Visual memory encompasses the set of processes we use to retain and access visual information across time, from fleeting glimpses that last milliseconds to detailed memories that persist for years. Cognitive scientists generally consider visual memory to have three interconnected but distinct stages: iconic memory maintains a brief, high-capacity representation of the entire visual field; visual working memory actively maintains a limited set of visual details through attentionally demanding maintenance processes; and visual long-term memory stores vast amounts of visual information with remarkable precision and capacity. Visual memory supports behaviors we do every day, like navigating through environments, recognizing objects and scenes, and remembering faces and expressions during social interactions. Its power is dramatically illustrated by memory experts who can use visual imagery techniques to achieve extraordinary feats of memory recall, such as memorizing the exact order of 48 decks of cards in under an hour (the Guinness world record). Understanding the basis of human visual memory capacity—and how each stage of memory can be influenced by factors like attention, meaning, and even verbal information—can reveal fundamental principles about how brains process and reconstruct visual experience over time.

History

Research on visual memory took off in the 1960’s when George Sperling demonstrated that observers briefly maintain access to much more visual information than they can consciously report, suggesting a large-capacity but rapidly decaying iconic memory. Building on George Miller's (1956) earlier work, researchers like Baddeley and Hitch (1974) distinguished this brief iconic store—which appears to have a nearly perfect visual snapshot of the world available—from working memory, which involved active maintenance processes through what they called a specialized visuospatial sketchpad and which has severe capacity limitations [see Working Memory].

Simultaneously, researchers were discovering the remarkable capacity of visual long-term memory. Pioneering studies by Roger Shepard (1967) and Lionel Standing (1973) demonstrated that people could remember thousands of pictures, suggesting people have a virtually unlimited visual memory capacity even when given just a few seconds to focus on encoding each image.

Early research also revealed that visual memory was not simply a passive recording system. Frederic Bartlett's (1932) early work on reconstructive memory showed that all memory involves active reconstruction rather than passive retrieval. For example, labeling the same image a barbell versus a pair of glasses results in different subsequent memories and drawings (Carmichael et al., 1932). Loftus's research on eyewitness testimony reinforced this insight, demonstrating that post-event verbal information could systematically alter visual memories of events (Loftus, 1975).

Core concepts

Visual memories undergo continuous transformation from the moment of perception. For a short time, the visual system maintains a highly detailed representation of the entire visual field, as demonstrated by people’s ability to report any cued item from a briefly presented display (Sperling, 1960). This rich but fleeting representation begins to decay within fractions of a second unless actively maintained through attention-like maintenance processes (Coltheart, 1980).

The transition from this initial representation to more stable forms of memory involves active maintenance processes. Attention acts as a gateway, determining which aspects of the initial representation will be maintained (Awh & Jonides, 2001; Chun et al., 2011) [see Attention]. Capacity-limited working memory maintenance processes can then preserve a limited subset of information with high precision. However, memory performance falls off rapidly if people are asked to actively maintain more than just a handful of objects actively in visual working memory (Bays et al., 2024; Zhang & Luck, 2008).

Memories do not simply transform into less precise versions over time, however. Visual long-term memory can be remarkably precise; for example, people can almost perfectly recall the color of familiar brand logos (Miner et al., 2020). This initially seems difficult to reconcile with working memory's severe capacity limitations and the findings of people’s remarkable inability to remember information about scenes even over short timescales, like change blindness (Simons & Levin, 1997). However, visual long-term memory seems to operate through fundamentally different processes. Although working memory actively maintains limited information over brief periods, long-term visual memory formation involves extended viewing that gradually accumulates detailed information about the objects we attend (Hollingworth, 2004). Encoding items for even a few seconds each enables storage of thousands of objects or scenes with remarkable detail (see Figure 1; Brady et al., 2008; Konkle et al., 2010), revealing that apparent contradictions reflect different mechanisms rather than a hierarchy of memory precision (Brady et al., 2024).

Questions, controversies, and new developments

Several key questions continue to drive research in visual memory. One central debate concerns the nature of capacity limitations in visual working memory, which has implications for understanding how the object-based nature of the visual world shapes what we remember. Although early research suggested discrete slot-based limits (e.g., that we can hold exactly four objects regardless of their complexity, as though we have four mental slots to insert the objects in; Luck & Vogel, 1997), more recent work has shown that people have much richer visual representations of items when holding in mind one object versus two (e.g., Wilken & Ma, 2004; Zhang & Luck, 2008), leading to a variety of new models in which cognitive resources must be split, and thus, precision decreases as more items are maintained (Bays et al., 2024).

The neural mechanisms supporting transformations from perception into visual working memory also remain actively debated. Sensory recruitment accounts propose that the same sensory regions involved in perception are used to maintain visual information in working memory via sustained neural activity (Adam et al., 2022). However, working memory maintenance may not always require sustained neural activity—instead, information may sometimes be maintained in activity-silent states (Stokes, 2015), and the sensory cortex may be nonessential for some visual working memory storage, with more abstract representations in the parietal and frontal cortex playing the dominant role (Xu, 2017) [see Visual Cognitive Neuroscience]. These debates highlight fundamental questions about how “visual” visual working memory representations actually are.

A growing area of investigation in visual long-term memory focuses on what makes some visual experiences more memorable than others. Research on memorability has revealed that certain visual features consistently make images more likely to be remembered across individuals (Isola et al., 2011). However, whether memorability is truly intrinsic to images (Bainbridge, 2019) or might emerge from their distinctiveness and semantic connections remains uncertain (Brady et al., 2024). Understanding the reliability patterns of memory across people is important because it has implications for educational design, advertising effectiveness, and more.

Broader connections

Visual memory research connects to numerous areas within cognitive science and beyond. For example, the discovery that visual working memory capacity limits predict performance on complex cognitive tasks has linked visual memory to broader theories of cognitive control and executive function (Cowan, 2001). Visual memory research has also informed educational theory, particularly the picture superiority effect and dual coding theory, which holds that people remember pictures better than words because they can store both verbal and visual memories of pictures but generally only store verbal memories of words unless engaged in active visual imagery (Paivio, 1991). Understanding visual working memory capacity has contributed to guidelines for managing cognitive load in multimedia instruction (Mayer & Moreno, 2003), although the translation from laboratory findings to classroom practice remains an active area of development. Research on constructive visual memory and post-event information effects has influenced legal procedures and expert testimony standards (Wells et al., 2020). Memory researchers have also worked to translate laboratory findings about visual recognition memory and confidence into guidelines for evaluating eyewitness reliability, focusing on distinguishing when visual memory is reliable and precise (and therefore can be trusted) and when it is not (Wixted & Wells, 2017).

Acknowledgments

T.F.B. was supported by National Science Foundation BCS-2141189 and National Science Foundation BCS-2146988.

Further reading

• Adam, K. C. S., Rademaker, R. L., & Serences, J. T. (2022). Evidence for, and challenges to, sensory recruitment models of visual working memory. In T. F. Brady & W. A. Bainbridge (Eds.), Visual memory (pp. 5-25). Routledge.

• Brady, T. F., Robinson, M. M., & Williams, J. R. (2024). Noisy and hierarchical visual memory across timescales. Nature Reviews Psychology, 3, 147-163. https://doi.org/10.1038/s44159-024-00276-2

• Schurgin, M. W. (2018). Visual memory, the long and the short of it: A review of visual working memory and long-term memory. Attention, Perception, & Psychophysics, 80, 1035-1056. https://doi.org/10.3758/s13414-018-1522-y