Jakobson, Fant & Halle (1952) proposed that listeners distinguish phonemes from every other phoneme in the system by making a series of ordered binary choices that correspond to the oppositions active in the language. They go so far as to state that the "dichotomous scale," which I will call a contrastive hierarchy (Dresher 2002, 2003), "is the pivotal principle of the linguistic structure." Jakobson & Halle (1956) suggest further that distinctive features are necessarily binary they are acquired through a series of "binary fissions". They propose that the order of these contrastive splits is partially fixed, thereby allowing for certain developmental sequences and ruling out others.

The notion of a contrastive hierarchy was subsequently applied in acquisition studies, where it is a natural way of describing developing phonological inventories (Pye, Ingram & List 1987, Ingram 1989, Dinnsen et al. 1990, Dinnsen 1992). However, following the arguments of Stanley (1967) it fell out of use in mainstream generative phonology. I will argue that the contrastive hierarchy is the only viable way to establish featural contrasts in segmental systems; in addition, it provides an organizing principle required by theories of markedness and underspecification. I will show further that the relevant contrasts must be imposed by the analyst and do not simply "emerge" from surface phonetics or from a grammar (cf. Itô, Mester & Padgett 1995, Kirchner 1997). I will also pursue some implications of the contrastive hierarchy for the acquisition of phonological systems.

First, acquiring phonological contrasts is not the same as acquiring phonetic contrasts. This is demonstrated most clearly by languages in which a phonological contrast is neutralized in all or most contexts: /u/ ~ /U/ in Classical Manchu (Zhang 1996), /i/ ~ /E/ in Nez Perce (Hall & Hall 1980), and /i: u:/ in Yowlumne (Kuroda 1967, Archangeli 1984) are three of many examples in this class. The dispersion phenomena discussed by Flemming (1995) operate at the phonetic level to enhance phonological contrasts (Stevens, Keyser & Kawasaki 1986, Rose 1993, Rice 2002).

It follows from the above that contrast acquisition requires learners to take into account phonological processes, and not just the local phonetics of individual segments (Dresher & van der Hulst 1995).

Third, a contrast-driven theory of phonological representations can account for several types of variation in child language. Within a language, the relatively underspecified nature of the developing system predicts greater variability in child language, as shown by Rice & Avery (1995) and Rice (1996a, b). Across languages, differences in the contrastive hierarchy itself are reflected in different paths of segmental elaboration (Dresher 1998).

Finally, child language displays positional effects in assessing contrast (Waterson 1987, Fikkert 1994). These effects are a further manifestation of the relative nature of phonological contrast, and the importance of establishing the proper scope of a contrast.