A recent study published in Nature Human Behaviour has found that certain visual properties of scenes — like their size, clutter, and memorability — can distort our perception of time. Specifically, larger and more memorable scenes appear to stretch time, while cluttered scenes seem to compress it.
The primary motivation behind this research was to explore a less-studied aspect of sensory perception: time. While time is integral to all sensory processes, its perception—how we gauge the passage of time and how it influences the processing of other sensory information—has remained somewhat enigmatic. This gap in understanding offers both a challenge and an opportunity to delve deeper into how time is encoded by the brain during sensory processing.
“My lab is interested in time perception generally; that is, how the brain measures and perceives intervals of time from hundreds of milliseconds to multiple seconds,” said study author Martin Wiener, an assistant professor of psychology at George Mason University. “Many previous studies have focused on how ‘simple’ visual stimuli, such as geometric shapes, dots, colors, etc. influence perceived time, whereas we wanted to look at ‘higher’ visual features like size, clutter, and memorability.”
The researchers designed four distinct experiments to investigate how various properties of images affect the perception of time.
In the first two experiments, participants engaged in a visual temporal categorization task, also known as a time bisection task, where they viewed images and decided whether each was presented for a ‘short’ or ‘long’ duration. The images for these experiments were sourced from the Size/Clutter database, a collection of images rated for scene size and clutter by previous participants. Experiment 1 utilized the images as provided, while Experiment 2 processed these images to grayscale and normalized their luminance to control for visual brightness and contrast effects.
The timing of the image presentation varied across six logarithmically spaced intervals from 300 to 900 milliseconds. Each participant viewed each image at each interval once, leading to a substantial total of trials per participant, interspersed with breaks. The order of images was randomized to prevent ordering effects from influencing the results. Participants were instructed to respond as quickly and accurately as possible, pressing designated keys to indicate if the image duration felt ‘short’ or ‘long’. Notably, they received no feedback on their performance to avoid influencing subsequent responses.
The results from the first two experiments demonstrated that scene size and clutter have significant but opposing effects on perceived duration. Larger scene sizes consistently caused participants to perceive the durations as longer (time dilation), while increased clutter led to the perception of shorter durations (time contraction).
The replication effort in Experiment 2 further confirmed the impact of scene size and clutter on time perception. This suggested that the effects were not merely due to color or brightness, which might otherwise draw more attention and skew perception.
Experiment 3 replicated the structure of the first two experiments but focused on the memorability of images. Images were drawn from the Large-Scale Image Memorability dataset (LaMem), which contains thousands of images rated for memorability — a measure of the likelihood that an image will be remembered. Images were selected to represent a range of memorability scores and were presented across seven log-spaced durations similar to the first two experiments.
Participants again participated in a time bisection task, categorizing each image duration as “short” or “long.” As with the previous experiments, no feedback was provided, and images were shown in a randomized order to maintain experimental integrity.
Wiener and his colleagues found that images with higher memorability scores were perceived to last longer than less memorable images. Additionally, these memorable images were not only perceived for longer durations but were also categorized as “short” or “long” with greater precision.
The final experiment required participants to engage in a duration reproduction task. Here, they viewed an image for one of seven linearly spaced time intervals (ranging from 500 to 1000 milliseconds) and then attempted to reproduce the duration by pressing and holding a button for what they felt was the equivalent amount of time. This part of the experiment aimed to assess how the memorability of an image might affect the accuracy of temporal reproduction.
After completing this task, participants returned the following day for a surprise memory recall test. They were shown the same images from the duration reproduction task, mixed with an equal number of new images, and asked to indicate whether they remembered seeing each image the previous day. This part of the study was designed to explore how the duration for which an image is remembered (or perceived to be remembered) influences actual memory recall.
Participants who perceived memorable images as lasting longer were more accurate in reproducing those longer durations and subsequently better at recalling those images. These results support a bidirectional relationship between time perception and memory: images perceived as lasting longer tend to be better remembered, and conversely, more memorable images are perceived to last longer.
This experiment highlighted the potential adaptive advantages of this perceptual mechanism. By extending the perceived duration of memorable images, the brain may enhance the encoding and subsequent recall of important or impactful information.
The findings indicate “that we have a ‘visual’ sense of time which can be warped or altered by what we’re looking at, and that this may allow the brain to gather more information when necessary,” Wiener told PsyPost. “Time has traditionally been seen as a ‘byproduct’ of perception. We integrate our senses (vision, hearing, touch), retain memories, form predictions, and then ‘time’ spills out of that process. Our work suggests time might be built at a much earlier stage in the process and for each sense independently (visual time, auditory time, etc.).”
To further investigate and explain the interaction between memorability of images and the perception of time, the researchers also utilized a recurrent convolutional neural network (rCNN) model, specifically designed to simulate the dynamics within the human ventral visual stream. This neural network, named “BLnet,” included layers with both forward and feedback connections, allowing the model to “unroll” over time, mimicking brain-like recurrent processing.
The neural network model revealed that more memorable images were processed faster than less memorable ones. The entropy of the softmax outputs — measuring the uncertainty or spread of possible categorizations — also decreased more quickly for more memorable images. This suggests a higher certainty in categorizing these images faster, which aligns with human data showing that more memorable images are perceived as lasting longer and with greater precision.
In other words, the results demonstrated that “a model of the visual system can explain these effects by the speed at which it processes the images,” Wiener explained.
The study provides unique insights into how visual properties influence time perception. However, like any research, it has its limitations that should be considered when interpreting the results. One primary limitation is the study’s relatively small participant pool (170 participants in total), which consisted mainly of undergraduate students from a single university. This demographic may not represent broader populations, potentially limiting the generalizability of the findings.
Future research could also expand on the neural network modeling used in the study. While the rCNN provided a mechanistic explanation for the findings, verifying these models against actual neurological data would strengthen the conclusions.
“We’d like to scale this up to more subjects to see how robust the findings are,” Wiener said. “We’d also like to use neuroimaging and brain stimulation technologies to get a better idea of the mechanism by which time dilation effects occur, and if we can alter it. Further, even though this study provides an answer to how more memorable images are dilated, it doesn’t say why. This will require more work and theory development.”
The study, “Memorability shapes perceived time (and vice versa),” was authored by Alex C. Ma , Ayana D. Cameron, and Martin Wiener.
