The pupil has long been more than a hole that lets light in. Beyond its reflexive constriction to brightness, pupil diameter tracks the brain’s internal arousal state, driven largely by phasic bursts of norepinephrine released from the locus coeruleus, a small brainstem nucleus that broadcasts neuromodulatory signals across the cortex. Decades of research have linked dilated pupils to better performance on visual detection and discrimination tasks, and the dominant theory holds that heightened arousal broadens the focus of attention. But there has been a stubborn problem with this literature: in near-threshold detection tasks, improvements could reflect genuine gains in perceptual sensitivity, strategic shifts in decision criteria, or changes in how signal-like the brain’s baseline noise feels. Teasing these apart has proven difficult, and it has remained unclear whether arousal only helps when stimuli are barely visible or whether it acts on something more fundamental—the raw availability of sensory information over time.

The Münster team designed their study to sidestep these confounds entirely. Instead of asking whether observers could barely detect a stimulus, they used a partial-report iconic memory task in which the stimuli are unambiguously visible and the challenge is purely temporal. Thirty-seven participants, drawn from an initial pool of sixty-one after standard exclusions, viewed displays of six light-grey circles arranged around a fixation point, each containing a small wedge cut out at one of eight possible orientations. The array flashed for just 40 milliseconds. At varying delays—ranging from 140 milliseconds before the display to 1,240 milliseconds after—a black line cue appeared, pointing to one of the six locations, and participants reported the orientation of the cued item using a number pad, followed by a confidence rating. Each participant completed 1,296 trials while an EyeLink 1000+ infrared eye tracker sampled their pupil diameter 1,000 times per second.

The logic of the task is elegant. When the cue appears simultaneously with the display, observers can immediately select one item and stash it in visual short-term memory, yielding near-perfect performance. As the cue is delayed, performance collapses because participants must rely on the rapidly fading iconic trace, which persists for only a few hundred milliseconds before detailed visual information becomes inaccessible. This produces a characteristic decay curve, and the researchers fitted it with an exponential decay function containing three parameters: a₀, capturing the asymptotic performance that reflects information successfully transferred into short-term memory; a₁, capturing the initial availability of the stimulus representation at short delays; and τ, the time constant describing how quickly the trace decays. Sensitivity itself was quantified as d′ using a standard formula for eight-alternative forced-choice data, ensuring that criterion shifts could not masquerade as perceptual changes.

To isolate the effect of arousal, the team computed pre-stimulus pupil dilation as the average pupil size from 500 to 2 milliseconds before the display appeared, baseline-corrected against an earlier interval and z-scored within each participant. Trials containing blinks or gaze deviations greater than 2.5 degrees of visual angle were discarded, as were pupil values exceeding two standard deviations. The remaining single-trial values were split into three quantile bins—small, medium, and large dilation—and the decay model was fitted separately to the accuracy data in each bin. The predictions were clean: if arousal amplifies the initial neural response to the stimulus, a₁ should increase, improving performance mainly at short cue delays; if arousal slows the decay of the trace, τ should lengthen, improving performance at intermediate delays.

The results delivered a striking double dissociation—just not the one many might have expected. Larger pre-stimulus pupils were indeed associated with significantly higher initial stimulus availability: the a₁ parameter differed reliably across bins (F(2,72) = 4.07, p = 0.021), with post hoc tests showing higher values for large versus small dilation (p = 0.046) and for medium versus small dilation (p = 0.002). But rather than slowing the fade-out of the trace, heightened arousal accelerated it. The time constant τ was significantly smaller for large than for small pupils (p = 0.003; overall F(2,72) = 5.136, p = 0.008), meaning the iconic snapshot died away faster when observers were more aroused. Meanwhile, the asymptotic parameter a₀—which indexes what makes it into durable short-term memory—was completely unaffected by pupil size, and confidence ratings showed no modulation whatsoever. Arousal, in other words, acts specifically on the transient sensory echo of the stimulus, not on the stable memory store downstream.

