|Title:||Testing for OO-Faithfulness in Artificial Phonological Acquisition|
|Comment:||To appear in the proceedings of BUCLD30. Note: this version is both brief and somewhat theory-sparse.|
|Abstract:||Phonological patterns often have systematic exceptions in morphologically-derived contexts. In OT, such patterns are often captured with constraints enforcing phonological similarity throughout a morphological paradigm -- e.g. Output-Output Faithfulness (Benua, 2000). In the OT learning literature, it has been argued that phonological learners must have an inherent bias for high-ranking OO-Faith to prevent the acquisition of superset grammars (McCarthy, 1998; Hayes, 2004).
The current study tested for such a bias, asking whether learners are preferentially OO-faithful at early stages of morphological acquisition. Twelve 4-year-old children learned the names of objects in an artificial language, including a novel plural suffix, and then played a 'wug-test' game (Berko, 1958). The wug-test compared participants' production of the same coda-onset clusters in two morphological contexts, where only one was protected by OO-faith. A pair-wise within-subjects t-test (p < 0.01) showed that codas were produced faithfully in fewer clusters where OO-Faith was not relevant (56/112; 50%) than in clusters where the coda was protected by OO-Faith (70/87; 80.4%). These initial results support the claim that children in the process of acquiring new marked structures prefer repairs that satisfy OO-faith, at the expense of other markedness and faithfulness pressures.