Listening to Limericks: a pupillometry investigation of perceivers' expectancy

Abstract

What features of a poem make it captivating, and which cognitive mechanisms are sensitive to these features? We addressed these questions experimentally by measuring pupillary responses of 40 participants who listened to a series of Limericks. The Limericks ended with either a semantic, syntactic, rhyme or metric violation. Compared to a control condition without violations, only the rhyme violation condition induced a reliable pupillary response. An anomaly-rating study on the same stimuli showed that all violations were reliably detectable relative to the control condition, but the anomaly induced by rhyme violations was perceived as most severe. Together, our data suggest that rhyme violations in Limericks may induce an emotional response beyond mere anomaly detection.

Figure 1. Pupil size per condition, from 0–2000 ms after critical word-onset.

Pupil size (Y-axis) is measured in log number of pixels per video frame, relative to a 200 ms pre-onset baseline. Time (X-axis) is sampled at 500 Hz.

Figure 2. Pupil size deviations from the control condition (no violation) as a function of time.

Time is plotted on the X-axis, from 0–2000 ms after critical word onset. The white curve in each plot refers to the mean difference between the relevant violation condition and the control condition (higher values mean more dilated pupils in the violation condition); the grey areas around the curves indicate 95% confidence bands for the difference. The green vertical line in each plot marks the average onset of the final word in the given violation condition, together with 95% confidence limits (green dotted lines). Red solid and dotted lines mark the average Limerick offset (end of auditory presentation) and corresponding 95% confidence limits. Left hand panels: by participants; right hand panels: by items. Top to bottom: semantic, syntactic, rhyme, and metric violation.

Figure 3. Pupil size data for extreme subsets of items.

Pupil size (Y-axis) is measured in log number of pixels per video frame, relative to 200 ms pre-onset baseline. Time (X-axis) is sampled at 500 Hz from 0–2000 ms after critical word-onset. “Strong rhyme” (blue curves, top) shows data for the six items with the highest anomaly ratings in the rhyme violation condition; “weak rhyme” (red curves, middle) shows data for the six items with the lowest anomaly ratings in the rhyme violation condition; “strong other” (green curves, bottom) shows data for the six items with the highest anomaly ratings in any of the non-rhyme violation conditions. For reference, the grand average curve for the no violation control condition is also included in each plot (black dotted lines).