Prudent error-driven learning with OT-CC ∗
Michael Becker · 2007
Highlights: • Theoretical problem: OT-CC offers a theory of GEN that allows us to derive outputs, but does so at the cost of a candidate set that grows factorial ly with the number of required repairs. • Under-reported empirical observation: Children prudently select the words that they attempt to produce, avoiding words that would surface too unfaithfully. The kinds of words that children attempt develop gradually, just like th e kinds of words that children produce, only a little earlier. • I propose that the factorial explosion is the cause for child ren’s selection of targets. Choosing targets that involve less repairs is a way to limit t he explosion of candidate chains. I offer a version of error-driven learning that inco rporates OT-CC’s GEN and the need to mitigate the factorial explosion.