Coding Strategies in Language

The problem of selecting a code to transmit four messages over the binary symmetric channel is studied in relation to two types of channel noise (“substitution≓ error and “deletion≓ error) and to two types of decoding strategy (maximum hit and minimum error). It is shown that mean Hamming distance is a good general guide to coding efficiency, except in the case of a minimum error strategy with a deletion error channel, where coding efficiency is critically dependent on noise level. An experiment in which subjects selected codes in an artificial language suggests that the process of recall from memory is similar to the process of transmitting over a deletion error channel with a minimum error strategy. A similar interpretation can be placed on the analysis of consonant systems in English, French, German and Welsh, where the sets of consonants of a given class in a given environment are considered as codes whose alphabet is the phonological distinctive feature system of Halle (1958)