I am developing a theory of how various elements of sound-based similarity (acoustic, allophonic, and phonemic) interact in determining overall judgments of cross-linguistic similarity between sounds in the first language (L1) and second language (L2).

My current thinking is that similarity at higher levels tends to outweigh similarity at lower levels. This kind of hierarchy between more abstract (e.g., phonemic) similarity and less abstract (e.g., psychoacoustic) similarity helps account for why the L1 and L2 sounds that are cognitively linked to one another by L2 learners are so often not the ones that are the closest acoustically and/or allophonically (Chang, 2015).