Pupillometry is one method to measure online, real-time processing of prosodic contours. Pupil dilatation is related to exertion of cognitive effort, including in speech comprehension. Previous work has shown that mismatches between prosodic contours and pragmatic context elicit increased pupil dilation, suggesting this technique is a useful method for exploring intonational meaning. Here, we use pupillometry to explore the intonational meaning of downstepped contours in Mainstream US English. Past work indicates that downstepped pitch accents mark accessible items, whereas high star pitch accents mark new items. Pairs of stimuli were created to be (1) accessible or (2) new in a two-sentence context. Participants listened to these contexts followed by a target sentence produced with either a high star (H* H* L-L%) or downstepped (H* !H* L-L%) contour while measuring pupil dilation. We predicted increased pupil dilation when the contour mismatches the pragmatic context (high star for accessible; downstep for new) relative to felicitous pairings. Results show increased pupil dilation for new information when uttered with a H* H* contour, potentially reflecting cognitive effort in adding items to the common ground, as well as for the two mismatched pairings (though to a lesser degree relative to the new-high star condition), potentially reflecting the effort related to processing infelicitous prosody. Finally, H* !H* in accessible contexts showed the least amount of dilation, potentially reflecting lower cognitive effort for processing accessible items when paired with felicitous prosody.