We collected weakly electric gymnotoid fish in the vicinity of Manaus, Amazonas, in the Solimoes river (white water). We tried to find out whether Electric Organ Discharges (EODs) are species specific which is essential for their presumed role in recognition of conspecifics and reproductive isolation. We considered at least 43 valid sympatric species, some of them unnamed. All of these displayed stable EOD waveform patterns, most of them clearly distinct from the other species' EODs. Eleven species are of the pulse EOD type, 32 of the wave EOD type (one of the latter is intermediate). The EODs of pulse species were analysed (1) by EOD repetition rate at rest (variation from £1 Hz to 60 Hz), (2) by Fourier amplitude spectrum analysis of single EODs (Fig. 1; in these spectra, frequencies of peak amplitude ranged up to 2300 Hz). There was a significant, positive correlation between both parameters (Fig. 2). Identification of pairs of species with similar EODs by these parameters does not appear to be possible because of inter-individual EOD variations. In wave species there is conclusive evidence that EOD fundamental frequencies (= repetition rate of a complete EOD period) do not allow species identification: twentyeight wave species displayed EOD fundamental frequencies flrom 300 to 1800 Hz (Fig, 3). This yields a hypothetical frequency band of p.09 octave to signal species identity; the actual value of EOD frequency variations in Eigenmannia is much greater (1.2 octaves). Seven species of the family Apteronotidae displayed a new signal type: the main energy of the signal was contained in higher harmonics, and not at the fundamental frequencies (Figs. 6 and 7). For all wave species there was a significant, positive correlation between their dominant frequency (= the strongest signal component) and harmonic content of their EOD although individual species deviated considerably from what was predicted by the regression line (Fig. 8 ) . Thus separation of species was greatly improved compared to the criterion of fundamental frequency (Fig. 3) but still appeared insufficient in a number of cases.
Therefore, in both wave and pulse species still other parameters must be involved in recognition of conspecifics.