Prediction of articulatory movement from phonetic input

Data obtained by a computer‐controlled x‐ray microbeam system are being analyzed [Kiritani et al., J. Acoust. Soc. Am. 57, 1516–1520 (1975)] in the form of predicting the movements of three pellets located along the tongue surface, and three other pellets located on the velum, the jaw, and the lower lip. The analyzed data include a set of isolated words and sentences spoken by two native speakers of American English. This communication describes the rules necessary for reconstructing with accuracy the position of the six pellets during the sequence /Cae(N)C/, where C represents an alveolar consonant. The high intraspeaker consistency makes it feasible to predict, for each speaker, the behavior of each of the pellets. The large interspeaker differences result in a set of speaker dependent rules. For example, the jaw pellet position and the tongue pellet position (located at about 1 cm behind the tongue tip) for an alveolar consonant depends mainly on the feature “tense‐lax,” for one speaker, and on the syllabic position of the alveolar consonant for the second speaker. The rules governing the articulators not directly involved in the production of the alveolar consonants generally show large interspeaker differences.