Our attention falters, which is commonly known as ‘attentional blink’ (AB), after spotting a target; as a result, the subsequent target often goes unnoticed. Attentional demand also affects the size of the pupillary aperture. In previous studies, differential pupillary dilation averaged across trials within a short critical epoch distinguished AB. In contrast, we examined the pupil dynamics on a trial-by-trial basis. Healthy young volunteers performed a task, wherein two visual targets embedded in a series of distractors were presented in a rapid succession. The participants were asked to identify the first target (T1) and report the second target (T2) that appeared occasionally in randomly interleaved trials. The average pupillary aperture across the population distinguished between blink (i.e., T2 not detected) and no-blink (i.e., T2 detected) trials. A pair of Gaussian distribution functions characterized the pupil dynamics in each trial, where the parameters of each fit reflected the peak, mean and standard deviation of the pupillary response, following T1 and T2 onset. We performed a principal component analysis (PCA) on shape-features of the fits. A multivariate analysis of variance suggested that the major clusters in the PC space for blink and no-blink trials were non-overlapping, and their centroids were significantly distant from each other, suggesting that the pupil dynamics can reliably forecast AB. Provided that the activity of locus coeruleus nor-epinephrine (LC-NE) neurons correlates pupil dynamics, our findings comply with the LC-NE account of AB that holds that the onset of T2 during the refractory-like period in LC-NE activity causes AB.