While the Ganong lexicality effect has been observed for phonemic and tonal categorization, the effects of frequency and markedness are less clear, especially in terms of tonal categorization. In this study, we use Mandarin Chinese to investigate the effects of lexicality, tone frequency and markedness. We examined Mandarin speakers’ tonal categorization of tokens on all possible tonal continua with one end being a word and the other being a tonotactic gap (i.e., an unattested syllable-tone combination). The results of a forced-choice identification experiment showed a general bias against the gap endpoints, with the noted exception of continua involving T4 (X51), the most frequent lexical tone. Specifically, when T4 served as the gap endpoint, no obvious bias against it was observed regardless of its lexical status. Moreover, on the T3–T4 continua, there was an apparent bias against T3 (X214), the tone with the most complex contour, again, regardless of lexicality, suggesting a strong markedness effect. Taken together, the results of this study show the individual effects of lexicality, tone frequency and markedness, as well as their interactions, which contribute to our understanding of tonal categorization in relation to lexical statistics (tone frequency) and phonology (markedness).

Mandarin is a tone language with four phonemic tones (i.e., high-level Tone 1 [55]), rising Tone 2 [35], falling-rising Tone 3 [214], and falling Tone 4 [51] and with the maximum (C)(G)V(G)/(C) syllable structure. However, not all syllables can be combined with each of the tones (i.e., accidental gaps). For example, the syllable [tsʰu] is allowed to be combined with T1 ([tsʰu]55 “coarse”), T4 ([tsʰu]51 “vinegar”), but not T2 and T3. A calculation of all the 391 allowable syllables showed that there are 131, 185, 155, 110 accidental gaps in each of the four tones. A one-way chi-square test revealed that the accidental gaps in T2 were over-represented (stats here). A further investigation into these gaps in T2 according to different syllable types (CV: 52, CVN: 43, CGV: 38, and CGVN: 29) showed that these gaps were marginally under-represented in CGVN (stats here). We attributed these findings partially to the marked status of contour tones and to the typological preference for complex tonal targets to complex rimes due to their inherent longer durations (cf. Zhang 2001).

Mandarin tones are shown to be produced with different lengths (i.e., from longer to shorter: T3 > T2 > T1 > T4) (cf. Wu & Kenstowicz, 2015). An AX rating experiment in which Taiwan Mandarin listeners were asked to rate the relative durations of syllables ([pa], [pi], [ta], [ti]) manipulated into five different duration steps (290 ms, 320 ms, 350 ms, 380 ms, and 410 ms) in Mandarin tones (high-level T1, rising T2, dipping T3, reduced low-level T3, and falling T4) compared with an anchor stimulus ([pa] with 350 ms in mid-level tone) showed that the complex contour tone (T3) was rated as longer than simple contour tones (T2 and T4) and simple contour tones were rated as longer than level tones (T1 and reduced T3). Between the simple contour tones, T2 was rated as longer than T4. Between the level tones, the reduced T3 was rated as longer than T1. The explanations to these tonal perceptual differences are tied to the typological correlation between rime duration and the complexity of tonal targets (T3 > T2/T4 > T1/T3) (e.g., Zhang 2001) as well as to the listeners’ experience to the durations of different tones (T2 > T4 and reduced T3 > T1).