Many modern communication systems transmit amplitude samples of the original input signal, the sampling being carried out according to the wellknown requirements of the Sampling Theorem. At the receiver the signal is reconstituted by low-pass filtering. This paper investigates the degree of privacy which can be introduced in a speech communication system by deliberate variation of the order of transmission of the samples. The distortion of the signal which results from the low-pass filtering of such 'scrambled' samples is investigated theoretically and practically, and intelligibility tests are used to assess the effectiveness of the system. Finally, some guidance is given as to the selection of the scramble sequence for maximum effectiveness.