE data were assigned into 10-19 dB HL, 20-39 dB HL or > or = 40 dB HL groups. In the second, TEOAE data were assigned into 10-29 > dB H >, 30-39 dB HL or > or = 40 dB HL groups. The most accurate prediction estimates were obtained from the second strategy scheme with a 90.9% accuracy in the 10-29 dB HL group, 82% in the 30-39 dB HL group and 71.4% in the > or = 40 dB HL group.