Predictions of bit error rate caused by a combination of impulsive and Gaussian noise are made for data communication systems using PSK, DPSK and FSK, with direct carrier modulation. The technique uses the amplitude probability distribution (APD) to characterise the noise at the input of the receiver demodulator. A direct numerical approach (with numerical integration, where appropriate) is used to produce predictions for the case where decisions are made by sampling the demodulated signal at times nominally at the centre of the received data symbols, there being a minimum amount of postdetector filtering. Error rates in mobile data communication systems have been measured at roadside locations and in the laboratory. In the latter case, an impulsive noise simulator capable of producing noise with specified statistical characteristics was used. Laboratory measurements have also been made with fading signals, and the prediction techniques have been extended to cover this case. Generally, agreement between theory and experiment is good, and it is concluded that the techniques ar applicable in any system where channel separations are well defined and, consequently, where IF filters are adequately standardised.