Possible words

• The set of possible words in natural languages, analogous to the set of sentences, is infinite.
• The set of possible syllables, however, is not; they can be calculated on the basis of how many phonemes can appear in which position of the syllable.
• An observationally adequate description of phonotactic data, then, could be in terms of a list of all actual or possible syllables.
• This set, although finite, is so large (around 250 000 based on experiments using a syllable network for German) that a list is no longer suitable.
• Furthermore, a list would not make any generalisation over structures other than the syllable itself. Syllable substructures such as onset (initial consonant clusters), peak (vowels) and codas (final consonant clusters) would have no relevance.
• Psycholinguistic experiments show that a native speaker respects the unity of onsets and rhymes; a native speaker recognises a syllable substructure rhyme (cf. Treiman, 1986 for evidence on English), which consists of a peak and coda, and that there is a closer relationship between these than between, say, onset and peak.