The article deals with the patterns of segmental adaptation of Polish voiceless affricates in initial and fi nal CC (consonant + consonant) clusters by native speakers of English. The data have been collected in an online loanword adaptation experiment in which 30 native speakers of Southern British English reproduced Polish words containing such sequences. The major problem posed by the data is the divergent adaptation of the post-alveolar /͡tʂ/ vs. the pre-palatal /͡tɕ/, with the former substituted mainly with the coronal plosive [t] and the latter realised as the palato-alveolar affricate [͡tʃ]. It is argued that these patterns of nativisation are due to the highlyranked IDENT-IO[dist] constraint, which militates against the modifi cation in the value of the feature [distributed]. Furthermore, it is demonstrated that the experimental results provide evidence in favour of the fundamental assumptions underlying the phonological approach to loan assimilation, namely the phonological input view as well as the faithful perception view.

Classic Optimality Theory treats phonological computation as a one-step parallel evaluation of output candidates. Such an approach does not handle well phonological opacity, in which the context of phonological processes is obscured on the surface. This article shows two examples of phonological opacity: Tiberian Hebrew Spirantization and Polish o-Raising. While these mappings are impossible to account for in classic OT, they are readily analyzed using the framework of Derivational OT.