Strong onsets and Spanish fortition

Spanish has in its phonetic inventory three sounds, transcribed here as [β], [δ], and [γ], often described as voiced fricatives in the literature. Based on my own observations and on the above description by Ladefoged, I take these sounds to actually be voiced approximants, with little or no frication at all. These approximants alternate with the voiced stops [b], [d], and [g], respectively, in what is standardly taken to be spirantization of underlying stops (Harris 1969). In this paper I argue that this alternation between stops and approximants is due to contextual hardening (fortition) of underlying approximants, following arguments originally made by Lozano (1979). The driving force behind fortition is a constraint, STRONG ONSET, interacting with other constraints in the grammar of Spanish (in the Optimality-theoretic sense of constraint interaction; see Prince & Smolensky 1993). I then argue that the strikingly similar distribution of the Spanish tap [r] and trill [rr] is a reflex of the basic fortition account laid out here, a relationship previous spirantization accounts must deny. The disparities between the two phenomena are shown to be the product of there being two phonological levels in Spanish: a word-level and a breath group-level.