Investigating the "Hidden" Structure of Phonological Systems *

Observed phonological systems are only a very small subset of what is theoretically conceivable considering the possibility of human vocal tract. In other words, given the number of possible phonetic features (corresponding to the vocal tract generative capacity), the number of attested segments is incredibly lower than what is combinatorily possible. The same observation is valid when comparing the number of attested systems with the set of theoretical ones predicted from the attested number of segments. Obviously, no doubt is possible about the fact that phonological systems are not just unorganized sets of segments picked up at random and, consequently, phonological theories and typological studies have focused on showing that these systems are structured according to various constraints from the perceptual, articulatory or cognitive levels (Troubetzkoy, 1929, Sedlak, 1969, Crothers, 1978, Maddieson, 1984). Yet the task of identifying these constraints and the way they interact is still at stakes today, even if numerous works have investigated part of them especially for vocalic systems (Liljencrants and Lindblom, 1972, Lindblom 1986, Stevens, 1989, Vallee, 1994). However, the constraints effecting on consonantal systems have been only partially investigated and our comprehension is still limited (Lindblom and Maddieson, 1988, Vallee & al. 2002). Lindblom and Maddieson, sketch out – in one of the first attempts to study consonantal systems – the basis of an “all-inclusive universal phonetic space” (hereunder UPS) which can be considered as the first study of phonological systems as a whole. Besides, UPSID (UCLA Phonological Segment Inventory Database, Maddieson, 1984, Maddieson and Precoda, 1990) provides a powerful material to