The preaspirated stop: a perceptually suboptimal phonological structure?

Ian D. Clayton, Boise State University

Abstract

The problem. The contribution of perceptual biases to phonological typology has been vigorously discussed in recent literature, e.g. Moreton (2008), Wilson (2006), Silverman (2003, 1997), and Bladon (1986). While the former two scholars have argued that much of typology must be attributed to cognitive biases, the latter two have argued instead that phonetic or aerodynamic features of certain phonological structures render them perceptually “suboptimal” in comparison with more abundant structures, thus rarer. For instance, preaspirated stops are scarcer than postaspirated stops because aspiration is a less viable cue before the stop closure than after the stop release. Preaspirated stops are certainly rare and postaspirated stops commonplace (Silverman 2003, Ladefoged and Maddieson 1996, Maddieson 1984), but the strength of the “suboptimality” hypothesis of Silverman and Bladon as an explanation for this rarity has not been empirically tested. A perception experiment was duly designed to perform this test.

Hypothesis: Preaspirated [hp ht hk] will be harder than postaspirated [ph th kh] to distinguish from unaspirated [p t k].

Experiment. Gaelic-like non-word stimuli were recorded with the assistance of 3 Gaelic speakers from the Isle of Lewis: in Lewis Gaelic, voiceless unaspirated [p t k] contrast with voiceless postaspirated [ph th kh] in initial position, and with voiceless preaspirated [hp ht hk] medially and finally. Stimuli featured the target stops [p t k], [ph th kh], or [hp ht hk] in initial, medial, or final position. Stimuli were presented in pairs of three types: the first featured one post- or preaspirated token and one unaspirated token in the target position; the second type featured two identical preaspirated, postaspirated, or unaspirated tokens; the third consisted of distracters. Participants were asked to respond “same” or “different” to each stimulus pair.

To control for L1 effects on perception, three participant populations were recruited: 9 native speakers of Scottish Gaelic; 12 native speakers of Polish, which lacks an aspiration contrast (Gussman 2007, Ruszkiewicz 1990); and 11 native speakers of English, to whom postaspiration but not preaspiration is familiar. Participants were equally distributed by gender and age.

Results. The hypothesis was not confirmed: there was no indication that preaspirated stops were harder for subjects to distinguish from unaspirated stops than postaspirated stops were. Rather, significant differences in confusion rates were a function of position, not aspiration type: for all three participant groups, confusion rates were highest for tokens in final position (preaspirated vs. nonaspirated). For two of the three participant groups (Gaels and Poles), the lowest confusion rates actually occurred in medial position (again, preaspirated vs. nonaspirated).

Conclusions. The hypothesis offered by Silverman (2003, 1997) and Bladon (1986) that preaspirated stops are rare due to suboptimal perceptibility is untenable. Instead, other explanations for the rarity of preaspirated stops must be sought. For instance, it may be that there are relatively few phonetic precursors from which preaspiration develops (Myers 2002). Alternatively, it may be that there are cognitive biases favoring postaspiration at the expense of preaspiration, as for other typological phenomena (Moreton 2008, Wilson 2006).