This paper provides evidence for the validity of the graded salience hypothesis, and argues against the priority of literal meaning hypothesis and the selective access view of context. The findings highlight that contexts which are biased towards less salient meanings do not prevent activation of salient meanings.