Phonological Knowledge beyond the Lexicon in Taiwanese Double Reduplication

In a recent trend, phonologists are becoming increasingly aware of the fact that many phonological processes are variable (Labov 1972, 1994), gradient (Hayes 2000, Frisch et al. 2004), and full of exceptions (Zimmer 1969, Zuraw 2000). This is not only because variability, gradience, and exceptions have been shown to be par for the course for many lexical patterns in phonology, but also because speakers have been shown to possess knowledge of variability, gradience, and patterns of exceptions (Albright and Hayes 2003, Frisch and Zawaydeh 2001, Hayes and Londe 2006, Pierrehumbert 2006, Zuraw 2000). For example, Frisch and Zawaydeh (2001) showed that Arabic speakers’ wordlikeness judgments of nonce words corresponded with the gradient consonant cooccurrence restrictions in the Arabic lexicon. Recent research has also shown that it does not suffice to study only the lexically manifested sound patterns observable from elicited data, as speakers’ phonological knowledge does not always match the lexical patterns. There are areas in which speakers seem to know more than the lexicon (Berent et al. 2007, Davidson et al. 2004, Zuraw 2007). For example, Berent et al. (2007) showed that English speakers have knowledge of onset sonority sequencing that their lexicon does not inform them of, as evidenced by syllable counting, identity judgment, and identity priming experiments. There are also areas where speakers seem to know less than the lexicon, particularly for phonologically opaque patterns (Sanders 2001). The opaque Taiwanese “tone circle” (see details in §2), for instance, is repeatedly shown to be largely unproductive in “wug” tests (Berko 1958) despite its exceptionlessness in the language itself (Hsieh 1970, 1975, 1976, Wang 1993, Zhang et al. 2006, 2007) The goal of this paper is two-fold. We first show that Taiwanese speakers’ knowledge of tone sandhi in double reduplication, as reflected in a wug test, is a combination of more than, less than, and exactly what their lexicon informs them of. Second, we argue that the “dual-listing/generation” theory of Zuraw’s (2000) can model the phonological knowledge of Taiwanese speakers.