A supplementary modeling analysis added nuance. When the researchers modeled the parameter values with linear and quadratic functions of pupil bin, the effect on a₁ was captured better by a quadratic model, with the improvement in fit approaching significance (χ²(1) = 3.84, p = 0.050)—a hint of the inverted U-shaped relationship between arousal and performance long predicted by theories of locus coeruleus function, in which intermediate arousal is optimal and extreme arousal degrades performance. The effect on τ, by contrast, was best described as linear (β = −0.051, p = 0.002), with no evidence of an optimum. The authors note that spontaneous pupil fluctuations in a seated laboratory task likely sampled only the lower and middle portions of the arousal–performance curve, never reaching the high-arousal territory where performance typically falls apart.

The combination of a stronger start and a faster fade might sound paradoxical, but the authors argue it reflects a fundamental trade-off in how perception handles time. Visual processing must balance temporal integration, which glues together information across successive moments to produce a stable world, against temporal segregation, which sacrifices stability to enable rapid updating when something new and potentially important happens. An aroused brain, the study suggests, tips this balance toward segregation. By amplifying the initial representation of a flash while hastening its decay, elevated arousal effectively narrows the temporal window over which old visual information lingers, clearing the stage faster for whatever comes next. Notably, the steepest performance decline in the data occurred between roughly 100 and 200 milliseconds, broadly consistent with the ~100-millisecond temporal integration windows measured in continuous stimulation paradigms—an intriguing, if tentative, hint that iconic decay places an upper bound on how long the visual system can blend successive inputs.

The findings also connect to a broader physiology of sensory persistence. Previous work has shown that emotional stimuli, whether positive or negative, enhance the initial readout of information from iconic memory, and since emotional images also trigger arousal, the new results suggest that spontaneous arousal shifts may partly underlie that effect. On the other side of the ledger, elevated cortisol—the stress hormone—has been shown to accelerate the decay of iconic traces, mirroring the fast-decay effect of dilated pupils observed here. And in an earlier study by the same team, spontaneous alpha-band oscillations in the EEG, another marker of cortical excitability, were found to extend the persistence of visual information without affecting decay. Together, these results sketch two distinct but potentially interacting mechanisms: alpha activity and pupil-linked arousal both boost the initial strength of the iconic trace, but only arousal changes how quickly it vanishes.

The absence of any effect on confidence is itself informative. In near-threshold detection tasks, larger pupils have been linked to inflated confidence, likely because arousal shifts the decision criterion—the internal threshold observers use to say “yes, I saw it.” But in the multi-alternative forced-choice task used here, confidence primarily tracks the strength of the selected representation and covaries with objective accuracy, leaving criterion shifts little room to operate. The fact that arousal altered accuracy without touching confidence in this paradigm strengthens the case that the effects are genuinely perceptual and mnemonic rather than artifacts of biased reporting. The authors acknowledge one residual caveat: a larger pupil admits more light onto the retina, a low-level optical factor that could contribute, though they argue that arousal remains the most plausible explanation given the task’s design and prior pupillometry literature.

What emerges is a compelling portrait of the brain’s arousal system as a temporal sculptor of experience. Moment to moment, without any instruction or awareness, fluctuations in norepinephrine signaling quietly decide how vividly a scene registers and how long its ghost lingers—prioritizing swift, up-to-the-millisecond updating over the comfort of a stable afterimage. The study was pre-registered as part of a larger project at the Open Science Framework, with data and analysis code publicly available, and the authors suggest that future work should relate these pupil effects to other physiological arousal markers such as skin conductance and respiration, which have each been tied to visual perception in their own right. For now, the message is clear: the fleeting world of iconic memory is not fixed. It breathes with your arousal, sharpening what you catch and shortening how long you hold it—an adaptive bargain that keeps perception fresh in a constantly changing world.