naturalness (n.)

A notion introduced into (especially GENERATIVE) LINGUISTIC theory to refer to the PHONETIC plausibility of an analysis, which is seen as an important criterion in evaluating analyses alongside such other criteria as SIMPLICITY. An analysis, it is argued, must make phonetic sense, if it is to have any explanatory role in relation to the speaker’s behavior, e.g. such factors as relative ease of ARTICULATION must be taken into account. One of the first steps in defining naturalness more formally is to recognize the notion of natural class. A set of SEGMENTS is said to constitute a natural class if fewer phonetic FEATURES are needed to specify the set as a whole than to specify any one member of the set. The set of VOICED PLOSIVE segments in English is a natural class, on this basis: /b/, /d/ and /g/ all share the features of VOICING, INSTANTANEOUS RELEASE and INTERRUPTED; but, to specify any one of these, further features would be required (e.g. /d/ would be CORONAL, in addition).

The term in this sense applies to any set of speech segments which can be shown to have a highest common factor in this way; but as it stands the criterion needs to be supplemented by others, as it is too general (e.g. it would allow for all sounds in a language to be considered a natural class, on the grounds that they are all PULMONIC EGRESSIVE). Several other relevant criteria have been suggested, e.g. that the set of sounds all turn up in the same PHONOLOGICAL RULES, undergoing similar processes together. Also, there are several difficulties in working with the notion in terms of features, e.g. the more natural solution is not always the simpler. The notion of naturalness has thus been developed to take into account the relative naturalness of (a) segments (mainly through the use of the MARKING convention), (b) sound SYSTEMS (by computing the relative complexity of its units, this being defined in terms of marking values) and (c) phonological rules (based on the tendency for some phonological processes to be more frequent and phonetically more expected than others, e.g. /i/ becoming /u/ rather than , or certain types of ASSIMILATION or SYLLABLE structures being preferred). These developments are continuing.