The Quantitative Syntagmatic Analysis of the Russian and Polish phonological Systems

The description of a language's phonological system is not complete without statistically reliable data pertaining to the varied syntagmatic relations occurring within that system. A measure of ftmctional load is needed to elucidate the efficiency, viewed from a cybernetic angle, of the particular set of phonemes. This approach opens up the way for various information-theoretic concepts such as entropy, relative entropy, information content and redundancy. These values are derived , with respect to the Russian and Polish phoneme inventories, ranging from zero to the third order. It is useful to consider the efficiency of various methods of information-theoretic coding procedures for phonemic strings. Huffman coding can achieve not only a high degree of efficiency but also a relatively trouble-free unique decoding routine. In the syntagmatic analysis of phonological systems, junctures of varying types must be taken into account and must be " retrievable " for statistical purposes. The presence of count details relating to juncture allow an acceptable " transfer modulus " to be derived which can be a useful parameter in the comparison of a language's phonemic and graphemic systems. The question of sample size is important in any statistical survey, not least in linguistics. Pilot tests have to be run in order to yield a general impression of a given phonemic distribution. Nearly all pho-nemic inventories show a Lexian or hyperbinomial distribution. Very often these distributions can be plotted logarithmically to give a reasonable fit to a theoretical exponential curve, derived from considerations similar to those postulated by Zipf in lexical statistics. Most phoneme systems contain rare phonemes-in Russian /~,/ is a case of such a phoneme-which present some difficulty, when probability intervals are to be estimated from relative frequencies. Using a 95 % confidence level and a 10 % error margin the sample size needed to satisfy these