Markedness and the Phonological Typology of Two-Height Tone Systems

Despite longstanding interest in defining what a tone system is and in contrasting tone, pitchand stress-accent systems, there have been surprisingly few attempts to sort out the relevant phonological properties which distinguish “true” tone systems from each other. In this paper, I explore a propertydriven typology of two-height tone systems, based on markedness. Drawing from a current database of ca. 600 tone systems (of which over 400 are two-height), I first confirm that two-height systems may be “equipollent” /H, L/, “privative” /H/ vs. Ø or /L/ vs. Ø. or both, /H, L/ vs. Ø. I then demonstrate a difference invoking tonal markedness: While the one tone of /H / or /L/ systems is unambiguously “marked”, contrasting with the absence of tone (Ø), the claimed univerally marked H tone of /H, L/ systems does not show a consistent tendency to be preserved (“faithful”) in outputs as proposed by Pulleyblank (2004) and de Lacy (2006). Two closely related Tibeto-Burman languages illustrate this point: In Kuki-Thaadow, the tone rules “conspire” to guarantee that every underlying /H/ will be realized on the surface. In closely related Hakha Lai, however, it is just the opposite: several rules apply in such a way as to guarantee that every input /L/ will be realized on the surface. One possible interpretation is that either /H/ or /L/ can be marked in /H, L/ systems, thus complementing recent proposals of language-specific markedness in segmental phonology (Hume 2003, Rice 2007). Given the polysemous and potentially contradictory notion of markedness (Haspelmath 2006), I suggest abandoning “markedness as faithfulness” in favor of the concept of “phonological activation” (Clements 2001, 2003; Hyman 2003): Which tone is activated (H? L? both?), where in the phonology (underlying? lexical output? surface?), and how?