My dissertation investigates interference effects in the comprehension and production of long-distant dependencies in adults, integrating insights from theories on formal syntax and cognitive psychology granting a key role to similarity-based interference effects as a source of processing difficulty and/or ungrammaticality. My research mainly focuses on wh-islands' acceptability and attraction effects in agreement, which can both be taken as windows on interference effects in sentence processing. My research suggests that: (i) only similarity on a narrow class of syntactic features has the potential to generate ungrammaticality, while similarity in terms of other features increases processing difficulty without changing the grammatical status of the sentence; (ii) attraction effects can be conceived as the result of similarity-based interference; (iii) encoding interference plays a key role in generating processing difficulties; (iv) self-organised sentence processing models offer a new promising way to implement interference effects in a rigorous mathematical framework.