Cognitive Connie
Understanding sperling — iconic memory
In 1960 George Sperling published one of the most elegant experiments in cognitive psychology. He showed participants a brief flash (50 ms) of a 3×4 grid of letters, then measured how many they could report. Whole-report accuracy averaged only 4–5 letters — implying poor memory. But participants consistently said they could "see" most of the grid during the flash. Sperling tested this claim with a deceptively simple change: instead of asking for all letters, he played a tone immediately after the flash cueing participants to report just one row. Accuracy jumped to near-perfect — they could report any row with ~76% accuracy, suggesting they had access to roughly 9–12 items. The trick was timing: as the delay between flash and tone increased beyond ~300 ms, the advantage evaporated.
Defining features
Key figures
Ulric Neisser
1928–2012Coined the term "iconic memory" in his influential 1967 textbook Cognitive Psychology, which synthesised Sperling's and related findings into a comprehensive model of early visual processing. Neisser's terminology gave the phenomenon its lasting name and placed it within the emerging cognitive psychology framework.
George Sperling
1934–American cognitive psychologist who devised the partial report paradigm and provided the first rigorous demonstration of the visual sensory buffer (iconic memory). His 1960 Psychological Monographs paper is one of the most methodologically precise and impactful studies in the history of cognitive psychology.
Methods
The whole-report and partial-report paradigm
Sperling built two conditions to compare — and the contrast between them was what revealed the sensory store.
Whole-report baseline
Participants viewed a 3×4 grid of letters for 50 ms, then wrote down as many as they could. Average recall: 4–5 of 12 letters (~36%). This seemed to suggest poor perception. But participants reported subjectively seeing more — the grid seemed clear during the flash but faded as they tried to recall it.
Partial-report technique
The key innovation: immediately after the grid disappeared, a tone indicated which row to report — high tone for top row, medium for middle, low for bottom. Participants did not know in advance which row would be cued. Row recall accuracy: ~76% (≈ 3.3 of 4.4 letters per row). Since the cue selected any row at random, this implies ~9–12 items available across the whole grid — far more than the 4–5 reported in whole-report.
Delay manipulation
Sperling systematically varied the delay between the grid offset and the tone cue: 0 ms, 150 ms, 300 ms, 500 ms, and 1,000 ms. Partial-report superiority declined progressively with delay and disappeared by ~500 ms — showing that the information was real but transient. By the time participants began reporting all items in whole-report, the rest of the buffer had decayed.
Pre-cue control condition
Sperling also tested partial report where the tone cue was presented before the grid (pre-cue). Performance was good but no better than delayed post-cue conditions, ruling out the alternative that the tones simply guided selective attention during the display. The superiority was specific to the immediate post-cue, confirming a persisting sensory trace rather than a perceptual selection artefact.
Findings
What Sperling discovered
High-capacity sensory store
The visual system briefly holds near-complete information about the scene — approximately 9–12 items from a 12-item grid — for ~150–500 ms after stimulus offset. This far exceeds the 4–5 items accessible to conscious report, showing that early visual processing captures more than attention can select before the trace fades.
Rapid decay (~500 ms)
The partial-report advantage decayed sharply between 0 and 500 ms post-stimulus. By 1,000 ms, partial-report performance equalled whole-report — the buffer had fully decayed. The icon is therefore an extremely brief but very rich representation.
Bottleneck is attention, not perception
The iconic store holds far more than can be transferred to STM. The limit on conscious awareness is attentional selection from the icon, not perceptual capacity. This finding became foundational for all subsequent attention research: perception is not the bottleneck — attentional transfer is.
Criticism
Limitations and later challenges
Is iconic memory really pre-categorical?
Sperling assumed the icon stores raw visual features before categorisation. But later work (Merikle, 1980; Coltheart, 1980) showed partial-report advantages for category cues (e.g., "report all letters" vs. "report all numbers"), suggesting some categorical processing occurs within the iconic store. The debate over how much semantic content the icon carries continues.
Ecological validity of 50 ms flashes
Real-world vision involves continuous saccadic eye movements and changing scenes. A brief artificial grid has no natural equivalent. Subsequent research on change blindness and inattentional blindness suggests that iconic memory is not straightforwardly preserved across saccades, complicating the generalisability of Sperling's results to natural perception.
Individual differences largely unexplored
Sperling used a very small number of participants (predominantly himself and one or two others in early pilots). Individual variation in iconic memory capacity and decay rate — now known to differ with age, cognitive load, and clinical status — was not addressed.
Relevance for today
Why it still matters
Foundation of the multi-store model
Sperling's sensory store became the "sensory register" in Atkinson and Shiffrin's (1968) multi-store model — the first stage in the most influential model of human memory. Without the empirical demonstration of a high-capacity, rapidly-decaying visual buffer, the multi-store model's architecture would lack its key first stage.
Change blindness research
The question of why we do not perceive change during saccades or flicker — despite Sperling showing we briefly hold a full visual representation — motivated an entire line of change blindness research. It turns out the iconic memory is largely reset at each saccade, which explains why we fail to notice changes across fixations despite the richness of the momentary icon.
Attention and visual working memory
Sperling's key insight — that the bottleneck is attentional selection from a rich sensory buffer — is now the foundation of visual working memory research. Contemporary models (Luck & Vogel, 1997; Cowan, 2001) distinguish the iconic sensory buffer from visual working memory and study how items are selected from one into the other.
Test your knowledge
Frequently asked questions
What is the partial report technique?+
A method where participants are cued (after the stimulus disappears) to report only a subset of items — for example, one row from a letter grid — rather than everything they saw. Because the cue arrives after the display, participants must have held the uncued items in a brief sensory buffer to report the correct row. The partial-report advantage over whole-report reveals how many items the buffer holds.
What is iconic memory?+
A very brief (≤500 ms), high-capacity (≈9–12 items) sensory buffer that holds a copy of the visual scene after stimulus offset. It is modality-specific (visual only), pre-attentional in the sense that more is held than attention can select, and its contents are available for attentional read-out but decay too rapidly for most items to reach conscious awareness or short-term memory.
Sources
Last reviewed August 2026- 1.
Sperling, G. (1960). The information available in brief visual presentations. Psychological Monographs: General and Applied, 74(11), 1–29. https://doi.org/10.1037/h0093759
+About this source
Original report of the whole-report and partial-report experiments; introduces the concept of a high-capacity, rapidly-decaying visual sensory store.
- 2.
Neisser, U. (1967). Cognitive Psychology. Appleton-Century-Crofts.
+About this source
Coined the term "iconic memory" and integrated Sperling's findings into the broader cognitive psychology framework.
- 3.
Luck, S. J., & Vogel, E. K. (1997). The capacity of visual working memory for features and conjunctions. Nature, 390, 279–281. https://doi.org/10.1038/36846
+About this source
Established the ~4-item capacity limit of visual working memory — the system that receives input from the iconic buffer — and distinguished it from the iconic store itself.