The causes of noise in electromagnetic flowmeters are reviewed, and some examples given of the equivalent thermal noise resistance likely to be found in practical flowmeters. Signal-to-noise ratio is examined in relation to both the unwanted transformer e.m.f.s and the type of magnet excitation used. It is shown that a sinusoidally excited flowmeter can have an effective noise bandwidth no greater than the output bandwidth, while flowmeters using complex exciting waveforms are, because of practical limitations, likely to have noise bandwidths several times this amount if large reduction of transformer e.m.f.s is to be achieved. In particular, a combination of complex exciting wave and peak detection can lead to greatly excessive noise, especially when half-wave detection is used. The best conditions for low noise and high rejection of transformer e.m.f.s when using complex waves are mean full-wave detection and squarewave excitation. Amplifier flicker noise is probably more important with complex wave flowmeters than with sinewave ones, for technical reasons. A description is given of the experimental measurement of effective noise bandwidth with different detectors, and of the corrections which had to be applied. The optimum output bandwidth is discussed in relation to the blood flow pulse.