|Abstract:||This thesis focuses on the most recent OT-based theory of opacity called Optimality Theory with Candidate Chains (OT-CC, see McCarthy 2007). To date very little attention has been dedicated to the problem of acquisition of OT-CC grammars and to the treatment of spontaneous opacity effects in the light of OT-CC. In this thesis we demonstrate that OT-CC grammars can be effectively learned by the BCD algorithm (Prince & Tesar 2004). Also, on the basis of evidence from obligatorily counterbleeding processes, NDEBs and non-target-like opacity effects, we propose to make certain changes to the status of Precedence constraints with the view to increase the descriptive adequacy of OT-CC. We show that our proposed adjustments allow OT-CC to account for emergence and subsequent loss of spontaneous opacity effects, as well as for the phenomenon of U-shaped learning and cross-subject variation in early production data.