The paper seeks to justify the contention that the efficiency of error-detection-coding techniques is frequently overestimated owing to insufficient appreciation of the interaction between the characteristics of the code and those of the signalling channel itself. The redundancy of an error-detection code (e.d.c.) requires an increase in modulation rate, and this causes an increase in element error rate due to reduction in normalised signal/noise ratio. On h.f. radio circuits, ionospheric effects can cause even greater increases, and current interleaving and convolution techniques only recoup a small proportion of the disadvantage. It is shown that, in many cases, these factors can outweigh the correcting capabilities of current codes. Comparison of various binary codes and the Piccolo m.f.s.k. system suggests that, where circumstances are such that extremely low error rates are not essential, a nonredundant unprotected code may be comparable in availability with an e.d.c., and Piccolo would probably be an improvement.