Not much is known about how people comprehend ironic utterances, and to date, most studies have simply compared processing of ironic versus non-ironic statements. A key aspect of the graded salience hypothesis, distinguishing it from other accounts (such as the standard pragmatic view and direct access view), is that it predicts differences between processing of familiar and unfamiliar ironies. Specifically, if an ironic utterance is familiar, then the ironic interpretation should be available without the need for extra inferential processes, whereas for unfamiliar ironies, the literal interpretation would be computed first, and a mismatch with context would lead to a re-interpretation of the statement as being ironic. We recorded participants' eye movements while they were reading (Experiment 1), and electrical brain activity while they were listening to (Experiment 2), familiar and unfamiliar ironies compared to non-ironic controls. Results show disruption to eye movements and an N400-like effect for unfamiliar ironies only, supporting the predictions of the graded salience hypothesis. In addition, in Experiment 2, a late positivity was found for both familiar and unfamiliar ironic materials, compared to non-ironic controls. We interpret this positivity as reflecting ongoing conflict between the literal and ironic interpretations of the utterance.